DOI
stringlengths 13
47
| Year
int64 2.01k
2.02k
| Journal
stringlengths 3
170
| Title
stringlengths 1
281
⌀ | Abstract
stringlengths 14
62.9k
⌀ | Full Text
stringlengths 22
5M
|
|---|---|---|---|---|---|
10. 1002/admi. 201300085
| 2,014
|
Advanced Materials Interfaces
|
Modulation of Biointeractions by Electrically Switchable Oligopeptide Surfaces: Structural Requirements and Mechanism
|
Understanding the dynamic behavior of switchable surfaces is of paramount importance for the development of controllable and tailor-made surface materials. Herein, electrically switchable mixed self-assembled monolayers based on oligopeptides have been investigated in order to elucidate their conformational mechanism and structural requirements for the regulation of biomolecular interactions between proteins and ligands appended to the end of surface tethered oligopeptides. The interaction of the neutravidin protein to a surface appended biotin ligand was chosen as a model system. All the considerable experimental data, taken together with detailed computational work, support a switching mechanism in which biomolecular interactions are controlled by conformational changes between fully extended (“ON” state) and collapsed (“OFF” state) oligopeptide conformer structures. In the fully extended conformation, the biotin appended to the oligopeptide is largely free from steric factors allowing it to efficiently bind to the neutravidin from solution. While under a collapsed conformation, the ligand presented at the surface is partially embedded in the second component of the mixed SAM, and thus sterically shielded and inaccessible for neutravidin binding. Steric hindrances aroused from the neighboring surface-confined oligopeptide chains exert a great influence over the conformational behaviour of the oligopeptides, and as a consequence, over the switching efficiency. Our results also highlight the role of oligopeptide length in controlling binding switching efficiency. This study lays the foundation for designing and constructing dynamic surface materials with novel biological functions and capabilities, enabling their utilization in a wide variety of biological and medical applications.
|
1. Introduction Materials and surfaces with stimuli-responsive properties, which mimic biology, 1 are being developed experimentally for biomedical applications. 2 Active and switchable surfaces comprising domains that confer tailored responsiveness are playing an important part in the development of tissue engineering scaffolds, 3 highly sensitive biosensors, 4 – 6 novel drug delivery systems, 7, 8 and highly functional microfluidic, bioanalysis, and bioseparation systems. 9 – 12 Physical stimuli, such as temperature, 13, 14 light, 15, 16 magnetic field 17 and electrical potential, 18 – 20 are able to alter and manipulate the surface properties and, thus, change and control function and activity of biomolecules on surfaces. Switchable self-assembled monolayers (SAMs) that can control biomolecular interactions using an electrical stimulus are particularly appealing because of their fast response times, ease of creating multiple individually addressable switchable regions on the same surface and low-driven voltage or electric field that are compatible with biological systems. 21 Electrically switchable SAMs have been demonstrated to display controllable switching properties that can modulate the interactions of surfaces with proteins, 18 – 20 DNA, 22, 23 and mammalian 21 and bacterial 24 cells. Particularly interesting are the electrically switchable oligopeptide mixed SAMs, which enable controlled protein interactions with surfaces. 20 These switchable mixed SAMs are composed of a two molecular components, (i) a positively charged 4-mer of lysine (K) that is functionalized at one end with biotin, which recognises the neutravidin protein, and at the other end with a cysteine (C), for binding to gold substrates (biotin-4KC), and (ii) an ethylene glycol-terminated thiol (to space out the oligolysine peptides). The salient feature of the oligolysine peptide mixed SAMs is that they exhibit protonated amino side chains at pH 7, providing the basis for the switching “ON” and “OFF” of the biological activity on the surface by an electrical potential. These SAMs have been shown to regulate the binding between the biotin ligand on the surface and neutravidin from solution. Although these SAMs have been demonstrated to be capable of switching upon exposure to an electrical potential and so possess unique advantages in terms of addressability, the nature of the conformational changes and the dynamics that regulate biomolecule activity are unknown. To design any reliable or predictable surface material based on such oligolysine molecular switch, it is essential to have a detailed understanding of the structural requirements and mechanism by which the oligopeptides on the surface alter the ligand function. This would allow the molecular architecture in a surface material to be optimized to maximise the performance and efficiency of the switching of biomolecular interactions. Toward these goals, we report here detailed studies performed by X-ray photoelectron spectroscopy (XPS), surface plasmon resonance spectroscopy (SPR) and molecular dynamic simulations on the oligopeptide:ethylene glycol-terminated thiol mixed SAMs. While SPR offers a system to retrieve information such as binding ability and binding switching efficiency, atomic molecular dynamic simulations provide molecular insights into the electrical-induced conformational changes of the oligolysines within the SAM. One of the components of the SAM is the end functionalised biotinylated oligolysine peptide described above – biotin-4KC ( Scheme 1 ). A tri(ethylene glycol)-terminated thiol (TEGT) is used as the second SAM component. Scheme 1 Chemical structures of the oligopeptides (biotin-2KC, biotin-4KC and biotin-6KC) and ethylene glycol-terminated thiols used in the mixed SAMs, and their molecular lengths in fully extended conformations. Here we consider how the ratio of these two components can affect the binding switching efficiency. We follow with experiments and theoretical studies detailing the role that the length of the oligolysine has on the switching ability of the biotin ligand, whereby we incorporate onto the oligopeptide two less or two more lysine residues, relative to biotin-4KC, namely, biotin-2KC and biotin-6KC (Scheme 6 ). 2. Results and Discussion In the search for an understanding of the relationship between oligopeptide composition in the mixed SAM and its switching ability, biotin-4KC:TEGT mixed SAMs on gold were prepared at different solution ratios ranging from 1:1 to 1:500 ( Table 1 ). X-ray photoelectron spectroscopy (XPS) confirmed the formation of the mixed SAMs, showing signals from C (1s), O (1s), S (2p) and N (1s) (Figure S1 in the Supporting Information). By integrating the area of the S (2p) and N (1s) peaks for the mixed monolayers (the biotin-4KC oligopeptide consists of 11 N atoms and 2 S atoms, whereas TEGT has no N and 1 S atom only), we were able to calculate the ratio of biotin-4KC to TEGT on the surface. The samples were found to be reproducible. As expected an increased amount of TEGT in the mixed solution has led to an increase of this component in the mixed SAM. Nevertheless, as reported previously for other mixed SAM systems, 25 – 27 the composition of the mixed solution does not directly equal the composition of the mixed SAMs on the surface. Correlation analysis indicated a logarithm relationship between the biotin-4KC:TEGT solution and surface ratios, with the biotin-4KC being significantly enriched in the mixed SAM in comparison to its solution composition. This has been a general trend observed when thiols with different chain lengths have been used to form monolayers, in which the longer chain component is preferentially adsorbed, suggesting a predominantly thermodynamic control of the adsorption. 25, 26 Table 1 Biotin-4KC:TEGT solution ratios and respective surface ratios calculated using XPS. Binding capacity and switching efficiency as determined by SPR analysis Biotin-4KC:TEGT ratio Binding capacity (RU) Switching efficiency (%) Solution Surface 1:0 1:0 3553 ± 258 7 ± 2 1:1 1:3 ± 3 3192 ± 164 27 ± 3 1:10 1:5 ± 2 3053 ± 69 34 ± 5 1:40 1:16 ± 4 2195 ± 161 90 ± 3 1:100 1:22 ± 8 1492 ± 72 62 ± 8 1:500 1:38 ± 6 1375 ± 75 60 ± 4 We next assessed the binding capacity and switching efficiency of the biotin-4KC:TEGT mixed SAMs, by analysing the binding events between the biotin ligand on the mixed SAM and neutravidin using SPR ( Figure 1 ). Neutravidin is a protein that consists of four identical subunits, each binding one biotin with extremely high affinity (K a ≈ 10 13 M −1 ). 28 In the SPR experiments, the mixed SAMs were exposed to a flow of PBS, to establish the baseline, followed by an injection of neutravidin in PBS into the SPR flow cell for 30 min. The SPR flow cell was then flushed with PBS to leave only the specifically bound neutravidin on the biotinylated mixed SAM. The binding capacity is defined as the difference in the SPR response units between the beginning of injection of protein and the end of washing with PBS. Figure 1 SPR sensorgram traces showing the binding of neutravidin (37 μg. mL −1 ) to the biotin-4KC:TEGT mixed SAMs at solution ratios of 1:0, 1:1, 1:10, 1:40, 1:100 and 1:500 under open circuit conditions (no applied potential), an applied positive (+0. 3 V) and negative (−0. 4 V) potential. The biotinylated mixed SAMs exhibited a high protein binding capacity, which upon a reduction of the amount of biotinylated peptide on the surface decreased significantly. Nevertheless, the reduction in the binding capacity was not proportional to the decrease in the amount of biotinylated peptide on the surface, indicating that steric hindrance and limited mobility of the densely packed biotin molecules might limit the protein binding to the biotin sites at low ratios of TEGT to Biotin-4KC. 29 It is important to note that the binding capacity is also dependent on the length of the ethylene glycol thiol. Experiments conducted on mixed SAMs comprising the Biotin-4KC and a longer ethylene glycol thiol— HEGT in Scheme 6 —has led to a greatly reduced binding of neutravidin to the biotinylated surface. Biotin-4KC:HEGT solution ratios were varied between 1:10 and 1:100, leading to surface ratios between 1:9 ± 4 and 1:19 ± 4, respectively. The neutravidin binding amount was essentially independent of the surface ratio used, with SPR signals in the range of 275–325 response units for all the surfaces. Taking into consideration that the lengths of the biotin-4KC and HEGT, in fully extended conformations, are 5. 2 nm and 3. 8 nm, respectively, to a certain extent the biotin functionalities are expected to protrude from a matrix of HEGTs even if most likely both molecules adapt a rather unstretched form on the surface. Nevertheless, and based on the above-mentioned SPR results, there is strong evidence that the biotin moieties are not accessible for binding. We propose therefore that the suppression of biorecognition with the biotin-binding pockets of neutravidin is a result of the biotin moieties not standing further away from the HEGT matrix, thus not allowing complete insertion of the biotin into the binding pockets. This reasoning is in line with previous studies that showed that increasing the length of the biotin linker in a mixed SAM increased the protein binding efficiency. 30 Our hypothesis is also consistent with X-ray crystallographic analysis that revealed that the biotin is buried quite deeply inside the neutravidin barrel, 31 indicating that the binding of biotin by neutravidin requires the complete insertion of the ligand into the binding pocket of the protein. TEGT in a fully extended conformation exhibits a length of 1. 7 nm, approximately three-fold shorter than the biotin-4KC, allowing complete insertion of the biotin into the binding pocket and efficient binding of the neutravidin to the biotinylated monolayer. Switching efficiency studies on the Biotin-4KC:TEGT mixed SAMs were conducted by monitoring, using electrochemical SPR spectroscopy, neutravidin binding to the biotinylated SAM to which a positive or negative potential was applied ( Figure 1 ). Previously, we have demonstrated that the bioactivity of oligopeptide mixed SAMs can be controlled by application of +0. 3 V (bioactive “ON” state) or –0. 4 V (bio-inactive “OFF” state), while not affecting the SAM integrity. Thus, similar electrical potentials were used in these studies. While applying +0. 3 V or –0. 4 V, the baseline for the SAM gold chip was established using PBS, following which the neutravidin was introduced. Data were collected for 30 min, after which the surface was rinsed with PBS. The switching efficiency was calculated by dividing the difference in binding capacity between +0. 3 V and –0. 4 V by the binding capacity at +0. 3 V. To begin with, we investigated the binding efficiency of the pure biotin-4KC monolayers ( Figure 1 ). There were, however, no significant changes in response units observed in both +0. 3 V and –0. 4 V compared to that of the OC conditions, indicating that regulation of biomolecular interactions is not possible with a very high density of biotinylated oligopeptides on the surface. At progressively lower densities of the oligopeptide, the application of +0. 3 V or –0. 4 V started to result in some control over the binding of the protein to the biotin ligand presented at the surface. Nevertheless, the binding switching efficiency was shown to be limited at higher biotin-4KC:TEGT solution ratios, i. e. <1:10 (Table 1 and Figure 2 ). These results illustrated that the oligopeptide should be presented at optimum ratio on the surface such that binding capacity and switching efficiency can be maximised. For the biotin-4KC:TEGT mixed monolayers, the optimum surface ratio was identified as being in the order of 1:16, i. e. a mixed ratio from solution of 1:40 as depicted in Figure 2. Figure 2 The binding capacity under open circuit conditions, an applied positive (+0. 3 V) and negative (−0. 4 V) potential as well as switching efficiency of the biotin-4KC:TEGT mixed SAMs at solution ratios of 1:0, 1:1, 1:10, 1:40, 1:100 and 1:500. To rationalize the binding and switching properties of the biotinylated mixed SAMs at different ratios, we propose that the activation/inactivation switching mechanism is related to conformational changes of the oligopeptide on the surface induced by an electrical potential, involving first among all reorganization of the biotin moiety on the surface. Interestingly, the lack of any binding switching ability in the pure biotin-4KC supports the hypothesis that there is a large steric hindrance around each oligopeptide, restricting conformational changes from taking place. As in the case of pure biotin-4KC, high ratios of biotin-4KC to TEGT limit the conformational change effects triggered by an electrical potential. The switching efficiency increased significantly ( Figure 2 ) as the proportion of biotin-4KC in the mixed monolayer decreases, reflecting the availability of local free volume for the conformational switching of the oligopeptides to occur on the gold surfaces. Interestingly, it is the fact that at low biotin-4KC:TEGT solution ratios, i. e. , 1:100 and 1:500, the switching efficiency is lower than that at solution ratio of 1:40. This behaviour might be attributed to the formation of a more organized EG matrix on the mixed SAMs prepared from solutions containing low concentrations of the oligopeptide. The presence of a more packed EG matrix might restrict the oligopeptide mobility and conformational flexibility. Another point of note is that these suggested conformational changes are not only taking place at –0. 4 V but also at +0. 3 V, since the binding capacity of the latter was shown to be superior to that of open circuit (OC) conditions (i. e. no applied potential). From this observation, and the fact that the gap distance between the biotin moiety and the EG matrix is an important factor in determining the accessibility of biotin for neutravidin binding, as discussed above, we may infer that the conformational changes might induce: i) gap distance variations between the biotin and the EG matrix, and/or ii) changes in the orientation of the biotin that could make it more or less favourable for binding to neutravidin. We consider these two points in greater detail when we discuss below the results of the molecular dynamic simulations. Next, we investigated the switching properties of the oligopeptide mixed SAM system as a function of the length of the oligolysine chain. A shorter (biotin-2KC) and a longer (biotin-6KC) oligopeptide than biotin-4KC were chosen as model systems (Scheme 6 ). Based on the switching studies performed on the biotin-4KC:TEGT mixed SAMs at different surface ratios, two oligopeptide:TEGT surface ratios were selected, namely 1:5 and 1:16. Apart from evaluating the switching abilities of the biotin-2KC and biotin-6KC, the use of an underperformed (1:5) and an optimum ratio (1:16) found for the biotin-4KC system allowed us to: i) evaluate the oligopeptides behavior in direct comparison and ii) unveil whether there is a relationship between the length of the oligolysine molecular switch and the molecular area that it requires for the switchable conformational changes to occur. Mixed SAMs of different solution concentration ratios of Biotin-2KC or Biotin-6KC and TEGT, ranging from 1:10 to 1:2000, were prepared and analyzed by XPS (Figure S2). Following a similar procedure as described for the biotin-4KC, we were able to determine that ratios of 1:40 and 1:100 of biotin-2KC:TEGT mixed solution were required to provide biotin-2KC:TEGT surface ratios of 1:6 ± 1 and 1:16 ± 2, respectively. In a similar manner, the XPS studies allowed us to establish that ratios of 1:40 and 1:2000 of biotin-6KC:TEGT mixed solution are needed to achieve biotin-6KC:TEGT surface ratios of 1:7 ± 2 and 1:17 ± 2, respectively. The binding capacity and binding switching efficiency for these four mixed SAM surfaces are summarized in Table 2 and Figure 3. A comparison of this experimental data with those obtained for the biotin-4KC:TEGT monolayers at same surface ratios, indicates that the binding capacity for the biotin-4KC and biotin-2KC systems are similar, while it is significantly impaired in the biotin-6KC:TEGT monolayers at a surface ratio of ≈1:16. We believe that such big difference in binding capacity is related to the unfavourable orientation of the biotin moiety in the biotin-6KC:TEGT mixed SAMs at low densities, caused by the long and flexible nature of the biotin-6KC. Table 2 Binding capacity and switching efficiency as determined by SPR analysis for biotin-2KC:TEGT and biotin-6KC:TEGT at different surface ratios Ratio Binding capacity (RU) Switching efficiency (%) Solution Surface Biotin-2KC:TEGT 1:40 1:6 ± 1 2634 ± 183 74 ± 4 1:100 1:16 ± 2 2295 ± 87 67 ± 8 Biotin-6KC:TEGT 1:40 1:7 ± 2 2822 ± 129 71± 7 1:2000 1:17 ± 2 229 ± 58 7 ± 4 Figure 3 SPR sensorgram traces showing the binding of neutravidin (37 μg. mL −1 ) to the biotin-2KC:TEGT mixed SAMs (solution ratios of 1:40 and 1:100) and biotin-6KC:TEGT mixed SAMs (solution ratios of 1:40 and 1:2000) under open circuit conditions, an applied positive (+0. 3 V) and negative (−0. 4 V) potential. In contrast to the low binding switching efficiency for the biotin-4KC:TEGT mixed SAM at a surface ratio of ≈1:5, the biotin-2KC:TEGT at the same surface ratio already demonstrated a great regulation of biomolecular interactions. This difference between the two oligopeptides can be rationalized by considering their molecular lengths (Scheme 6 ), and the need for a greater local free volume in the biotin-4KC:TEGT for the conformational switching of the oligopeptide to occur on the gold surface. Remarkably, the switching efficiency for the biotin-6KC:TEGT at a ≈1:5 surface ratio is also higher than for the biotin-4KC:TEGT, suggesting that for the longer oligopeptide (biotin-6KC) the conformation changes may induce intercrossing between oligopeptide chains leading to lower accessibility of biotin for neutravidin binding. In the case of the biotin-6KC:TEGT at a ≈1:16 surface ratio, the switching efficiency is significantly lowered relative to the two other systems, biotin-2KC:TEGT and biotin-4KC:TEGT, indicating that the length of the lysine switching unit exert great influence in the surface ratio that could provide maximum binding capacity and switching efficiency. These findings also suggest that a long switching unit can place constraints on the rearrangement of the biotin moiety on the surface, such that it can be available or not for neutravidin binding. In order to gain insight into the mechanism of the switching behavior, we performed molecular dynamics (MD) simulations. The performance of MD simulations depends mainly on the force field selected, and thus we tested three different force fields, namely cvff (consistent-valence force field), compass (condensed-phase optimized molecular potentials for atomistic simulation studies) and pcff (polymer consistent force field, see Supporting Information for details). The cvff force field performed best according to the test, and thus it was adopted in our simulations. The simulation models are shown in Scheme 2. Two dimensional rhombic periodic boundary condition and slab models were applied throughout our simulations. Water molecules and chloride ions were adopted to simulate the PBS solution. Detailed model parameters are summarized in Table S1. External electric fields were applied to model the electric potentials used in the experiment. In order to consider the polarization caused by the electric field, density functional theory-derived partial charge was used. We carried out simulations for all the above-mentioned systems, including biotin- n KC:TEGT ( n = 2, 4, 6), biotin-4KC:HEGT and pure biotin-4KC. Scheme 2 The surface models used in the MD simulations. The purple, blue and dark green parts of the biotin- n KC chain represent the biotin motif, lysine and cysteine residues, respectively. The orange dots, light green balls, yellow balls and short grey chains denote water molecules, chloride ions, gold atoms and TEGT, respectively. The results are shown in Figures 4 – 5 and detailed MD simulation snapshots are listed in the Supporting Information. For the biotin-2KC:TEGT (surface ratio 1:8) and biotin-4KC (surface ratio 1:15), an evident switching behavior is observed. The oligopeptide chain extended fully and the biotin head (the purple part of the long central chain) was totally exposed when an electric field E z was applied, corresponding to the “ON” state. In contrast, when E -z was applied, the chain adopted a collapsed conformation and the biotin head was partially concealed by TEGT chains, thus showing no bioactivity (“OFF” state). The “OC” state took an intermediate conformation and the intercrossing between oligopeptide chains was more probable to occur, which would lower the chance of binding to neutravidin and result in moderate bioactivity. Figure 4 The conformational changes of biotin-2KC, biotin-4KC and biotin-6KC under different electric fields, along with the MD simulation snapshots. The black arrows show the directions of the applied electric fields. Water molecules and hydrogen atoms are omitted for clarity. d is defined as the gap distance variation between the biotin and TEGT matrix. Figure 5 The conformational changes of biotin-4KC:HEGT (up) and pure biotin-4KC (down) under different electric fields, along with the MD simulation snapshots. L is defined as the gap distance variation between the biotin and gold surface. These theoretical results are in fair agreement with the data obtained with the SPR experiments for the optimum ratios of biotin-2KC:TEGT and biotin-4KC:TEGT, which clearly revealed high binding at +0. 3 V, reduced binding at −0. 4 V, and intermediate binding in the OC conditions. The electrostatic interaction between the applied electric field and the positively charged lysine side chains, whose amino groups were fully protonated, is proposed to be the main driving force for these switchable conformational changes. From Figure 4, we can see that in the 1:15 biotin-6KC:TEGT SAM the biotin-6KC chain was so long that even in the folded conformation the TEGTs could not conceal the biotin head. Another feature of note in Figure 4 is that, when an electric field E z was applied to the biotin-6KC:TEGT mixed SAM, the biotin moiety is highly available for binding. This can be seen from the gap distance variation between the biotin and TEGT matrix ( d in Figure 4 ), which is more than 1. 5 nm. These results agree with the experimental data and confirm that the use of a long switching unit is not ideal for controlling ligand activity under an electrical stimulus. In the case of biotin-4KC:HEGT ( Figure 5 ), the ethylene glycol chains were long enough to cover partially the biotin in the OC condition ( d < 0. 5 nm). The biotin-4KC chain would extend to about 5. 2 nm and reach neutravidin only when the E z field was applied ( d > 1. 4 nm). This finding supports the interpretation that low neutravidin binding to the biotin-4KC:HEGT mixed SAM is an effect related with the biotin moieties standing to close to the ethylene glycol matrix, sterically shielding the biotin and making it inaccessible to the neutravidin. This hypothesis is also consistent with the conformational structures obtained for the biotin-2KC and biotin-4KC under E -z and the low experimental binding obtained under a negative potential. The intercrossing between oligopeptide chains was more probable to occur in the OC condition, which would lower the chance of binding to neutravidin. Therefore, its bio-activity was lower than that in the ON condition. Thus, we now have a rationale for explaining how the switching mechanism present in an oligopeptide:TEGT mixed SAM can control binding capacity on the surface. At this point, it is of interest to ask how oligopeptide density can affect the switching mechanism, and as a consequence, the binding switching efficiency. From Figure 5, it is noticeable that for the pure biotin-4KC SAM, the chains were closely packed on the surface and not sufficient space was left for the chains to collapse. The biotin heads were always exposed, leading to a persistent bioactivity, an observation that is consistent with the similar binding capacities at OC conditions, and applied negative and positive potential. We can infer from these findings that a basic criterion in the design of the switching surfaces is to provide sufficient freedom for conformational transitions of surface confined oligopeptide chains. In our MD simulation, the electric field we adopted is an electrostatic field and no current was involved in the system. In the experiment, however, a circuit was formed and current was observed. The electrostatic field in our simulation was an approximation to the experimental electrical potential. The intensity of the adopted electric field was in the range of 10 0 V/nm, which is commonly used in such kind of electro-switchable systems. 3. Conclusion The switching mechanism on electrically switchable oligopeptide surfaces is based on conformational changes between collapsed (“OFF” state) and fully extended (“ON” state) oligopeptide structures. The principles behind the bioactivity OFF/ON switch can be understood in terms of the proximity of the biotin ligand to the EG matrix. When the oligopeptide is in its collapsed conformation, the biotin moiety draws closer to the EG matrix, hindering molecular recognition with the biotin in the binding pockets of neutravidin. In contrast, in the fully extended conformation, the biotin is largely free from steric hindrance effects and is able to efficiently bind to neutravidin. Our experimental and computational results strongly suggest that steric hindrances aroused from the neighboring surface-confined oligopeptide chains exert a great influence over the conformational behaviour of the oligopeptides, and as a consequence, over the binding switching efficiency. Our results also highlighted the relevance of the length of the switching unit to ensure maximum binding switching efficiency. Equipped with this kind of intimate understanding of the structure-switching property relationship, we are now better able to design and develop new dynamic surface materials for biomedical applications ranging in diversity from—but not limited to—in vitro model systems for fundamental cellular studies, all the way through to sophisticated drug delivery systems.
|
10. 1002/admi. 201400026
| 2,014
|
Advanced Materials Interfaces
|
Direct Observation of Reversible Biomolecule Switching Controlled By Electrical Stimulus
|
No abstract available
|
Control and reversibility of biomolecular interactions at engineered interfaces presents opportunities to develop highly efficient substrates and devices for a wide range of biomedical applications. 1 – 4 A major challenge nowadays in the field of stimuli-responsive interfaces is to acquire a molecular understanding of the changes occurring at the biointerface upon external stimulation. Herein, we used in situ Sum-Frequency-Generation (SFG) spectroscopy to study changes in molecular orientations in electrically switchable biofunctionalized self-assembled monolayers (SAMs). The bioactivity of a mixed SAM on gold consisting of a biotin-terminated positively charged oligopeptide (biotin-KKKKC) and a tri(ethylene glycol)-terminated thiol is shown to be related to a switch between upward exposure and random orientation of the biotin group in response to positive and negative applied potentials, respectively. The findings reported here support the mechanism by which charged biomolecules control biomolecular interactions, for example, protein binding affinities, and lay the foundation for future studies aiming to explore molecular conformational changes in response to electrical stimuli. Dynamic surfaces are particularly attractive for biomedical applications and are playing an increasingly important part in the development of highly sensitive biosensors, 5 – 7 novel drug delivery systems, 8 and tissue engineering scaffolds. 9 Stimuli-responsive SAMs have garnered much interest since they can provide a high level of molecular organization and control over the surface properties. 10 To date, stimuli-responsive SAMs have been able to selectively respond to external inputs such as electrical, 11 – 13 temperature, 14, 15 pH, 16 and light. 17, 18 Switchable SAMs used to control biomolecular interactions via an electrical stimulus are particularly appealing because of their fast response times, ease of creating multiple individually addressable switchable regions on the same surface, as well as low-driven voltage and electric fields that are compatible with biological systems. 19 Electrically switchable SAMs have been demonstrated to modulate the interactions of surfaces with proteins, 11 – 13 DNA, 20, 21 and mammalian 19 and bacterial 22 cells. For instance, an electrically switchable mixed SAM on gold that comprised a positively charged oligopeptide (biotin-KKKKC) and a shorter tri(ethylene glycol)-terminated thiol (TEGT) ( Figure 1 ) was previously 13 demonstrated by us to be able to control the bioactivity of biotin on the surface and its binding to a specific protein, Neutravidin. High protein binding was observed for an applied positive potential (+0. 3 V, bio-active state), while minimal binding was detected for an applied negative potential (-0. 4 V, bio-inactive state). Figure 1 Schematic representation of the biotin-KKKKC:TEGT SAM and chemical structure of the biotin-KKKKC oligopeptide and TEGT. While the charged molecular backbone or the end group on the structure of the reported electrically switchable SAMs are the hypothesized basis for controlling such biomolecular interactions, 23 the mechanistic principles underpinning these electrically-driven systems are not fully experimentally proven. Without such proven mechanistic detail, designing novel biologically relevant surfaces, and understanding the potential and limitations of present ones is haphazard at best. Herein, we address this challenge by studying the biotin-KKKKC:TEGT mixed SAM as a model system and report the first observation of biotin orientations in the charged mixed SAM in response to an applied potential, using in situ SFG spectroscopy. This technique takes advantage of even-order non-linear optical selection rules that precludes signals from isotropic environments or molecular arrangements that possess inversion symmetry. 24 At interfaces, the symmetry is necessarily broken making even-order processes, such as second-order SFG, intrinsically surface specific. 25 In a typical SFG experiment, IR and visible laser pulses are overlapped in time and space at an interface to eventually generate a sum-frequency signal. Spectra are recorded as a function of IR frequency; SFG signals are resonantly enhanced when the IR light is exciting vibrational states that are both IR and Raman active. In the realm of SAMs used in electrochemical settings, SFG spectroscopy has been applied to study chemical, orientational, and conformational changes within SAMs at electrified interfaces. 26 – 35 While the majority of these studies focus on the use of aliphatic and aromatic compounds, herein we use for the first time in situ SFG spectroscopy to investigate changes in molecular orientations in charged biofunctionalized SAMs in response to an applied electrical potential. This is particularly challenging due to the comparably lower density of charged biomolecules on the surface and a resulting greater degree of conformational freedom. Both effects decrease SFG signals, which are depending on density and order. The SFG characterization was performed with an easy to assemble and purpose built electrochemical cell that allowed recording of SFG spectra while applying an electrical potential. Furthermore, the conformational changes of the biotin-KKKKC:TEGT SAMs observed by SFG at negative and positive potential are compared with densely packed biotin-KKKKC SAMs to understand the importance of the presence of TEGT as a spacer group. SPR data are also discussed to further strengthen our findings. Finally, the reversibility of the conformational changes is verified. The switching properties of the mixed SAMs were investigated by applying an electrical potential as an external stimulus. SAMs of biotin-KKKKC:TEGT were formed directly onto a 15 nm gold coated side of an equilateral CaF 2 prism (size 25 mm). The prism was then placed onto a purpose built Teflon electrochemical cell ( Figure 2 ) containing PBS buffer (pH = 7. 4). Applied potentials were measured versus a Ag/AgCl reference electrode (FLEXREF, WPI, USA). Reference and counter electrode (Pt wire) were placed about 3–5 mm below the gold coated prism, which was used as the working electrode. Details on the SAM preparation and picosecond SFG setup (EKSPLA, Lithuania) can be found in the Supporting Information. Figure 2 Schematic representation of the spectro-electrochemical cell. SFG measurements were performed at static potentials. First, a potential of +0. 3 V was applied to the biotin-KKKKC:TEGT SAM and the corresponding SFG spectra was recorded. Subsequently, a potential of –0. 4 V was applied to the same substrate while the SFG setup was left unchanged. The corresponding spectra (normalized to the IR and visible intensities of the incoming beams) are shown in Figure 3 for the region between 3150 and 3350 cm −1. While changing the potential from positive to negative values, SFG signals nearly overlap, except for spectral contributions around 3245 cm −1. While NH vibrations of the peptide are occurring at frequencies centered above 3280 cm −1 36 and CH vibrations are significantly lower (below 3000 cm −1 ), the peak around 3245 cm −1 can be attributed to molecular vibrations within the heterocyclic imidazole moiety of the biotin group. 37 Figure 3 SFG spectra biotin-KKKKC:TEGT at +0. 3 V and –0. 4 V. Differences between the spectra are marked in green. The illustrations are interpretations of the corresponding molecular arrangement at positive and negative surface potentials. Besides the narrow band visible in Figure 3, a rather steep incline in the overall spectral shape is observable that is associated to non-resonant signals from electronic transitions within the Au substrate, Fresnel coefficients that change with frequency, and typically broad OH contributions from water. 38 SFG control studies performed with the single component biotin-KKKKC SAM show that no switching occurs, presumably due to the high level of packing of the oligolysine chains that are constrained in one conformation (Supporting Information). Further evidence of this non-switching behaviour is provided by electrochemical SPR (Supporting information). An analysis of the relative phase in between the resonant and non-resonant signals can retain information on the orientation of this particular group. A dip in the spectrum is related to a mean orientation of the corresponding transition dipole moment (TDM) away from the substrate (destructive interference), while a peak is indicating an orientation towards the substrate (constructive interference). Applied to the spectra in this present research, the biotin moiety contributing in this spectral region has its mean TDM orientation pointing away from the substrate at positive potential resulting in a dip in the spectrum. In this scenario, the positively charged peptide chain is prone to adopt a conformation that will extend itself away from the substrate due to electrostatic repulsion resulting in an anisotropic upright orientation of the biotin group. At negative potential, the peptide chains are likely to adopt a collapsed folded conformation due to electrostatic attraction between the negative potential of the surface and the positive charges on the peptide backbone, which appears to have resulted in a disordered biotin group since the SFG signal is no longer visible (isotropic molecular ordering cannot generate SFG signals), Figure 3. Oftentimes in situ SFG spectra of biomolecules at surfaces are rather complex and relatively weak due to the isotropic nature coming with a less tightly packed arrangement of molecules. The mixed biotin-KKKKC:TEGT SAM used in our study has been previously 13 characterised by X-ray photoelectron spectroscopy (XPS) and an average ratio on the surface of 1:16 ± 4 was observed. It is remarkable that at such a small surface coverage, SFG signals of biomolecules in solution still deliver a significant contribution above the noise ratio. So, our experiments are highlighting the importance of following changes in SFG spectra while changing external parameters such as surface potentials. In this respect, the combination of electrochemistry and SFG, as applied in this study, provides a powerful platform when it comes to in situ spectral analysis utilizing SFG in the context of biointerfaces. Furthermore, the applied potential can be switched back and forth to reproducibly cycle between the two spectral states. Moreover, the observed spectral features might be only slightly above noise level, but the reproducibility strengthens evidence and provides statistical means to an otherwise only singular event. Cycling the external parameter also allows investigating the reversibility of molecular conformations as discussed in the following paragraphs. Figure 4 a shows baseline corrected normalized SFG spectra that have been recorded at +0. 3 V, −0. 4 V, and back to +0. 3 V applied potential. The ability to turn on and off the upwards orientation with the applied potential allows us to monitor the molecular reorientation of the biotin group. The reappearance of biotin peaks at positive potential shows that the biotin group can be reversibly switched from being isotropically oriented at negative potential towards an anisotropic orientation at positive potential. 39 The corresponding fitted intensities for the spectra shown in Figure 4 a can be found in Figure 4 b. Positive and negative potentials are clearly separated demonstrating that the identification of 2 states (upright and random orientation) is above noise level. Additionally, mean values at repeating positive potential are within the error of the spectral fit quantifying the reversible nature of the switching process. Figure 4 a) Normalized and baseline corrected SFG spectra of the Biotin-KKKKC:TEGT SAM (grey lines) and their corresponding fits (purple lines) for +0. 3 V, –0. 4V, and returning to +0. 3 V. b) Sum of the fitted resonant SFG intensity (represented through amplitude divided by width) at switching surface potentials. c) SPR sensorgram traces showing the binding of Neutravidin (37 μg mL −1 ) to the biotin-KKKKC:TEGT mixed SAMs at a solution ratio under an applied positive (+ 0. 3 V) and negative (−0. 4 V) potential. After Neutravidin binding for 30 min, the surfaces were washed with PBS for 20 min to remove any nonspecifically adsorbed Neutravidin. The conformational change of the biotin-KKKKC peptide is further illustrated by SPR measurements of the binding events between the biotin end-group of the biotin-KKKKC:TEGT SAM and the neutrally charged protein Neutravidin at different applied potential ( Figure 4 c). For biotin-KKKKC:TEGT SAM, the binding process is favoured at +0. 3 V when the oligolysine backbones are in an extended conformation and the biotin end-groups are exposed. In contrast, at –0. 4 V the interaction between the biotin and the Neutravidin is prevented due to the folded conformation of the backbones which makes the biotin unavailable. On the other hand, when the same experiment is performed on a pure biotin-KKKKC SAM no difference in binding events is observed (see Supporting Information), indicating that despite the different potentials applied no molecular conformational changes are occurring. Although it is important to study the switching behaviour of SAMs with different experimental detection techniques, no information at the molecular level can be gained by SPR. Furthermore, the reversibility of the switching monitored by SPR implies the use of specific analytes which could lead to several issues such as the occurrence of non-specific binding and the irreversible chemical bonds between the analyte and the ligand. Both these circumstances have an impact on the reversibility performance and therefore on its analysis. However, these difficulties are overcome by using SFG spectroscopy where no extra binding processes are needed ( Figures 4 a and b). We have demonstrated that SFG spectroscopy is a highly sensitive tool able to provide an in depth characterisation of the reversibility of electrically switchable biotin-KKKKC:TEGT SAMs. By studying the orientation of the SFG peak's characteristics of the biotin end-group, the determination of the structural orientation under electro-induced switching was ascertained. The switchable process and its reversibility were assessed by repeatedly switching between positive and negative surface potentials. Monolayers of single component biotin-KKKKC SAMs exhibited no conformational change upon the application of an electrical potential, indicating that due to the high level of packing the oligolysine chains are constrained to one extended conformation (anisotropic molecular ordering). On the contrary, when a negative potential of –0. 4 V is applied to the mixed biotin-KKKKC:TEGT SAM, a change in conformation occurs owing to the presence of the spacer, TEGT, presumably allowing the folding of the oligolysine backbones (isotropic molecular ordering). Furthermore, the reverse phase of the SFG signal in the region between 3200–3300 cm −1 at –0. 4 V suggest that the biotin end-group is facing in the opposite direction compared to the initial measurement (+0. 3 V). Such information has the potential to positively impact the field of biosensors as well as surface engineering where the direct knowledge of the structure and the geometry of the molecules at interfaces are vitally important.
|
10. 1002/admt. 201900592
| 2,020
|
Advanced materials technologies
|
Stereolithographic 3D Printing for Deterministic Control over Integration in Dual-Material Composites
|
This work introduces a rapid and facile approach to predictably control integration between two materials with divergent properties. Programmed integration between photopolymerizable soft and stiff hydrogels was investigated for their promise in applications such as tissue engineering where heterogeneous properties are often desired. Spatial control afforded by grayscale 3D printing was leveraged to define regions at the interface that permit diffusive transport of a second material in-filled into the 3D printed part. The printing parameters (i. e. , effective exposure dose) for the resin were correlated directly to mesh size to achieve controlled diffusion. Applying this information to grayscale exposures led to a range of distances over which integration was achieved with high fidelity. A prescribed finite distance of integration between soft and stiff hydrogels led to a 33% increase in strain to failure under tensile testing and eliminated failure at the interface. The feasibility of this approach was demonstrated in a layer-by-layer 3D printed part fabricated by stereolithography, which was subsequently infilled with a soft hydrogel containing osteoblastic cells. In summary, this approach holds promise for applications where integration of multiple materials and living cells is needed by allowing precise control over integration and reducing mechanical failure at contrasting material interfaces.
|
No full text available
|
10. 1002/admt. 202000683
| 2,021
|
Advanced materials technologies
|
Microfluidic Printing of Tunable Hollow Microfibers for Vascular Tissue Engineering
|
Bioprinting of vascular tissues holds great potential in tissue engineering and regenerative medicine. However, challenges remain in fabricating biocompatible and versatile scaffolds for the rapid engineering of vascular tissues and vascularized organs. Here, we report novel bioink-enabled microfluidic printing of tunable hollow microfibers for the rapid formation of blood vessels. By compositing biomaterials including sodium alginate, gelatin methacrylate (GelMA), and glycidyl-methacrylate silk fibroin (SilkMA), we prepared a novel composite bioink with excellent printability and biocompatibility. This composite bioink can be printed into hollow microfibers with tunable dimensions using a microfluidic co-axial printing. After seeding human umbilical vein endothelial cells (HUVEC) into the hollow chambers via a microfluidic prefusion device, these cells can adhere to, grow, proliferate, and then cover the internal surface of the printed hollow scaffolds to form vessel-like tissue structures within three days. By combining the unique composite bioink, microfluidic printing of vascular scaffolds, and microfluidic cell seeding and culturing, our strategy can fabricate vascular-like tissue structures with high viability and tunable dimension within three days. The presented method may engineer in vitro vasculatures for the broad applications in basic research and translational medicine including in vitro disease models, tissue microcirculation, and tissue transplantation.
|
No full text available
|
10. 1002/admt. 202100551
| 2,021
|
Advanced materials technologies
|
Biologically enhanced starch bio-ink for promoting 3D cell growth
|
The excellent rheological property has legitimated the suitability of starch hydrogel for extrusion-based 3D printing. However, the inability to promote cell attachment and migration has precluded the non-modified starch hydrogel from direct applications in the biomedical field. Herein, we develop a novel 3D printable nanocomposite starch hydrogel with highly enhanced biocompatibility for promoting 3D cell growth, by formulating with gelatin nanoparticles and collagen. The rheological evaluation reveals the shear-thinning and thixotropic properties of the starch-based hydrogel, as well as the combinatorial effect of collagen and gelatin nanoparticles on maintaining the printability and 3D shape fidelity. The homogeneous microporous structure with abundant collagen fibers and gelatin nanoparticles interlaced and supplies rich attachment sites for cell growth. Corroborated by the cell metabolic activity study, the multiplied proliferation rate of cells on the 3D printed nanocomposite starch hydrogel scaffold confirms the remarkable enhancement of biological function of developed starch hydrogel. Hence, the developed nanocomposite starch hydrogel serves as a highly desirable bio-ink for advancing 3D tissue engineering.
|
No full text available
|
10. 1002/admt. 202101636
| 2,023
|
Advanced materials technologies
|
Latest Advances in 3D Bioprinting of Cardiac Tissues
|
Cardiovascular diseases (CVDs) are known as the major cause of death worldwide. In spite of tremendous advancements in medical therapy, the gold standard for CVD treatment is still transplantation. Tissue engineering, on the other hand, has emerged as a pioneering field of study with promising results in tissue regeneration using cells, biological cues, and scaffolds. Three-dimensional (3D) bioprinting is a rapidly growing technique in tissue engineering because of its ability to create complex scaffold structures, encapsulate cells, and perform these tasks with precision. More recently, 3D bioprinting has made its debut in cardiac tissue engineering, and scientists are investigating this technique for development of new strategies for cardiac tissue regeneration. In this review, the fundamentals of cardiac tissue biology, available 3D bioprinting techniques and bioinks, and cells implemented for cardiac regeneration are briefly summarized and presented. Afterwards, the pioneering and state-of-the-art works that have utilized 3D bioprinting for cardiac tissue engineering are thoroughly reviewed. Finally, regulatory pathways and their contemporary limitations and challenges for clinical translation are discussed.
|
No full text available
|
10. 1002/admt. 202101696
| 2,022
|
Advanced Materials Technologies
|
Collagen‐Tannic Acid Spheroids for β‐Cell Encapsulation Fabricated Using a 3D Bioprinter
|
Abstract Type 1 Diabetes results from autoimmune response elicited against β‐cell antigens. Nowadays, insulin injections remain the leading therapeutic option. However, injection treatment fails to emulate the highly dynamic insulin release that β‐cells provide. 3D cell‐laden microspheres have been proposed during the last years as a major platform for bioengineering insulin‐secreting constructs for tissue graft implantation and a model for in vitro drug screening platforms. Current microsphere fabrication technologies have several drawbacks: the need for an oil phase containing surfactants, diameter inconsistency of the microspheres, and high time‐consuming processes. These technologies have widely used alginate for its rapid gelation, high processability, and low cost. However, its low biocompatible properties do not provide effective cell attachment. This study proposes a high‐throughput methodology using a 3D bioprinter that employs an ECM‐like microenvironment for effective cell‐laden microsphere production to overcome these limitations. Crosslinking the resulting microspheres with tannic acid prevents collagenase degradation and enhances spherical structural consistency while allowing the diffusion of nutrients and oxygen. The approach allows customization of microsphere diameter with extremely low variability. In conclusion, a novel bio‐printing procedure is developed to fabricate large amounts of reproducible microspheres capable of secreting insulin in response to extracellular glucose stimuli.
|
1 Introduction Type 1 Diabetes Mellitus (T1DM) is an autoimmune disease characterized by an insulin deficiency caused by pancreatic β‐cell destruction, which leads to persistent high blood glucose levels. [ 1 ] Currently, a therapy based on daily exogenous insulin administration is the primary treatment for patients with T1DM. However, this temporal treatment does not mimic the real‐time insulin secretion of β‐cells resulting in loss of capacity to control glucose homeostasis and leading to chronic complications such as cardiovascular diseases, nephropathy, and retinopathy. [ 2 ] Nowadays, islet transplantation combined with immunosuppressive drugs is a promising way to provide new β‐cells to diabetic patients. [ 3 ] However, several drawbacks limit its exploitation in the clinical fields, such as the scarcity of compatible cadaveric donors, the requirement of immunosuppressors, and the islet loss due to the lack of vascularization and oxygenation. [ 4 ] Another common drawback of direct infusion of insulin‐secreting cells is the poor retention of cells following implantation. One solution to cellular retention at the injection site has been the use of biomaterials to encapsulate cells within a microenvironment before implantation. [ 5 ] In this line, encapsulation of beta‐cells within protective semipermeable biomaterials has emerged to overcome the challenges posed by islet transplantation. The biocompatible cell‐laden material then should have several functions: 1) suitable porous structure with mechanical stability, 2) excellent cytocompatibility to enhance cellular activity, 3) provide a physical barrier to protect embedded cells from cell to cell contact with the host's immune system. [ 6 ] Furthermore, the semipermeable membrane should provide bidirectional transport of insulin, glucose, oxygen, and nutrients preventing central necrosis. [ 7 ] In recent years, encapsulated‐based microspheres, or spheroids, are emerging as an optimal engineered tissue for graft implantation [ 5, 8 ] and in vitro drug testing. [ 9, 10 ] The 3D spheroids have the advantage of encapsulating large amounts of cells in a reduced volume, which helps deliver oxygen and nutrients to the core, simplified handling in a fluid suspension, and precise spatial and dosing control to fit in vivo applications. [ 11 ] The standard techniques employed to fabricate hydrogel spheroids are currently based on microfluidics and electrostatic droplet generation. The first method is suitable for the rapid fabrication of spheroids by emulsion. However, the need for an oil phase containing surfactants, which negatively affects cell viability, is still its main limitation. [ 7, 12 ] On the other hand, electrostatic droplet formation avoids using the cytotoxic oil phase, but it requires biomaterials with high viscosity to maintain a spherical shape that is challenging to handle. [ 13 ] Additionally, both methods are laborious, require considerable skill, and in addition, it is difficult to obtain uniformly sized spheroids. Therefore, there is an urgent need for more efficient and cost‐effective strategies to produce cell‐laden spheroids with high cell viability, controlling the diameter sizes and maintaining the spherical shape structure. These technologies have widely used alginate hydrogel for appealing physical properties, such as lack of biodegradability, robustness, and economic cost. [ 14 ] The alginate spheres are formed by combining the alginate solution with ionic cross‐linking agents, such as divalent cations (i. e. , Ca 2+ ). However, ionically cross‐linked alginate hydrogels present limited long‐term stability in physiological conditions, because they can be dissolved into the surrounding media due to the release of the divalent ions caused by the exchange reactions with monovalent cations. [ 15 ] In addition, the calcium ions released from the gel may promote hemostasis, while the gel serves as a matrix for the aggregation of platelets and erythrocytes. [ 16 ] Moreover, alginate revealed low biocompatible properties and limited control of mechanical properties. [ 17 ] In this regard, collagen, the main supportive protein in the ECM, is getting recognized as an optimal biomaterial for cell‐laden bioink. In the pancreas, collagens I and IV are abundant and widely expressed in mature human islets of Langerhans, where they support cell attachment and cohesiveness while also stimulating cell surface receptors to influence pancreatic cell proliferation and function. [ 18 ] However, the self‐assembled fibrillar structure is challenging to handle and is mechanically weak for direct use. A way to improve its mechanical stability is the further crosslink by a chemical agent such as glutaraldehyde, [ 19 ] genipin, [ 20 ] methacrylate, [ 21 ] and by enzymatic crosslinker such as transglutaminase. [ 22 ] Although these methods can improve some of the collagen's physical properties, their plethora of disadvantages range from cytotoxicity to expensive cost and excessive crosslinking time. Therefore, the formation of β‐cell spheroids encapsulated in a non‐cytotoxic collagen‐based bioink is still needed. To overcome the challenges in developing nontoxic biomimetic spheroids in an automatic way, we developed a high‐throughput methodology using 3D bioprinter to encapsulate β‐cells in a collagen bioink crosslinked with tannic acid (TA). This study aims to develop a new engineering strategy for the fabrication of an ECM‐like environment that more closely mimics native islet conditions promoting cell‐matrix interactions. We used collagen type I as a biocompatible encapsulation material that provides a physiological environment and mechanical support for the β‐cell insulin production. In this study, rat insulinoma INS1E cells were encapsulated into 3D collagen‐based microcapsules to investigate the substrate architecture's effect on ′cellular spatial organization, survival, and function. After spheroid fabrication, we examined cell viability and metabolic activity on days 1, 10, and 30. The results show that collagen crosslinked with TA scaffolds maintained cell viability for up to 30 days indicating preserved nutrient and oxygen diffusion. We also show that collagen fibrils treated with TA enhanced cell retention inside the spheroid compared to the pure collagen spheroids. In addition, INS1E spheroids improved glucose‐induced secretion relative to cells maintained in 2D cultures. In summary, these results demonstrate that the high‐throughput approach facilitates the fabrication of a cell encapsulation system that ensures good survival and functionality of insulin‐producing β‐cells. This strategy can help improve clinical outcomes of β‐cell transplantation strategies for diabetes treatment. 2 Results and Discussion 2. 1 Collagen‐Based Hydrogel Characterization TA is a specific class of hydrolyzable tannins containing a pentagalloylglucose core in which all hydroxyl functional groups are esterified by an additional gallic acid molecule ( Figure 1 a ). The TA molecule interacts hydrophobically with collagen type I, forming hydrogen bonds with amino acid residues (Gly, Met, Glu, Ser, Arg, and Pro) in the collagen chain (Figure 1b ). [ 23 ] Additionally, TA possesses pleiotropic effects, such as anti‐oxidant, anti‐inflammatory, anti‐bacterial, and anticancer properties. [ 24 ] Figure 1 General overview of the collagen hydrogels crosslinked with two different weight fractions of TA 1 wt% (ColTA 1x), and 3 wt% (ColTA 3x) solutions used for this study. a) Chemical structure of TA. TA contains a pentagalloylglucose core in which all hydroxyl functional group are esterified by an additional gallic acid molecule. b) Hydrogen bond interactions between TA molecule and amino acid residues from the collagen peptide chain. c) Fabrication process scheme of the disc‐shaped hydrogels used for the characterization tests. To generate cylinder‐like scaffolds, a PDMS mold was filled with collagen prepolymer solution. Then, the mold was placed at 37 °C. Once the material crosslinks thermally, the scaffolds were submerged in TA solution for 1 min. Scale bar = 10 mm. d) Scanning electron microscope (SEM) images of the hydrogel porosity. First row scale bar = 50 μm. Second row scale bar = 5 μm. We fabricated cylindrical collagen hydrogels without cells to evaluate the TA effects on the scaffolds. These scaffolds were generated at 37 °C to induce collagen fibril formation. One main property of TA crosslinking is the possibility to modulate the scaffold pore size by changing the concentration of TA used. Thus, collagen scaffolds were submerged into TA solution at 1% w/v (ColTA 1x) or 3% w/v (ColTA 3x) for 1 min (Figure 1c ). Collagen samples were fragile and difficult to handle compared to ColTA hydrogels, which presented a more stable structure and were easy to handle and manipulate. At the ultrastructural level, in SEM images acquired after dehydration and critical point dry, we observed that TA generated significant changes in the collagen fibers within the hydrogels (Figure 1d ). Specifically, TA crosslinking induced fusion of collagen fibrils, increasing their thickness and decreasing the pore size compared with non‐crosslinked collagen hydrogels. Further, a uniform surface appeared when crosslinked for 10 min (Figure S1, Supporting Information). We determined the average pore size by analyzing SEM images. In the case of ColTA 1x, the average porosity was 0. 09 µm, while in ColTA 3x was 0. 07 µm, and in collagen, control was 0. 12 µm ( Figure 2 a ). By comparing different pore diameter ranges (Figure 2b ), we noted an increment in the percentage of small pores related to the concentration of TA. The ability to decrease pore diameter by simply increasing TA concentration and/or crosslinking time is an interesting feature of our approach. While the porous mesh allows bidirectional transport of small molecules as insulin and glucose (0, 0027 and 0, 0015 µm, respectively), [ 25 ] cells embedded in the hydrogel can be protected from cell‐to‐cell contact with human macrophages (21 µm) [ 26 ] and T lymphocytes (7 µm). [ 27 ] However, molecules as cytotoxic cytokines released by the host immune system, such as interleukin 1b (0, 0024 µm), [ 28 ] are still permeable and could contribute to the death of the transplanted β‐cells. Current live‐cell encapsulation technologies face a trade‐off among several major considerations. Enhancement of one consideration may sacrifice the other. For example, a thick membrane with good immune isolation and mechanical stability often leads to poor nutrient/oxygen supply. A more in‐depth study is required to reach a good compromise between the preservation of β‐cell functional features, i. e. , insulin secretion and cytokine protection. Figure 2 Mechanical characterization of the scaffold. a) Pore distribution using different concentrations of the TA crosslinking solution. b) Percentage of pores in different diameter ranges according to the concentration of the TA. Of note, the higher the concentration, the higher the percentage of pores between 0–0. 03 μm of length. c) Swelling ratio. d) Rheological stiffness results. e) Rheological loss modulus results. f) Percentage of weight loss using the collagenase type I to examine the degradation of collagen‐based scaffolds. Results are expressed as the mean ± SEM from at least three independent experiments ( n = 9). **** p <0. 0001. Next, we checked scaffold swelling properties. To measure the diffusion of the nutrients indirectly through the scaffolds, we analyzed the swelling ratio. Swelling is the water uptake capability of a hydrogel, an indirect measurement of pore interconnectivity. [ 29 ] After only 24 h, the hydrogels reached an equilibrium, with a swelling ratio of 53. 63 ± 6. 11% for collagen, 35. 73 ± 2. 44% for ColTA 1x, and 30. 81 ± 1. 94% for ColTA 3x (Figure 2c ). Although this property was higher in collagen than in ColTA conditions, as expected, all the scaffolds presented permeability. These percentages of swelling indicate that the scaffold's structures were highly interconnected, as water could colonize all the structures. The reduction in swelling ratio of ColTA conditions could be due to the decrease in the number of hydrophilic groups in the presence of TA, which increased the hydrophobicity of the hydrogels. Even so, the strong point of this approach was the capability of the scaffolds to be permeable, allowing the diffusion of oxygen and nutrients through the hydrogels. Moreover, in other approaches, the swelling ratio of pure collagen hydrogel was set as ≈50%, [ 30 ] a value that fits with the scaffold swelling achieved. Another feature that we studied was scaffold stiffness. The ECM supports cells and plays an important role in regulating cell behaviors such as cell spreading, migration, proliferation, and differentiation. [ 31 ] It is composed mainly of collagen fibers which cells are in contact with and sense through mechanotransduction. [ 32 ] Therefore, the final stiffness contribution mainly depends on collagen abundance and organization. [ 33 ] As we wanted to mimic the ECM and the pancreatic islet environment, the collagen scaffolds stiffness was an essential property to analyze. Moreover, knowing that cells modulate their behavior in different substrate stiffness, [ 34, 35 ] maintaining a similar stiffness as soft tissues should help to the viability and the functionality of the embedded cells. Rheometer analysis was performed to measure the stiffness of the scaffolds. Rheology studies the relationship between force (stress) and deformation (strain) under a constant frequency. [ 36 ] This dynamic analysis was performed by applying a viscoelastic strain in cycles repeatedly over time to yield elastic (storage) and viscous (loss) moduli. The storage modulus is related to elastic deformation of the material ( G ′), whereas the loss modulus represents the energy dissipated by internal structural rearrangements ( G ′′). The results obtained confirmed the increase in stiffness after crosslinking hydrogels with TA as demonstrated by storage modulus (Figure 2d ). The G ′ value in the collagen condition was 0. 9 kPa and increased to 8, 6 to 15 kPa with TA solution 1x and 3x, respectively. The frequency‐dependent measurements of our hydrogels from all formulations showed that the storage modulus ( G ′) was always higher than the loss modulus, showing that the hydrogels were predominantly elastic (Figure 2e ). The stiffness achieved correlates well with the proper stiffness defined for soft tissues which range from 0. 1 to 100 kPa. [ 32 ] In addition, our values are in good agreement with previously reported values for collagen and collagen crosslinked with TA hydrogels which range from 0, 61 to 67, 13 kPa. [ 37 ] Moreover, the increment in stiffness in ColTA conditions was in accordance with the increase of collagen fiber fusion and the diminution of pore size shown previously in Figure 2a, b. To assess whether the increased stiffness of hydrogels crosslinked with TA confers higher resistance to enzymatic degradation, we incubated the collagen, ColTA 1x, and ColTA 3x hydrogels with Collagenase type‐I (see Experimental Section). Collagen hydrogels treated with TA showed higher resistance to collagenase degradation compared with pure collagen. Figure 2f shows the percentage of weight loss using collagenase type I to examine the degradation of the collagen hydrogels in different crosslinking conditions. Non‐crosslinked scaffolds underwent a complete degradation in 8 h, while ColTA hydrogels remained stable against collagenase and did not show any degree of degradation over the 7 days of the experiment. Thereby, the TA molecule is bound strongly to collagen and stabilized collagen. One explanation is that the hydrophobic interaction of TA with the amino acid residues of collagen cause that collagenases fail to recognize their cleavage site in collagen structure. [ 21, 38 ] We can conclude that our scaffolds satisfy all the mechanical and physical needs of the β cells. The in vivo degradation of collagen‐based spheroids was evaluated by spheroid transplantation into immunodeficient NOD Scid Gamma (NSG) mice. ColTA 1x and 3x spheroids were implanted on the omentum surface, considered a favorable environment for cell survival and function, [ 39 ] and fixed by folding the tissue. After 15 and 30 days, the animals were sacrificed to assess the integrity of the scaffold. As shown in Figure S2 (Supporting Information), the crosslinked spheroids persisted at days 15 and 30, allowing the spheroids constructs to be readily identified and retrieved. Overall, these results show a well‐defined and reproducible method to afford non‐degradable and microporous cell‐supportive scaffolds. 2. 2 Tunable Stiffness of Collagen‐Based Hydrogels Numerous studies have described the effects of matrix stiffening on cell behavior using 2D synthetic surfaces; however, less is known about the effects of matrix stiffening on cells embedded in 3D in vivo‐like matrices. A primary limitation in investigating the effects of matrix stiffness in 3D is the lack of materials that can be tuned to control stiffness independently of matrix density. Recent efforts to modulate 3D matrix stiffness have included modifying matrix density of natural proteins [ 40, 41, 42 ] and alginate, [ 43 ] using synthetic polymers with tunable crosslinking densities such as poly(ethylene) glycol (PEG) to create hydrogel scaffolds. [ 44, 45 ] Although these modifications are capable of generating 3D scaffolds with tunable mechanical properties, they also increase the density of the matrix, causing changes in the porosity and the number of cell binding sites. Here we were interested in developing mechanically tunable collagen‐based microspheres. To study the stiffness gradient across the spheroid radius due to the TA diffusion, we designed an experimental setup approximation. This approximation consisted of a rectangular prism inscribed in a sphere with a radius of 5 mm ( Figure 3 a ). We used a PDMS mold containing an empty pool at one end to fabricate the rectangular hydrogel. Then, we filled the empty pool with TA 1x solution for 1 min. Finally, we measured the local stiffness along the rectangular hydrogel length by AFM (Figure 3b ). The end of the rectangular hydrogel in contact with the TA pool exhibited high stiffness values which were declining along with the hydrogel, generating a stiffness gradient (Figure 3c ). Thereby, we were able to tune the stiffness of the collagen hydrogel in a range between 100 to 2 kPa. Additionally, confocal reflection images revealed a higher presence of a brighter structure at the end of the rectangular stripe in contact with TA solution than the other end (Figure 3d ). ColTA hydrogels exhibited thicker fiber bundles than non‐treated collagen in a time‐dependent manner. Overall, we have described a method to modulate the 3D stiffness of fibrillar collagen scaffolds, minimally altering the inherent fiber structure without the addition of synthetic materials. Figure 3 Characterization of tannic acid diffusion across the collage‐based stripe in a gradient mode. a) Rectangular model approximation based on a rectangular prism inscribed in a radius of 5 mm, to study tannic acid diffusion along the x ‐axis. b) The microrheology of the hydrogels was probed with an Atomic Force Microscope (AFM). Note that stiffness decreases along the rectangular stripe confirming that tannic acid diffuses while crosslinking the hydrogel. c) Representative reflection microscopy image of collagen‐based rectangular stripe used for the AFM analysis. Scale bar = 1000 μm. Results are expressed as the mean and SD ( n = 3). d) Reflection microscopy images of collagen and ColTA 1x hydrogels crosslinked for 1 min period. Note that thicker fiber bundles appear in collagen treated with TA (red arrows). Scale bar = 50 μm. 2. 3 Collagen‐Based Hydrogel can be Used as A Suitable Bioink for Spheroid Fabrication Using 3D Bioprinter Technology To generate a functional 3D structure to support β cells, we encapsulated INS1E cells into the collagen hydrogel spheroids using the 3D bioprinter (see Experimental Section). The fabricating process of cell‐laden spheroids is shown in Figure 4 a. Briefly, the inkjet/valve printhead was loaded with the cell‐laden collagen bioink. The hydrophobic petri dish was placed on the system working stage. A square array of 60 spheroids was fabricated with a cell density of 7 × 10 6 cells mL −1 (Figure 4d ). Spheroids were thermally crosslinked at 37 °C for 10 min. As before, 1% and 3% (wt/v) TA concentrations were used to examine their effects on cellular activities. Figure 4 Cell‐laden spheroids generated by the novel 3D bioprinter technique. a) Schematic depiction of INS‐1E‐laden spheroids fabrication crosslinked using different concentrations of TA solution. b) Contact angle differences between a hydrophilic and hydrophobic substrate. Optical images of the cell‐laden spheroids generated by the inkjet/valve printhead of the 3D bioprinter. c) Spheroid diameter range generated using the inkjet/valve printhead. Of note, a shorter obturation time of the valve translates into a lower volume of hydrogel deposited onto the superhydrophobic surface, thereby reducing the diameter of the spheroids Results are expressed as the mean and SD ( n = 20). d) Images of a square array of 60 spheroids of 0. 83 mm diameter, fabricated using the 3D bioprinter. Scale bar = 10 mm. Collagen crosslinked with TA spheroid compared to collagen control spheroid. Scale bar = 10 mm. First, to characterize the hydrophobicity of the plate, we measured the contact angle of the spheroid with the surface. As shown in Figure 4b, the hydrophobic surface maintained the spherical shape of the hydrogel drop, showing a contact angle of 141. 5°. In contrast, the non‐treated plate allowed the droplets to spread out with a smaller contact angle of 37. 8°. In conclusion, the superhydrophobicity of the pre‐treated plates allowed the spherical shape of collagen hydrogel drop. On the other hand, a common drawback in spheroid cultures is the inadequate supply of nutrients and oxygen to the core of the spheroid. Previous studies demonstrated that increasing spheroid diameter decreased glucose and oxygen concentration inside the spheroid [ 46, 47, 48 ] and resulted in the inefficient removal of discarded metabolites. [ 7 ] For this reason, we aimed to fabricate cell‐laden spheroids in a wide range of sizes using the 3D bioprinter. For in vivo long‐term applications, it is crucial to decide on an appropriate spheroid diameter with suitable diffusion properties that fit with the implementation site. We took advantage of the inkjet/valve printhead to produce a range of different diameter sizes, which incorporates a piezoelectric valve. The operator can control the opening and closing time of the valve. A shorter obturation time translates into a lower volume of hydrogel deposited onto the superhydrophobic surface and vice versa. The bioprinter system was then applied to produce controlled droplet arrays that can fabricate 100 spheroids/min with a well‐controlled diameter size. We fabricated spheres of up to 5 different diameters using different hydrogel volumes. We successfully produced spheroids with a diameter ranging from 0. 46 ± 0. 06 to 14. 9 ± 0. 03 mm by changing the valve‐opening duration from 2. 5 to 25 ms, respectively (Figure 4c ). Overall, we produced spheroids with tunable sizes by using the 3D bioprinter technique. One more exciting aspect is that after being crosslinked with TA, spheroids increased the stability of the entire structure, allowing us to obtain perfect microspheres that tolerate intense mechanical manipulation without losing their morphology (Figure 4d ). 2. 4 Collagen‐Based Hydrogels Crosslinked with TA Maintain Cell Viability and Promote Cell Retention Inside the Spheroids Next, we sought to establish cell viability during long‐term in vitro culture of the cell‐laden spheroids. Cell viability in spheroids of 0. 83 mm diameter was assessed at days 1, 10, 30 by Live/dead assay. A qualitative examination of the fluorescence images revealed homogeneous cell distribution within the spheroids ( Figure 5 a ). On day 1 after encapsulation, spheroids presented lower viability as compared to days 10 and 30. The shear stress present at the nozzle during the fabrication process is the main cause of cell damage and loss. [ 49 ] Remarkably, we found similar viability in collagen and ColTA spheroids at all times studied (Figure 5b ). These data indicate that 0. 83 mm diameter spheroids represent a good compromise between cell viability, ease of handling, and visibility and hence were used in subsequent experiments. Figure 5 In vitro cell viability and analysis of cell release from cell‐laden spheroids a) Representative fluorescent images of INS1E Live/Dead at day 10 after encapsulation. Live cells are marked with Calcein AM in green, and dead cells are marked with EthD‐1 in red. Scale bar = 500 µm b) Cell viability assay of INS1E cells at day 1, 10, and 30 encapsulated into the spheroids. c) Alamar blue test of collagen‐based shperoids after day 1, 10, 30 of encapsulation. Almar blue assay was detected at 570 nm. d) Optical images of cell release test after 1, 10, and 30 days of culture. e) Number of cells that were able to escape from the spheroids at day 1, 10, and 30 of culture. Results are expressed as the mean and SD ( n = 10). ** p < 0. 01 *** p < 0. 001 **** p < 0. 0001. Changes in metabolic activity can be easily quantified with the Alamar blue assay. While metabolic activity was similar in collagen and colTA spheroids at day 1, it was significantly lower in colTA spheroids compared to collagen spheroids at day 10 (Figure 5c ). Surprisingly, on day 30, results showed increased metabolic activity of ColTA 1x spheroids compared to collagen and ColTA 3x spheroids. These results demonstrated that TA did not affect cell viability inside the spheroids for up to 30 days in culture. Another exciting feature that we wanted to study was the ability of the collagen crosslinked with TA hydrogels to retain the cells inside the microcapsules. The ability of the cells to migrate or proliferate outside the biomaterial is a critical point in the tissue replacement therapies field regarding the biosafety of the patient. Biomaterials must keep cells inside the structures to prevent their spread in the body, eliminating the possibility of teratoma formation. [ 48, 49 ] It has previously been reported that density and spatial alignment of the collagen architecture can guide cell migration and leaking. To confirm that, the whole cell‐laden collagen scaffolds were monitored to check cell release/leaking from the spheroids over time. In the collagen control condition, single INS1E cells were detected outside the spheroids on days 1, 10, and 30. In contrast, free cells were rarely found outside the ColTA 1x and ColTA 3x spheroids (Figure 5d ), suggesting a marked increase in cell retention after the TA treatment. Indeed, collagen spheroids at day 1 released 359 ± 117 cells compared to 4 ± 1 cell in the ColTA 1x and 3 ± 2 in ColTA 3x, reaching 92, 05% less cell leak between collagen and ColTA 1x. Over time, collagen spheroids decreased the number of leaked cells from 172 ± 70 cells at day 10, to 70 ± 16 cells at day 30, while the ColTA spheroids maintained a cell scaping of 3 ± 1 cells in ColTA 1x and 2 ± 1 cell in ColTA 3x for the 30 days. These results show that the crosslinked collagen fibrils treated with TA promoted cell retention inside the spheroid compared to the pure collagen spheroids. The cell retention effects of this highly compacted collagen network correlate with the physical changes in stiffness and porosity described in biomaterial characterization (Figure 2a–d ). 2. 5 ColTA 1x Spheroids Secrete Insulin and Can be A Suitable Model for the Study of the β‐Cell Function We performed confocal image analysis of insulin and DAPI expression in INS1E‐containing spheroids. As expected, all cells in spheroids expressed insulin, and, in agreement with images in Figure 6 a, cells were homogeneously distributed in all conditions at day 10 after encapsulation. The different z‐stacks of spheroids per condition showed uniform distribution of insulin staining all over the spheroids (Figure 6a ). Interestingly, in both collagen and ColTA hydrogels, cells were able to proliferate, as we qualitatively analyzed by immunodetection of the ki67 proliferation marker (Figure S3, Supporting Information). Figure 6 Collagen crosslinked with TA enables insulin secretion. a) Representative images of INS1E cells encapsulated in the spheroids at day 10 insulin (red) and nuclei (DAPI). First column scale bar = 500 µm and others 100 µm. b) Glucose stimulation insulin secretion (GSIS) assay at day 8 in pristine collagen and colTA spheroids. For the GSIS assays, cells were incubated for one hour at 2. 8 × 10 −3 m glucose, followed by 16. 7 × 10 −3 m glucose. Results are expressed as the percentage of insulin secreted related to the total insulin content of the corresponding sample ± SEM from three independent experiments, each one including at least 3 different replicates per condition. **** p < 0. 001. It has been described that cell‐matrix interactions improve β‐cell survival and insulin secretion in 3D culture. [ 50 ] Lastly, to determine whether encapsulated INS1E cells into 3D hydrogel exhibited increased β‐cell function, we tested the insulin secretory response to glucose of the cell‐laden spheroids compared to 2D monolayer culture. A GSIS assay, which defines the ability of β‐cells to secrete a suitable amount of insulin in response to extracellular glucose stimuli, was performed in all conditions to check islet functionality. As shown in Figure 6b, collagen and ColTA 1x spheroids improved the insulin secretion stimulation index in response to glucose compared to 2D monolayer culture. Moreover, both conditions presented similar insulin secretion values. Indeed, on day 7, INS1E cells seeded in a 24 well‐plate presented a 2. 62 ± 0. 4 fold‐increase of insulin secretion when cells were challenged with 16. 7 × 10 −3 m compared to cells incubated with 2. 8 × 10 −3 m glucose. Collagen spheroids showed a 3. 77 ± 1. 2 fold‐increase. Interestingly, we reached a fold increase of 3. 95 ± 1. 3 of insulin secretion when ColTA 1x spheroids were challenged with 16. 7 × 10 −3 m glucose. By contrast, insulin secretion by ColTA 3x spheroids was nearly undetectable, revealing that this TA concentration hinders proper insulin output. This observation might be explained by the difficulty for insulin to break through the small‐sized pores from the fused collagen fibers. Therefore, our study validates that a 3D mesh structure mimicking the in vivo ECM ensures an optimal insulin response of embedded β‐cells provided a TA concentration < 3x. 3 Conclusions The field of tissue engineering has experienced remarkable progress with the incorporation of novel technologies as 3D bioprinting. This technology opened new opportunities to generate more replicable, customized, and cost‐optimized engineered tissues. Otherwise, collagen‐based materials have been recognized as promising to accomplish an ideal mimetic bioink for cell encapsulation with high cell‐activating properties. However, despite the wide range of outstanding biological properties, the use of collagen has been limited due to its characteristic low mechanical strength. In this study, we described a novel methodology to 3D‐bioprint collagen solution in the form of cell‐laden spheroids. The spheroids were post‐treated with TA crosslinking solution, which provided structural consistency and protected the embedded cells from the degradation of collagenase I. This hydrogel possessed a well‐organized microstructure with adjustable mechanical stiffness and porosity. Moreover, viability tests demonstrated that the ColTA spheroid is an excellent encapsulating material that permits good gas and nutrient exchange. Also, we demonstrated that the optimal concentration of TA to ensure glucose‐induced insulin secretion was 1 wt% (ColTA 1x condition). Significantly, the reduced spheroid volume and the increment of surface‐area‐to‐volume ratio could improve the diffusion distance compared to large volumes leading to a significant decrease in time response to blood glucose changes. Finally, this protocol allows the encapsulation of a large number of cells in a short period (less than 1 min), preventing them from hypoxic stress that can cause cell function loss. Furthermore, in vivo studies showed that transplanted structures did not present any sign of degradation, demonstrating its non‐biodegradability property and highlighting the feasibility of using this strategy as a potential therapy for T1D. In summary, the present study provides a foundation toolset to generate cell‐laden spheroids using a 3D bioprinter approach, which can be helpful for future advanced functional studies. (e. g. , with more clinically relevant cells such as iPSC‐derived β cells or dissociated primary islets). This technology can be applied to encapsulate a wide range of transplantable cell types. Also, by adding components of the extracellular matrix to the collagen‐based hydrogel, it could be possible to obtain a more cell‐laden matrix similar to the native environment found in vivo. 4 Experimental Section Preparation of Collagen‐Based Solutions Type I collagen from rat tail (Corning) at 8. 43 mg mL −1 was used for the standard working solution. Collagen was dissolved with 10x PBS (Sigma) at the ratio of 1:10 and neutralized with 1 m NaOH (PanReac‐AppliChem) to achieve a pH of 7. 5. The resulting hydrogel solution was dissolved with RPMI 1640 medium (Gibco) to reach the final concentration of 4 mg mL −1. Scanning Electron Microscopy Collagen hydrogel solution was poured in a cylindrical PDMS (DOW Corning, SYLGARD 184) mold of 10 mm diameter and 3 mm height (Figure 1c ). Collagen hydrogel was thermally crosslinked after 10 min at 37 °C. Next, collagen hydrogels were detached from the mold and submerged in a TA solution (Sigma) at 1% w/v (TA 1x) and 3% w/v (TA 3x) in PBS 1x, for 1 min period. Finally, the collagen‐based hydrogel crosslinked with TA 1x (ColTA 1x) and TA 3x (ColTA 3x) were washed twice with 1x PBS. After that, dehydration was carried out by sequential immersion in graded ethanol solutions in Milli‐Q water: 30%, 50%, 70%, 80%, 90%, and 96% v/v for 10 min each and twice for 100% ethanol. Then, samples were placed in a critical point dryer (K850, Quorum Technologies, UK), sealed, and cooled. Ethanol was replaced entirely by liquid CO 2 and by slowly heating. After critical point drying, hydrogels were covered with an ultra‐thin coating gold and imaged by ultrahigh‐resolution scanning electron microscopy (SEM). Pore diameters were quantified with ImageJ version 1. 52b software (National Institutes of Health). Swelling Swelling is the water uptake ratio by a scaffold. Hydrogels were fabricated as explained previously and were d weighted after fabrication to measure this. Next, cryogels were submerged into MilliQ water for 24 h when they reached equilibrium and weighted again. The swelling ratio was calculated as follows: (1) Swelling ratio = We-Wf Wf x 100 Where We is the weight in equilibrium and Wf is the weight after fabrication. Three hydrogels per condition were measured in this assay. Stiffness Characterization Mechanical properties of hydrogels were assessed using a parallel plate rheometer (Discovery HR‐2 rheometer, TA instruments, Inc. , UK). Hydrogels were fabricated in a cylindrical shape (1 mm thick, 8 mm diameter), and bulk modulus ( G ′) and viscous modulus ( G ″) measurements were recorded at a frequency range of 1–10 Hz at room temperature using 8 mm aluminum plate geometry. The gap was adjusted starting from the original sample height and compressing the sample to reach a regular force of 0. 3 N. Rheological measurements were made on hydrogels after 24 h post crosslinking. Degradation Cylinder‐shaped hydrogels, 8 mm in diameter, were fabricated as described above for the degradation analysis. Hydrogels were submerged in 1x PBS solution and left swelling for 1 day. A total of 2 mL of 0. 25 U mL −1 Type‐I‐Collagenase from Clostridium histolyticum (Sigma) dissolved in 5 × 10 −3 m CaCl2 with TBS1x was added to the hydrogels. Then, hydrogels were weighed after 1, 2, 4, 24, 48 h, 72 h, 96 h, and 7 days. The percent hydrogel remaining (%Wr) was determined by the following equation: (2) % Wr = Wt Wi · 100 Where W t and W i are the weights of hydrogels before and after the collagenase incubation. TA Diffusion The collagen type I hydrogel solution was prepared using a PDMS mold of 8 × 4 × 1 mm (Figure 3b ). A volume of 50 µL of the hydrogel solution at 4 mg mL −1 was poured and allowed to crosslink thermally at 37 °C for 10 min. The rectangular hydrogel was cut to create a pool. The pool was filled with a TA 1x solution for 1 min, performing a crosslinking gradient across the hydrogel. Then the hydrogel was removed from the mold and rinsed with PBS. The hydrogels micro‐rheology was probed with an Atomic Force Microscope (AFM) NanoWizard 4 Bioscience AFM (JPK Instruments) mounted on the stage of an inverted optical microscope (Nikon Eclipse Ti‐U). Silicon‐Nitride V‐shaped cantilevers with a constant force of 0. 08 N m −1, resonance frequency of 17 kHz, and cantilever length of 200 µm (Nanoworld innovative technologies, PNP‐TR‐50) were used to analyze samples. The force‐distance curves obtained were fitted to obtain the elastic modulus, using the Hertz model for a pyramidal tip (JPK data analysis software). Reflection Microscopy Images Fiber differences between collagen and ColTA hydrogels were visualized using 35 mm glass‐bottom dishes (MatTek). A volume of 50 µL of cold collagen (4 mg mL −1 ) was deposited over the glass bottom. Next, dishes were centrifugated at 2000 rpm for 3 min and incubated at 37 °C for 20 min. The collagen hydrogel was submerged in 500 µL of TA 1x solution for 1 s, 1 min, and 10 min. The resulting hydrogels were imaged using confocal reflection microscopy inverted Zeiss LSM‐780, with a 32x water immersion objective with a numerical aperture of 0. 85. Superhydrophobic Surface Ultra‐Ever Dry (SE 7. 6. 110) solutions based on two‐part coating compounds (bottom and top) were used to create hydrophobic surfaces. The surface activation of standard Petri dishes (Thermofisher) was achieved following the manufactured instructions. The bottom coat solution was briefly shacked in a fume hood and applied over the Petri dish to obtain a 1. 0 mil film. After 20 min, the topcoat was applied until a translucent white surface was seen. The plates became hydrophobic after 2 h. Spheroid Fabrication Using 3D Bioprinter The process of β‐cell spheroid formation is illustrated in Figure 4a. This work used a fixed concentration of collagen type‐I (4 mg mL −1 ) and INS1E cells as a bio‐ink. Collagen was mixed with the cell density of 7 × 10 6 cells/ mL. A bioprinter platform (3DDiscovery, regenHU Ltd) with inkjet/valve printhead (Microvalve CF300, MVJ‐D0. 1S0. 06) was used to fabricate the spheroids. The hydrophobic petri dish was placed on the system working stage, and the syringe was filled with the cell‐laden hydrogel. Square array patterns of 50 points were designed using the BioCAD v1. 0 software (regenHU Ltd) and launched to the bioprinter platform (Figure 4a ). The optimal printability was achieved with a nozzle diameter of 0. 15 mm, pneumatic pressure of 0. 2 bar, and printhead temperature of 6 °C. To obtain collagen spheres with different diameters: 1. 49, 1. 03, 0. 83, 0. 65, and 0. 46 mm, a valve‐opening duration of 25, 15, 10, 5, and 2. 5 ms was applied, respectively. It was decided to work with a diameter size of 0. 83 µm, equivalent to a valve‐opening duration of 10 ms. After printing, the cell‐laden collagen spheroids were thermally crosslinked at 37 °C for 10 min. The collagen spheroids without crosslinking process were used as a control, and the spheroids were crosslinked with two different concentrations, 1x and 3x, for 1 min. After several washes with PBS 1x, the spheroids were placed in a 24 non‐treated MW plate (Costar) and cultured in 3D suspension in constant stirring, with low growth medium based on RPMI 1640 medium (Gibco) supplemented with glucose (5. 5 × 10 −3 m ), 5% fetal bovine serum (FBS) (v/v), 10 × 10 −3 m HEPES, 2 × 10 −3 m L‐glutamine, 1 × 10 −3 m sodium‐pyruvate, 0. 05 × 10 −3 m 2‐mercaptoethanol and 1% penicillin/streptomycin (v/v). Cell Culture Rat pancreatic β‐cell line INS1E cells were cultured in high growth medium based on RPMI‐1640 (Sigma) with 11. 1 × 10 −3 m glucose, supplemented with 10% fetal bovine serum (FBS) (v/v) (Thermofisher), 10 × 10 −3 m HEPES (Gibco), 2 × 10 −3 m L‐glutamine (Gibco), 1 × 10 −3 m sodium‐pyruvate (Gibco), 0. 05 × 10 −3 m 2‐mercaptoethanol (Thermofisher), and 1% penicillin/ streptomycin (v/v) (Thermofisher). When cells reached confluency, cells were trypsinized with 0, 25 Trypsin/0. 1% EDTA and plated in a new flask at 1:4 density. Cells were maintained in an incubator at 37 °C and 5% CO 2. In Vivo Biodegradability Study NSG mice were used to assess the in vivo evaluation of collagen‐based spheroid biodegradability. Acellular collagen spheroids crosslinked with TA were fabricated using the 3D bioprinter. Subsequently, the spheroids were crosslinked with TA 1x and 3x for 1 min. For each condition, 3 spheroids were transplanted per mouse. The selected location was the bursa omentalis. The in vivo transplant was evaluated after 15 and 30 days. The local ethical committee for animal experimentation of the University of Barcelona approved all animal experiments and procedures. 345/18‐P3 (CEA UB), 10239P3 (Generalitat Catalunya). Live/Dead Viability assays were performed with the Live/Dead assay kit (Thermofischer) according to manufacturer instructions. The assays were performed on days 1, 10, and 30 of culture after cell encapsulation. Briefly, the spheroids were washed 5 min with PBS three times to replace the culture medium and incubated with the working solution (12 × 10 −6 m EthD‐1, 3 × 10 −6 m Calcein AM, and 12 × 10 −6 m Hoechst) for 25 min at 37 °C in agitation. Then spheroids were washed three times with PBS. Finally, confocal images were taken using a Zeiss LSM 800 confocal microscope. The quantification of the Live/Dead assay ratio was calculated as follows: (3) Live ratio = # Live cells # Live cells+Dead cells x 100 Alamar Blue AlamarBlue (Thermofisher) was performed according to manufacturer specifications. Shortly, each spheroid was placed in a 96‐well plate throughout the experiment ( n = 10). The medium was removed from the well plate and replaced by 11. 1 × 10 −3 m of glucose medium RPMI‐1640 with a 1:10 dilution of AlamarBlue. After 3 h incubation, 100 μL of each condition were placed in a well of 96 well‐plate and read in a Power wave X microplate spectrophotometer at 570 nm. Releasing Cell Test The releasing cell assay was performed to evaluate the number of cells able to escape from the biomaterial. Briefly, single spheroids were placed in 96 well‐plate and incubated for 1, 10, and 30 days. The escaped cells were counted both in medium and cells attached to the bottom of the plate. For the cells attached, 50 µL of trypsin‐EDTA (0. 025%) was added to the well and incubated for 10 min. The pellet was resuspended in 10 µL and mixed with 10 µL of trypan blue 0. 4% (ratio 1:1) (15250061, Thermo Fischer) and counted using an automated cell counter Countess (Fisher scientific). Immunostaining For confocal analysis, stained cell‐laden spheroids were used on day 10. After culturing, the spheroids were washed with PBS and fixed with 10% formalin solution (Sigma) for 30 min. Then, spheroids were washed with Tris Buffered Saline (TBS) (Canvax Biotech) and permeabilized with 0. 1% Triton X‐100 (v/v) (Sigma) solution in TBS for 30 min, under agitation. Spheroids were blocked with 0. 3% Triton X‐100 (v/v) and 3% Donkey serum (v/v) (Sigma) into TBS for 2 h in shaking conditions. Spheroids were incubated overnight with primary antibody against mouse anti‐insulin (1:500, Acris) in a blocking solution at 4 °C in shacking conditions. The following day, hydrogels were washed with permeabilization solution and incubated with secondary antibody for 2 h at room temperature (Alexa‐Fluor 647 conjugate goat anti‐mouse 1:200) under agitation. DAPI (1:1000 Thermofisher) was used to stain nuclei. Finally, spheroids were washed with TBS for 15 min and stored at 4 °C until confocal microscopy acquisition. Images were taken using an LSM 800 from Zeiss. Glucose‐Stimulated Insulin Secretion (GSIS) Encapsulated spheroids were preincubated with Krebs‐Ringer bicarbonate HEPES buffer solution at day 7 after fabrication (115 × 10 −3 m NaCl, 24 × 10 −3 m NaHCO 3, 5 × 10 −3 m KCL, 1 × 10 −3 m MgCa 2 ·6H 2 O, 1 × 10 −3 m CaCl2·2H 2 O, 20 × 10 −3 m HEPES and 0. 5% BSA, pH 7. 4) containing 2. 8 × 10 −3 m glucose for 30 min. Three spheroids per condition were used for the GSIS study. Then, spheroids were incubated at low glucose (2. 8 × 10 −3 m ) for 90 min, then incubated at high glucose (16. 7 × 10 −3 m ) for 90 min. After each incubation, supernatants were collected, and cellular insulin contents were recovered in acid‐ethanol solution. Insulin concentration was determined by ELISA (Crystal Chem). Statistical Analysis Results are expressed as mean values ± standard error of the mean (SEM), and all statistical analyses were performed using GraphPad Prism 8. 3. 0. One‐way analysis of variance was performed followed by Dunnet's or Tukey's multiple comparison post hoc tests, to determine individual differences between the groups. The representation of statistical significance is as follows: * p <0. 05, ** p <0. 01, and *** p < 0. 001, **** p <0. 0001. Conflict of Interest The authors declare no conflict of interest. Authors Contributions L. C. ‐F. and F. C. contributed to the study design, performance of experiments, data analysis, and writing and review of the manuscript. J. C. contributed to the performance of hydrogels micro‐rheology analysis and A. G. ‐A to the in vivo experiment. J. R. ‐A. and R. G contributed to the study design, data analysis, and report and manuscript review. J. R. ‐A is the guarantor of this work. It had full access to all the data in the study and took responsibility for the data integrity and data analysis accuracy. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/admt. 202300026
| 2,023
|
Advanced materials technologies
|
Spatial-Selective Volumetric 4D Printing and Single-Photon Grafting of Biomolecules within Centimeter-Scale Hydrogels via Tomographic Manufacturing
|
Conventional additive manufacturing and biofabrication techniques are unable to edit the chemicophysical properties of the printed object postprinting. Herein, a new approach is presented, leveraging light-based volumetric printing as a tool to spatially pattern any biomolecule of interest in custom-designed geometries even across large, centimeter-scale hydrogels. As biomaterial platform, a gelatin norbornene resin is developed with tunable mechanical properties suitable for tissue engineering applications. The resin can be volumetrically printed within seconds at high resolution (23. 68 ± 10. 75 μm). Thiol–ene click chemistry allows on-demand photografting of thiolated compounds postprinting, from small to large (bio)molecules (e. g. , fluorescent dyes or growth factors). These molecules are covalently attached into printed structures using volumetric light projections, forming 3D geometries with high spatiotemporal control and ≈50 μm resolution. As a proof of concept, vascular endothelial growth factor is locally photografted into a bioprinted construct and demonstrated region-dependent enhanced adhesion and network formation of endothelial cells. This technology paves the way toward the precise spatiotemporal biofunctionalization and modification of the chemical composition of (bio)printed constructs to better guide cell behavior, build bioactive cue gradients. Moreover, it opens future possibilities for 4D printing to mimic the dynamic changes in morphogen presentation natively experienced in biological tissues.
|
1 Introduction 3D printing technologies have rapidly become fundamental tools for biomedical research and personalized implant generation. These technologies have exceptional ability to generate biomaterial-based constructs with customized architecture and precise spatial patterning of different biocompatible materials and living cells (i. e. , via biofabrication technologies including bioprinting). , [ 1, 2 ] Key applications of biofabricated structures that mimic salient features of native tissues include patient-specific in vitro models for drug discovery, and implantable constructs for regenerative medicine. [ 3 ] A main limitation of current bioprinting technologies is the lack of using stimuli-responsive materials and shape memory polymers as building blocks. These approaches have often been defined as 4D printing, with time being the fourth dimension. [ 4 ] Typically, these strategies include the induction of predictable and desired changes in stiffness, architecture, or size of constructs postprinting upon exposure to heat, [ 5 ] ions, [ 6 ] ultrasounds, [ 7 ] or electromagnetic fields. [ 8 ] These geometrical changes are especially useful in the field of soft robotics, [ 9 ] and to mechanically stimulate cells during tissue culture. [ 10 ] On the other hand, time-dependent, on-demand modifications of the biochemical properties of the printed structure remain particularly challenging. Precise spatial control over the biochemical composition of a construct would allow for the gradual presentation of different growth factors and morphogens into local cell environments, thus enabling local control of (stem) cell fate, mimicking environmental changes naturally occurring during developmental, healing, and degenerative processes. In the field of 3D printed hydrogel for tissue engineering applications, capturing the physicochemical composition of the native extracellular matrix (ECM) remains an important objective. In fact, tissue ECM displays unique region-dependent mechanical properties, and it also acts as a depot of biologically active biochemical signals. These are both in the form of peptide sequences embedded in the ECM polymeric backbone, as well as through the release and presentation of growth factors tethered to specific domains in ECM proteins and glycans. Several elegant hydrogel-based systems have been designed to be readily functionalized with such bioactive sequences during their preparation. Often, these systems result in the isotropic distribution of bioactive cues that are effective in steering cell behavior. [ 11, 12 ] Alternatively, postcuring in photopolymers can be leveraged to graft molecules of interest onto prefabricated structures, homogenously in specific regions in which the still reactive material is present, as shown, for example, in studies processing nonhydrogel materials. [ 13, 14 ] To date, spatioselective chemical grafting of bioactive molecules has been typically performed in tissue cultures exploiting the contactless nature of light-based fabrication technologies, for example, with lithographic techniques, [ 15 – 17 ] which permit projections of 2D patterns, and via multiphoton lithography. [ 18, 19 ] The latter, albeit showing exceptionally high resolution (<1 μm), is limited by the working distance of the objective used in the device, which rarely exceeds 1 mm, thus preventing photochemical editing of larger objects. In the present study, we introduce a new visible-light-mediated technology to precisely imprint volumetric 3D patterns of fluorescent moieties and biological molecules within cubic-centimeter-scale hydrogels, leveraging the potential of tomographic printing. This approach enables the generation of geometrically defined patterns of biologically active species for directing cell behavior, which can be introduced arbitrarily at any time point after hydrogel cross-linking and printing. Previously, our group demonstrated the possibility to produce complex, hydrogel-based cell-free or cell-laden constructs of clinically relevant size in mere seconds, via volumetric bioprinting (VBP). [ 20 ] This novel light-based printing method, inspired by computed tomography, generates whole objects in a layerless fashion (as opposed to conventional layer-by-layer 3D printing). [ 20 – 22 ] This permits high-speed printing, while still achieving printing resolution in the range of ≈40–50 μm, even when printing in the presence of cells and complex cellular aggregates like organoids. [ 23 ] In volumetric printing (VP), also called volumetric additive manufacturing, a digital micromirror device shapes (visible) light into filtered backprojections of the object to be printed, as instructed by a tomographic reconstruction algorithm. The projections are sent to a rotating volume of a photoresponsive material at specific angles, and the resulting light dose accumulation allows for selective cross-linking of the resin into the desired 3D object. While this concept has been so far applied for photo-cross-linking and 3D printing, volumetric printing can be more broadly envisioned as a technique to spatially confine any light-triggered chemical reaction. In addition, as long as the printing resin is sufficiently transparent to the desired wavelength, the photoreaction could be conducted at any point in time post manufacture of a given object, in a noninvasive and biocompatible manner. To demonstrate this concept, in this study, we selected gelatin as a base material due to its known biocompatibility and possibility to source it with low endotoxin content, which makes it potentially translatable for medical and pharmaceutical use. [ 24 ] Moreover, gelatin allows for a broad array of chemical modifications, to accurately modulate its degradation profile and mechanical properties. [ 25, 26 ] As a platform material, we prepared and characterized a thiol–ene photo-cross-linkable norbornenemodified gelatin (gelNOR), which enables the generation of covalent hydrogel structures displaying complex geometries via volumetric printing. Thiol–ene click chemistry has gained increasing attention in the field of 3D printing and tissue engineering, as it yields hydrogels with highly homogenous network composition and mesh size. As the thiol–ene reaction progresses via a step-growth mechanism, the physical characteristics of the network can be reproducibly controlled by selecting the thiol-bearing cross-linker (length, molecular weight, number of reactive groups), network density, degree of functionalization (DoF), and the thiol-to-norbornene ratio. [ 27 ] Given the accurate control over the cross-linking kinetics and reaction termination upon removal of light irradiation, it is also readily possible to contextually control the amount of unreacted norbornene groups, which remain available for secondary reactions (i. e. , in this case, photografting) even after network percolation. Next, we subjected the volumetrically printed object to a second volumetric printing step in the presence of desired thiolated biomolecules of interest, a precise chemical editing can be performed locally at any point in time, decorating the hydrogel construct with 3D convoluted photopatterns ( Figure 1 ). To ensure high spatial resolution during the photografting process, the interaction between tomographic light dose, initiator concentration, and inhibiting antioxidant compounds was thoroughly characterized. As a proof of concept of biological functionality, a hydrogel chip with a perfusable channel was produced and assessed for cell adhesion, spatioselective proliferation, and promoted self-assembly of endothelial cells seeded within the channel, in response to volumetrically grafted patterns of vascular endothelial growth factor (VEGF), a key chemokine in angiogenesis. Overall, this novel approach paves the way toward the production of future tissue culture scaffolds and biofabricated constructs that can be gradually modified to match the evolving, dynamic requirements of cells during tissue culture and maturation, thus offering a new toolbox toward the engineering of functional living tissues. 2 Results and Discussion As a starting point, gelNOR was selected as it satisfies multiple requirements, namely: i) the compatibility with light-based 3D printing processes, to provide architectural control over 3D printed scaffolds of an arbitrary geometry, ii) the printing of structures with tunable mechanical properties across a broad range of stiffness relevant for tissue culture, via screening different dithiolated molecules, and iii) the ability to be used for photografting of any molecule bearing a free thiol group, such as those found in cysteine residues in native proteins. First, we screened an array of hydrogel formulations by varying cross-linker lengths and thiol–ene ratios, while keeping a constant 5 w/v% gelNOR concentration with 80% degree of norbornene functionalization, to maximize the amount of norbornene groups ( Figure S1, Supporting Information ). In addition, a relatively high degree of functionalization maximizes the number of reactive groups available for post-cross-linking during the volumetric photografting process. We thoroughly characterized a broad library of hydrogel formulations with tunable mechanical properties, by introducing two cross-linkers displaying different lengths: dithiothreitol (DTT) and a dithiolated diethyleneglycol (DEG), at different thiolto-norbornene ratios ( Figure 2 ). To evaluate the general trend in the cross-linking kinetics of these hydrogel formulations, photorheology was performed ( Figure 2A ). We noticed that for all formulations, cross-linking of the hydrogels starts immediately upon the moment of light exposure and progresses with similar kinetics. This suggests that, in the range tested herein, the cross-linker length does not significantly influence the cross-linking kinetics of the hydrogel, which is in line with previous research. [ 28 ] However, varying the thiol-to-norbornene ratios provided a clear difference in cross-linking kinetics between the samples, with 1:1 thiol-to-norbornene ratio yielding the fastest cross-linking kinetics (under 12 s to reach 80% of complete cross-linking). For hydrogels with a 4:5 thiol–ene ratio, the cross-linking was achieved in under 24 s and for the 3:5 thiol–ene ratio, cross-linking was achieved in under 36 s. The step-growth mechanism of gelNOR is known to provide rapid cross-linking, resulting from ring strain relief, especially when compared with hydrogels formed with a chain-growth mechanism, like gelatin methacryloyl (gelMA). [ 28, 29 ] Consequently, this allows to control the mesh size of the formed hydrogel network without greatly affecting the reaction kinetics, by changing the length of the cross-linker while maintaining constant the thiol–ene ratio. The soluble fraction (sol-fraction), which represents the amount of un-cross-linked polymer that washes out of the hydrogel network, of varying gelNOR formulations showed no significant difference for different cross-linkers or thiol-to-norbornene ratios ( Figure 2B ). Hydrogels formed with DTT as cross-linker displayed a sol-fraction of 12. 58 ± 4. 95%, 10. 76 ± 4. 03%, and 6. 04 ± 2. 39% for thiol–ene ratios 1:1, 4:5, and 3:5, respectively. Hydrogels formed with DEG as cross-linker provided a sol-fraction of 9. 94 ± 4. 04%, 7. 00 ± 2. 05%, and 4. 78 ± 3. 78% for the thiol–ene ratios 1:1, 4:5, and 3:5, respectively. All these measurements showed no significant difference. Dynamic mechanical analysis (DMA) was performed to determine the compressive modulus of the different hydrogel formulations ( Figure 2C ). Notably, a significant decrease in stiffness was observed as the thiol-to-norbornene ratios decreased in DTT samples (6. 30 ± 0. 29, 4. 94 ± 0. 70, and 3. 53 ± 0. 79 kPa for the thiol–ene ratios 1:1, 4:5, and 3:5) and between the 1:1 thiol–ene ratio (5. 26 ± 0. 13 kPa), the 4:5 ratio (3. 62 ± 0. 12 kPa), and the 3:5 ratio (2. 52 ± 0. 08 kPa) for DEG-cross-linked samples. This was to be expected since a 1:1 thiol–ene ratio would provide a maximal cross-linking of the polymer network and thus the stiffest gels, while at 4:5 and 3:5 thiol–ene ratios, there is an excess of norbornene groups that do not participate in the network. As for the effect of cross-linker length on the compressive properties of the hydrogels, a significant increase in stiffness (1. 36-fold) was observed in DTT samples compared to DEG at the 4:5 thiol-to-norbornene ratio. This mechanical versatility supports previous data shown for this bioresin and demonstrates that the mechanical properties of gelNOR hydrogels can be easily tailored to specific needs by adjusting either the thiol–ene ratio, and/or the length of the thiol cross-linker. [ 28 – 30 ] Considering the wide range of biomechanical requirements for culturing cells from different native tissues and organs, the mechanical versatility exhibited by gelNOR is of great interest to create stable, mechanically competent scaffolds for different tissue engineering applications. [ 31 ] In terms of the stress relaxation response of the materials, all the formulations showed a predominantly elastic behavior, in line with the characteristics of covalent hydrogels, with minimal relaxation, and high retention of the peak stress upon application of a constant strain ( Figure 2D and Figure S2 (Supporting Information) ). The swelling ratio of the hydrogel formulations differed significantly, both for cross-linker length and thiol-to-norbornene ratio ( Figure 2E ). The swelling ratios for the hydrogels with DTT as cross-linker were 14. 47 ± 0. 62, 16. 84 ± 0. 42, and 20. 18 ± 1. 07 for thiol–ene ratios 1:1, 4:5, 3:5, respectively. For the hydrogels with DEG as cross-linker, we measured the swelling ratio to be 17. 71 ± 1. 28, 19. 66 ± 0. 33, and 23. 02 ± 0. 87 for the thiol–ene ratios 1:1, 4:5, and 3:5, respectively. These results show that the hydrogels with a 1:1 thiol–ene ratio have a significant difference with varying cross-linker lengths, where the longer DEG has a higher swelling ratio than the shorter DTT cross-linker, probably also because DEG has a more hydrophilic profile than DTT. Furthermore, the measurement showed a significant difference in varying thiol–ene ratios for the formulations with DTT as cross-linker, where we see that the lower the cross-linker density, the higher the swelling ratio. This indicates the higher cross-linking density to be effectively formed for the 1:1 thiol–ene ratio, as compared to the other thiol–ene ratios. To confirm that the tunability of mechanical and physical properties of the hydrogels did not hinder sample stability over time, the rate of degradation of the different gelNOR formulations was evaluated in the presence of low collagenase concentrations ( Figure 2F ). The results of this accelerated degradation test showed that all hydrogel formulations could be completely enzymatically degraded with a similar kinetics in a 60 min timeframe, therefore suggesting the potential for cultured cells to remodel the gelatin matrix. Next, having available this set of photoresponsive hydrogels, the potential for shaping them into complex architectures through volumetric printing was investigated. For this, the formulation yielding 1:1 thiol–norbornene ratio and DTT as a cross-linker was used, as it was the one showing the highest mechanical stability and stiffness, thus allowing to maximize the ease of handling during printing and photografting. Light-based biofabrication technologies, such as stereolithography, [ 32 ] digital light projection printing, [ 33 ] and multiphoton lithography, [ 34 ] enable printing at higher resolution (nanometers to tens of micrometers) [ 35 ] and superior freedom of design compared to extrusion printing. In fact, being nozzle-free, light-based techniques sculpt photoresponsive materials, enabling the production of convoluted geometries recurrent in biological tissues (i. e. , templates of vascular networks) that cannot be readily produced with conventional extrusion techniques. With the recent introduction of volumetric printing, such complex geometries can now also be produced with a resolution in the range of few tens of micrometers, while printing centimeter-sized objects in less than 20 s ( Figure 3 ). To date, this technology has been applied to produce architecturally complex objects made of light-sensitive hydrogels, [ 20, 23, 29 ] polymeric acrylic and thiol– ene resins, [ 21, 36 ] elastomers, [ 37 ] nanoparticle-laden materials, [ 38 ] and glass. [ 39 ] In this study, we successfully achieved high printing resolutions with the selected gelNOR formulation of 23. 68 ± 10. 75 μm for positive features (e. g. , spikes), and of 176. 01 ± 36. 34 μm, printing open, perfusable channels within a soft hydrogel matrix ( Figure 3A, B ). These findings show the highest printing resolution of positive features to date, and complement the high-speed, high-resolution printing of gelatin norbornene materials previously reported using this printing technique. [ 29 ] Based on these printing conditions, more complex scaffolds were accurately resolved, from a mathematically derived gyroidal structure to torus-knot-shaped channels ( Figure 3D, E ). These highly convoluted structures were printed in less than 15 s, further underlining the ability of volumetric printing and of the gelNOR bioresin to rapidly and consistently produce architecturally complex, porous 3D structures. Printing accuracy was shown to be extremely high for both positive and negative feature constructs, showing no significant difference in volume between the digital model and the printed object itself (gyroid: 69. 40 mm 3 model vs 71. 95 ± 2. 11 mm 3 print; torus knot channels: 154. 73 mm 3 model vs 169. 85 ± 13. 34 mm 3 print) ( Figure 3F ). Building on the high-resolution printability of the gelNOR resin, we then investigated the potential to functionalize the printed constructs by covalently cross-linking single thiol-bearing molecules on the gelatin backbone in a spatioselective fashion across centimeter-scale objects. During the volumetric printing process of hydrogels, light irradiation is on purpose prematurely stopped to avoid cross-linking of out-of-target regions of the build volume, which could lead to printing artefacts. Consequently, the hydrogel reaches enough network percolation to be considered stable, however the maximum cross-linking density is not achieved, and if necessary, can be reached only with a postcuring process. [ 20, 23 ] This feature is especially desirable for enabling secondary reactions postprinting, such is the case of photografting onto still available norbornene groups. As a first step, we thoroughly characterized the photografting process and how to modulate its accuracy, taking advantage of both the tomographic printing principle and the reactivity of the photoresin with thiols ( Figure 4 ). For this purpose, we selected as a model molecule a fluorescent Cy3-tagged polyethylene glycol (PEG) chain functionalized with a single thiol moiety, which could be easily visualized and analyzed to determine photografting accuracy and intensity and exhibited stable fluorescence levels over time (Cy3–PEG–SH; 5 kDa; Figure S3, Supporting Information ). As first step, it was first confirmed that the Cy3–PEG–SH compound could be covalently bound to the gelNOR network. To assess this, gelNOR cylinders were infused with the grafting cocktail (containing Cy3–PEG–SH and lithium phenyl-2, 4, 6-trimethylbenzoylphosphinate (LAP) as photoinitiator) and were either irradiated with light from the volumetric printer (2000 mJ cm -2 ), or left in the dark. As shown by fluorescence imaging, the photoexposed samples retained a stable level of fluorescence intensity over multiple days of incubation in phosphate-buffered saline (PBS). Conversely, the Cy3–PEG–SH rapidly diminished over time in non-photoexposed samples ( Figure 4A ). A quantitative assessment of the fluorescence measured in the PBS used to wash the hydrogels further corroborated this observation, displaying sixfold higher fluorescence signal in the eluates from the nonilluminated controls already after 1 day of incubation, showing a rapid release of the PEG probe as opposed to a stable incorporation facilitated by the volumetric printer ( Figure S4, Supporting Information ). Furthermore, quantitative analysis of the grafted and nongrafted samples demonstrated that for the grafting conditions selected in this experiment, the tethered monothiolated Cy3–PEG–SH was found to be in the range of 30. 92 ± 2. 06 μM concentration, and samples infused in the grafting cocktail but not photografted showed nondetectable dye concentrations ( Figure S5, Supporting Information ). Next, in order to ensure spatial control over the 3D patterns imparted during volumetric photografting, a thorough characterization of the reaction was performed. During the tomographic printing process, in fact, it is important to keep in mind that the whole hydrogel volume is exposed to light, by delivering an anisotropic, 3D dose distribution. With the aim to correctly confine the photografting reaction within the desired region dictated by the standard triangle language (STL) file, an optimal process would show high grafting specificity, which is a parameter measuring the contrast between in-target binding and off-target binding. As testing platform, gelNOR cylinders previously infused with a grafting cocktail were exposed to a series of disk-shaped tomographic projections using the volumetric printing setup and delivering to each disk a different light dose (750–2000 mJ cm -2 ), to screen grafting specificity, intensity, and degree of off-target grafting ( Figure 4B and Figure S6 (Supporting Information) ). It was initially observed that by simply adjusting the light dose delivered to the printed construct (750–2000 mJ cm -2 ) and photoinitiator concentration (0. 6–1. 0 w/v%), covalent photografting could be achieved, but the Cy3 dye was detected at nearly equal amounts everywhere across the light path traversing the hydrogel with low spatial specificity, likely due to the high reactivity of the gelNOR system ( Figure S7, Supporting Information ). We therefore hypothesized that slowing down the reaction kinetics by adding a free-radical inhibitor to the grafting cocktail could help minimize unwanted off-target events. In this study, we chose (2, 2, 6, 6-tetramethylpiperidin-1-yl)oxidanyl (TEMPO) as inhibiting compound, since it has been previously used to enhance resolution in volumetric printing in combination with norbornene-based, nonhydrogel resins. [ 40 ] At relatively high concentrations, TEMPO can act as a prooxidant and elicit cytotoxicity on bacterial and mammalian cells, [ 41 ] however, this compound has been also proven to induce a protective effect for cells from oxidation-induced cell death, [ 42 ] and to act as a reactive oxygen species (ROS) scavenger, [ 43 ] when used in the safe concentration range also tested in our study (0. 006–0. 01 w/v%). [ 41, 44 ] At the lower average light dose tested (750–1250 mJ cm -2 ), regardless of the TEMPO concentration, low specificity ratios were still observed (0. 804 ± 0. 05–1. 825 ± 0. 08) and correlated with low grafting overall (both in- and off-target, Figure 4C–E and Figure S8 (Supporting Information) ), the latter being indicative of limited reaction efficiency, in line with the inhibiting action of TEMPO. At higher light doses, instead, sufficient free radicals can be generated within the region of interest in the hydrogel, resulting in an improved contrast over the surrounding regions, which instead receive a lower dose as programmed by the tomographic algorithm and are therefore more affected by the presence of TEMPO ( Figure 4F–H and Figure S8 (Supporting Information) ). specifically, for the highest tested light dose (2000 mJ cm -2 ) and using the formulation consisting of 1. 0 w/v% LAP and 0. 008 w/v% TEMPO, grafting specificity of 2. 388 ± 0. 06 (2. 1-fold higher than what was found without TEMPO) could be achieved, while also showing the highest in-target fluorescence intensity (4. 473 ± 0. 11 times higher than the background), and a low off-target intensity of 1. 568 ± 0. 26 (with 1 being the value of the native autofluorescence of the hydrogel). Altogether, these measurements showed that the grafting cocktail consisting of 1. 0 w/v% LAP and 0. 008 w/v% TEMPO allows for the most specific photografting to be achieved, while exhibiting dose-dependent intensity changes and greatly reducing off-target grafting. Having optimized the grafting cocktail to achieve highly specific spatial patterning of our fluorescent molecule, we explored the potential to photograft more complex architectures, and assessed the effect of light dose grafting specificity of the Cy3–PEG–SH compound. To assess this, a tubular spiral was grafted surrounding a central channel within a printed cylinder ( Figure 4I, J ). This structure was successfully patterned and visualized in 3D ( Figure 4I, J ). Previously, it has been shown in several studies employing the VP approach that different architectures, depending on their feature heterogeneity and size, require different light doses to be accurately resolved using this tomography-based approach. [ 23 ] In the case of photografting of complex objects, a light dose sweep was performed to determine whether grafting specificity was in any way affected by light dose. We showed that, albeit the highest grafting specificity for the spiral pattern was found at 1500 mJ cm -2, there was no significant difference for the other tested selected light doses, which also managed to resolve the spiral structure. This could suggest that the optimal formulation of the infusion cocktail may yield a broad, robust window for grafting such convoluted geometry at high specificity ( Figure 4K ). On top of this large grafting window at different doses, our gelNOR photografting system also yielded high resolutions of the grafted objects within our volumetric prints. A grafted spiral starting at 5 mm in width (5. 04 ± 0. 08 mm grafted resolution) that gradually became thinner in width until reaching a resolution of 1 pixel in the digital file reached a fully grafted resolution of 57. 20 ± 1. 66 μm ( Figure 4L ). This high level of resolution could be of particular impact and interest to produce patterns of bioactive molecules mimicking the microscale organization of biochemical components found in native biological tissues even at a scale close to the size of a single mammalian cell. Having established a successful protocol for photografting structures at high resolution, a range of different structures with varying feature sizes and degrees of complexity were accurately grafted ( Figure 5 and Figure S9 (Supporting Information) ). A highly tortuous, mathematically derived gyroidal structure surrounding a central hollow channel within a printed gelNOR cylinder ( Figure 5A-i ), a spiral structure surrounding a hollow channel ( Figure 5A-ii ), an interlocked chain structure with subunits in different axial orientations ( Figure 5A-iii ), the name of our research lab “Levato” spelled vertically along a gelNOR cylinder ( Figure 5A-iv ), and a random vessel structure ( Figure 5A-v ) were successfully grafted using the previously optimized grafting cocktail. Moreover, since most tissues present highly diverse types of proteins and growth factors critical for tissue function that are heterogeneously distributed along the same area, the possibility to graft multiple compounds in a spatially defined regions within the same printed object were also investigated. Here, a spiral shape was first grafted with Cy3–PEG–SH. Subsequently, another grafting process was performed, using a Cy5–PEG–SH, which was imprinted in the shape of vertically aligned cylinders ( Figure 5A-vi ). Grafting specificity of these complex geometries was measured for the gyroid (4. 04 ± 0. 70), spiral (3. 54 ± 0. 64), interlocked chain (2. 09 ± 0. 45), “Levato” (2. 18 ± 0. 33), and the random vessel (1. 88 ± 0. 69) ( Figure 5B ). Furthermore, grafting intensity of the complex geometries was measured for the gyroid (4. 27 ± 0. 80), spiral (3. 80 ± 0. 44), interlocked chain (4. 05 ± 0. 36), “Levato” (3. 98 ± 0. 29), and the random vessel (3. 99 ± 0. 14) ( Figure 5C ). The fact that both sets of values are within the same range as those observed in Figure 4 for simpler structures, further supports our previous observation that when using the optimized grafting cocktail, this process is extremely reproducible and as shown here, applicable to a wide range of architectures ( Figure 5B ). Variations in grafting specificity shown in Figure 5B are a phenomenon dependent on the tomographic reconstruction algorithm used for volumetric printing. As described in the previous literature, [ 22 ] when delivering light doses from multiple angles following a Radon transform and filtered-backprojection-based algorithm, the exact light dose delivered in every voxel oscillates around the average light dose set by the user. As a result, regions at the borders of the construct, especially in presence of sharp corners, tend to receive slightly higher doses and react faster. Printing (and herein, grafting) artefacts caused by this phenomenon could be resolved with dedicated corrections of the tomographic algorithm, as previously shown. [ 22 ] Despite this phenomenon, we demonstrate the possibility to accurately photograft complex patterns, even within more convoluted 3D printed structures, like an Atlas statue ( Figure 5D ) and a mathematically derived gyroid ( Figure 5E ). All in all, this fast method of grafting complex 3D patterns of several thiol-functionalized molecules can greatly increase the possibility of editing large hydrogel-based constructs in a spatiotemporally controlled fashion via sequential volumetric printing. Noteworthy to mention that thiol–ene chemistry is not the only possibility for photografting small molecules into a hydrogel system. In this study, thiol– ene chemistry was chosen since gelNOR is mechanically tunable and in many biological molecules there are cysteine residues capable of forming covalent networks through this thiol–ene chemistry. Other photochemistries could be studied for covalently grafting molecules to a hydrogel, i. e. , dityrosine oxidation, [ 45 ] photolysis of aromatic azides, [ 46 ] or selectively cleaving areas in a gelatin hydrogel, [ 28, 47 ] which could further expand the library of functionalizing compounds that are usable with this volumetric photografting approach, to further enhance the biochemical profile of bioprinted scaffolds. Besides the tethering of fluorescent compounds for easy visualization and optimization of the photografting process, this approach can also be used to covalently attach proteins or growth factors within the printed structures for guiding cell fate with spatiotemporal control. While covalent grafting of biomolecules could have an effect on protein bioactivity, the use of norbornene moieties for thiol–ene photoclick chemistry has previously been shown to enable thiolated protein immobilization, with several growth factors showing maintained bioactivity post immobilization. [ 48 – 52 ] As a proof of concept, we volumetrically grafted VEGF within a tissue-engineered macrochannel, aiming to improve the adhesion and sprouting capacity of human umbilical vein endothelial cells (HUVECs) within an uncoated lumen. VEGF expresses synergistic interactions with the integrin adhesion receptors guiding vessel growth and maturation, as well as endothelial cell survival by the regulation of antiapoptotic factor expression in these cells in vivo. [ 53 – 55 ] This proangiogenic growth factor is routinely used in endothelial growth culture medium to selectively enhance vascularization in in vitro engineered models as well. [ 56 – 60 ] VEGF has an uneven amount of cysteine residues and can be covalently coupled to a free norbornene onto the gelNOR network through the optimized photografting approach presented here. A volumetrically printed vascular chip, consisting of a central lumen of 1. 5 mm in diameter, was fabricated to assess the effect of photografted VEGF on seeded HUVEC adhesion, interconnectivity, and sprouting capabilities ( Figure 6A ). Given the short half-life of recombinant VEGF protein, cell performance in the grafted and nongrafted regions of the printed samples was evaluated after 3 days to ensure the proangiogenic effects of the tethered growth factor were captured, in the presence of either VEGF-free (VEGF - ) or VEGF-supplemented (VEGF + ) medium. To ensure that only the effects of VEGF incorporation were analyzed, the nongrafted regions of the prints were postcured at the same light dose that was used to graft the VEGF, resulting in homogenous mechanical properties throughout the construct ( Figure S10, Supporting Information ). After 3 days of culture, clear differences in HUVEC adhesion and interconnectivity were observed in the VEGF-photografted regions cultured in both VEGF + and VEGF - culture media ( Figure 6B, C and Figure S11 (Supporting Information) ). In VEGF -, HUVEC adhesion in the grafted regions of the lumen was significantly enhanced compared to nongrafted areas, as shown by the increased average cell coverage in the grafted (84. 06 ± 6. 36%) versus nongrafted regions (35. 72 ± 2. 43%) ( Figure 6B ). In samples cultured with VEGF + medium, the difference in the average HUVEC area coverage was less pronounced but showed significant differences between VEGF-grafted (83. 71 ± 3. 87%) and nongrafted (54. 46 ± 8. 57%) regions ( Figure 6B ) suggesting that in terms of cell adhesion to the hydrogel, the tethered VEGF provides a superior stimulation compared to free VEGF. Further, VEGF supplementation in the media did not significantly increase the coverage in the grafted regions of the construct, suggesting the absence of, or weak cumulative effects of the grafted and free soluble VEGF. Similarly, VEGF grafting had a significant effect on the cell interconnectivity, showing a higher number of intercluster junctions compared to regions lacking the covalently bound VEGF molecules in the absence of VEGF in the culture medium (309. 00 ± 104. 65 in VEGF-grafted region vs 145. 40 ± 49. 07 in nongrafted regions) ( Figure 6C ). These effects were conserved across VEGF + and VEGF - conditions. These observations suggest that the covalently tethered VEGF may provide a better support for HUVEC adhesion and growth compared to supplementation of soluble VEGF, at least in these initial stages of culture. The VEGF grafting could potentially be repeated over time to steer the vascular growth volumetrically printed constructs in real time in order to obtain more controlled multiscale vascular structures. Further, after only 3 days, the photografted VEGF facilitated HUVEC infiltration into the printed hydrogel, as shown by the significantly higher spanning depth of the cells from the inner edge of the lumen into the bulk hydrogel (130. 77 ± 25. 83 μm in VEGF-grafted regions vs 35. 42 ± 7. 73 μm in nongrafted regions) ( Figure 6D, E ). This observed cell infiltration was observed across the whole perimeter of the lumen ( Figure 6F ) and across the entire length of the printed channel ( Figure 6G ) in the grafted regions, while being completely absent in the non-grafted regions of the vascular chip model. These observations are encouraging, given that a proangiogenic effect is clearly seen in the VEGF-grafted regions of the lumen, where the growth factor acts as a chemoattractant capable not only of enhancing cell adhesion to an uncoated printed lumen, but also facilitates cell sprouting into the bulk hydrogel in the early days of culture. Since these effects are seen in both VEGF - and VEGF + medium, this strongly suggests that the covalently bound VEGF molecules retain sufficient bioactivity to steer the behavior of the seeded HUVECs. This grafting step could potentially be repeated over time to achieve different degrees of vascular growth volumetrically printed constructs and obtain more controlled multiscale vascular structures. Overall, this study proves the feasibility of grafting biologically functional compounds like growth factors, allowing these to maintain their bioactivity and guide cell fate with exceptional spatiotemporal control. Despite the demonstrated potential of this volumetric photografting technique, the infancy of this approach leaves room for future developments and exploration. To further boost the potential of the photografting process, a wider library of chemically editable bioresins suitable for VBP should be developed, and the stability and long-term functionality of different grafted compounds should be elucidated in more depth. Importantly, in terms of future perspectives, the biocompatibility of the grafting conditions toward cell viability and function should be assessed in long-term culture conditions to further evaluate translatability for potential regenerative medicine applications. Furthermore, with such future developments, the possibility to continuously edit the printed construct with different bioactive molecules during culture (i. e. , to replenish the growth factor content over time, or to change its localization over time), can also be explored to more closely mimic certain developmental and tissue repair processes. 3 Conclusions In this study, we demonstrated a new technological solution to create volumetric, 3D patterns of biological molecules within large, centimeter-scale hydrogels via tomographic printing and using visible light and bio-orthogonal thiol–ene chemistry. The selected material platform, gelNOR, was shown to possess highly controllable mechanical properties (through the adjustment of cross-linker length and resin DoF), and was shown to be printable via VBP, achieving a high printing resolution (20–30 μm for positive features). We demonstrated that these versatile gelNOR bioresins are suitable for the photografting of complex shapes onto volumetrically printed hydrogel constructs, as demonstrated by the controlled grafting of fluorescent dyes within gelNOR prints using tomographic projections, therefore allowing to both sculpt the architecture of the hydrogel and locally edit its chemical composition with high resolution (in the range of 50 μm). Through the extensive optimization of the grafting cocktail formulations (containing the thiolated compounds, cross-linking inhibitor TEMPO, and LAP photoinitiator) and the light dose delivered to the printed object, we achieved, for the first time, effective and precise photografting of both small dyes and large bioactive molecules, achieving micrometer-scale resolution of the grafted structures within centimeter-scale constructs while using a single-photon approach. As a proof of concept, we further applied this photografting principle to covalently tether the bioactive, proangiogenic growth factor VEGF to selectively guide and confine endothelial cell growth in the grafted, biofunctionalized areas. Improved cell adhesion and early formation of endothelial cell connections were observed preferentially in the biofunctionalized regions of the printed chip construct. Given that these observations match those of cells exposed to unbound VEGF molecules, this study indicates that the grafting process preserves bioactivity of the growth factors and opens the door for further characterization and tissue engineering applications. Overall, this work takes the first step in the characterization and development of smart materials that allow spatiotemporally precise biochemical editing. In combination with the ultrafast VBP technique, this photografting approach holds great promise to bring about the creation of biofabricated scaffolds that can better guide cell fate and behavior and therefore more closely mimic the complex biochemical environment of native tissues and organs. 4 Experimental Section Materials Gelatin from porcine skin (type A, X-Pure low endotoxin content) was kindly provided by Rousselot Biomedical (Ghent, Belgium). Commercial grade gelNOR (type B, bovine hide, DoF 60%) was kindly provided by BIO INX BV (Zwijnaarde, Belgium). Cellulose dialysis membrane tubes (molecular weight cutoff = 12 kDa) were purchased from Sigma-Aldrich. LAP was purchased from Tokyo Chemical Industry (Tokyo, Japan). Cy3–PEG–SH and Cy5–PEG–SH ( M w = 5 kDa) were purchased from Biopharma PEG (Watertown, USA). All other chemicals were obtained from Sigma-Aldrich unless stated otherwise. GelNOR Synthesis Type A gelatin was dissolved in a carbonate– bicarbonate buffer (pH 9, 0. 1 M concentration) to reach a 10 w/v% concentration. This solution was heated to 50 °C for the gelatin to dissolve and kept at this constant temperature throughout the synthesis. To reach a desired DoF, 0. 2 g (1. 2 mmol) of carbic anhydride (CA) per gram of gelatin was used in the reaction. The CA was added every 10 min for a total of 5 times starting at t = 0. After every addition of CA, the pH of the reaction was stabilized with 5 M NaOH to reach a pH of 9. After 240 min (DoF 80%) from the first addition of CA, the reaction was stopped by centrifuging the solution at 4000 rpm at room temperature for 5 min. Afterward, the pH was stabilized to 7. 4 using 1 M HCl. To benchmark the custom-synthesized gelNOR, a commercial grade gelNOR kindly supplied by BIO INX BV (Zwijnaarde, Belgium) was used which exhibited comparable mechanical properties ( Figure S12, Supporting Information ) and grafting accuracy ( Figure S13, Supporting Information ) as the custom-synthesized hydrogel. The solution was diluted to reach a 5 w/v% concentration of gelatin and dialyzed against MilliQ water for 4 days at 4 °C. After the dialysis, the solution was further diluted with MilliQ to reach a final concentration of 2. 5 w/v%. The solution was then heated to 50 °C and sterile filtered. Next, the solution was frozen at -80 °C, and lyophilized in a freeze dryer (Alpha 1-4 LSCbasic, Chris) to yield the dry product. Degree of Functionalization Quantification 2, 4, 6-trinitrobenzene sulfonic acid (TNBSA) assay was performed for the quantification of the amount of free amine groups present in the gelatin before and after functionalization. A glycine standard curve, to determine the amino group concentration, was prepared with concentrations of 0, 0. 8, 8, 16, 32, 64 μg mL -1. Gelatin samples were dissolved in 1. 6 mg mL -1 of 0. 1 M NaHCO3 buffer. Subsequently, 0. 5 mL of the sample was mixed with 0. 5 mL of a 0. 1 w/v% TNBSA solution in the buffer and incubated at 37 °C for 2 h. Next, the reaction was stopped by the addition of 0. 25 mL of 1 M HCl and 0. 5 mL of 10 w/v% sodium dodecyl sulfate. The absorbance of the samples was measured by a CLARIOstar Plus (BMG Labtech, Germany) plate reader at λ = 335 nm. The amount of free amines was calculated to be 0. 3371 mmol per gram of gelatin, based on the TNBSA results ( n > 5). Sample Preparation for Hydrogel Cross-Linking Unless stated otherwise, all experiments were conducted using gelNOR hydrogel supplemented with the following components to achieve photo-cross-linking. GelNOR stock solutions were made in PBS at a 10 w/v% concentration. LAP stock solution was made in PBS at a 1 w/v% concentration. A stock solution of DTT or 2, 2′-(ethylenedioxy)diethanethiol was prepared in PBS at a 100 mm concentration. To facilitate complete dissolution, all stock solutions were heated to 37 °C. Afterward, the stock solutions were mixed and diluted with PBS to reach a final concentration of 5 w/v% gelatin-based material, 0. 1 w/v% LAP, and the tunable ratio of thiol cross-linker to norbornene as needed for each experiment (1:1, 4:5, or 3:5 thiol–ene ratio). Mechanical Analysis GelNOR solutions from different aliquots of the same synthesis batch were casted in a cylindrical mold (6 mm diameter, 2 mm height), and cross-linked for 10 min (Cl-1000, Ultraviolet Cross-Linker, λ = 365 nm, I = 8 mW cm -2, UVP, USA). Samples were washed in PBS at 37 °C overnight to reach equilibrium swelling. To assess the compressive properties, the samples ( n = 5) were subjected to a strain ramp at 20% min -1 strain rate until 30% deformation using a dynamic mechanical analyzer (DMA Q800, TA Instruments, The Netherlands). The compression modulus was calculated as the slope of the stress/strain curve in the 10– 15% linear strain range. To assess the viscoelastic properties, the samples ( n = 5) were subjected to a strain recovery measurement at a constant 20% strain for 2 min and then left for recovery for 1 min, with a preload force of 0. 0010 N. The elasticity index was calculated as the ratio between the recovered stress and the maximal stress under constant strain. Photorheology Photorheology experiments on gelNOR precursor solutions to determine the cross-linking kinetics were assessed using a DHR2 rheometer (TA Instruments, The Netherlands). Time sweep experiments were performed at a frequency of 10. 0 Hz, angular frequency of 62. 83 rad s -1, with 5. 0% constant strain at 21 °C ( n = 3). A volume of 100 μL of gel was used with a gap size of 300 μm. A 20. 0 mm parallel stainless steel electrically heated plate (EHP) was used as geometry. 30 s after the start of the measurement, the light source was activated (1200 mha, AOMEES, China, λ = 365 nm, intensity of 24 mW cm -2 for the remaining 2. 5 min). Soluble Fraction and Swelling Ratio The sol-fraction and swelling ratio experiment was performed according to a recent publication. [ 29 ] Briefly, to assess sol-fraction of the gelNOR hydrogel formulation, cylindrical samples produced from different aliquots of the same synthesis batch (6 mm diameter, 3 mm height, n = 5) were weighed immediately after cross-linking for their initial mass. Next, samples were placed in PBS and placed in the incubator at 37 °C overnight. The next day, the hydrogel samples were weighed again, and their mass was measured as masswet, t 0. Subsequently the hydrogels were lyophilized, and the dry mass (massdry, t 0) was measured. The samples were stored in PBS again to ensure swelling of the dry gels and placed in the incubator at 37 °C overnight. The wet mass of the hydrogels was measured as masswet, t 1. The samples were lyophilized, and the mass of the dry samples was measured as massdry, t 1. The sol-fraction formula of the hydrogel formulations for analysis of the cross-linking properties of the gelNOR formulations was calculated with the following formula (1) Sol − fraction [%]= mass dry, t 0 − mass dry, t 1 mass dry, t 0 × 100 The swelling ratio of the hydrogel formulations for analysis of the swelling behavior of the gelNOR hydrogel formulations was calculated with the following formula (2) Swelling ratio ( q ) = mass wet, t 1 mass dry, t 1 Enzymatic Degradation Assay GelNOR hydrogels were swollen in PBS overnight and subsequently incubated in a 0. 2 w/v% collagenase type II in Dulbecco’s modified Eagle medium (31966, Gibco, The Netherlands) supplemented with 10 v/v% heat-inactivated fetal bovine serum (FBS, Gibco, The Netherlands), and 1 v/v% penicillin and streptomycin (Life Technologies, The Netherlands) at 37 °C. Samples were removed from the enzymatic solution at different time points (15, 30, 45, and 60 min, n = 3 independent samples per time point). The mass of the hydrogel samples was measured and compared to the initial mass of the hydrogels before enzymatic incubation to determine the degradation rate of samples over time. Volumetric Printing GelNOR solutions were dispensed into cylindrical borosilicate glass vials ( Ø 10 mm), which were then loaded into a commercial volumetric 3D printer (Tomolite V1, Readily3D, Switzerland), equipped with a 405 nm laser, set to deliver an average light intensity of 11. 98 mW cm -2 within the printing volume. Prior to printing, the samples were cooled to 4 °C to achieve physical gelation of the gelatin-based materials. Custom-designed STL files were loaded into the printer software (Apparite, Readily3D, Switzerland). After the printing process, the vials were heated to 37 °C and washed gently with 37 °C PBS to retrieve the prints. To ensure homogenous cross-linking, the sample was submerged in 0. 1 w/v% solution of LAP in PBS and irradiated for 1 min in a UV oven. Volumetric Photografting Printed constructs were subjected to a second printing step to induce spatioselective photografting. Samples were printed at equimolar amounts of thiol to norbornene at a 5 w/v% gelNOR concentration. Next, the printed samples were washed with PBS overnight, and infused with a fluorescent probe molecule, Cy3–PEG–SH (0. 06 w/v%). To characterize the photografting reaction, several formulations of the infusion mix were prepared containing varying amounts of LAP (0. 6, 0. 8, or 1. 0 w/v% concentration) and TEMPO (0, 0. 006, 0. 008, or 0. 01 w/v% concentration), as inhibitor of the thiol–ene grafting reaction. The printed constructs were infused with the infusion mix at 4 °C for 2 h. Next, the gels were placed back into the printing vials with a small amount of gelatin (5 w/v% in PBS) to ensure thermal gelation and fixation of the construct inside the vial. The grafting process was performed in the printer, by loading STL files of the pattern to be grafted into the Apparite software, and performing a new tomographic light exposure step, to induce the 3D patterning of the fluorescent Cy3–PEG–SH in the programmed geometry. For the characterization of the volumetric grafting reaction, an array of vertically aligned cylindrical disks (3 mm diameter, 1 mm height) were grafted within a gelNOR cylinder (6 mm diameter, 20 mm height), with every disk exposed to a different dose (dose range: 250, 750, 1250, 1500, 1750, and 2000 mJ cm -2 ) ( Figure 4A ). The accuracy of the photografting process was assessed imaging cross-sections of these samples with a fluorescence microscope (Leica Microsystems, Germany), and the fluorescence intensity within the grafted regions of interest was compared to that of off-target areas. To assess the accuracy of photografting, 3 different ratios were calculated using the following formulas. Grafting specificity formula for analysis of grafted GelNOR hydrogels with fluorescent dyes (3) Grafting specificity = Fluorescence of interest region Fluorescence of side bands Grafting intensity formula for analysis of grafted GelNOR hydrogels with fluorescent dyes (4) Grafting intensity = Fluorescence of interest region Hydrogel autofluorescence (background) Off-target grafting formula for analysis of grafted GelNOR hydrogels with fluorescent dyes (5) Off - target grafting = Fluorescence of side bands Hydrogel autofluorescence (background) Using optimized grafting parameters, complex, arbitrary 3D patterns of the Cy3–PEG–SH were imparted within custom designed, 3D printed objects. Finally, the constructs were washed with PBS for a maximum of 5 days, until the un-cross-linked dye was completely removed from the gel. Subsequently, the photografted constructs were imaged with a light-sheet microscope. To demonstrate the possibility of grafting multiple molecules in a sequential fashion, a second grafting process was also performed using Cy5–PEG–SH as a fluorescent dye, using the same components of the grafting cocktail. Volumetric Grafting of VEGF and Cell Culture Assays Cylindrical constructs with a perfusable channel spanning through the center of the construct were volumetrically printed as described above, and a proangiogenic growth factor was photografted on the bottom half of these constructs ( n = 4 replicate samples, single HUVEC donor line). To ensure homogenous mechanical properties in the grafted and nongrafted regions, the bottom half (nongrafted) of the construct was postcured immediately after printing at the same light dose that was subsequently used during photografting (750 mJ cm -2 ). Samples were then washed and incubated overnight at 4 °C in an infusion mix of LAP (1 w/v%), TEMPO (0. 008 w/v%), and recombinant human vascular endothelial growth factor (1000 ng mL -1 ; VEGF165, PeproTech). The volumetric photografting process was conducted as described above to deliver an average light dose of 750 mJ cm -2 to the top half of the construct and generate constructs with anisotropic VEGF patterning. The constructs were then washed at 37 °C for 5 h to remove excess, nongrafted VEGF. Green-fluorescent-protein (GFP)-tagged human umbilical vein endothelial cells (GFP–HUVECs, Angio-Proteomie, Boston, MA, USA, passage 5) were seeded into the channel within the printed construct at a concentration of 10 7 cells mL -1. To ensure homogenous seeding through the round channel, the samples were placed in rectangular polydimethylsiloxane (PDMS) molds and rotated 90° every 15 min for the first hour of culture. Cell-seeded constructs were cultured in endothelial cell growth medium-2 (EGM-2) containing endothelial basal medium-2 + SingleQuots (except VEGF), 100 U mL -1 –100 μg mL -1 PenStrep, and 10% heat-inactivated FBS. Samples were cultured at 37 °C and 5% CO 2, medium was refreshed every day. To assess the effect of nongrafted VEGF, the full EGM-2 medium (including VEGF) was used for control samples. On day 3, fluorescent images of the GFP–HUVEC growing along the printed channels were acquired via confocal laser scanning microscopy (SPX8, Leica Microsystems, The Netherlands). The HUVEC area coverage and cell spanning depth (distance from inner side of the lumen to the outer edge of the lumen, or sprouting cells) were measured with Fiji, [ 61 ] and junction numbers were analyzed using the vessel analysis software AngioTool. [ 62 ] Statistics Results were reported as mean ± standard deviation. Statistical analysis was performed using GraphPad Prism 9 (GraphPad Software, USA). Comparisons between experimental groups were assessed via one or two-way analysis of variances (ANOVA), followed by post hoc Bonferroni correction to evaluate differences between groups. When normality could not be assumed, nonparametric tests were performed. differences were found to be significant when p < 0. 05. Supplementary Material Supplementary file
|
10. 1002/adom. 201800419
| 2,018
|
Advanced Optical Materials
|
Additive Manufacturing: Applications and Directions in Photonics and Optoelectronics
|
Abstract The combination of materials with targeted optical properties and of complex, 3D architectures, which can be nowadays obtained by additive manufacturing, opens unprecedented opportunities for developing new integrated systems in photonics and optoelectronics. The recent progress in additive technologies for processing optical materials is here presented, with emphasis on accessible geometries, achievable spatial resolution, and requirements for printable optical materials. Relevant examples of photonic and optoelectronic devices fabricated by 3D printing are shown, which include light‐emitting diodes, lasers, waveguides, optical sensors, photonic crystals and metamaterials, and micro‐optical components. The potential of additive manufacturing applied to photonics and optoelectronics is enormous, and the field is still in its infancy. Future directions for research include the development of fully printable optical and architected materials, of effective and versatile platforms for multimaterial processing, and of high‐throughput 3D printing technologies that can concomitantly reach high resolution and large working volumes.
|
1 Introduction Additive manufacturing (AM) is experiencing a rapid evolution in its relevant technologies, which are constantly shifting from object prototyping at small scale to much more comprehensive fabrication platforms. Currently available AM techniques encompass various methods, categorized in seven groups highlighting the basic mechanism used for material processing. 1, 2 Most of additive technologies exploit either curing of a photosensitive material by exposure to light, or the direct deposition of inks or solutions with suitable viscoelastic properties, to build 3D objects in a layer‐by‐layer fashion, starting from a digital model. Materials printable by AM methods include polymers, 3 metals, 4, 5 ceramics, 6 nanocomposites, 7 and soft matter. 8 In this way, components with unprecedented complex geometries can be obtained, with potential impact in numerous fields, such as aerospace and automotive, 9, 10 tissue engineering, 11, 12 microfluidics, 13, 14 and energy storage. 15 The interest toward AM methods is mainly motivated by their capability of generating 3D items with elaborate architecture, and by the possibility of remarkably simplifying the design of devices, making it more stable and robust. AM technologies also allow material waste to be greatly reduced compared to conventional top‐down manufacturing. Furthermore, at variance with conventional optical or electron‐beam lithographies, AM methods allow devices to be printed or manufactured on nonplanar surfaces, curvilinear substrates, and tortuous 3D patterns. 16 For instance, through preliminary scanning of real surfaces, their shape can be precisely considered in the digital design of the desired object, as well as in the manufacturing process. In turn, such flexibility leads to new opportunities for system integration, allowing different components of a multipart functional system to be assembled and interconnected along the out‐of‐plane direction, that is, along the third (vertical) spatial dimension. All these aspects are of paramount importance in the fields of photonics and optoelectronics, where most of components (light sources, lasers, lenses, waveguides, and photodetectors) have been traditionally pushed in either planar geometries or, when presenting simple 3D shapes (e. g. , cylindrical or spherical, or combination of them) have been mostly assembled in planar configurations. In the last decade, instead, many truly 3D photonic components have emerged, featuring a range of novel or improved optical properties or device characteristics. These systems include 3D photonic crystals and metamaterials, 17, 18 where the light propagation and light–matter interaction can be precisely tailored to enhance reflection, absorption, and amplification in specific spectral bands, 3D interferometers, 19, 20 and omnidirectional optical couplers for improved injection and extraction of photons. 21 Furthermore, the manipulation of light by 3D free‐form optical components and lenses with aspheric surfaces is expected to enhance optical resolution and to compensate aberrations in vision components for robots, drones, and, more generally, autonomous movable systems. 22 Finally, 3D printing might critically support the integration of optoelectronics in soft, nonplanar systems. These can include gloves, bioimplants, endoscopes, catheters, contact lenses, glasses, and optical fibers, leading to new approaches for real‐time monitoring of physiological parameters in medical applications, 23, 24, 25 production of injectable devices for optogenetics, 26 and enhancement of the sensorial experience in augmented and virtual reality. 27, 28, 29 However, the development of 3D photonics and optoelectronics has been slowed down by a few stringent requirements associated to these applications, in terms of i) optical properties of the used materials, ii) spatial resolution needed to be achieved by the fabrication processes, and iii) necessary uniformity of the interfaces between diverse materials. All these issues greatly affect the performance of optical devices. For instance, the transport of light along a passive waveguide is affected by propagation losses made up by the sum of several contributions: 30 the absorption of the used material at the frequency of the propagating light, the Rayleigh scattering by inhomogeneities of the bulk (i. e. , local variations of the refractive index, density, composition), the Rayleigh scattering by defects at the waveguide surface, and leakages due to light coupling with the surrounding materials or environment. In order to realize these components additively, first printable materials have to be developed with low absorption at the target wavelengths. In addition, these materials should exhibit well‐suited optical properties (e. g. , luminescence, optical gain, nonlinear optical response) in order to build active and functional photonic elements. Preserving the function of active materials during printing might be challenging, since the use of UV radiation or processing at high temperature might modify the absorption and emission bands, decrease the emission quantum yield, and suppress optical gain in many organic compounds. 31, 32 Finally, a tight process control is needed to print with high uniformity both in the bulk and at interfaces, thus reducing undesired losses due to light scattering. The quality of interfaces is also critical for those devices, such as light‐emitting diodes (LEDs) or solar cells, where charge injection or extraction is present and where optically active layers are adjacent to metals or to hole‐ and electron‐transporting layers, because the presence of defects and trap states greatly decreases the efficiency of the current‐to‐light or light‐to‐current conversion. This progress report presents the recent progress in AM technologies specifically addressed to the fabrication of photonic and optoelectronic devices, and a few perspective directions for this vibrant field. The most promising approaches are reviewed, with focus on available printable optical materials, resolution achievable, generated geometries, and on the additive fabrication of light sources, waveguides, photonic crystals and metamaterials, optical sensors, and microscale optical components. In the last section, future challenges and opportunities of 3D printing technologies for photonics and optoelectronics are presented and discussed. 2 Light Exposure‐Based Additive Technologies Light interacting with photosensitive materials can trigger a series of chemical reactions, which can be exploited to produce patterns with micrometric or sub‐micrometric resolution. 3 This is what is commonly done in photolithography, where UV light is used to pattern a resist by texturing a beam through a photomask, which contains the motifs to be transferred. Similarly, in direct laser writing, 33, 34 a laser beam is directed on the resist, modifying locally the chemicophysical properties of the organic layer, and the desired pattern is realized by scanning the beam. UV photolithography and laser writing allow 2D microstructures to be generated, and they are widely used for industrial production in microelectronics, optoelectronics, and photonics. The extension of photopolymerization methods to the fabrication of complex, 3D items is accomplished by implementing a layer‐by‐layer approach, in which the desired object is sliced in various 2D layers which are built consecutively, each individual layer being obtained by exposing a photosensitive material to structured light or to a scanning laser beam, similarly to photolithography and to laser writing. 2. 1 Stereolithography The invention of stereolithography (STL) by C. Hull in 1986 35 has opened the field of 3D printing, enabling the fabrication of 3D components through the photopolymerization of UV‐sensitive resins. In a typical STL process, an UV laser beam is focused on a liquid prepolymer, composed of suitable monomers and/or oligomers, and photoinitiators triggering the polymerization reaction. Curing can occur through the production of free radicals, the generation of cationic species, and other pathways. 36, 37, 38 Critical process parameters are typically given by the concentration of photoinitiators, the intensity of the laser and its in‐plane scanning speed. 39 For instance, a characteristic light dose is required to initiate the formation of the polymer network, which is related, among other variables, to the absorption coefficient of the photoinitiators and to the concentration of dissolved oxygen (which generally inhibits the polymerization process). Similar considerations hold for the photopoly‐merization rate. 39 Recently, geometrical constraints, confining the volume of photocurable polymers, have also been found to modify the curing kinetics, 40 slowing down polymerization and introducing a complex temporal behavior (polychromatic process). This aspect might be pivotal for building objects made by nanocomposite materials, where the addition of nanoparticles in the prepolymeric resin can introduce geometrical constraints and heterogeneous interfaces, significantly affecting the development of polymer networks during photopolymerization. For example, the addition of silica nanoparticles to a mixture of acrylate‐functionalized oligomers leads to an increase of the light dose needed to fully polymerize the nanocomposite, 41 and quantum dots (QDs) are found to decrease the polymerization rate of acrylate monomers. 42 In addition, the inclusion of nanoparticles to prepolymers raises some concerns for the potential increase of opacity and the spatial variations of the refractive index, leading to light scattering. Preventing particle aggregation and improving the stability of suspensions in photocurable resins through organic surface modifiers greatly help in producing transparent 3D printed nanocomposites (optical transmission > 90% in the 400–1100 nm spectral range). 41 Used with laser STL, this approach also reduces the shrinkage and distortions in printed structures, improving the resulting mechanical properties. 41 The choice of the monomers and oligomers in resins for STL depends on the final desired application. Acrylate‐based compounds are widely used in photonics due to their good optical transparency. 3 However, the variety of relevant materials printable by UV STL is continuously increasing, an effort which is also motivated by some disadvantages related to the use of acrylates such as yellowing upon aging, limited mechanical properties, and poor biocompatibility. 3 In such framework, glass is the material of choice for many optical components (lenses, optical fibers, optical windows). Recently, a STL approach for processing fused silica glass has been introduced, 43 which is based on using a nanocomposite prepolymer with amorphous silica nanoparticles dispersed in a photocurable matrix of the monomer hydroxyethylmethacrylate ( Figure 1 a). The nanocomposite prepolymer solidifies in 3D upon STL, and a two‐step, postprinting thermal treatment is carried out to convert the printed object to fused silica glass. At first, the organic matrix is decomposed by heating up to 600 °C, then sintering at 1300 °C leads to printed glass (Figure 1 b), which exhibits optical transparency comparable to those of commercially available equivalents (Figure 1 c), and can be colored by adding metal salts in the prepolymer (Figure 1 d). Figure 1 3D printed transparent glass. a) Schematics of the printing process of fused silica glass: first amorphous silica powder is mixed with an UV‐curable monomer, and the resulting nanocomposite resin is patterned by STL. The polymerized material is then converted in fused silica glass through debinding and sintering thermal treatments. Scale bar: 7 mm. b) Example of a 3D‐printed and sintered glass structure. Scale bar: 5 mm. c) UV–visible transmission spectrum of printed and sintered glass (black continuous line). The spectrum of commercial fused silica is shown for comparison. d) Example of colored fused silica glasses obtained upon doping with metal salts. Reproduced with permission. 43 Copyright 2017, Macmillan Publishers Limited. A major limit of laser STL, which is an intrinsically serial process, is in its low printing speed. Substantial variants have been introduced in this respect, to cure individual layers in a single step by projecting patterns onto the prepolymer instead of scanning by a laser beam. Projection is performed by using an UV‐illuminated digital system, such as a digital micromirror array device 44, 45, 46 or a liquid‐crystal phase mask. 47 The versatility of such approach, known as digital light processing (DLP) is exemplified in Figure 2 a–c, which displays a pH sensor printed on the surface of a tapered optical microfiber. 46 The pattern is made of the ionic hydrogel, poly(acrylic acid) (PAA), which shows pH‐dependent swelling when immersed in solutions with pH in the range 2–7. Upon printing, the PAA micropads shrink, thus applying a periodic strain on the fiber and consequently coupling the fundamental guided mode and cladding modes by resonant scattering, 46 which leads to a pH‐dependent dip in the fiber transmission spectra (Figure 2 d). The sensors show reversible response, with characteristic time of about 10 2 s (Figure 2 e). Figure 2 DLP‐printed pH optical sensor. a) Scheme of 3D printing system used to fabricate the PAA microstructures on the surface of an optical tapered fiber. b) pH‐sensing device. c) Optical microscopy image of a tapered optical fiber with a diameter of 30 µm. The insets c 1, c 2, and c 3 show confocal microscopy images of PAA micropads printed on the fiber, with sizes: 325 × 100 µm 2 (c 1 ), 325 × 300 µm 2 (c 2 ) and 325 × 600 µm 2 (c 3 ). Insets in (c 1 –c 3 ): photo of the devices. d) Reversible shift of the wavelength of the optical transmission dip versus pH. The inset shows the transmission spectra of the printed device at various values of pH. e) Dynamic response of the printed optical sensor (i. e. , wavelength shift, left vertical scale) to different pH solutions (right vertical scale). Reproduced with permission. 46 Copyright 2016, Wiley‐VCH. Vertical growth rates for objects made by DLP can be speeded up to 1 m h −1 by inhibiting free‐radical polymerization through O 2 diffusion, 48 a method where almost continuous printing is achieved by letting a thin layer of uncured prepoly‐mer between the last polymerized layer and the transparent substrate through which light exposure occurs. This is accomplished by using an O 2 ‐permeable substrate, by which the resin photopolymerization is locally inhibited, and a continuous renewal of reactive, liquid resin is obtained in the layer exposed to UV light. The spatial resolution of laser STL and projection technologies is typically in the range 10–100 µm. Optical diffraction sets fundamental limits to the minimum size of a focused light beam and, consequently, to the in‐plane resolution, whereas the vertical resolution is limited by the capability of depositing very thin layers of prepolymers in between the exposure steps of consecutive slices. For UV light at 405 nm (currently used in commercially available laser STL 3D printers), the nominal, minimum spot size of a laser beam focused by low‐numerical aperture (NA) optics (NA < 0. 1) is ≥2 µm, according to the Abbe relation. Another important aspect consists in the possibly anisotropic physical properties of structures printed layer‐by‐layer and with different in‐plane and vertical spatial resolutions. 49 Little is known about the eventual anisotropy of the achieved optical properties. Instead, various studies are focused on methods for improving the characteristic pixelated and step‐case surface roughness, that is also detrimental for applications in optics. For instance, in‐plane roughness can be compensated by overexposure in STL, and by defocusing projected layer images in DLP. A new approach takes advantages of grayscale photo‐polymerization 50 and meniscus equilibrium postcuring, 51 improving significantly the finish of curved surfaces printed by DLP (achieved surface roughness < 7 nm), and allowing millimeter‐scale lenses to be obtained with imaging capability in line with standard optical components. 52 Overall, spatial resolution and surface finishing remain issues, which are especially relevant for realizing optical components and devices operating at visible and near‐infrared (NIR) wavelengths. In this respect, other STL methods are being engineered, which rely on nonlinear optical absorption 53 to improve the spatial resolution well below diffraction limits. Exploiting nonlinear absorption also allows the layer‐by‐layer concept to build objects to be overcome, enabling the continuous production of arbitrary 3D geometries. 2. 2 Multiphoton Stereolithography Multiphoton stereolithography (MP‐STL) is a laser‐based AM technique with resolution reaching a few tens of nanometers. 54, 55, 56 When an ultrafast laser (pulse duration of tens or hundreds of femtoseconds) is focused inside the volume of a transparent, photosensitive material, the high intensity of the beam can induce multiphoton absorption within the focal voxel, causing local curing. Moving the beam along a continuous trajectory allows high‐resolution 3D structures to be obtained, in a free‐form manner and without the need to build the objects layer‐by‐layer. This significantly suppresses the presence of morphological or optical anisotropies between in‐plane and vertical directions. As a result, 3D objects with smooth surfaces can be realized, 57 a property that makes MP‐STL particularly suitable for the fabrication of optical and optoelectronic components. Like in any other lithographic technique, in order to recover the 3D pattern one needs to “develop” the sample, by immersing it into an appropriate solvent. However, MP‐STL has characteristic capabilities that no other technique can provide. On one hand, classic 3D printing such as STL cannot make structures with resolution better than a few micro‐meters, as mentioned above. On the other hand, high‐resolution methods such as electron‐beam lithography and focused ion beam can only pattern the surface layer of films; multiple layers and complex multistep depositions are required to fabricate anything resembling a 3D structure. The basic principles, challenges, and applications of MP‐STL are described extensively in several reviews and books and are beyond the scope of this progress report. 58, 59, 60 Here, we will concentrate on the applications of MP‐STL in photonics, namely in the fabrication of micro‐optical elements and electromagnetic metamaterial structures. Hybrid material technology deserves to be introduced in this framework, since this is prominent in MP‐STL addressing photonic applications. The first materials employed in MP‐STL were acrylate photopolymers and the high aspect ratio, negative photoresist SU‐8. 61, 62, 63 However, the photonics community very quickly switched to using organic–inorganic hybrid materials prepared by sol–gel processes. 64, 65, 66 These methods are based on the phase transformation of a sol of metallic oxide or alkoxide precursors to form a wet macromolecular hybrid network structure. The gel formed is subsequently reacted through MP‐STL to give a product similar to glass. The sol–gel technology provides a powerful tool for the development of optical materials by MP‐STL, as the resulting samples show high optical transparency and quality, postprocessing chemical and electrochemical inertness, and good mechanical and chemical stability. In addition, various functional compounds or nanosystems can be incorporated by employing guest–host or side chain–main chain strategies. Examples are metal alkoxides, 67, 68 nonlinear optical molecules, 69 QDs, 70 and conducting ionogels and graphene. 71 The most widely used hybrid materials are the silicate OrmoComp (and its different formulations) developed by the Fraunhofer Institute of Silicate Research 72 and commercially supplied by Microresist Technology, 73 and the zirconium silicate SZ2080. 74, 75, 76, 77, 78 Examples of 3D components made by the latter material are shown in Figure 3 a, b. Figure 3 a) A photonic crystal and b) a complex concentric microstructure made using SZ2080. (a) Reproduced with permission. 75 Copyright 2008, American Chemical Society. (b) Reproduced with permission. 78 Copyright 2015, American Vacuum Society. c–e) Freestanding lens compound optical systems: (c) device schemes, (d) simulated images by using a USAF 1951 resolution test chart, (e) actual devices; scale bars: 20 µm. (c–e) Reproduced with permission. 79 Copyright 2016, Macmillan Publishers Limited. f–i) A system of four different compound lenses on the same CMOS image sensor, combining different fields of view in one single system: (f) device schemes, (g) corresponding pixel size, evidencing an increased resolution toward the center of the image, (h, i) actual devices printed by MP‐STL. (f–i) Reproduced under the terms of the CC‐BY‐NC license. 22 Copyright 2017, The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. Various miniaturized photonic structures have been developed by MP‐STL, taking advantage of the sub‐micrometer spatial resolution of the process and of the availability of materials with excellent transparency, as highlighted above. The obtained micro‐optical elements can be used for beam focusing and collimation, shaping, and imaging. For instance, MP‐STL is highly suitable for fabricating complex systems of lenses, as they can be made in a single step, without requiring further assembly. 54 Freestanding compound optics, such as micro‐objective lenses (Figure 3 c–e), 79 pave the path for other relevant miniaturized systems such as endoscopes and microillumination elements. Compound optics can be even built directly on functional surfaces or devices as shown in Figure 3 f–i, where lenses with different fields‐of‐view have been printed onto a single complementary metal–oxide–semiconductor (CMOS) sensor. 22 Similarly, MP‐STL might be used to fabricate optical systems on fiber ends directly ( Figure 4 a, b), eliminating also in this case the need for further assembly and consequent errors. There are optical fibers bearing ultra‐compact lens objectives (spherical, cylindrical, toric lenses) with numerous refractive surfaces, 74, 79, 80 parabolic microreflectors, 81 free‐form optical elements for donut and top‐hat beam shaping, 82, 83 axicons, 84 and sensors. 85, 86 Figure 4 a) A complex lens system on the end of an optical fiber. Scale bar: 25 µm. Reproduced under the terms of the CC‐BY license. 82 Copyright 2016, The Authors. Published by Macmillan Publishers Limited. b) A fiber‐end Fabry–Perot gas microsensor, built directly on the top of a fiber. Reproduced with permission. 86 Copyright 2015, IEEE. c–g) Examples of 3D metamaterial structures made using MP‐STL and selective electroless plating. (c)–(f) Reproduced with permission. 98 Copyright 2017, Wiley‐VCH. (g) Reproduced with permission. 99 Copyright 2015, American Chemical Society. Another class of systems which can be manufactured by MP‐STL are metamaterials. These photonic elements do not exist in nature; their name derives from the Greek word “meta, ” which means “beyond. ” They are artificially engineered metallo‐dielectric structures whose electromagnetic properties are due to their architecture and not to the chemistry of their base components. 87 In metamaterials, an assembly of structures can replace the role that atoms and molecules have in conventional materials, and elementary units are usually arranged in periodic patterns, at scales smaller than operating wavelengths. By appropriately designing and building metamaterial architectures, many new and unusual optical properties have been demonstrated such as negative refractive index, magnetism at optical frequencies, perfect absorption, and enhanced optical nonlinearities, in turn leading to original applications including ultrahigh‐resolution imaging, compact polarization optics, and cloaking devices. 88, 89, 90, 91 However, the vast majority of materials printable with MP‐STL are dielectrics, so metal 3D metamaterials cannot be produced directly yet. There have been a few attempts employing multiphoton reduction of metals, but neither the achieved resolution nor the structural integrity was adequate. 92, 93 The most common (and successful) approach to date is to make a dielectric scaffold using MP‐STL, and then cover it with silver or gold. The thickness of the metal layer should be thicker than the skin depth of the operating wavelength, so that the optical response of the structure is comparable to that of solid metals. To make metallic coats, electroless plating and chemical vapor deposition 94, 95, 96 can be used. These techniques lack selectivity, leading to “blanket” metal deposition covering the substrate as well as the structures, which limits their application to reflective modes, or to bigger, freestanding structures. Infrared metamaterials can be also made using a positive photoresist and electroplating. 97 In this case, a positive mold of the structure is made, which is filled with metal. Despite the excellent results, this method is not yet widely used, as accessible geometries are limited and the highest resolution achieved to date is of a few hundreds of nanometers. It is also possible to employ MP‐STL resins with metal‐binding sites, allowing metamaterial operation in transmission mode and much finer features to be obtained (Figure 4 c–g). 98, 99, 100 A few commercial MP‐STL systems are currently available, with typical in‐plane resolution of few hundreds of nanometers and vertical resolution of a few micrometers. 101, 102, 103 Similarly to other AM technologies, one of the main issues is the time required to build a 3D object. The fabrication of a millimeter‐scale object with a resolution of a few hundreds of nanometers still requires building times of the order of days. 52 New approaches are being developed aimed at achieving reasonable building times as well as sub‐micrometer feature size. 104, 105, 106, 107 3 3D Printing by Extrusion The AM technologies which are based on light exposure provide the highest achievable spatial resolution, but they are limited by available photocurable compounds, a feature that restricts the possibility to build objects with heterogeneous parts and diverse composition. This issue is addressed by methods which rely on extrusion of materials, generally in form of continuous filaments through a nozzle, while 3D objects are still built by consecutively depositing many layers. Here, extrusion is promoted by either melting thermoplastic polymers as in fused deposition modeling (FDM), or applying a pressure to shear‐thinning viscous inks (as in direct writing methods). Extrusion technologies have enabled multimaterial printing, with resolution around hundreds of nanometers. 108 In FDM, 3D plastic objects are generated by the deposition of polymer filaments. 109, 110 A thermoplastic wire (polymethylmetacrylate (PMMA), polylactic acid (PLA), polysterene, polycarbonate, acrylonitrile butadiene styrene (ABS), etc. ) is directed through a hot extrusion head, and the fused polymer is deposited in a continuous way, solidifying as soon as the filament temperature falls below the glass transition temperature. Typical extrusion temperatures are in the range 200–400 °C, depending on the melting behavior of the used polymer. Also the target substrate can be heated to ≈100 °C, in order to promote better adhesion between adjacent layers. When building optical components, the main limit of FDM is in its spatial resolution (100–400 µm), which is largely determined by the size of the nozzle (0. 1–0. 4 mm). This might lead to quite rough surfaces and opaque printed structures, where coupling and propagation of UV, visible, and NIR light can be suppressed in a severe way. Nevertheless, FDM can be still exploited for optoelectronic device packaging. 111 A relevant example is the fabrication of stretchable LEDs, realized by embedding liquid metal‐based electrical interconnects in printed, cylindrical hollow polymer structures. Such flexible systems can power LEDs even upon twisting of the polymer envelope up to 540°. 111 Furthermore, FDM resolution would match values needed to realize components for use with terahertz and microwave radiation. 112, 113 Compact terahertz laser sources 114 are available which can be exploited as highly selective and noninvasive tools for environmental sensing and security. Hence, the field would significantly benefit from the development of specific methods to print windows, waveguides, lenses, and filters operating at terahertz frequencies. In this respect, issues are related to the absorption of most of the available polymers (e. g. , ≥10 cm −1 for PLA at 0. 5 THz). 112 While attempts with high density polyethylene and polypropylene, which are transparent at terahertz frequencies, did not lead to satisfactory results, polystyrene turned out to be a suitable material in terms of both printability by FDM and low absorption losses (0. 5 cm −1 at 0. 5 THz). 112 In addition, graded refractive index lenses for microwaves have been made by FDM of materials with different permittivity at 15 GHz (ABS possibly doped with 28% v:v SrTiO 3 particles). 113 3. 1 Direct Ink Writing Direct Ink Writing (DIW) relies on the controlled delivery of inks through a nozzle by an applied pressure. 108 The ink then rapidly solidifies due to evaporation, phase changes, or gelation. The nozzle is mounted on a 3D translation stage, which moves along patterns defined by the object architecture and by its digital model. DIW allows single or multiple inks to be released in the form of filaments, and objects with well‐controlled shapes and composition to be created. Various parameters have to be accounted for during process engineering, related to the viscoelastic properties of the ink and to specific DIW variables. 108, 115 ] Figure 5 a, b shows the typical rheology of DIW inks: 116, 117 they must exhibit well‐defined viscoelasticity to flow through the nozzle while retaining shape upon deposition, and to feature shear thinning thus favoring extrusion. Viscoelasticity is particularly critical for avoiding collapse in structures suspended through a gap, such as freestanding dielectric waveguides with air cladding. The characteristic dependence of the ink elastic shear modulus ( G′ ) on the shear stress is shown in Figure 5 a: the rapid decrease of the shear modulus upon increasing the shear stress around values close to the shear yield is necessary for establishing a continuous flow through the nozzle upon applying a pressure, whereas G′ values above 10 4 Pa at low shear stress provide self‐supporting structures with superior stiffness. 108 The insets of Figure 5 a display examples of two gap‐spanning structures, imaged 1 h after production, highlighting a more stable shape for the structure printed from a stiffer ink. To fit with these rheological requirements, a compromise between high stiffness and moderate shear yield has to be found when formulating the composition of the ink. For instance, this can be accomplished by adding solid particles to the ink. 116, 117 Figure 5 b shows results for inks with different content of silica nanoparticles: a particle amount of 20 wt% leads to joining of the deposited filaments, creating a compact and monolithic structure with excellent optical transparency. Filaments extruded with slightly higher content of silica particles (23 wt%) can retain their gap‐spanning shape after extrusion but form an opaque structure. Therefore, the filamentary, layer‐by‐layer morphology of the printed elements can be controlled by tailoring the ink composition, with more uniform and homogeneous structures achieved by less viscous inks. The approach introduced in ref. 117 is very interesting, because in combination with postprinting thermal treatments it allows free‐form 3D glass components to be obtained, with optical properties (transparency and refractive index) well matching those of commercially available glasses. 118 Figure 5 Material and process properties in DIW. a) Storage shear modulus, G′, versus applied shear stress for a fugitive ink filled with nanoparticles (fill circles) and for an organic ink doped with 40 wt% of microcrystalline wax (empty circles). The insets show optical images of corresponding suspended filaments (length 10 mm and diameter about 1 mm), acquired 1 h after printing. Reproduced with permission. 116 Copyright 2005, Wiley‐VCH. b) Viscosity versus shear rate for two inks with 20 wt% (circles) and 23 wt% (squares) of silica nanoparticles. The photographs show the corresponding printed samples which have very different optical transparency. Low silica content favors fusion of printed filaments (left pictures), whereas increasing silica content allows stable gap‐spanning features, but leads to lower optical transparency (right pictures). Scale bar: 4 mm. Inset scale bar: 0. 5 mm. Reproduced with permission. 117 Copyright 2017, Wiley‐VCH. c) Schematic representation of a typical setup for 3D printing by DIW. d) Images of elementary structures printed by DIW for various process parameters, H* and V*, corresponding to the diverse printing modes. (c), (d) Reproduced with permission. 115 Copyright 2018, Wiley‐VCH. DIW can be applied to many other inks including concentrated polymers, viscous elastomers, fugitive organics, and filled epoxies, 8, 108 with minimum feature size as low as 600 nm. 119 Importantly, the process is carried out in ambient conditions, a feature relevant to preserve the optical properties of active components which can be embedded in the inks. Furthermore, the shear forces exerted during extrusion might induce an anisotropic arrangement of molecular constituents and of the embedded fillers, favoring their alignment along the length of the deposited filament. Recent experiments have highlighted a preferential alignment of elongated fillers (fibers, rods, and nanotubes) along the longitudinal axis of filaments, with resulting anisotropic swelling behavior and mechanical properties. 120, 121 Additional work will be needed to investigate achieved optical properties (absorption, refractive index, photoluminescence, and light scattering), including their anisotropy (dichroism, birefringence, and polarization effects) in active or passive layers printed by DIW. Figure 5 c schematizes a few relevant printing parameters, such as the applied pressure, P, the nozzle inner diameter, D, and translation velocity, V, the ink extrusion speed, C, and filament diameter, d, and the working distance, H. Recently, Yuk and Zhao have developed an insightful model of the process, highlighting various DIW regimes characterized by different shapes of the deposited filaments. 115 These regimes are summarized in Figure 5 d, together with the corresponding adimensional parameters: H* = H / aD and V* = V / C, where a is the so‐called die swelling ratio which accounts for the broadening of the filament diameter upon printing ( a > 1). For instance, upon increasing the nozzle velocity, one can obtain coiling structures, then continuous filaments with progressively decreasing diameter, and ultimately discontinuous structures (Figure 5 d). 115 This rich phenomenology can be highly relevant for 3D printing of photonic components, a potential not yet fully exploited. The filamentary composition of materials printed by DIW is inherently well‐suited for the fabrication of optical waveguides. Figure 6 a–c shows exemplary top and cross‐sectional views of waveguides made of silk, 122 a biopolymer with refractive index ( n ) 1. 54 at visible wavelengths (633 nm), that can sustain light propagation when deposited on substrates with lower refractive index, such as glass ( n = 1. 52) or quartz ( n = 1. 48). These waveguides are fabricated by DIW with an ink composed of an aqueous silk‐fibroin solution (concentration 28–30 wt%), using a 5 µm glass nozzle immersed in a methanol‐rich reservoir, that induces filament solidification by favoring the transformation of amorphous random coils to stiff β‐sheets in the silk fibroin. 122 Both straight and curved silk waveguides can be printed (Figure 6 a, b), with smooth and defect‐free surface (inset of Figure 6 b) and trapezoidal cross‐section (Figure 6 c) due to a partial spread of the as‐deposited ink before complete solidification. The centimeter‐long silk waveguides can be coupled to a red laser, highlighting effective light transport (Figure 6 d–f). Propagation losses are 0. 81 dB cm −1 for wavy waveguides, and decrease to 0. 25 dB cm −1 for straight structures. 122 Being biocompatible, these printed components could be interfaced with living cells and integrated in implantable and biodegradable photonic devices. Figure 6 Waveguides printed by DIW. a) Optical microscopy image of printed silk waveguides. b) Higher magnification optical microscopy image of the curved region of the silk waveguide, as highlighted by a dashed box in (a). Inset: SEM image of the printed silk waveguide. c) Cross‐sectional optical image of a cleaved silk waveguide. d, e) Photographs showing diffused laser light coupled in straight (d) and curved (e) silk waveguides. f) Cross‐sectional view of light diffused by the cleaved surface of a silk waveguide. (a)–(f) Reproduced with permission. 122 Copyright 2009, Wiley‐VCH. g) Photograph of curved waveguides made of OrmoClear. h) Optical propagation losses versus the radius of curvature of the waveguides. i) Photograph of a network of six OrmoClear waveguides with coupled LED light at three different colors. (g)–(i) Reproduced with permission. 124 Copyright 2011, Wiley‐VCH. The DIW printing head and nozzles can be modified to generate more complex structures, 123 such as core–shell filaments. 124 This is particularly advantageous for patterning compounds with poor viscoelasticity. Indeed, such materials can be embedded in the core of filaments, with a sacrificial shell featuring viscoelastic properties suitable for DIW. Following printing, suitable treatments can be carried out to solidify the core material while dissolving or removing the sacrificial shell. Figure 6 g displays curved waveguides fabricated by printing a liquid core made of a photocurable resin (Ormoclear) with low absorption in the visible–NIR, in a shell from a viscoelastic aqueous solution of the triblock copolymer Pluronic F127. After printing and phopolymerization of the core resin, the shell is removed either by immersion in deionized water at room temperature, or by cooling at 5 °C and water rinsing. 124 The curved waveguides exhibit optical losses ≤ 1 dB cm −1 for visible light, which decrease upon increasing the radius of curvature (Figure 6 h), and they can be used to assemble printed photonic networks (Figure 6 i). Furthermore, photonic crystals can be obtained by DIW, as illustrated in Figure 7 a. Photonic crystals can be used in many device applications because of their high spectral selectivity, e. g. , in spectral filters, high efficient reflectors, optical sensors, components for enhancing the radiative rate of embedded chromophores, and cavities for lasers. 125 Polymeric and organic inks used for DIW often have relatively low refractive index ( n = 1. 5–1. 8), which hardly leads to index contrast high enough to achieve good spectral selectivity. The refractive index of the printed structure can be enhanced, however, by either depositing high‐ n materials on the surface of filaments (Figure 7 b–d) 126 or embedding high‐ n particles in the inks (Figure 7 e–h). 127 In the first method, the printed item is used as a template for delivering other materials, such as Si and oxides, which also enhances mechanical stability. Figure 7 b–d shows a photonic crystal realized by first printing a woodpile structure composed by rods with 1 µm diameter, spaced by 2. 8 µm or 4. 0 µm, and deposited from a concentrated polyelectrolyte ink. 126 This structure is then coated by a 100 nm thick SiO 2 layer. Afterward, the original organic template is removed by heating at 475 °C for 3 h. The resulting structure is composed by hollow SiO 2 rods, which are finally coated by a 100 nm thick Si layer by chemical vapor deposition (Figure 7 d). So‐produced samples feature a reflection peak at about 2 µm, in agreement with calculated spectra. 126 Figure 7 Photonic crystals by DIW. a) Scheme of a woodpile photonic crystal structure printed by DIW. b) SEM image of a Si/SiO 2 /Si hollow woodpile structure, realized by c) the multistep method schematized. d) SEM image of an individual hollow filament of the Si/SiO 2 /Si photonic crystal. (a)–(d) Reproduced with permission. 126 Copyright 2006, Wiley‐VCH. e) A BaTiO 3 /PDMS nanocomposite 3D photonic crystal upon bending. f) Scheme of the variation of photonic crystal morphology under elongational stress. g, h) SEM images of 3D nanocomposite photonic crystals as viewed from the top (g) and in cross‐section (h). i) Measured transmission terahertz spectra of 3D nanocomposite photonic crystals, realized by printing inks with various BaTiO 3 content. j) Transmission terahertz spectra of the 3D nanocomposite photonic crystals upon varying the relative elongation as schematized in (f). (e)–(j) Reproduced with permission. 127 Copyright 2017, Wiley‐VCH. The possibility of tuning the response of photonic crystals is frequently desired. In flexible photonic crystals, spectral properties can be tuned by applying elongational strains which modify the lattice geometry (Figure 7 e, f). These crystals might be printed from a nanocomposite ink, made of an elastomer (polydimethylsiloxane, PDMS) and BaTiO 3 nanoparticles added at various concentrations (10–40 wt%) to tailor refractive index as well as rheology. 127 The printed hexagonal photonic crystals in Figure 7 g, h have a bandgap around 0. 4 THz, which can be shifted to lower frequencies by increasing the amount of loaded nanoparticles (Figure 7 i). A redshift of the bandgap is also found upon elongation (Figure 7 j), due to the varied spacing distances between the rods forming the crystal. 127 This approach would allow the intensity of light signals reflected or transmitted by soft photonic crystals to be modulated upon stretching and releasing alternatively the printed structure. Processing materials at ambient temperature and without using harsh chemicals, DIW is especially suitable for use with active optical compounds and light‐emitting organics. Figure 8 shows an interesting example of a fully printed optoelectronic device realized by DIW of multiple materials. 128 This LED utilizes core–shell CdSe/ZnS QDs in its active layer, emitting green (550 nm) and orange (618 nm) light, and various other functional layers are printed in order to provide efficient charge injection and transport. More specifically, the printed QD LED encompasses a hole transport layer made of poly[ N, N′ ‐bis(4‐butylphenyl)‐ N, N′ ‐bis(phenyl)‐benzidine] (poly‐TPD), a transparent anode of poly(ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) surrounded by a ring of sintered Ag nanoparticles, which provides the necessary electrical contact, and a cathode made of eutectic gallium indium liquid metal (EGaIn). Different layers are deposited consecutively by DIW, through a selection of suitable orthogonal solvents. 128 The printed devices exhibit turn on voltages of few Volts and maximum brightness of 250 cd m −2 at 5 V and 70 cd m −2 at 8 V for the green‐emitting and the orange‐emitting LEDs, respectively. Also important is the possibility to print the LEDs on curved substrates (Figure 8 a–c). To this aim, a contact lens was first scanned (Figure 8 b), and once digitalized, it was utilized as blueprint for optimizing the conformal morphology of the various LED layers. The QD LED printed on top of the contact lens (Figure 8 c) features well‐behaved current–voltage characteristics and bright emission. By this approach, 3D‐printed optoelectronics (LEDs, photodetectors, photovoltaic components) could be integrated in complex and soft optical components (lenses, flexible biosensors, optical fibers) for medical applications, optogenetics, and components for augmented/virtual reality. Furthermore, as mentioned above, a key aspect of AM is in its capability of assembling many functional units in an interconnected and 3D fashion, well beyond standard planar architectures of optoelectronic boards and devices. Among other advantages, this would lead to artificial systems much more closely resembling biological functions. This is exemplified in Figure 8 d–f, showing a matrix of 2 × 2 × 2 QD LEDs made by printing several different material in a unique DIW process. 128 LEDs can be individually addressed in the 3D matrix (Figure 8 g). Overall, this work has evidenced the unequalled potentialities of AM for photonics and optoelectronics, in terms of materials that can be combined in a single device, use of nonflat substrates, and truly 3D integration. Figure 8 3D printed QD LEDs. a) 3D digital model of a QD LED to be printed on a curved substrate, whose scanned surface is shown in b). c) Current density versus the driving voltage of a QD LED printed on top of the surface of a contact lens. Inset: device image. Scale bar: 1 mm. d, e) Design of a 2 × 2 × 2 array of QD LEDs and of their electrical interconnects. f) Schematic illustration of the device architecture and of the layers of each QD LED component. g) Photographs of QD LED devices of the top layer (i)–(iii) and of the bottom layer (iv)–(vi). The letters (i)–(vi) highlight the assigned LED position in the 3D array, as illustrated in (d). Scale bar: 1 cm. Reproduced with permission. 128 Copyright 2014, American Chemical Society. 3. 2 Printing of Fibers by Electrospinning The direct deposition of optical materials can be also performed by spinning viscoelastic polymer solutions by means of electrostatic fields. Electrospinning (ES) 129, 130, 131, 132 utilizes an applied bias to stretch a polymer solution that is injected at a controlled flow rate through a metallic nozzle. When electrostatic forces are smaller than forces which stabilize the fluid body (viscoelasticity, surface tension), a pendant drop is formed at the nozzle and shaped into a Taylor cone. Upon increasing the applied voltage up to values high enough to overcome stabilizing forces, a continuous polymer jet is formed, which is elongated and stretched during its propagation from the nozzle to a second metallic element (collector). In the meanwhile, the solvent evaporates leading to the formation of solid and continuous filaments than can be deposited as individual fibers, as ordered 2D or 3D arrays, or as disordered networks ( Figures 9 and 10 ). 133, 134, 135, 136, 137 The main requirement for a successful ES process is having a sufficient amount of entanglements of the polymer chains in the spun solution (viscosities are in the range 10 −1 –10 3 Pa s), such to guarantee the formation of a continuous jet. 138, 139 Figure 9 Direct printing by ES. a) Schematic illustration of a typical setup used for printing by ES. b) Optical microscopy image of an array of parallel printed filaments made by poly(9‐vinyl carbazole) (PVK). Inset: SEM image of a PVK filament. Scale bar: 200 nm. c) Cross‐sectional view of a PVK fiber imaged by SEM. (a)–(c) Reproduced with permission. 133 Copyright 2013, Macmillan Publishers Limited. d, e) Confocal fluorescence microscopy image of an array of parallel filaments (d) and of a single fiber (e) made of MEH‐PPV/PEO. f) Plot of the fluorescence intensity transported through a printed fiber waveguide versus the distance, d, from the excitation spot. The continuous line is a fit to the data by an exponential decay function. Inset: fluorescence microscopy image of a MEH‐PPV/PEO fiber. The red circle highlights the excitation focused laser. Scale bar: 2 µm. (d)–(f) Reproduced under the terms of the CC‐BY license. 134 Copyright 2013, The Authors. Published by The Royal Society of Chemistry. Figure 10 Examples of photonic devices made by ES. a) SEM image of a freestanding polymer waveguide doped with single QDs. b) Fluorescence image of a waveguide embedding an isolated QD, which is highlighted by a dashed box. c) Plot of the second order correlation function, g 2 ( t ), measured by continuous wave laser excitation. The continuous line is a fit to the data by: g 2 ( t ) ∼ 1 − (1/ N )exp(| t |/τ), t is time and τ is the fluorescence decay time. The fit gives a value of g 2 ( t ) = 0. 1 at t = 0 delays, that is indicative of single photon emission. (a)–(c) Reproduced under the terms of the CC‐BY license. 135 Copyright 2016, The Authors. Published by American Chemical Society. d–i) Examples of random laser devices. SEM images of an array of polycaprolactone (PCL) filaments (d) and of PCL filaments with silica nanoparticles deposited by biomineralization (g). Scale bars: 5 µm. The corresponding emission intensity maps and emission spectra, showing the variation of the photoluminescence as a function of the pumping fluence, are shown in (e), (f) and (h), (i), respectively. Insets in (e), (h): sample photographs. Scale bar: 5 mm. (d)–(i) Reproduced under the terms of the CC‐BY‐NC‐ND license. 136 Copyright 2018, The Authors. Published by Wiley‐VCH. j) Bright light and k) UV‐excited emission photographs of an array of uniaxially aligned polymer filaments, doped with an UV dye. l–n) Photographs of emission from blue‐, green‐, and red‐emitting chromophores, respectively, in solutions excited by the UV light beam emitted by the array of polymer filaments. o) Images of the beam spot emitted by the polymer filaments at various distances, z, from the light source (top image: z = 0 mm, bottom image z = 1 mm). The beam divergence is 16. 5 mrad. (j)–(o) Reproduced under the terms of the CC‐BY‐NC license. 137 Copyright 2015, The Authors. Published by American Chemical Society. The collector of fibers can be either static or dynamic, and either uniform or patterned. Disordered 3D networks of fibers are additively deposited on static and uniform collectors, whereas dynamic (rotating and translating collectors, or combination of them) or patterned target substrates 140 allows ordered architectures to be manufactured. An important process parameter affecting the geometry achieved in deposited fibers is the distance between the nozzle and the collector. This aspect is strictly related to the complex dynamics of electrified jets. 141 In fact, the jet trajectory is roughly straight only at short distance from the nozzle (typically, a few millimeters). At longer distances, oscillations of the spinneret, atmosphere turbulence, etc. , can deviate the jet from a collinear configuration triggering the onset of large bending and whipping instabilities. 141, 142, 143 As a consequence, two different ES approaches have been developed. In conventional ES, 129, 130, 131 the collector is at large distance from the nozzle (many centimeters), and the jet instabilities prevent the precise positioning of individual filaments, which are instead collected as 3D, nonwoven mats of fibers. In the so‐called near‐field ES, 132, 133 the initially stable, collinear jet is exploited, which is obtained by greatly reducing the nozzle–collector distance compared to the conventional method, and a scanning system is used to deposit the polymer in ordered 2D and 3D structures (Figure 9 a). Overall, in ES, the order and morphology of the manufactured structures can be tuned by a proper choice of the process variables. In addition, the very high strain rates applied to polymer jets (≥10 3 s −1 ) have two main effects: the first is a considerable decrease of the jet diameter upon increasing the distance from the nozzle, roughly following a power‐law relation 139 (1) a a 0 − 2 ≅ 1 + z z 0 2 κ where a is the polymer jet radius, a 0 is the initial radius, z stands for the axial coordinate, and κ is a constant of the order of unity. The length scale of the jet radius decrease is therefore z 0 (≈ 10 −3 m), which is in turn related to the solution viscosity and conductivity, the applied flow rate, and the electric field. Solid state fibers with diameters down to 100 nm or less can be easily deposited, which are suitable for exploring phenomena occurring when light interacts with or propagates in subwavelength systems. The high strain rate also induces a preferential alignment of the polymer macromolecules along the jet, and then along the length of the deposited filaments, 144 as found by polarized Raman spectroscopy, 145 transmission electron microscopy, 146 and scanning near‐field microscopy. 147 This effect makes possible to obtain materials with highly anisotropic properties, such as optical dichroism and birefringence, 146, 148 polarized emission, 147, 149 and high charge mobility. 150 Figure 9 b, c shows ordered electrospun filaments with circular cross‐section, made of conjugated polymers. 133 Arrays of fibers can be also realized by light‐emitting polymers or blends, as in arrays obtained by near‐field ES of poly[2‐methoxy‐5‐(2‐ethylhexyl‐oxy)‐1, 4‐phenylene‐vinylene] (MEH‐PPV) mixed with polyethylene oxide (PEO, Figure 9 d, e). 134 Similarly to waveguides printed by DIW, individual electrospun filaments transport light along their longitudinal axis, up to typical distances of tens of micrometers. When made with light‐emitting polymers, or doped with low‐molar‐mass chromophores or nanoparticles, electrospun fibers can guide photons emitted by the embedded sources, with propagation losses in the range 10 1 –10 3 cm −1 (Figure 9 f). 134, 151 A recent application of these electrospun waveguides is in combination with single‐photon sources (Figure 10 a–c). 135 Here, single CdSeTe QDs are incorporated in polymer filaments, suspended across gaps to enhance confinement of light in the subwavelength fiber. The embedded QDs emit single photons with typical lifetime of 137 ns, which are coupled to the fundamental mode of the waveguide with efficiency up to 31%. 135 Such hybrid systems are relevant for building 2D and 3D networks of quantum emitters, possibly interacting through arrays of polymer waveguides. Another application of electrospun optical materials, shown in Figure 10 d–i, is based on mats of disordered, light‐scattering filaments in flexible random lasers. 136 Here, an entirely additive method has been developed for the controlled fabrication of hybrid organosilica on top of electrospun fibers, by surface functionalization by silicatein and in vitro biosilicification. The comparison of the light scattering and random lasing properties of pristine electrospun fibers (Figure 10 d–f) and of fibers decorated with silica particles (Figure 10 g–i) evidences the onset of random lasing only in those samples where silica particles are grown with a specific spatial density. 136 As a result, while pristine fibers can only support amplified spontaneous emission of a superimposed dye‐doped polymer layer (Figure 10 e, f), the substrates with silica particles might exhibit random lasing modes (Figure 10 h, i). Other devices for disordered photonics based on electrospun nanofibers exploit polymer filaments internally doped with either metallic 152 or dielectric 153 nanoparticles, as well as deterministic ring‐ 154 and cylindrical‐shaped 155 lasing resonators with randomly distributed cavity sizes. Figure 10 j shows an array of uniaxially aligned filaments of PMMA doped with an UV‐emitting dye. These flexible sources can be deposited conformally on nonplanar surfaces (Figure 10 k). 137 Upon optical excitation, the light emitted by the fibers can efficiently excite other fluorophores (Figure 10 l–n), while propagating in free space with a divergence of 16. 5 mrad (Figure 10 o). 137 Overall, ES turns out to be able to preserve the optical and electronic properties of organic materials and inorganic fillers. As a gentle AM process to realize 1D, 2D, and 3D photonic components with varied order, ES is enabling a novel range of nanowire‐based photonic devices where light emission, scattering, amplification, and transport can all be engineered. 156, 157 4 Additive Fabrication by Material Jetting Additive technologies which allow optoelectronic devices to be built with heterogeneous materials include methods based on jetting. Inkjet printing has been widely applied to realize optoelectronic devices with organic semiconductors 158 or 2D materials. 159 In inkjet printing processes, droplets are ejected through a nozzle on demand, by applying a suitable force (mechanical or electrical), and delivered onto on a substrate to define patterns. 160, 161, 162 The ink must be properly engineered to lead to stable jetting, namely to produce single droplets without the formation of secondary fluid beads. To this aim, one can consider the inverse of the Ohnesorge number (Oh): Z O = 1 /Oh = γ ρ D n / η, where D n is the nozzle diameter, and γ, ρ, η are the surface tension, density, and viscosity of the ink, respectively. Stable jetting is typically reported for 1 < Z O < 14. 159, 160, 163 The morphology of deposited inks depends on their rheology, on the substrate wettability, on the properties of the printing apparatus, and on the applied external field. Various strategies allow such morphology to be directed with accuracy, including the suppression or the exploitation of coffee‐ring effects, the use of liquid substrates or of patterned solids, and the control of the extent of droplet coalescence. 164 Since materials are deposited from solutions, the solvent evaporation is critically important to determine the ultimate morphology, crystallinity, and thickness uniformity of layers. These can be tailored, for instance, through in situ thermal treatments, by substrate heating. In combination with an optimized chemical formulation of the ink, such approach allows solar cells based on perovskites to be realized, showing power conversion efficiency up to 12. 3%. 165 The versatility of the inkjet technology in terms of solvents and chemical additives, which can be chosen to better preserve optoelectronic properties of active compounds, and the availability of commercial desktop printers, make this method widely used for the deposition of layers in photovoltaic devices, 166, 167 visible and infrared photodetectors, 163, 168, 169, 170, 171 rewritable structural color displays, 172 saturable absorbers, 163 luminescent patterns, 168, 173 nanocavities, 174 microring resonators, 175 and microlasers. 176 Interested readers are referred to outstanding reviews in the field. 160, 161, 162, 177, 178 Structures realized by inkjet technique are mainly 2D. Interestingly, 3D optical components can be realized by the polyjet technology, combining inkjet printing of photocurable inks and local UV photopolymerization. Example of optical components realized by the polyjet method include 3D, mechanically tunable gyroid photonic crystals, and waveguides for terahertz radiation. 179, 180 5 Other Solution‐Based and Hybrid Technologies The variety of AM technologies is continuously increasing, and new methods or combinations of exposure‐ and extrusion‐based approaches are being developed to fulfill the demanding requirements of photonics and optoelectronics. These include techniques for 3D printing of glass, such as selective laser melting/sintering 181 and binder jetting processes. 182 The structures obtained in this way are often opaque, because of trapped bubbles that limit optical transparency by scattering of light. Recently, filament‐fed laser heating has been developed for 3D printing of transparent quartz and soda lime glass. 183, 184 In this approach, a glass filament is continuously fed, locally heated by a focused CO 2 laser, and once molten, it is deposited layer‐by‐layer. The use of a solid and continuously fed glass filament limits the formation and the entrapment of bubbles in the printed structures, improving significantly the transparency of the final objects. 181 By this approach, quartz flat and curved components with transparency exceeding 90% (corresponding to an extinction coefficient of 13. 2 m −1 ) are demonstrated upon surface polishing. 184 In addition, refractive index has been measured in printed quartz windows and found to be homogeneous. 184 A recently developed approach for the fabrication of 3D optical interconnects is the so‐called meniscus‐guided 3D writing, where the meniscus of a monomer or polymer solution is drawn by an atomic force microscope tip 185 or by a micropipette 186, 187 to form filamentary structures ( Figure 11 a). This method is very effective for the fabrication of freestanding 3D waveguides (Figure 11 b) with subwavelength diameter and low propagation losses (1–10 dB mm −1 ). 187 The out‐of‐plane design allows optical losses due to light coupling into the substrate underneath to be greatly suppressed. Interestingly, stacked waveguides and optical interconnects can be printed with no contacts (Figure 11 c), as well as interconnects spanning through gaps (Figure 11 d) or over steps (Figure 11 e). Figure 11 f, g illustrates a potential application of meniscus‐guided 3D writing, that is printing of 3D, out‐of‐plane optical connections between nanostructured photon sources. Here a polystyrene waveguide (P1) is realized to bridge two ZnO nanorods (Z1 and Z2). Upon laser excitation, the printed waveguide transports light emitted by Z1 to Z2, with no coupling with the underlying substrate (Figure 11 h). Freestanding multiterminal and branched optical interconnects can also be printed (Figure 11 i, j), with complete merging and continuous connection between touching waveguides. By this method, optical splitters and multiplexers can be printed at high density, fully exploiting the volume out of the substrate plane. 187 Figure 11 3D optical waveguides. a) Schematics of the 3D printing process based on the stretching of the meniscus of a polymer solution. b) SEM image of a polystyrene 3D waveguide. Insets: zooms of the sample regions highlighted by the white and black box, respectively. c) SEM image of crossed 3D waveguides. The insets show magnified views of the crossing area. d, e) Examples of 3D waveguides printed through a gap (d) and a step (e). f) Schematic illustration of a 3D‐printed interconnect between ZnO nanorods. g) SEM image of a printed waveguide (P1) connecting two ZnO nanorods (Z1 and Z2) and h) corresponding fluorescence map, obtained by exciting the Z1 nanorod by a focused laser as schematized in (f). The fluorescence map evidences the optical coupling between Z1 and Z2 through P1, whereas no signal is observed from Z3. i) Schematics of a multibranched 3D optical interconnect. j) SEM image of a freestanding, multibranched, printed waveguide. The insets show a magnification of the branching point and of the area of coupling to nanoscale photon sources (NPSs), as highlighted by the white and black dashed boxes, respectively. Scale bars: 2 µm. Reproduced with permission. 187 Copyright 2016, Wiley‐VCH. A different approach for building 3D optical components is given by contact printing, where a suitable ink is transferred to a substrate by means of a rigid or elastomeric patterned mold. A relevant example is the well‐known microcontact printing, 188 where the mold is inked with the material of interest (surface modifiers, organic molecules, proteins, etc. ), and placed in conformal contact with a target substrate. The transfer of the ink to the target occurs in correspondence of the protruding features of the mold, provided that the interactions between the ink mole‐cules and the substrate are stronger than the adhesion between the ink and the mold. Elastomeric molds are typically used for the spatially selective delivery of molecules, which allows conformal contact with the target substrate to be easily obtained and elements to be fabricated by soft lithography at very low cost. 188 Many substrates can be used, including metals, inorganic semiconductors, and oxides, 188, 189 and printing can be carried out sequentially in a layer‐by‐layer fashion in order to build 3D structures. 190 Interestingly, selective deposition of different molecules can be accomplished, using diverse inks simultaneously. Applications include printing of colloidal crystals and nanostructures for antireflection coatings, 191 and bilayers of QD donor–acceptor systems with tight spatial control of the nonradiative energy transfer. 192 The laser‐induced forward transfer (LIFT) is a noncontact 3D printing and assembly process, based on the transfer of a material from a transparent donor (carrier) to a receiving surface, as schematized in Figure 12 a. 193, 194 The material to be transferred can be a highly viscous composite paste, a polymer film, a complex fluid, a metal oxide, a semiconductor, etc. , 194, 195 whereas the shape of the elementary transferred building block (voxel) is determined by the spatial profile of the used laser, which can be changed in real‐time by using physical masks or spatial light modulators. Both the carrier and the receiving substrates can be translated along prescribed paths in order to print complex 2D or 3D structures (Figure 12 b–e). In a typical LIFT process with a solid donor layer, above a threshold fluence the incident laser generates a gas pocket between the carrier and the donor film. During expansion, the stress at the edge of the irradiated area increases, and a mechanical fracture of the donor film takes place when its ultimate stress is surpassed. At this point, the delaminated area is propelled toward the receiving surface. Typical requirements for the donor film are therefore a high absorption coefficient at the incident laser wavelength and high mechanical resistance upon bending to avoid internal rupture of the irradiated area during the deflection caused by expansion of the gas. 194, 196, 197 Microdisks of a positive photoresist (S1813) have been transferred onto flat and curved surfaces to form arrays of microlenses. 196 To this aim, the microdisks have been heated up to their glass transition temperature, which allow the formation of plano‐convex microlenses through thermal reflow (Figure 12 f–i). LIFT is highly versatile, allowing materials with diverse physicochemical properties to be assembled, and the geometries of the unitary building block to be varied during device building at variance with most AM methods, where the unitary building block is fixed by the laser spot size, the nozzle, or the printhead shape. LIFT can be also exploited for assembling fully functional optoelectronic devices, for instance, by printing interconnects on top of LEDs (Figure 12 j–n). 193 Polymer light‐emitting diode (PLED) pixels, with size 0. 6 × 0. 5 mm 2, have been realized by using LIFT and a forward transfer process applied to a multilayer. 197 The multilayer was composed by the Al cathode and the active layer (MEH‐PPV) of the devices, while the receiving surface was either indium tin oxide (ITO) or ITO coated with PEDOT:PSS. 197 PLED pixels can be also fabricated by sequential LIFT applied to the active layer and the Al cathode. 198 Other relevant examples of photonic devices made by LIFT include terahertz metamaterials, realized by forward laser printing of Ag split‐ring resonators on Si substrates. 199 Figure 12 Laser‐induced forward transfer of optoelectronic and micro‐optical components. a) Scheme of the experimental setup for 3D printing by LIFT. b–e) SEM of various structures realized by 3D printing of voxels composed by a silver paste, and assembled to form a bridge on a Si channel (width 100 µm) (b), a multilayer scaffold (c), a pyramid (d), and high‐aspect ratio pillars (e). (a)–(e) Reproduced with permission. 193 Copyright 2010, Wiley‐VCH. f–i) Examples of arrays of microlenses manufactured by laser transfer. SEM images of an array of closed‐packed microlenses (f), of periodically arranged dimers of microlenses (g), and of an array of lenses printed on the surface of a bent substrate (h) and of a glass capillary (i). (f)–(i) Reproduced with permission. 196 Copyright 2018, Wiley‐VCH. j, k) Images of a LED embedded in a polyimide substrate before (j) and after (k) printing of the interconnects. l, m) SEM images of the printed interconnects. n) Image of the manufactured LED under operation. (j)–(n) Reproduced with permission. 193 Copyright 2010, Wiley‐VCH. The capability to manipulate fluids by micro‐ and nanofluidics is another route for the controlled delivery of materials with high precision. This can be done through networks or arrays of micrometric or sub‐micrometric channels, with interesting implications for building optical and photonic components additively. Examples include the fabrication of arrays of mesostructured optical waveguides with either straight or curved geometry, by using a patterned PDMS mold in conformal contact with a flat substrate. 200 These channels can then be filled by sol–gel block copolymers placed at one of the open ends and penetrating by capillary flow. Full cross‐linking and consolidation of the silica network occur in about 12 h. After that time, the PDMS mold can be removed and an array of waveguides is formed. Upon doping with rhodamine and optical pumping, amplified spontaneous emission at 577 nm was observed from these systems. 200 Microfluidics methods are especially effective for the fabrication of photonic components with anisotropic properties, because flows in channels can favor molecules in adopting an elongated, almost 1D conformation. Such molecular alignment has been observed for conjugated polymers flowing within channels with sub‐100 nm transversal size, and the resulting nanostructures have been found to feature anisotropic optical properties, namely to emit light polarized in a direction parallel to their length. 201 Finally, it is worth mentioning that hybrid approaches combining microfluidics and UV and two‐photon photopolymerization have been developed for the continuous production of 3D fluorescent micro‐objects. These methods, known as continuous flow lithography, 202, 203, 204 are schematized in Figure 13 a, b. A prepolymer doped with fluorescent particles or chromophores is cured locally by the use of a light beam during its flow through the channel. Either UV exposure through a mask (Figure 13 a), or two‐photon polymerization (Figure 13 b) can be used. The geometry of the printed micro‐objects can be changed in real time by shaping the light beam through a digital mirror device, 205 instead of using fixed physical masks. 3D micro‐objects with complex shapes can be realized in this way, including fluorescent helical particles (Figure 13 c, d) with surface roughness < 10 nm. 204 In addition, various prepolymer solutions, featuring different emission wavelengths, can be combined in a microchannel where the laminar coflow provides sharp interfaces between them due to poor, diffusion‐limited mixing. 206 By photopolymerization across the interface of coflowing solutions, 3D particles with various encoded optical patterns can be realized (inset of Figure 13 a). 203 For instance, lanthanide ion nanocrystals, featuring distinct, visible light emission by upconversion upon NIR optical pumping, have been incorporated in such particles. The resulting large anti‐Stokes shifts leads to straightforward spectral separation and minimal interference with the excitation radiation, and improved signal‐to‐noise ratios. 203 Figure 13 e shows an example of application of these microparticles as anticounterfeiting barcodes, which can be embedded in pharmaceutical packages, banknotes, credit cards, ceramic objects, artworks, and high‐temperature cast items made of poly‐styrene (top panels in Figure 13 e). The particles can be easily imaged with a smartphone and a 20× objective. In absence of NIR excitation, no fluorescence signal is detected (lower panels in Figure 13 e), whereas bright, NIR‐excited luminescence (middle panels) allows the identification of the embedded barcodes unambiguously. Figure 13 3D light‐emitting microparticles. a) Schematic diagram of a continuous flow lithography system, utilizing multiple coflows of monomers with lanthanide‐doped fluorescent nanoparticles. The inset shows fluorescence images of various manufactured particles. Reproduced with permission. 203 Copyright 2014, Macmillan Publishers Limited. b) Schematics of two‐photon continuous lithography. c) Bright field optical microscopy image of a 3D helical particle made by two‐photon continuous flow lithography. d) The corresponding fluorescence microscopy image is shown. (b)–(d) Reproduced with permission. 204 Copyright 2012, Wiley‐VCH. e) Fluorescent microparticles as encoded barcodes in anticounterfeiting. The imaging system utilized a portable detector (Apple iPhone 4S with a 20× objective), as shown in the left image. The particles were embedded in various objects as shown in the top panels (from left to right: blister packs, banknotes, credit cards, curved ceramic objects, artwork, and high‐temperature‐cast polystyrene objects). The middle and bottom panels show the corresponding acquired images upon excitation with a 980 nm laser (excitation power: 1W) and without laser excitation, respectively. Reproduced with permission. 203 Copyright 2014, Macmillan Publishers Limited. 6 Conclusion and Outlook AM technologies are introducing new design and production rules for both passive and active optical materials, for the fabrication of photonic and optoelectronic devices, and for the interwoven integration of photonic components. The progress of this field is supported by the continuous expansion of the range of materials that can be processed, by the introduction of new fabrication methods, and by the improvement in the achievable spatial resolution, as summarized in Table 1. The variety of 3D‐printable optical materials, and the possibility to shape them across many lengthscales (from hundreds of nanometers to centimeters) are enabling the design of photonic devices (such as photonics crystals and waveguides) with frequency response ranging from the visible to terahertz and microwave. Recent relevant breakthroughs include the possibility of printing glass‐based components, with transparency, thermal and mechanical resistance comparable to commercially available glass. 43, 117, 118, 183, 184 Fully 3D printed LEDs have also been demonstrated, by intriguing approaches for manufacturing the various needed layers, as well as photonic devices conformable to complex, nonplanar surfaces. 128 Combination of different technologies turns out to be a valid route for multimaterial processing, and for high‐throughput production of 3D objects with calibrated optical properties. Currently, fluorescent 3D microparticles can be produced at a rate of 10 4 –10 5 particles h −1, 203, 206 by a combination of microfluidics and either UV or two‐photon photopolymerization. Table 1 Additive technologies developed for manufacturing of photonic and optoelectronic components Additive technology Commercial printers available Building volume a) [cm 3 ] Feature size [µm] Printing speed a) [mm s −1 ] Optical/electronic material demonstrated Multimaterial capability Demonstrated photonic and optoelectronic devices Laser stereolithography Yes 10 3 –10 6 10–100 a) – Acrylates, epoxies, nanocomposites No Optical windows Digital light processing Yes 10 2 –10 4 30–100 a) – Acrylates, glasses b) No Lenses, optical windows, optical sensors Continuous digital light processing Yes 10 2 –10 3 30–100 a) 0. 03 Acrylates No – Multiphoton stereolithography Yes 3 × 10 −5 –3 × 10 2 0. 2–10 a) 0. 1–10 Acrylates, epoxides, nanocomposites No Microlenses, optical sensors, photonic crystals, waveguides, metamaterials Fused deposition modeling Yes 10 3 –10 6 100–500 a) 10–10 3 Thermoplastic polymers, nanocomposites Yes Microwave and terahertz lenses Direct ink writing Yes 10 2 –10 3 10–100 a) 0. 1–10 2 Glasses, b) acrylates, epoxies, silk, elastomers, nanocomposites, conjugated polymers, conductive inks Yes Waveguides, photonic crystals, LEDs, optical windows Electrospinning Yes 1–10 2 0. 1–1 a) 1 Thermoplastic polymers, conjugated polymers, nanocomposites Yes Waveguides, single‐photon sources, optical amplifiers, optical sensors, lasers Inkjet Yes Typical printing area 10 2 cm 2 20–80 a) – Conjugated polymers, conductive inks, perovskites, block copolymers, 2D materials Yes Photodiodes, solar cells, organic LED, saturable absorbers, optical resonators, lasers, color displays Polyjet Yes 10 4 16–30 a) – Acrylates Yes Waveguides, photonic crystals Filament‐fed laser heating process No – 300 c) – Glasses No Optical windows Meniscus‐guided 3D writing No – 0. 1–1 d) – Thermoplastic polymers No Waveguides Microcontact printing Yes Typical printing area 10 2 cm 2 0. 05 a) – Quantum dots, colloids, proteins Yes Luminescent patterns Laser‐induced forward transfer No – 0. 5–10 e) – Photoresists, conductive pastes, conjugated polymers Yes Microlenses, metal contacts, terahertz metamaterials, polymer LEDs Micro‐ and nanofluidics No – 0. 07 f) – Conjugated polymers, sol–gel, block copolymers Yes Waveguides, lasers Continuous flow lithography No – 0. 4 g) – Acrylates, nanocomposites Yes Luminescent microparticles, barcodes a) Data are for commercially available printers b) Postprocessing thermal treatment needed c) Ref. 184 d) Ref. 187 e) Ref. 194 f) Ref. 201 g) Ref. 204. John Wiley & Sons, Ltd. Despite such progress, some challenges are still open, as highlighted in Table 2, and AM technologies for photonics and optoelectronics still remain matter of research rather than production tools. To move from laboratories to factories, first throughput has to be improved. For instance, while achieving high spatial resolution and good surface finishing, multiphoton processes proceed by scanning point by point in an individual layer, building up the 3D shape a single layer at a time. 207 The efficiency of this process depends on the laser power available, the speed of the scanning system, and also the photopolymer. Since the technological opportunities opened by MP‐STL are unique, a lot of research is currently focused on the improvement of its throughput by developing more efficient photoinitiators, as well as higher power lasers and faster motion systems. In addition, efforts are devoted to improve the throughput of the technique by developing high aspect ratio MP‐STL, 104 by using multiple beams, 105, 106 and by moving toward holographic MP‐STL. 107 Other issues to be addressed in the near future are those associated with i) multiscale printing, ii) multimaterial processing, and iii) multidimensional integration of optoelectronic and photonic devices. Table 2 Main challenges of AM technologies for fabrication of photonic and optoelectronic devices Additive technology Optical properties Spatial resolution Interface uniformity Uniformity of structure Laser stereolithography Postprocessing treatment needed Low – Rough surface and postprocessing treatment needed Digital light processing Postprocessing treatment needed Low – Rough surface and postprocessing treatment needed Continuous digital light processing – Satisfactory – Smooth surfaces Multiphoton stereolithography Homogeneous optical properties of printed structures Good Satisfactory for components with heterogeneous materials Smooth surfaces Fused deposition modeling Postprocessing treatment needed Low – Rough surface and postprocessing treatment needed Direct ink writing Properties of pristine materials preserved Satisfactory Printing of functional multimaterials in individual devices demonstrated Tailorable upon control of process variables Electrospinning Anisotropy of optical properties observed Satisfactory Satisfactory for multimaterial processing Tailorable upon control of process variables Inkjet Properties of pristine materials preserved Low In situ and/or postprocessing annealing needed Tailorable upon control of process variables Polyjet Properties of pristine materials preserved Low Satisfactory for components with heterogeneous materials Tailorable upon control of process variables Filament‐fed laser heating process Homogeneous optical properties of printed structures Low – Rough surface and postprocessing treatment needed Meniscus‐guided 3D writing Properties of pristine materials preserved Good Satisfactory for components with heterogeneous materials Smooth surfaces Microcontact printing Properties of pristine materials preserved Good Satisfactory for components with heterogeneous materials Smooth surfaces Laser‐induced forward transfer Properties of pristine materials preserved Low Postprocessing annealing needed Smooth surfaces after postprocessing treatment Micro‐ and nanofluidics Anisotropy of optical properties observed Good Satisfactory for components with heterogeneous materials Smooth surfaces Continuous flow lithography Properties of pristine materials preserved Good Sharp interfaces Smooth surfaces John Wiley & Sons, Ltd. Multiscale Printing : Photonics and optoelectronics have stringent requirements in terms of minimum size of the device features and they are currently fully accessible only by MP‐STL. At the same time, most of optical systems include components with macroscale size, thus needing AM technologies which can shape materials at different spatial scales. The resolution achieved nowadays by MP‐STL is at the nanometer scale, mainly due to techniques employing radical depletion, the highest resolution reported is an impressive 9 nm. 56 However, this is done using high NA, oil‐immersed objectives with working distance below 200 µm, limiting both the area that can be patterned and the height of the structures to be built. This issue has been partly addressed by techniques such as “dip‐in” lithography 208 or MP‐STL with wider objective working range (WOW), 104 where the objective is immersed inside a liquid photoresist, making the working distance irrelevant. However, these techniques have their drawbacks: dip‐in lithography requires that the photopolymer refractive index matches the n value of the objective, limiting dramatically the materials that can be used. This is particularly important when the desired structures need to have a specific bulk functionality, such as nonlinearity or metal affinity. WOW MP‐STL, which uses a sealed oil‐immersed objective, can be used only with liquid resists, while the majority of materials used in photonics and optoelectronics are gels. 64 Other pulse‐shaping techniques have produced high aspect ratio structures with adequate resolution. 209 Recently, a new method employing simultaneous spatiotemporal focusing of femtosecond laser pulses has been used to make centimeter‐scale structures with 10 µm resolution. 210 The fabrication of macroscale structures with sub‐micrometer resolution using materials with specific functionality remains elusive. Conversely, macroscale systems can be built by other AM approaches, tough with poor resolution, often not suitable for optical applications. One can image that in the future novel methodologies or combinations of current AM processes will be conceived, hopefully with integrated in situ process‐sensing elements, 211 which will allow macroscale optical components with high spatial precision and uniformity to be realized. In this respect, a paradigm shift is probably necessary, more specifically methods no more based on the layer‐by‐layer approach, but instead based on continuous printing such as the continuous liquid interface production 48 and the continuous flow lithographies. 202, 204 While enhancing the printing speed, continuous processes have been demonstrated to be effective in improving the surface roughness of the manufactured objects, down to tens of nanometers. 204 Multimaterial Processing : Simultaneous processing of diverse materials is of paramount relevance in photonics and optolectronics, where metal contacts, semiconducting active layers, transparent windows or waveguides, and dielectric or metal mirrors have to be frequently combined in complex systems. While DIW can currently manage different materials in a unique process, 128 research efforts are still required in order to develop inks with advanced optoelectronic properties, such as optical gain, optical nonlinearity, efficient light‐to‐current (and current‐to‐light) conversion. For exposure‐based printing processes, the route is harder to some extent, because these methods can normally process one material at a time (typically passive and with limited optical functionality). Printed materials with graded composition and optical properties are still lacking. All such issues are critical for the future development of environmentally responsive optical systems, namely adaptive components which can self‐tune their properties in response to external stimuli. 120, 212, 213 Multidimensional System Integration : Photonics and optoelectronics have been traditionally developed in planar configurations, with unavoidable limitations in terms of device architecture and density. AM technologies can potentially extend optoelectronic systems out of plane, but little has been done in this direction so far. This requires new design rules for single devices, as well as for their interconnects. Fully 3D waveguides with bent geometries can in principle lead to a more efficient connection of multiple optically active elements. 187 Overall, there is already a lot of work on employing AM in a variety of photonic and optoelectronic applications, and the main challenge remains to move from prototyping to production. One of the most exciting areas in research is related to architected materials with targeted optical properties, that are defined by material architecture and not by material chemistry or composition. These systems defy common associations, such as mechanical strength and density, or optical absorption and bandgap. Photonic crystals and electromagnetic metamaterials presented above are only a small part of the current revolution in architected materials. More efforts can be done in this direction, as currently being pursued in metamaterials with targeted mechanical properties. 214, 215, 216, 217, 218, 219 There are still numerous opportunities for enhancing the properties and interconnectivity of optoelectronic devices, by exploring novel, fully 3D architectures and AM of optically active materials. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/adsu. 202000167
| 2,020
|
Advanced sustainable systems
|
Strategies for Fabricating Protein Films for Biomaterials Applications
|
Proteins are naturally occurring functional building blocks that are useful for the fabrication of materials. Naturally-occurring proteins are biodegradable and most are biocompatible and non-toxic, making them attractive for the fabrication of biomaterials. Moreover, the fabrication of protein-based materials can be conducted in a green and sustainable manner due to their high aqueous solubility. Consequently, the applicability of protein-based materials is limited by their aqueous and mechanical instability. This review summarizes strategies for the stabilization of protein films, highlighting their salient features and potential limitations. Applications of protein films ranging from food packaging materials, tissue engineering scaffolds, antimicrobial coatings etc. are also discussed. Finally, the need for robust and efficient fabrication strategies for translation to commercial applications as well as potential applications of protein films in the field of sensing, diagnostics and controlled release systems are discussed.
|
No full text available
|
10. 1002/adtp. 202000203
| 2,020
|
Advanced Therapeutics
|
Targeted Nanotherapeutics for Respiratory Diseases: Cancer, Fibrosis, and Coronavirus
|
Abstract Systemic delivery of therapeutics for treatment of lung diseases has several limitations including poor organ distribution of delivered payload with relatively low accumulation of active substances in the lungs and severe adverse side effects. In contrast, nanocarrier based therapeutics provide a broad range of opportunities due to their ability to encapsulate substances with different aqueous solubility, transport distinct types of cargo, target therapeutics specifically to the deceased organ, cell, or cellular organelle limiting adverse side effects and increasing the efficacy of therapy. Moreover, many nanotherapeutics can be delivered by inhalation locally to the lungs avoiding systemic circulation. In addition, nanoscale based delivery systems can be multifunctional, simultaneously carrying out several tasks including diagnostics, treatment and suppression of cellular resistance to the treatment. Nanoscale delivery systems improve the clinical efficacy of conventional therapeutics allowing new approaches for the treatment of respiratory diseases which are difficult to treat or possess intrinsic or acquired resistance to treatment. The present review summarizes recent advances in the development of nanocarrier based therapeutics for local and targeted delivery of drugs, nucleic acids and imaging agents for diagnostics and treatment of various diseases such as cancer, cystic fibrosis, and coronavirus.
|
1 Introduction Lung diseases are one of the main causes of death among both men and women worldwide. The mortality rates for lung diseases have been increasing by each year. [ 1, 2 ] Therefore, methods of developing new therapeutic solutions as well as improving the current therapies for the common lung diseases such as asthma, cystic fibrosis, chronic obstructive pulmonary disease, lung cancer, and coronavirus infections remain the main focus in the fields of targeted drug delivery. The widely utilized conventional drug delivery methods usually induce adverse side effects. [ 3, 4 ] Recent development of nanoscale‐based systems opens a door for a better delivery of therapeutics by addressing the limitations of conventional therapy. Nanocarrier‐based drug delivery systems can increase bioavailability of poorly water‐soluble therapeutics and address other barriers and shortcomings of traditional drugs. [ 5 ] While majority of the marketed drugs are poorly water soluble, which limits their administration at high doses, [ 6, 7 ] nanoscale‐based drug delivery systems were found to improve solubility and increase therapeutic efficacy of free non‐bound drugs. [ 8, 9 ] Similarly, biomacromolecular therapeutics such as nucleic acids (DNA, small interfering RNA, antisense oligonucleotides, etc. ) are usually degraded in the biological fluids and difficult to deliver at their target site. [ 10, 11 ] However, delivery of nucleic acids via nanocarrier based systems increased their stability and concentration at the target site. [ 12, 13, 14, 15, 16 ] Moreover, nanoscale drug delivery systems can be administered via different routes, such as intravenous, [ 17, 18 ] oral, [ 19, 20 ] and inhalation [ 15, 16, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30 ] routes. Furthermore, nanoscale delivery systems are less toxic and immunogenic than the traditional viral vector‐based gene delivery systems. [ 31, 32, 33 ] Because of such handful advantages, researchers around the world have been applied significant efforts in recent years to develop various nanosized carriers for targeted delivery of therapeutics. Over the decade, several nanocarrier‐based therapeutics were also approved by FDA for clinical application. [ 34, 35 ] Thus, the recent development of wide spectrum nanoscale systems introduced a new way in diagnosis, treatment and prevention of diseases. Previously, we formulated main requirements and basic concept of effective drug and nucleic acid delivery systems for effective treatment of diseases including cancer. [ 36 ] To enhance treatment effectiveness, the advanced system should provide for a 1) protected delivery of active components in order to prevent their degradation during its journey to targeted cells; 2) targeted transport specifically to the site of action with the aim of limiting adverse side effects of treatment upon healthy organs, tissues, and cells; 3) modulation of pump drug resistance with the purpose to prevent drug efflux from the diseased cells; 4) suppression of nonpump resistance in order to overcome other resistance mechanisms non related to drug efflux pumps; and 5) controlled release of active components in a predefined desired manner. [ 36, 37, 38, 39, 40, 41, 42, 43, 44, 45 ] These main requirements concretized for cancer treatment and general composition of advanced proapoptotic drug delivery system are shown in Figure 1. In this review, we summarize recent reports on the development of various nanotherapeutics including nanocarrier‐based drugs and nucleic acids for the treatment of lung diseases with emphasis on lung cancer, idiopathic pulmonary, and cystic fibrosis and recent coronavirus infections. [ 15, 16, 23, 24, 46, 47, 48, 49, 50 ] Figure 1 A design principle and typical structures of an advanced targeted proapoptotic drug delivery systems for effective treatment cancers with limiting adverse side effects upon healthy organs, tissues, and cells. 2 Targeted Nanotherapeutics for Lung Diseases: Active and Passive Targeting Targeted drug delivery of therapeutics is aimed at transporting of the administered active component predominately at a desired site of action limiting its accumulation in healthy organs and tissues. This goal can be achieved by both passive and active targeting of drugs. [ 37, 52, 53, 54 ] In passive targeting, high molecular weight substances are accumulated in targeted cells because of specific pathophysiological characteristics of the diseased cells and surrounding microenvironment. For instance, passive targeting to solid tumors is dependent on the enhanced permeability of vessels that supply blood, oxygen, and nutrients to the tumor and limited lymphatic drainage from the tumor environment. This phenomenon was termed as the enhanced permeability and retention (EPR) effect. [ 55 ] A schematic representation of the EPR effect is displayed in Figure 2. However, the efficiency of passive targeting is limited and the conditions responsible for the EPR effect is not attributed for all diseased tissues which often develop specific mechanisms for resisting traditional treatment approaches. [ 56 ] On the other hand, active targeting is achieved mainly by decorating the surface of the nanocarriers with targeting moieties such as antibodies, [ 57 ] proteins, [ 58 ] peptides, [ 59 ] aptamers, [ 60, 61, 62 ] lectins, carbohydrates, and glycoproteins, [ 63 ] small molecules, [ 64, 65, 66, 67, 68 ] etc. which have strong affinity to their cellular binding partners such as tumor antigens, cell surface receptors, tumor vasculature. [ 69 ] Such active targeting nanocarriers are designed to increase the accumulation of therapeutics at their site of actions with limiting exposure to other healthy organs thereby reducing the risk of adverse side effects. [ 70, 71, 72, 73, 74, 75 ] Nanocarrier‐based passive and active targeting of cancerous cells and their advantages are presented in Figure 3. Several receptor proteins (such as receptors specific for folate, LHRH, transferrin, etc. ) are often overexpressed in tumor cells. A schematic demonstration of various overexpressed cell surface receptors in lung under pathological conditions is outlined in Figure 4. Western blot analysis revealed the presence of the targeting receptor proteins such as folate receptor alpha (FRA), epidermal growth factor receptor (EGFR), integrin etc. in different lung cancer cells as shown in the right panel of Figure 4. Folic acid, transferrin etc. were widely explored as affinity ligands for targeting many tumors cells. [ 76, 77, 78 ] For example, recently, researcher developed a gold nanocarrier loaded with drug Aurimmune CYT‐6091 and functionalized with tumor necrosis factor alpha (TNF‐α) using polyethylene glycol (PEG) linker for treating lung cancer. TNF‐α was used as both targeting and therapeutic agent. [ 79 ] Bind‐014 is another nanocarrier based therapeutics which was investigated in phase II clinical trial for patients with non‐small cell lung cancer. [ 80 ] Bind‐014 is a polylactic acid (PLA) based nanocarrier system wherein the anticancer drug docetaxel was entrapped. The surface of this system was coated with PEG and targeting ligands against prostate‐specific membrane antigen (PSMA) which is usually abundant in prostate cancer cells as well as in the non‐prostate cancers, such as NSCLC. [ 81 ] Figure 5 A represented a schematic illustration of BIND‐014 composed of a hydrophobic PLA polymeric core and a hydrophilic PEG corona decorated with small‐molecule targeting ligands, and an encapsulated anticancer drug docetaxel. [ 82 ] The CT scans acquired from a patient suffering from a primary cholangiocarcinoma revealed regression of the lung metastases after two cycles of BIND‐014 treatment (Figure 5B ). [ 83 ] These results indicated that Bind‐014 was clinically active and nontoxic for NSCLC. Figure 2 A schematic representation of the Enhanced Permeability and Retention (EPR) effect. The leaky vasculature and dysfunctional lymphatics of tumors allow the preferential accumulation and retention of high molecular weight nanoparticles in solid tumors. Reproduced with permission. [ 51 ] Copyright 213, Hindawi. Figure 3 Passive and active targeting cancer cells. Passive targeting depends on the enhance permeability and retention (EPR) effect. Active targeting may be achieved by enabling the uptake of nanotherapeutics by receptor mediated endocytosis. For instance, nanoparticles may be decorated with a ligand targeting receptors (or other molecules) overexpressed on the plasma membrane of cancer cells. Figure 4 Different plasma membrane receptors overexpressed in lung cancer cells. Western blots on the right panel show the expression of these receptors in different lung cancer cells. Reproduced with permission. [ 48 ] Copyright 2017, Springer. Figure 5 The impact of the nanotherapeutic BIND‐014, a polymeric nanodrug carrier, encapsulating docetaxel, and targeting prostate specific membrane antigen on human drug metastases from a primary cholangiocarcinoma. A) A schematic representation of BIND‐014 composed of a biodegradable and hydrophobic PLA polymeric core and a hydrophilic PEG corona decorated with small‐molecule targeting ligands, and an encapsulated anticancer drug (docetaxel). Adapted with permission. [ 82 ] Copyright 2013, Dovepress. B) Representative axial images from contrast‐enhanced CT scans obtained from a patient with lung metastases at baseline and at day 42 after two treatment cycles of BIND‐014. Red circles indicate locations of metastatic lesions observed in the baseline CT scan. Adapted with permission. [ 83 ] Copyright 2012, American Association for the Advancement of Science. 3 Nanotherapeutics for Lung Diseases 3. 1 Lung Cancer Lung cancer is the prime cause of cancer death worldwide. Chemotherapy, radiation therapy and their combination with surgery are the current therapeutic options for all type of lung cancers. [ 84 ] In most cases of lung cancer chemotherapy, drugs are often administered intravenously, and they can circulate throughout the body affecting both normal and cancer cells. Over the last two decades, various nanoparticles (such as metal‐based nanoparticles, lipid‐based nanoparticles, polymeric nanoparticles, etc. ) were explored for targeted therapeutic delivery and diagnostic applications (or combination of both in one theranostic system). [ 51, 85, 86 ] However, the translational application of naked metallic nanoparticles (as imaging contrast agents) is limited due to their toxicity. [ 87, 88 ] Therefore, lipid and polymer‐based nanoparticles have received more attention of the researchers for drug delivery and theranostics applications. In this review, we will summarize recent reports on the development of lipid and polymer based nanocarriers for targeted delivery of drugs and nucleic acids for the treatment of lung cancer. 3. 1. 1 Drug Delivery Drugs can be either encapsulated physically or bonded chemically through linker with the nanocarriers and can be delivered to almost all organs because of their small size and ease of penetration of many biological barriers. Over the years, a broad range of nanocarriers were evaluated for targeted delivery of several anticancer drugs for lung cancers. Few examples of lipid‐based nanoparticles as well as polymeric nanoparticles for targeted drug delivery applications have been summarized below. Lipid‐Based Nanoparticles Lipid‐based nanoparticles possess unique benefits necessary for drug delivery application. Lipid‐based nanoparticles have an advantage of being the least toxic among other nanocarriers with a substantial progress in the fields of drug and nucleic acid delivery using lipid‐based nanoassemblies. [ 89 ] Here, we have summarized recent reports of several lipid‐based nanosystems including liposome, nanostructured lipid carriers (NLCs), and micelles and their application as targeted drug delivery systems. Liposomes : Liposome is a type of lipid‐based nanoparticle with a bilayer structure comprised of phospholipids, phosphatidylcholine, cholesterol, etc. ( Figure 6 ). Liposome can incorporate lipid‐soluble drugs in its lipid bilayer structure as well as encapsulate water‐soluble drugs in its inner aqueous core. [ 90 ] Liposomes were widely investigated for drug delivery applications because of their hydrophobic and hydrophilic drug loading capability as well as their biocompatibility properties. [ 16, 23, 26, 27, 44, 91, 92, 93, 94, 95 ] Doxil is the first liposomal drug which got FDA approval as an anticancer nanotherapeutics in 1995. [ 96 ] Over the last two decades, researchers have explored various liposomal formulations (such as temperature‐sensitive liposomes, [ 97 ] cationic liposomes, [ 98 ] and archaeosomes, [ 99 ] etc. ) for drug and gene delivery applications (Figure 6 ). For example, Song et al. developed a multifunctional liposome based complex system for in vivo treatment of NSCLC. The authors loaded an anticancer drug epirubicin inside the aqueous core of the liposome and an anti‐metastatic drug Honokiol into the lipid bilayer of the formulation. The surface of the liposome was further conjugated with a somatostatin targeting peptide octreotide which can binds to somatostatin receptors overexpressed in cancer microenvironment and facilitate targeted drug delivery. This complex liposomal system showed improved in vivo anticancer activity. [ 100 ] Cisplatin is one of the widely used drugs for the treatment of lung cancer. However, it showed nephrotoxicity in patient with high doses. [ 101 ] Devarajan et al. reported a liposomal formulation of cisplatin (namely Lipoplatin) which showed less nephrotoxicity when compared with free non‐bound cisplatin. [ 102 ] Paclitaxel is another anticancer drug which was widely used for the treatment of various cancers including lung cancer. In a phase I clinical trial in NSCLC patients, a liposomal formulation of paclitaxel showed enhanced therapeutic efficacy. [ 103 ] In a recent study, researchers prepared a liposomal formulation containing both cisplatin and paclitaxel for the treatment of lung cancer. In a phase III trial, this liposomal formulation displayed improved therapeutic activity and reduced nephrotoxicity in NSCLC patients. [ 104 ] In the last decade, researchers have developed a special type of liposomes–archaeosomes, which are made with ether lipids unique to the domain of Archaeobacteria. [ 105 ] Achaean‐type lipids consist of archaeol (diether) and/or caldarchaeol (tetraether) core structures. The membrane of this type of liposomes is made with both conventional phospholipids and bipolar lipids (Figure 6 ). Archaeosomes can be made using standard procedures used for liposome preparation including a sonication of hydrated film, extrusion, and detergent dialysis. In contrast to conventional liposomes, archaeosomes remain stable at acidic pH, high temperature and pressure, resist oxidative degradation and can be sterilized by autoclaving. A stability of archaeosomes at a wide physiological (or lower) temperature range opens a possibility of encapsulating thermally stable compounds. However, the uptake of archaeosomes by phagocytic cells can be up to 50‐fold greater when compared with conventional liposomes, which may be considered as their disadvantage. Figure 6 Examples of lipid‐based nanoscale‐based drug and nucleic acids delivery systems (not to scale). Typical structures of neutral, cationic and temperature‐sensitive liposomes, archaeosome (liposome made with one. or more ether lipids that are unique to the domain of Archaeobacteria), PEGylated nanostructured lipid carrier, and micelle with a single hydrocarbon chain in aqueous solution are shown. Nanostructured Lipid Carriers : NLCs are another widely used lipid‐based nanocarrier for targeted drug and gene delivery applications. NLCs are composed of biodegradable and biocompatible lipids and usually prepared by mixing a liquid lipid mixture containing unsaturated lipid or oils to a solid lipid (Figure 6 ). Among the lipid‐based nanocarrier, NLCs have the advantages of ease of manufacturing processes, drug protection during the storage, low toxicity, biodegradation, etc. These and other benefits of NLCs made them a promising therapeutic delivery system in recent years. [ 15, 25, 29, 106 ] Here, we have summarized few examples of NLC‐based therapeutic nanomedicines for drug and gene delivery. For instance, Guo et al. [ 107 ] prepared a dual drug paclitaxel (PTX) and 5‐Demethylnobiletin (DMN) loaded and cetuximab (CET) functionalized CET‐PTX‐DMN‐NLCs for the combination therapy of lung cancer. The treatment of the CET‐PTX‐DMN‐NLC inhibited the growth of the lung cancer cells as compared to the single PTC‐NLC and DMN‐NLC treatments. The authors also observed remarkable inhibition of in vivo lung tumor for the treatment of this dual drug containing CET‐PTX‐DMN‐NLC. [ 107 ] Wang et al. [ 108 ] developed a dual drug loaded NLC system for the treatment of lung cancer. This dual drug loaded PTX/DOX NLC displayed 3× higher activity as compared to that of single drug PTC‐NLC and DOX‐NLC treatment as revealed in cytotoxicity assay in NCL‐H460 cells. Also, in vivo study of this dual drug NLC on a non‐small cell lung cancer mice model showed improved the anticancer activity. [ 108 ] Micelles : Micelles consist of lipid molecules arranged in a spherical form in polar solvents (e. g. , water). Polar groups of lipids form an outer shell of the nanoparticles in polar solvent system while lipid hydrophobic tails create an inner core of micelles. In contrast to liposomes, micelles usually have a single hydrocarbon chain (Figure 6 ). In 2007, Kim et al. prepared nanosized micelles (Genexol‐PM) loaded with the anticancer drug paclitaxel for the treatment of NSCLC. [ 109 ] This nanotherapeutic was found to deliver higher paclitaxel dose with reduced drug toxicity as well as exhibited significant antitumor activity in the treatment of advanced NSCLC. A series of nanocarrier systems which included liposomes, micelles, quantum dots, mesoporous silica nanoparticles, dendrimers, and PEG polymers were prepared and examined in our laboratory in order to find out the suitable nanocarrier for local and inhalation delivery of anticancer drugs to the lungs. We investigated organ distribution and retention of all these nanocarriers in the lung and observed higher accumulation of liposomes and micelles based nanocarriers in lungs as compared to that of mesoporous silica nanoparticles, quantum dots, and dendrimers. [ 26 ] We found a significant enhancement of anticancer activity of doxorubicin when it was delivered to mice bearing lung tumor by inhalation by liposome‐based system. This study revealed that lipid‐based nanocarriers such as liposomes with higher accumulation of drug in lungs and longer retention time were more suitable and effective than non‐lipid‐based nanocarriers in treating lung cancer by inhalation. [ 26 ] Polymeric Nanoparticles Polymeric nanoparticles are usually prepared by self‐assembly of various block‐copolymers with alternate hydrophobic unit between blocks. Various biodegradable polymers such as poly (lactic‐co‐glycolic) acid (PLGA), polycaprolactone, poly (lactic acid) (PLA), chitosan etc. were widely used for the preparation of polymeric nanoparticles mainly because of their biocompatibility and controlled release properties. While core‐shell of such polymeric nanoparticles can encapsulate hydrophobic drugs, the surface of the polymeric nanoparticles can be modified for receptor targeted drug delivery. [ 110, 111 ] In the past two decades, many polymeric nanoparticles have been investigated for treatment of different diseases including lung cancers and enhancing efficacy of anticancer drugs. [ 112 ] In a recent report, Hu et al. [ 113 ] developed paclitaxel encapsulated polycaprolactone/poly (ethylene glycol)/polycaprolactone nanoparticles for the combined treatment of lung cancer with chronomodulated chemotherapy. This combination treatment showed better inhibition of tumor growth in vivo. Wang et al. [ 114 ] developed a new strategy for delivering drug loaded polymeric nanoparticles at the disease site using mesenchymal stem cells (MSC) as carrier. Docetaxel was encapsulated in the nanoparticles and tested in vivo. Inhibition of the tumor growth was observed in the animal's experiment, which also revealed the translocation of nanoparticles from MSC to cancer cells. Researchers developed PEG modified and taxane encapsulated polylactic acid nanoparticles for lung cancer treatment. The authors observed significant improvement of the efficacy of chemoradiation therapy in an A549 lung tumor xenograft model. [ 115 ] In another report, investigators developed a polymeric nanoparticle system comprised of block copolymers of PEG and polylactic acid and encapsulated paclitaxel and cisplatin for treatment of lung cancer. [ 116 ] Tseng et al. prepared gelatin based polymeric nanoparticles conjugated with biotinylated EGF (bEGF) motif for EGFR‐targeted drug delivery. [ 117 ] The authors observed enhanced cellular uptake of this polymeric nanoparticle in EGFR overexpressing cancer cell lines such as lung cancer cells. Zhang et al. [ 118 ] studied glutathione stimuli responsive organic PLGA‐SS‐PEG nanocarriers for targeted delivery of anticancer agent homoharringtonine (HHT) to lung cancer cells. A relatively high level of glutathione reduced (GSH) in cancer cells caused disulfide‐bond (SS) breakage in a reductive manner and release of HHT inside cancer cells. This nanocarrier showed a reasonable biocompatibility and was further decorated with an epidermal growth factor receptor (EGFR) aptamer as a targeting moiety. After endocytosis of this nanosystem by lung cancer cells, the high level of glutathione in tumor cells stimulated the release of the loaded drug. Finally, this multifunctional and stimuli responsive nanocomplex inhibited the growth of human lung cancer cells and displayed better therapeutic efficacy when compared with the free non‐bound anticancer agent ( Figure 7 ). Figure 7 Glutathione stimuli responsive organic PLGA‐SS‐PEG nanocarriers for targeted delivery of anticancer agent homoharringtonine (HHT) to lung cancer cells. A) Core–shell structured nanoparticles were synthesized using the solvent evaporation approach of oil‐in‐water. The PLGA nanomedicine release was triggered by reduced glutathione (GSH) overexpressed in tumor cells after receptor mediated endocytosis. B) In vivo antitumor efficacy of PLGA nanomedicine. Treatment with PLGA‐SS‐PEG nanoparticles reduced the tumor volume significantly when compared with non‐treated and treated with free non‐bound HHT groups. Adapted with permission. [ 48 ] Copyright 2017, Springer. Combined Lipid‐Polymer Nanoparticles In addition of nanocarriers prepared mainly with lipids or polymers, complex anisotropic nanoparticles containing lipid and polymeric structures (as well as polymers with different properties) were prepared. [ 119, 120, 121, 122, 123, 124 ] As Pierre‐Gilles de Gennes pointed in his 1991 Nobel Prize lecture, [ 125 ] similarly to the ancient Roman god of gates Janus who was portrayed with two faces—one facing the past, and one facing the future, Janus particles also have two distinct parts with antagonistic properties ( Figure 8 ). One such Janus structure with two faces (lipid and polymeric) was tested in our laboratory for inhalation lung delivery of a mixture of lipo‐ and hydrophilic drugs namely curcumin and doxorubicin. [ 28 ] These Janus particles were synthesized from binary mixture of biodegradable and biocompatible materials and evaluated for cytotoxicity and genotoxicity. The inhibition of lung tumor growth by the combination treatment was significantly higher when compared with either free drugs or nanoparticles containing only one drug. This study showed that such Janus particle could be explored for the simultaneous co‐delivery of hydrophilic and hydrophobic drugs. Figure 8 Suppression of lung tumor growth in mice treated by inhalation with Janus nanoparticles containing anticancer drug(s). A) Ancient Roman god of gates Janus who was portrayed with two faces (photo by A. Kokorin on Behance). B) Representative optical, C) scanning electron and D) fluorescence microscope images of anisotropic biodegradable biphasic polymer/lipid Janus nanoparticles. Polymeric phase of nanoparticles was labeled with FITC (green fluorescence); lipid phase was labelled with DiR (red fluorescence). E) Representative optical and F) magnetic resonance images of untreated (control) and treated mice four weeks after tumor instillation. G) Changes in lung tumor volume after beginning of the treatment with nanoparticles containing doxorubicin (DOX), curcumin (CUR), and both drugs. Mice were treated twice per week. Means ± SD are shown. Reproduced with permission. [ 28 ] Copyright 2014, ACS Publications. Carbon Nanomaterials Carbon nanomaterials are a new class of nanosized materials comprised of sp2 hybridized carbon atoms with hexagonal structure. Common carbon nanomaterials include 0D fullerenes, 1D carbon nanotubes (CNTs), and 2D graphene such as graphene oxide. [ 126 ] Among the carbon nanomaterials, CNTs have received significant attention in the past two decades for their high surface area, biocompatibility and drug‐loading capacity which made them suitable for wide range of applications such as drug delivery, tissue engineering, biosensors, cosmetic products etc. [ 127, 128 ]. There are mainly three types CNTs namely single wall‐, double wall‐ and multi wall‐ carbon nanotubes with diameter up to 100 nm and lengths up to microns size. [ 129, 130 ] Surface of these CNTs can be functionalized with hydrophobic and hydrophilic drugs for their targeted delivery applications. Carbon nanomaterials such as CNTs were explored as an attractive systems in targeted drug delivery application. In 2017, Kim et al. developed PEG‐coated carbon nanotube system loaded with small molecule BCL‐2 inhibitor ABT‐737 for its targeted delivery to lung cancer cells. [ 131 ] The authors investigated cellular uptake, apoptosis, and cytotoxicity this PEG‐CNT‐ABT737 nanotube system in lung cancer A549 cells and observed BCL‐2‐mediated apoptosis of lung cancer cells. The PEG‐CNT‐ABT737 system also excreted improved cytotoxic activity in A549 cells when compared with treatment by free non‐bound ABT737. The drug loaded nanotubes represented an effective system for inducing BCL‐2‐mediated apoptosis in lung cancer cells. In 2019, Cirillo et al. reported a pH‐responsive nanohybrid system comprised of multi‐walled carbon nanotubes and chitosan for delivery of methotrexate to lung cancer cells. [ 132 ] This chitosan coated CS‐MWCNT nanohybrid system displayed its pH‐responsive behavior and showed faster and higher release of the drug methotrexate in acidic (pH 5. 0) versus neutral (pH 7. 4) environments. Such a nanohybrid system showed reduced drug toxicity in normal lung MRC‐5 cells while it exerted anticancer activity in lung cancer H1299 cells. Mesoporous Silica Nanoparticles Nanosized silica particles also known as mesoporous silica nanoparticles (MSNs) have been investigated in the past two decades for various drug and gene delivery application. Because of large pore size, high surface area, good chemical stability, biocompatibility, and ease of surface modification with targeting ligands, MSNs based nanomaterials were extensively studied for various therapeutic delivery applications. [ 6, 30, 131, 133, 134 ] For example, Wang et al. designed a nanosized drug delivery system containing anticancer drug paclitaxel into the core‐shell of mesoporous silica nanoparticle (PAC‐csMSN) for the treatment of lung cancer. [ 135 ] This csMSNs formulation improved the adsorption of the poorly water‐soluble drug paclitaxel. The authors found PAC‐csMSN system was more effective in promoting cell apoptosis in A549 lung cancer cells than the free drug. This PAC‐csMSN system was administered for three consecutive days in animals and no indication of inflammation was observed in the lung biopsy. All these results indicated that such PAC‐csMSN system has the potential for inhalation delivery of paclitaxel for the treatment of lung cancer. 2018, Jing‐Hua Sun et al. prepared another MSNs system for co‐delivery of a photosensitizer chlorin e6 (Ce6) and a drug doxorubicin (Dox) for both photodynamic therapy and chemotherapy of lung cancer. [ 136 ] The anticancer drug doxorubicin was encapsulated into the pores of MSNs system while Ce6 was conjugated with the MSNs through covalent bonding. Treatment with these Dox@MSNs‐Ce6 hybrid nanoparticles increased the level of cellular reactive oxygen species and exerted synergistic therapeutic effect in lung cancer A549 cells when compared with treatment by each individual component. Gold Nanoparticles Gold nanoparticles (AuNPs) are one of the extensively used inorganic nanocarriers for various biomedical applications including drug and gene delivery. Because of high atomic number, stable nature and surface plasmon resonance properties, AuNPs can serve as stable contrast agents for photothermal therapy and medical imaging. [ 137, 138 ] Moreover, AuNPs possess unique physiochemical characteristics such as high biocompatibility and low‐toxicity as well as AuNPs are non‐immunogenic which made gold nanoparticles as an attractive nanocarriers for various biomedical applications. [ 139, 140, 141 ] In 2014, Qian et al. conjugated Cetuximab (C225), a targeting agent for epidermal growth factor receptor (EGFR) with AuNPs for the treatment of EGFR positive non‐small cell lung cancer (NSCLC). [ 142 ] This C225‐AuNPs inhibited proliferation and migration of A549 cells and accelerated apoptosis in A549 cells as compared to treatment with free C225 alone. The activity of C225‐AuNPs was higher in A549 cells with higher EGFR expression than in H1299 cells with low EGFR expression. Treatment of nude mice bearing tumor xenografts with C225‐AuNPs showed significant suppression of tumor size. Such EGFR‐targeted AuNPs system can be a promising strategy for targeted delivery of therapeutics in EGFR positive NSCLC cells. In 2018, Ramalingam et al. conjugated doxorubicin on the surface of gold nanoparticles through polyvinylpyrrolidone linker for the treatment of human lung cancer cells. [ 143 ] These Dox‐PVP‐Au nanoparticles inhibited the growth of human lung cancer cells more effectively than both the PVP‐AuNPs and free drug. Treatment with these nanoparticles induced early and late apoptosis as well as upregulated expression of tumor suppressor genes in the human lung cancer cells. This Dox‐PVP‐Au nanoparticle system represents a promising drug delivery approach for lung cancer therapy. Cell Based Drug Carriers Biocompatibility, biodegradability, and cytotoxicity of synthetic nanocarriers represent a substantial problem. Moreover, exogenous carriers potentially may induce immune responses. Consequently, drug carriers prepared from human live cells or their derivatives attract a considerable attention in recent years. Such carriers demonstrate native targeting mechanisms and controlled release of the encapsulated drug molecules. Several types of human cells have been considered for a targeted drug delivery for treatment of cancers and variety of other pathological conditions, such as cardiovascular and inflammatory diseases. [ 144, 145 ] Major characteristics of different cell types used for drug delivery are presenting in the Figure 9. Figure 9 Various cell types employed for drug delivery. Redrawn from. [ 144, 145 ] Erythrocytes are usually used for a systemic drug delivery and do not possess intrinsic tropism. In contrast, platelets, neutrophils, adipose cells, macrophages, and stem cells can be used for targeted delivery of therapeutics because of their native tropism to tumors, circulating tumor cells, sites with inflammation, and hypoxic conditions as well as microorganisms. [ 145 ] A strategy of drug loading into human cells used for therapeutic delivery is selected based on the nature and properties of drugs, loading capacity and release mechanisms. Therapeutic agents can be loaded into cellular cytoplasm of carrier cells, attached to their surface by a membrane insertion, nonspecific noncovalent or targeted interaction as well as covalent coupling methods. The release of payload may be continuous (e. g. , after slow hydrolysis of a prodrug and exocytosis) or triggered by various internal in vivo signals (glucose, hormones, cytokines, and other biomolecules, pH, and changes in cell shape) or external stimuli (light, ultrasound, magnetic field, temperature, etc. ). [ 145 ] A comprehensive list of cell based carriers designed for intraperitoneal, intratracheal, intravenous, subcutaneous and left anterior descending delivery of therapeutics developed during the last decade is presented in the open access review by Lutz et al. [ 144 ] 3. 1. 2 Delivery of Nucleic Acids Gene therapy has become a promising therapeutic option in recent time for lung cancer. Several different nanocarriers (such as dendrimer, micelles, gold nanoparticles, liposomes, lipid nanoparticles, auroliposome etc. ) were explored as carriers of nucleic acids with effective results. [ 14, 15, 23, 25, 27, 29, 42, 43, 45, 92, 134, 146, 147, 148, 149 ] Examples of carriers used for the delivery of nucleic acids are presented in Figure 10. Nucleic acids used as gene therapeutics are negatively charged because they are composed of few, several or many nucleotides with phosphate backbones carried one negative charge per residue. [ 150 ] Consequently, they often delivered as conjugates formed with positively charged (cationic) carriers. Such conjugation not only protects gene material from degradation in the blood stream and improves pharmacokinetics of the resulting complex, but also neutralizes positive charge of highly toxic anionic carriers limiting their cyto‐ and genotoxicity. [ 151 ] However, an encapsulation inside nanocarriers or direct conjugation of native or modified nucleic acids via different (preferably cleaved inside targeted cells, e. g. S─S) bonds are also used to form a stable system for an effective gene delivery. Small chunks of nucleic acids can be modified to decrease their negative charge and encapsulated inside nanocarriers. In our laboratory, the DNA backbone of all bases in antisense oligonucleotides (ASO) was P‐ethoxy modified in order to make the entire ASO neutral and increase their incorporation efficacy into liposomes. [ 149 ] Such modification also enhanced nuclease resistance of ASO. Liposomal ASO were successfully used to suppress pump and nonpump resistance of cancer cells. [ 27, 41, 43, 91 ] It should be stressed, that treatment of cancer with nucleic acids alone (e. g. , siRNA, antisense oligonucleotides) in most cases demonstrate a pretty limited anticancer effect. However, a combination of anticancer drug(s) with nucleic acids targeted to the drug efflux pumps, antiapoptotic, and other cancer cell defensive proteins/mRNAs is expected to substantially enhance anticancer efficacy of both anticancer drugs and nucleic acids. Such a concept of advanced proapoptotic anticancer delivery system was first developed and tested in the laboratory of Professor Minko at Rutgers University almost 20 years ago. [ 36, 37 ] Possible structures of such multifunctional nanoparticles are presented in Figure 1. In several such systems, where nucleic acids (siRNA and antisense oligonucleotides) were used as suppressors of drug efflux pumps (pump drug resistance) and antiapoptotic cellular defense (non‐pump resistance). [ 54, 91 ] Figure 10 Commonly used nanoparticles for the delivery of nucleic acids (not to scale). Negatively charged nucleic acids are usually integrated with nanoparticles by electrostatic interactions or chemical conjugation. Depending on the size of nucleic acid, it may be fully or partially entrapped inside one or several nanoparticles or bound to the surface of nanoparticle(s). Lipid‐Based Nanoparticles Lipid‐based nanoparticles such as liposomes, NLCs, and various polymeric nanoparticles are widely used to protect and deliver nucleic acids (Figure 10 ). Few examples of both liposome and NLC based nucleic acid delivery systems are discussed below. Liposomes Gopalan et al. prepared DOTAP/cholesterol based nanocarrier system for direct delivery of tumor suppressive gene at the tumor site. This system was effective and non‐immunogenic. [ 152 ] In an early study, our group prepared a multicomponent liposomal delivery system for improving anticancer activity of doxorubicin against multidrug‐resistant human non‐small‐cell lung cancer cells. This multi‐component liposomal system was included doxorubicin as an anticancer drug, antisense oligonucleotide (ASO) as a suppressor of pump resistance for MRP1 mRNA and another ASO as a suppressor of nonpump resistance for BCL2 mRNA. [ 27, 41, 43, 92, 149 ] Antisense oligonucleotides were P‐ethoxy modified to decrease their charge, enhance nuclease resistance, and increase incorporation efficacy into liposomes. We reported successful intracellular delivery of both doxorubicin and ASOs to lung cancer cells. Also, this liposomal treatment increased anticancer efficacy of doxorubicin and inhibited synthesis of both MRP1 and BCL2 proteins. This multicomponent liposomal system displayed 10‐fold higher cytotoxicity as compared to both free and liposomal doxorubicin treatment against the resistant lung cancer cells and could be used for the enhancement of anticancer activity of doxorubicin against multidrug‐resistant lung cancer cells. In a similar study, we used a complex liposomal drug delivery system containing anticancer drug doxorubicin and both MRP1 and BCL2 targeting antisense oligonucleotides for inhalation treatment in lung cancer cells. [ 27 ] While empty liposome, free antisense oligonucleotides and their combination treatment showed almost no influence on viability of lung cancer cells; liposome targeted to both MRP1 mRNA and BCL2 mRNA significantly inhibited the growth of the lung cancer cells. We evaluated this complex liposomal system on an orthotopic murine model of human lung cancer and the results revealed its higher chemotherapeutic efficacy with lower side effects as compared to that observed for individual treatment of each component. [ 27 ] Nanostructured Lipid Arriers Garbuzenko et al. reported a multi‐functional nanostructured lipid carrier (LHRH‐NLC‐siRNAs‐TAX) composed of an anticancer drug paclitaxel (TAX), a peptide analog targeted to luteinizing hormone‐releasing hormone (LHRH) receptor and a pool of siRNAs as inhibitors of different types of EGFR‐TKs. This LHRH‐NLC‐siRNAs‐TAX nanoparticle system was investigated in various human lung cancer cell lines and in vivo on an orthotopic NSCLC mouse model and displayed good organ distribution, stability, solubility, and improved anticancer activity when compared with free individual drugs and non‐targeted therapy. [ 15, 25, 29 ] Han et al. developed a multifunctional BLC system for delivery of anticancer drug doxorubicin (DOX) and green fluorescent protein plasmid (pEGFP) DNA as a prototypical nucleic acid in lung cancer cells. The authors prepared DOX and pEGFP encapsulated NLC and modified its surface with transferrin‐targeting motif. Transferrin‐modified and DOX and pEGFP co‐encapsulated NLC system showed higher in vitro and in vivo transfection of plasmid DNA than the other control treatment. These results also indicated that such multifunctional NLC system could be an effective method for both drug and gene delivery for the treatment of lung cancer. [ 153 ] In another work, Han et al. prepared an NLC system for delivery of plasmid‐containing enhanced green fluorescence protein (pEGFP) in lung cancer cells. The authors prepared the pEGFP‐loaded NLC and decorated its surface with transferrin (Tf) targeting ligands. This Tf‐NLC/pEGFP showed higher transfection efficiency as observed in in vitro and in vivo studies than the non‐targeted NLC/pEGFP–suggesting this NLC system could be a promising vehicle for gene therapy in lung cancer. [ 154 ] Polymeric and Hybrid Nanoparticles Chitosan‐based polymeric nanoparticles were widely used for the delivery of nucleic acids. Because of ionizable sidechain amino groups in chitosan, it has cationic nature which made chitosan polymer a good vehicle for the delivery of anionic agents such as siRNA, DNA etc. Okamoto et al. developed a chitosan‐based nanoparticle for delivery of pCMV‐Luc gene into the lung cancer cells. [ 155 ] Nafee et al. prepared a chitosan based nanocarrier system loaded with an antisense oligonucleotide such as 2‐ O ‐Methyl‐RNA for the treatment of lung cancer. The authors evaluated the inhibitory function of telomerase after treating the nanocarrier in lung cancer cells and observed 50% reduction of the telomerase activity in A549 lung cancer cells. These chitosan nanoparticles were safe and effective for the treatment of lung cancer. [ 156 ] Dhananjay et al. developed a polymeric nanoparticle system comprised of PEI‐PEG copolymer for the delivery of Akt1 shRNA in lung cancer cells. [ 157 ] Both service and internally anionic polypropylenimine tetrahexacontaamine (PPI) and polyamidoamine (PAMAM) dendrimers were successfully used for the delivery of nucleic acids into cancer cells. [ 158, 159, 160 ] It was found that dendrimers and additional caging of resulting nanoplexes protected nucleic acids from degradation and effectively delivered genetic material inside cancer cells. Delivered siRNA demonstrated high intracellular activity and effectively knocked down gene expression and synthesis of targeted proteins. Carbon Nanomaterials Carbon nanomaterials such as CNTs decorated with positively charged polymers were investigated as gene carrier in recent years for specific delivery of nucleic acids. [ 161, 162 ] Podesta et al. prepared amino‐functionalized MWCNT system for delivery of siRNA and tested it using an animal xenograft and orthotopic breast tumor models. [ 137 ] Treatment with this MWNCT‐siRNA complex significantly delayed the growth of tumor and increased the survival of the tumor bearing animals. Varkouhi et al. developed MWCNT systems decorated with cationic PEI (CNT‐PEI) and CNT–pyridinium for the delivery of siRNA to lung cancer cells. [ 163 ] Both these CNTs displayed cytotoxicity effect and gene silencing activity in H1299 human lung cancer cells, while non functionalized CNTs did not show any such effects. Mesoporous Silica Nanoparticles Mesoporous silica nanoparticles have been investigated as a delivery vehicle of various cargo molecules including nucleic acid therapeutics. For example, Dilnawaz et al. developed MSN based system for co‐delivery of anticancer drug doxorubicin and siRNA in lung cancer cells. [ 164 ] This combinational treatment enhanced in vitro cellular uptake, cytotoxic effect in A549 lung cancer cells. In 2020, Song Yinxue et al. developed a complex MSN system comprised of a polyphenolic drug Myricetin (Myr), siRNA specific to multidrug resistance protein (MRP‐1) and a targeting ligand folic acid (FA) in order to improve delivery efficiency of Myr in NSCLC cells. [ 165 ] This targeted Myr‐MRP‐1/MSN‐FA nanoparticles showed significant cellular uptake and reduced viability of A549, NCI‐H1299 lung cancer cells when compared with free drug and other controls. In vivo results revealed that this system was more effective in suppressing the tumor growth and it might be an attractive therapeutic strategy for the treatment of NSCLC. Gold Nanoparticles Gold nanoparticles possess good biodistribution, physiological stability and low cytotoxicity which made them an attractive vehicle for delivery of various payloads including large biomolecules such as nucleic acids. [ 166 ] Over the years, researchers have explored both non‐covalent and covalent conjugation of the nucleic acids such as siRNA, oligonucleotides etc. on the surface of AuNPs for their effective transportation to the target cells. For instances, Conde et al. developed a PEG modified gold nanoparticle system by conjugating of RGD peptide and c‐myc siRNA on the surface of gold nanoparticles and tested this system on mice bearing CMT/167 lung carcinoma tumors. [ 167 ] The authors observed downregulation of the c‐myc oncogene and significant inhibition of lung tumor growth after the treatment with si‐RNA/RGD AuNPs. Recently, an innovative hybrid formulation (so‐called auroliposomes) consisting of liposomes loaded with 20‐nm gold nanoparticles (AuNPs) was developed and used for the siRNA delivery (Figure 10 ). It was found that auroliposomes modulated the intracellular uptake and silencing efficacy leading to the enhanced suppression of tumor growth in vivo when compared with conventional liposomes. [ 168 ] 3. 1. 3 Diagnostics and Theranostics Nanocarriers have the potential to enhance the diagnosis of diseases. Recently, nanobased materials and methods have emerged as novel diagnostic tools for several diseases. Over the years, various nanocarriers were explored for the delivery of imaging (or both imaging and therapeutic) agents for diagnosis of many diseases including lung cancer. [ 169, 170, 171 ] For example, researchers designed folic acid functionalized dendrimers containing gold nanocarrier as cancer‐targeted imaging probes for computed tomography (CT) imaging of lung cancer cells. [ 172 ] CT imaging after nanoparticle uptake revealed the presence of these gold nanocarriers in the lysosomes of lung adenocarcinoma cells. In another study, researchers developed ultra‐small (3. 0 ± 0. 1 nm) Gadolinium containing nanoparticles (so called ultra‐small rigid platforms or USRPs) for enhancing Ultrashort Echo Time (276 ms) proton MRI of the lung. [ 173 ] These nanoparticles were prepared using 1, 4, 7, 10‐tetraazacyclo‐dodecane‐1, 4, 7, 10‐tetraacetic acid (DOTA) as a chelator and was delivered by the intratracheal instillation. The authors observed the substantial (>250%) enhancement of MRI signal in the lungs for almost 2. 5 h after instilling the solution of the nanoparticles. Erten et al. prepared a dextran core‐based stealth PEGylated liposomes containing anticancer drug doxorubicin, iron oxide as an MRI contrast agent and Boron dipyrromethene (BODIPY) fluorescence stain for imaging theranostics applications. [ 174 ] The authors observed strong ability of these liposomal nanoparticles of enhancing both types of imaging in the in vivo murine model of Lewis lung cancer. In another report, Lowery et al. labeled a tumor targeted doxorubicin loaded liposomes with Alexa Fluor 750 for imaging of lung tumor. [ 175 ] An HVGGSSV peptide with a selective binding to irradiated tumors was used as a targeting moiety in order to deliver the anticancer drug and imaging agent specifically to irradiated tumors limiting their accumulation in the normal tissues. These liposomes (100 nm) contained maleimide and amine functionalized PEG chains for the conjugation of the cysteine containing peptide and the N ‐(Succinyl)‐fluorophore, respectively. Doxorubicin in theranostic liposomes was loaded by the pH gradient. The authors studied these fluorophore labeled irradiated tumor targeted liposomes in murine model of Lewis lung cancer. They found that such a radiation‐guided tumor‐targeted delivery of liposomes enhanced the delivery of the fluorophore and anticancer drug specifically to irradiated tumors in the lungs, effectively induced cell death and limited cell proliferation within lung tumors finally inducing a delay in tumor growth and destruction of tumor blood vessels and increase of apoptosis in lung tumor cells. HVGGSSV targeting peptide also increase the accumulation of an entire system in irradiated tumors enhancing imaging quality. 3. 2 Idiopathic Pulmonary and Cystic Fibrosis Cystic fibrosis is an inherited disorder, caused by mutations of the cystic fibrosis transmembrane conductance regulator (CFTR) gene. Over‐production of mucous in the lungs causes airway obstruction resulting in infectious diseases such as cystic fibrosis (CF). [ 176 ] Typically, heterogeneous and large molecular weight oligomeric gel‐forming mucin glycoprotein are produced in CF. Gene therapy is the mainstay therapy to inhibit the mutation of CFTR protein. Gene therapies involve delivery of siRNA, DNA etc. into cells to rescue the function of the defective CFTR gene. Usually, viral and non‐viral vectors are employed to transfer correct copies of CFTR gene in the effected cells in lungs. [ 177 ] Because of small size, nanocarriers have emerged as an effective vehicle for delivery of gene through the mucus barriers. Nanoscale carriers can be used as vectors for gene therapy due to their less immunogenic and good gene transport capacity. [ 178 ] Nanocarrier based non‐viral vectors are easy to prepare as compared to that of viral vectors. [ 179 ] In recent years, researchers attempted to develop various nucleic acid based nanocarrier to affect mutation of CFTR gene to change the composition of mucin as well as to minimize the mucin production. [ 180, 181 ] In an early research, Konstan et al. developed a DNA nanoparticle for the treatment of cystic fibrosis and observed effective transfer of vector gene. [ 182 ] In order to overcome the mucus barrier, recently, Suk et al. prepared a densely PEG‐coated DNA nanoparticle system, which can penetrate extracorporeal human cystic fibrosis to deliver its payload. [ 183 ] This nanocarrier displayed better gene transfer after intranasal administration to mice as compared to other carriers. Minko et al. prepared a liposomal‐α‐tocopherol (LAT) formulation for the treatment of hypoxic lung injury in rats. The authors evaluated antioxidant and antiapoptotic activity of this LAT in rats with severe hypoxia and observed significant antihypoxic effects. [ 93 ] It was found that, treatment with LAT of rats under severe hypoxic conditions (breathing of 6% of oxygen within two hours) normalized lung phospholipid composition, inhibited lipid peroxidation, suppressed genes responsible for the development of lung damage and improved breathing pattern. Finally, such a treatment two‐times decreased the mortality of the animals under severe hypoxic conditions. Lately in our lab, a similar liposomal system for inhalation delivery of prostaglandin E2 (PGE2) was developed for treatment of pulmonary fibrosis. [ 16 ] This liposomal system was evaluated for local delivery of PGE2 using a standard bleomycin‐induced murine model of idiopathic pulmonary fibrosis. The results revealed that liposomes were accumulated in higher amount in lungs after inhalation delivery when compared with intravenous administration. Besides, this inhalation treatment reduced fibrotic injury in the lung tissues. These data probed that the inhalation administration of liposomal form of PGE2 can be an effective therapy for cystic fibrosis in the lungs. To further improve inhalation treatment of idiopathic pulmonary fibrosis (IPF) by liposomal PGE2, siRNAs targeted to major proteins responsible for the lung damage under IPF (MMP3, CCL12, and HIF1A) were added to the NLC based nanoparticles containing PGE2 and tested on the similar experimental model of lung fibrosis using inhalation delivery. [ 23 ] This enhanced advance system was more effective in the treatment of IPF when compared with siRNA and PGE2 delivered separately. Another combination of drugs in one NLC‐based nanoparticle system was recently tested for the treatment of lung manifestation of cystic fibrosis (CF). [ 24 ] The system included lumacaftor for the correction of correct p. Phe508del mutation (the loss of phenylalanine at position 508) and CFTR potentiator ivacaftor for increasing the open probability of CFTR channels. This system was tested in vitro using CF cells and in vivo on homozygote/homozygote bi‐transgenic mice with spontaneously developed CF. The system was delivered in vivo by inhalation. The results showed a high efficacy of the proposed treatment of the lung manifestation of CF. Wang et al. prepared rapamycin and azithromycin loaded polymeric nanocarrier via nanoprecipitation method. [ 184 ] Nanocomposite microparticles (nCmP) were formulated from this nanoparticle for the inhalation delivery of antibiotics in the form of dry powder aerosols. These, nanocomposite microparticles displayed aerosol dispersion characteristics indicating their deposit in the lungs. 3. 3 Coronavirus Diseases Viral infections in respiratory systems such as in lungs have become a worldwide public health threat in recent years. Several emerging positive‐stranded RNA coronaviruses [ 185, 186 ] such as Severe Acute Respiratory Syndrome Coronavirus (SARS‐CoV), [ 187, 188, 189 ] Middle East Respiratory Syndrome Coronavirus (MERS‐CoV) [ 191 ] etc. not only threatened public health, but also caused international epidemics in the past two decades. Recent outbreak of coronavirus infection caused by the severe acute respiratory syndrome‐coronavirus‐2 pathogen has seriously threatened public health all over the world. The taxonomic name “severe acute respiratory syndrome coronavirus 2” (SARS‐CoV‐2) given by the International Committee on Taxonomy of Viruses (ICTV) became official to refer to this virus strain. On February 11, 2020 the World Health Organization (WHO) officially named the “coronavirus disease 2019” as “COVID‐19”. [ 192 ] The genome of SARS‐CoV‐2 is a 29 903 bp with single‐stranded RNA (ss‐RNA). The complete genome sequence of SARS‐CoV‐2 is available in the National Center for Biotechnology1 (NCBI) database, with ID NC_04 5512. [ 193, 194 ] COVID‐19 is characterized by severe respiratory disease along with mild to high fever, cough, and shortness of breath. COVID‐19 has been considered as an emerging disease and on March 11, 2020 the outbreak of this disease has been declared as global pandemic by the WHO. [ 195, 196 ] This virus has been found to spread from person to person mainly through respiratory droplets, cough, sneeze, etc. [ 197 ] causing severe acute respiratory distress syndrome (ARDS). As of September 20, 2020, this virus has already infected more than thirty million people and caused 950000 deaths with billions of people are at risk around the world. [ 198 ] Despite repeated outbreaks of SARS‐CoV in 2003 and MERS‐CoV in 2012, no potent vaccines and anti‐viral drugs are commercially available against these viral infections—mainly due to the fact that the outbreaks of these viruses were rapidly contained and did not reappear. [ 199 ] Therefore, there are no effective treatment for the ongoing pandemic of COVID‐19, a close subtype of SARS‐CoV. [ 200 ] Because of constant emergence of new viruses including current SARS‐CoV‐2 infection, there is an urgent need for the development of potent and broad‐spectrum vaccines and antiviral drugs for effective control of viral diseases. Since the first report of SARS‐CoV‐2 infection in late December in 2019, both researchers and clinicians have been attempted clinical trials of several known antiviral drugs, their combination as well as development of vaccine in patients with confirmed COVID‐19 disease. This review is mainly focused for summarizing recent developments of nanotherapeutics for respiratory diseases including SARS‐CoV, MERS‐CoV, and COVID‐19. Therefore, other types of therapeutic and diagnostic methods such as small molecule antiviral therapeutics, anti‐SARS‐CoV‐2 antibody treatments, convalescent plasma therapy etc. [ 201, 202, 203, 204 ] which have been discussed elsewhere are out of the scope of this review. Briefly, we will summarize recent innovation of nanobased diagnostics such as nanoparticle‐based PCR and anti‐body test as well as nanoparticle‐based therapeutic approaches for COVID‐19. 3. 3. 1 Diagnostic Approaches Diagnostic tests are essential not only for monitoring every stage of a disease, but also to identify new patients with that illness–especially for an outbreak of viral disease. Typical diagnosis methods for viral diseases include nucleic acid detection of the viral genome in clinical samples. Currently, COVID‐19 has been diagnosed by real‐time polymerase chain reaction (RT‐PCR) test for the detection of viral genome, serological, and immunological assays for the detection of anti‐SARS‐CoV‐2 antibody in patient samples as well as chest computed tomography (CT) imaging for screening abnormal observations in chest scans. [ 205, 206 ] However, most of these methods are laborious and time‐consuming processes. Therefore, there is an urgent need for developing time‐economic, easily performed and point‐of‐contact test for the detection of this virus. Because of similar size and shape of SARS‐CoV‐2 virus with the synthetic nanoparticles, researchers have attempted to develop nanoparticle based diagnostic methods for COVID‐19. For examples, Huang et al. developed a rapid, easily operated and cost‐effective detection of the IgM antibody produced in serum sample of patient with COVID‐19. [ 207 ] The authors prepared a colloidal gold nanoparticle‐based lateral‐flow (AuNP‐LF) system composed of various low‐cost inorganic nanomaterials. The AuNP‐LF strip was developed for the sample test by coating an analytical membrane with the SARS‐CoV‐2 nucleoprotein followed by conjugating anti‐human IgM antibody. This method can detect the SARS‐CoV‐2 virus in 15 min using only 20 µL of serum sample of the patient. The authors have evaluated the specificity of this detection method against the results of widely used TR‐PCR's test. This AuNP‐LF assay has a great potential for large‐scale and fast detection of COVID‐19 disease specially during this pandemic period. [ 207 ] In another report, Moitra et al. developed gold nanoparticle based colorimetric assay for naked‐eye detection of SARS‐CoV‐2 virus present in patient samples. [ 208 ] The authors decorated the gold nanoparticles (AuNPs) with thiol‐modified antisense oligonucleotides (ASOs) which are specific for N‐gene (nucleocapsid phosphoprotein) of SARS‐CoV‐2 virus. The use of RNaseH in this detection helps to cleave the RNA–DNA hybrid resulting in the visually detectable precipitation from the experimental solution. This AuNP system can detect the presence of SARS‐CoV‐2 stain in the isolated RNA samples within 10 min. Thus, this method can be a very promising for visual detection of COVID‐19 positive patient without the use of typical instrumental procedures as outlined in Figure 11. [ 208 ] Figure 11 Schematic representation for the selective naked‐eye detection of SARS‐CoV‐2 RNA mediated by the suitably designed ASO‐capped gold nanoparticles. This naked eye detection method involves the isolation of the viral RNA from the clinical swab sample from COVID‐19 patient and then incubation of the viral RNA samples with ASO capped gold nanoparticles for 5 min. At the end, RNase H is added in the viral composite of ASO‐capped gold nanoparticles and the resulting mixture is incubated for 5 min at 65 C to get the visual precipitate. Reproduced with permission. [ 208 ] Copyright 2020, ACS Publications. 3. 3. 2 Therapeutic Approaches There are no clinically approved therapeutics by the U. S. Food and Drug Administration (FDA) to prevent or treat the COVID‐19 disease. However, clinical trials of many known anti‐viral drugs are ongoing in patient with confirmed COVID‐19 disease. [ 209, 210 ] For instance, a newly developed antiviral drug Remdesivir previously effective against Ebola virus diseases which interferes with viral RNA polymerase (RdRp) and arrests viral replication [ 211 ] was repurposed for the treatment of COVID‐19 infections. [ 212 ] Human trials showed promising clinical improvements in ≈70% of patients. [ 213 ] Recent studies showed that the SARS‐CoV‐2 virus has similar size range (50–150 nm) and spherical shape like the synthetic nanoparticles. [ 214 ] Therefore, nanosized spherical therapeutics can be promising to detect and neutralize coronaviruses as previously evaluated for SARS‐CoV, MERS‐CoV etc. Various nanoparticle based therapeutic approaches have been discussed here for MERS‐CoV, SARS‐CoV, and COVID‐19 diseases. For instance, Huang et al. prepared gold nanorod complex of heptad repeat 1 (HR1) peptide inhibitors for Middle East respiratory syndrome coronavirus (MERS‐CoV) disease. [ 215 ] This gold nanorod complex was biocompatible and metabolically stable and displayed 10‐fold higher inhibition of membrane fusion between host cells and MERS‐CoV via HR1/HR2‐mechanism as compared to that of the free inhibitor treatment. This the gold nanorod based anti‐viral system showed a great promise in treating MERS‐CoV infection. Lin et al. prepared a virus‐like hollow nanoparticle comprised of biodegradable polymer and a viral antigen along with an adjuvant. [ 216 ] This plasmid like nanoparticle was capable of delivering of both antigens and stimulator of interferon genes agonist adjuvant to induce potentiation to the immune cells. The authors observed that this nanoparticle‐based MERS‐CoV vaccine was effective against a lethal dose of MERS‐CoV infection as compared to control treatment in a MERS‐CoV‐permissive transgenic mouse model. The potency of this nanoparticle‐based vaccine for MERS‐CoV was demonstrated, and this study provides a new outline for developing nanocarrier based vaccine for viral pathogen. Loczechin et al. prepared seven different carbon quantum dots (CQDs) and investigated their anti‐viral activity against the human coronavirus HCoV‐229E infections. [ 217 ] These CQDs displayed a concentration‐dependent virus inactivation and CQD produced from 4‐aminophenylboronic acid showed better activity with EC 50 5. 2 ± 0. 7 µg·mL −1. The mechanistic studies revealed that interaction of the surface functional groups of the CQDs with entry receptors of the HCoV‐229E virus resulted in inhibition of the infection. These results suggested such CQDs systems might be explored for developing anti‐viral therapeutics for other coronavirus infections ( Figure 12 ). [ 217 ] Polymeric nanoparticles consisting of poly (ethylene glycol)‐block‐poly(lactide‐coglycolide) (PEG‐PLGA) were developed for delivery of diphyllin, a novel vacuolar ATPase blocker for its antiviral activity against the feline coronavirus infection. [ 218 ] Treatment with these nanoparticles significantly reduced toxicity and enhanced antiviral effect of diphyllin. These nanoparticles were well tolerated as revealed in animal study in mice following high‐dose intravenous administration. The results of the study indicated that such diphyllin nanoparticles could be explored as effective host‐targeted antiviral therapeutics for other coronavirus infections. Coleman et al. developed a novel strategy for preparing spike nanoparticles which in combination with adjuvants produced high titer anti‐bodies in mice against both the severe acute respiratory syndrome coronavirus (SARS‐CoV) and Middle East Respiratory Syndrome Coronavirus (MERS‐CoV) infections. [ 219 ] The results showed that these spike nanoparticles were able to neutralize the antibody responses in mice—suggesting a step towards nanovaccine development. [ 219 ] Recently, fabric material‐based face mask containing hydrophilic absorbent layers and hydrophobic barrier layers was constructed. [ 220 ] This fabric masks were found to show equivalent or better filtration and adsorption of nanoparticle like aerosols than the commercial N95 respirators. The aerosols were composed of fluorescent labeled virus like nanoparticles for tracking their transmission through the fabric masks. The authors evaluated 70 different combinations of common fabric materials using forced convection air flux with pulsed aerosols. This fabric masks can be used to protect from the inhalation of such viruses. In an early study, Li et. al reported that treatment based on RNA interference (RNAi) exhibited antiviral immunity in mammals. [ 221 ] In 2018, Sohrab S. et al. designed and developed a series of lipid and polymeric nanoparticles for delivery of therapeutic siRNA for the treatment of MERS‐CoV infection. [ 222 ] Since SARS‐CoV‐2 is a single strand RNA virus like other coronaviruses, therefore inhibiting the life cycle of SARS‐CoV‐2 viruses via silencing their viral mRNA in the host cells by RNA interference might be an effective therapy for COVID‐19. Recently, both researchers and clinicians have been working in developing nanoparticle encapsulated mRNA‐based vaccines for COVID‐19. For examples, vaccine candidate mRNA‐1273 (Moderna) is a lipid nanocarrier based mRNA vaccine that encodes spike protein of SARS‐CoV‐2 virus. In an early study (clinical trial identifier: NCT04283461), Jackson et al. conducted a phase I open label clinical trial of the mRNA‐1273 in 45 adults with 15 people in each group of 25, 100, and 250 µg dose. [ 223 ] Each group of people received second vaccination after 28 days of the first vaccination. Systematic side effects such as headache, fatigue, pain at the injection site etc. were observed in half the participants particularly after second vaccination with higher dose treatment. Initial results revealed that treatment of mRNA‐1273 induced immune responses in all participants and the antibody responses were higher for the higher dose treatment participants. Recently, Corbett et al. studied this mRNA‐1273 vaccine in nonhuman primates and observed that treatment with this vaccine candidate increased neutralizing antibody levels that were higher than in human convalescent‐phase serum sample. [ 223 ] The mRNA‐1273 vaccine is now under further evaluation for COVID‐19. BNT162b1 is another nanoparticle encapsulated mRNA vaccine candidate that encodes receptor binding domain (RBD) of spike glycoprotein of SARS‐CoV‐2. Recently, Mulligan et al. conducted a placebo‐controlled Phase 1/2 clinical trial of nucleoside‐modified mRNA vaccine which was formulated in a lipid‐based nanoparticle system for targeting RBD of spike glycoprotein of SARS‐CoV‐2 virus (clinical trial identifier: NCT04368728). [ 223 ] The authors performed a randomized and placebo‐controlled trial of BNT162b1 vaccine in 45 healthy adults. There were 12 participants for each of the dose level 10, 30, and 100 µg of the vaccination and nine participants in placebo with BNT162b1 increased the SARS‐CoV‐2 neutralizing titers with dose level in the serum sample. Thus, nanosized therapeutics such as nanoformulation of anti‐viral drugs, nanovaccines, etc. can be promising options for the diagnosis and treatment of such viral infection. Figure 12 Influence of carbon quantum dots, prepared by hydrothermal carbonization, on binding of HCoV‐229E virus to cells: A) inhibition of protein S receptor interaction, and B) inhibition of viral RNA genome replication. Boronic acid adducts based carbon quantum dots (CQDs) derived from phenylboronic acid and 4‐aminophenylboronic acid showed antiviral activity mainly by inhibiting the entry of HCoV‐229E virus as well as inhibiting the replication step of the viral genome. Reproduced with permission. [ 217 ] Copyright 2019, ACS Publications. 4 Conclusions and Outlook As summarized in this review, the past two decades have witnessed substantial amount of work in the development and applications of nanocarrier based systems for targeted delivery of drugs, gene, imaging agents etc. as well as nanoparticle‐based diagnostics for various respiratory diseases. Several preclinical and clinical investigations revealed that nanocarrier‐based systems address many limitations of conventional therapy not only by site‐specific delivery of therapeutics at the lung tissue, but also reducing the drug availability into other organs thereby reducing adverse side effects. Besides, nanocarrier based systems demonstrated sustain and control release of the therapeutics than the burst release observed in systematic delivery of therapeutics in lungs. Moreover, nanosized carriers have a potential to overcome the mucus barrier and poor lung penetration associated with various respiratory diseases. Similarly, recent development of various nanoparticle‐based detection methods for coronavirus infection showed a great promise in the development of time‐economic diagnosis of COVID‐19 disease. While such nanoscale systems promise new therapeutic options for respiratory diseases, still the most challenging task is their safety assessment. Preparation of an appropriate size of nanoparticles in each batch of synthesis is also challenging. Many such challenges need to be overcome in order to translate the nanotherapeutics into clinical practice. Though nanotechnology can find a way for further application of nanotherapeutics against the COVID‐19 diseases, still more research needs to be conducted for evaluation of nanosized therapeutics for COVID‐19. It has been considered that the effective vaccine against COVID‐19 will be available in 12–18 months. Therefore, early diagnosis, effective treatment etc. are essentials to mitigate the spread of this infection before any clinically approved vaccine comes in the market. Finally, the following should be mentioned. Because the major result of COVID‐19 infections is acute respiratory distress syndrome, the discussed nanomedicines for treatment lung hypoxia and fibrosis potentially can be used for the treatment of COVID‐19 in combination with other antiviral actions. Such pilot investigations have been recently initiated in our laboratory. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201400010
| 2,015
|
Advanced Science
|
Nanoparticle–Hydrogel Composites: Concept, Design, and Applications of These Promising, Multi‐Functional Materials
|
New technologies rely on the development of new materials, and these may simply be the innovative combination of known components. The structural combination of a polymer hydrogel network with a nanoparticle (metals, non‐metals, metal oxides, and polymeric moieties) holds the promise of providing superior functionality to the composite material with applications in diverse fields, including catalysis, electronics, bio‐sensing, drug delivery, nano‐medicine, and environmental remediation. This mixing may result in a synergistic property enhancement of each component: for example, the mechanical strength of the hydrogel and concomitantly decrease aggregation of the nanoparticles. These mutual benefits and the associated potential applications have seen a surge of interest in the past decade from multi‐disciplinary research groups. Recent advances in nanoparticle–hydrogel composites are herein reviewed with a focus on their synthesis, design, potential applications, and the inherent challenges accompanying these exciting materials.
|
1 Introduction Wichterlie and Lim 1 reported the first synthetic hydrogels with control of properties such as swelling and shrinking over several orders of magnitude. This initial discovery provided the foundation for stimuli‐responsive systems. The cross‐linked polymer network was highly sensitivity to stimuli such as solvent composition, solutes, pH, temperature, electric field, and light. This behavior of synthetic hydrogels has been reviewed elsewhere. 2 Parallel to these developments in hydrogels and their eventual reach of the biomedical and consumer care market, several types of nanoparticle and their composites have also been being developed over decades various material research groups. Today, nanoparticles find their applications in common consumer products and appliances due to their differences in properties compared to bulk materials. This trend has generated public debate on the safety of nanoparticle technology and regulatory authorities have intervened in several countries. 3 The challenges in the application of nanoparticles could potentially be overcome by incorporation into hydrogels resulting in decreased risks to human health and the environment. In addition, innovative combination of these two entirely different types of materials was thought to generate not only structural diversity but also a plurality of property enhancements. Such property enhancements were the main focus of research on hydrogel‐nanoparticle composite materials that resulted in improved mechanical strength and stimuli response. For example, recently reported silica nanoparticle‐hydrogel composite made of silica nanoparticles and modified poly ethylene glycol demonstrated remarkable improvements in tissue adhesive property, mechanical stiffness and bioactivity compared to hydrogel without nanoparticles. 4 Similarly significant changes in mechanical property and thermal response were observed in poly N ‐isopropyl amide hydrogels when gold nanoparticles immobilized in the gel. 5 Thus the benefits of the combination of two different materials viz. , nanoparticle and hydrogels lead to advanced materials with unique properties absent in the individual components. This uniqueness has catalyzed intense research activity at the interface of nanoparticle hydrogel composites looking forward to numerous applications over the past decade. One of the earliest investigations of such materials was reported by the Willner group, 6 in which gold nanoparticles (Au‐NPs) were immobilized in polyacrylamide (PAAm) by swelling the dehydrated gel in the presence of Au‐NP solution, resulting in uniform distribution of the Au‐NPs in the gel matrix. Consequently, various approaches for realizing these structurally unique dispersions have been reported by researchers investigating potential applications in biomedicine and optics, optics etc. Structurally similar nano‐dispersions of Au in PAAm hydrogels were obtained by in situ reduction of gold (III) chloride in the hydrogel network. 7 Three different supramolecular hydrogel‐nanoparticle designs can be proposed. They are: a) micro or nano‐gels stabilizing single/multiple nanoparticles, b) nanoparticles non‐covalently immobilized in an hydrogel matrix, and c) nanoparticles covalently immobilized in an hydrogel matrix ( Figure 1 ). There are reviews on type (a) structures covering their applications in biomedicine (particularly drug delivery), catalysis and electronics. 8 Bulk hydrogels of type (b) and (c) have been reviewed by Schexnailder and Schmidt, 9 but since then there have been few accounts covering only very specific applications. 10 Figure 1 Concept for combination of nanoparticles and hydrogel to form new functional materials. Three different structural designs exist: a) micro‐ or nano‐sized hydrogel particles stabilizing inorganic or polymer nanoparticles, b) nanoparticles non‐covalently immobilized in a hydrogel matrix, and c) nanoparticles covalently immobilized in hydrogel matrix. The current review categorizes various approaches to the design of structurally diverse nanoparticle‐hydrogel composites. It will hopefully provide a tool for selecting an appropriate method for achieving a desired nanoparticle‐hydrogel composite. The potential applications and challenges in developing these materials are presented in parallel. 2 Design, Synthesis, and Properties A diverse range of nanoparticle–hydrogel composites have been developed with varying types of nanoparticle embedded in a bulk hydrogel framework. Five main approaches have been used to obtain a uniform distribution: 1) hydrogel formation in a nanoparticle suspension, 2) physically embedding the nanoparticles into hydrogel matrix after gelation, 3) reactive nanoparticle formation within a preformed gel, and 4) cross‐linking using nanoparticles to form hydrogels, and 5) gel formation using nanoparticles, polymers, and distinct gelator molecules. A schematic illustration of these five approaches is given in Figure 2. The approach chosen will be, in part, determined by the final application of nanoparticle‐hydrogel composite. Figure 2 Five main approaches used to obtain hydrogel‐nanoparticle conjugates with uniform distribution: 1) hydrogel formation in a nanoparticle suspension, 2) physically embedding the nanoparticles into hydrogel matrix after gelation, 3) reactive nanoparticle formation within a preformed gel, 4) cross‐linking using nanoparticles to form hydrogels, 5) gel formation using nanoparticles, polymers, and distinct gelator molecules. 2. 1 Hydrogel Formation in a Nanoparticle Suspension The simplest approach to forming a nanoparticle–hydrogel composite is the gelation of a suspension of pre‐formed nanoparticles in a hydrogel‐forming monomer solution. This approach has been utilized to form optically responsive opto‐mechanical nanoparticle‐hydrogel composites. 11 In an example of this approach, S. R. Sershen et al. 11 prepared gold nanoparticle (Au‐NP) hydrogel composites by adding nanoshell Au‐NPs into a solution of monomers (95/5 molar ratio of N ‐isopropylacrylamide/acrylamide (NIPAAm/AAm) followed by addition of the gelation initiator ammonium persulfate (APS) and accelerator tetramethylethylenediamine (TMEDA). The composite was subsequently cured. Ravi et al. used this approach to incorporate three different types of nanoparticles: i) proteo‐mimetic PAAm nano‐gel, ii) bovine serum albumin, and iii) hydrophilized silica (Si), into a hydrogel matrix for intra‐ocular lens applications. 12 Other groups have adopted this simple approach to obtain hydrogels containing Au or Si‐NPs and thereby prevent aggregation. 13 This approach has certain drawbacks including the leaching of nanoparticles out of the hydrogel matrix if the cross link density is low. 14 An advanced application of an hydrogel‐nanoparticle composite using a similar protocol was reported by Liu et al. for synthesizing photo‐modulable thermo‐responsive hydrogels using unilamellar titania nanosheets (TiNSs) as photocatalytic cross linkers ( Figure 3 ). 15 In this synthesis, it was noted that the nanoparticles act as photo‐catalysts rather than cross‐linking agents and the use of bisfunctional monomer N, N '‐methylenebisacrylamide (MBAAm) was necessary for the formation of mechanically durable hydrogels. Figure 3 a) Schematic of titania nano sheets (Ti) acting as photo catalyst for gelation (difference in energy level of valence band and conduction band is shown). UV radiation with wavelength below 260 nm will produce hydroxyl free radicals causing gel formation reaction with vinyl monomers. b) List of vinyl monomers used for photo induced hydrogelation. c) Pictures before and after the hydrogelation using IPAAm (10. 0 wt%) as monomer. The examples show hydrogel formation in a nanoparticle suspension. The nanoparticles are retained in the hydrogel by non‐covalent interactions. Reproduced with permission. 15 Copyright 2013, Macmillan Publishers Ltd. 2. 2 Physical Incorporation of Nanoparticles into a Hydrogel Matrix after Gelation In order to study the solvent switchable electronic properties of hydrogel‐Au‐NPs, Pardo‐Yissar et al. 6 incorporated Au‐NPs into a PAAm gel after the electro polymerization formation of the hydrogel. Electro polymerization cannot be performed in the presence of the Au‐NPs because they readily aggregate under the influence of an electric field. To counter this problem, nanoparticles were introduced into the gel matrix after the gel had already been formed via a “breathing in” mechanism. PAAm gels are highly swollen in aqueous solution, but shrink dramatically in an aprotic solvent such as acetone. Introduction of nanoparticles into the gel “breathing” consisted of three steps, which were repeated several times to obtain the desired nanoparticle density ( Figure 4 ): a) the swollen gel was placed in acetone for 2 min, causing the gel to collapse with the expulsion of water (breathing out). b) the shrunken gel was then placed in an aqueous solution of citrate‐stabilized 13 nm diameter Au‐nanoparticles (ca. 5 nM) for 2 min. This aqueous solution caused swelling of the gel in solution (breathing in), including breathing in the suspended nanoparticles. Finally, the gel was washed thoroughly with water to remove any weakly surface‐adsorbed nanoparticles. Upon the next “breathing out” cycle, the nanoparticles remain attached inside the gel, possibly due to i) physical entanglement and ii) hydrogen bonding interactions between the polymer chains and the citrate surface of the nanoparticles. These researchers confirmed the increasing presence of Au‐NPs in the gel with each “breathing” cycle by X‐ray photoelectron spectroscopy (XPS) and atomic absorption spectroscopy (AAS) measurements. 16 Guo et al. used a similar “breathing in” approach to form highly dispersed Au‐NPs in porous anodic aluminium oxide (AAO) films by incorporating Au and platinum (Pt) NPs in PAAm hydrogels, followed by calcination. 17 Alternatively, embedding of nanoparticles into pre‐formed colloidal micro‐gels could be achieved by repeated heating, centrifugation and re‐dispersion followed by annealing. 18 In a typical procedure, a solution containing various ratios of micro‐gel and colloidal Au was placed into a centrifuge tube and centrifuged. Au‐NP solution was added to the micro‐gel pellet obtained and re‐dispersed throughout the micro‐gel pellet by repeated heating, agitating, and sonication. The suspension was then re‐centrifuged and the protocol repeated before a final annealing. Figure 4 The construction of a gold‐nanoparticle/hydrogel composite at the electrode interface by switching between its swollen and shrunken states. a) Electrochemical formation of acrylamide hydrogel film at the gold electrode showing its shrinking behavior in acetone and swelling behavior in water. b) Use of alternate cycles of swelling of the hydrogel gold nanoparticle suspension in water followed by shrinking in acetone to physically entrap nanoparticles in the hydrogel. Adapted with permission. 6 2. 3 Reactive Nanoparticle Formation Aided by the Hydrogel Network Langer's group developed this approach, which involves loading nanoparticle precursors into a gel, rather than preformed nanoparticles. 7 In a typical procedure, crosslinking NIPAAm and co‐monomers containing thiol groups formed a hydrogel network containing embedded Au (III) ions. The thiol‐functionalized hydrogel matrix enabled the modulation of Au‐NP formation when a reducing agent such as sodium borohydride was added. The resulting hydrogel contained un‐aggregated nanoparticles throughout the matrix. An improved procedure of nanoparticle formation within a gel without the use of thiol or phenol containing monomers was reported by Saravanan et al. ( Figure 5 ). 19 Free‐radical cross‐linking polymerization of acrylamide monomer in an aqueous medium containing Ag + ions was conducted. The Ag + ions functionalized‐PAAm hydrogel matrix was then hydrolysed to yield Ag NPs within the hydrogel network. The size of the Ag NPs in the hydrogel were estimated to be 4–7nm and well‐dispersed. The intensity of surface plasmon bands display a symmetrical configuration with no band shift and proportional increases with the increase of Ag+ ion concentration. This showed that the size of Ag nanoparticles did not vary on changing the Ag+ ion concentration and aggregation of particles are not present. These conclusions were confirmed with the electron microscopy results. This protocol has also been used for generating silver nanoparticles in interpenetrating network hydrogels. 20 Adopting a similar method, Xiang et al. synthesized pH‐responsive Ag‐NP/poly(2‐hydroxyethyl methacrylate (HEMA)‐poly(ethylene glycol) methyl ether methacrylate (PEGMA)‐methacrylic acid (MAA)) composite hydrogels by reducing Ag + ions anchored to free carboxylate groups in the matrix. 21 In a recent study, Varaprasad et al. reduced Ag + ions using green reducing agent (mint leaf extract) in a Carbopol 980 NF and acrylamide gelation mixture to produce composite hydrogels for antibacterial applications. 22 In another recent example, Gema, M. et al. utilized redox active catechol side chain in acrylamide‐NIPAAm hydrogels to reduce gold precursor to nanoparticles to form nanoparticle‐hydrogel composites that mimic mussel adhesive protein. 5 A schematic of this approach is presented in Figure 5 (b). These hydrogels were readily formed without an external reducing agent, and great reinforcement of mechanical property was observed for the resulting composite hydrogel. Figure 5 a) Preparation of Ag/PAAm hydrogel nanocomposite without the use of thiols. Reproduced with permission. 19 Copyright 2007, Elsevier. b) Hydrogel formation and functionalization with Au NPs to obtain catalytic hydrogels by redox active catechol groups. Adapted with permission. 5 Copyright 2014, American Chemical Society. 2. 4 Cross‐Linking using Nanoparticles to Form Hydrogels One particularly interesting example in the development of nanoparticle‐hydrogel composites involves the use of cross‐linking groups present on the nanoparticle surface. Souza et al. produced bacteriophage molecular networks by the spontaneous assembly of phage with Au‐NPs. The resulting hydrogel network preserved the cell surface receptor binding and internalization attributes of the peptide. 23 The spontaneous arrangement of these networks could be further manipulated by incorporation of imidazole (Au–phage–imid), which induced changes in morphology, fractal structure and near‐infrared optical properties. The capacity to form multiple bonds within the gel networks (multivalency) would be a major advantage of nanoparticles as cross‐linkers, rather than the two covalent bonds of a traditional hydrogel formation reaction. This concept was well demonstrated with collagen gel formation using the 1‐ethyl‐3‐(3‐dimethylaminopropyl)carbodiimide (EDC) crosslinking reaction of mercaptopropionyl glycine protected Au‐NPs. 24 Similarly, Zhao et al. used co‐polymerization of vinyl functionalized Au‐NPs to synthesize well‐dispersed Au‐NP‐PNIPAAm hydrogel composites with thermo‐switchable electrical properties. 25 The transition temperature of the composite could be adjusted from 0 °C to 40 °C by changing the concentration of Au‐NPs, the degree of cross‐linking or the stoichiometry of the composite. Prestwich et al. exploited the multivalency and thiophilicity of Au‐NPs to crosslink commercially available thiolated hyaluronic acid into printable, extrudable, cytocompatible and biodegradable hydrogels. 26 In a very recent study, Zhang et al. synthesized semiconductor nanoparticle‐based hydrogels by self‐initiated polymerization under light irradiation. 27 The system consisted of four components: i) water, ii) water soluble semiconductor nanoparticles of zinc oxide (ZnO), titanium dioxide (TiO 2 ), iron(III) oxide (Fe 2 O 3 ), tin dioxide (SnO 2 ), zirconium dioxide (ZrO 2 ), cadmium selenide (CdSe) or cadmium telluride (CdTe), iii) N, N ‐dimethylacrylamide (DMAA), and iv) clay nanosheets. The semiconductor nanoparticles functioned as inorganic initiators for the polymerization of DMAA. The authors demonstrated that CdSe and CdTe were able to form stable gels even under irradiation by visible light. The mechanism of gel formation and their visual and microscopic properties is shown in Figure 6. Recently, incorporation of silica nanoparticles were shown to result in increased interfacial binding between network and nanoparticles leading to increased stiffness as well as excellent energy dissipation capability[[qv: 4b]] with orders of magnitude improvement in fracture resistance in compressive loading. The versatility of using nanoparticles as cross‐linking agent was further developed by Rose et al. for adhesion between two hydrogels. 28 The authors demonstrated that strong, rapid adhesion between two hydrogels could be achieved at room temperature by adding a droplet of a nanoparticle solution on to the surface of the gel and bringing the other gel into contact with it. This method relied on i) the nanoparticles' ability to be adsorbed onto the polymer gels, ii) to act as a connector between polymer chains and iii) on the ability of the polymer chains to reorganize and dissipate energy under stress when adsorbed onto the nanoparticles. Silica nanoparticles (Si‐NPs), surface‐modified carbon nanotubes or cellulose nano‐crystals were suitable for this purpose. Two samples of biological tissue (e. g. , cut pieces of calf's liver) could be glued together using Si‐NPs. This technology could have potential applications in surgery or tissue engineering. Figure 6 Crosslinking using nanoparticles to form nanoparticle‐hydrogel composites. Cross‐linking using clay nanostructure (Clay‐NS) to form nanoparticle‐hydrogel composites with enhanced mechanical properties. The semiconductor NPs, monomer, and Clay‐NS are homogeneously dispersed in water. Upon photo activation semiconductor nanoparticles produce free radicals initiating polymerization and crosslinking through clay‐NS. The photograph of the vials depicts optical images of the hydrogelation process. A mixture solution of ZnO nanoparticles, DMAA (N, N‐dimethylacrylamide), and Clay‐NS before gelation, hydrogelation after 1 h of irradiation and the resultant elastic ZnO nanocomposite hydrogel taken out of the vial are shown. Adapted and reproduced with permission. 15 Copyright 2014, Macmillan Publishers Ltd. 2. 5 Using Nanoparticles, Polymers and Distinct Gelator Molecules Wu et al. has reported incorporation of Si‐NPs into a conducting polymer hydrogel for Si‐based anodes. 29 The hydrogel was polymerized in situ to produce a well‐connected three dimensional network structure consisting of Si‐NP coated with the conducting polymer. This hydrogel framework combined multiple positive features, including a continuous electrically conductive polyaniline network, binding with the Si surface through either hydrogen bonding with phytic acid or electrostatic interaction with the positively charged polymer, and porous space for volume expansion. Formation of this hydrogel was achieved with a scalable solution phase synthesis by mixing Si NPs with phytic acid and aniline in water followed by the addition of an oxidizer (for example, ammonium persulphate). The aniline oxidised rapidly and polymerized to form cross‐linked polyaniline and the mixture formed a dark green viscous gel due to the presence of the phytic acid gelator. The viscous gel was then bladed onto a copper foil current collector and dried to form a uniform film over a large area for electrochemical applications ( Figure 7 ). Figure 7 Nanoparticle‐Hydrogel composite with gelator molecules to form electrodes. a) Each (silicon‐nanoparticle) Si‐NP is encapsulated within a conductive polymer surface coating and is further connected to the highly porous hydrogel framework. b) Dispersion of Si‐NPs in the hydrogel precursor solution containing the crosslinker (phytic acid), the monomer aniline and the initiator ammonium per sulphate. c) Cross‐linked viscous gel formed after several minutes of chemical reaction, d) the hydrogel gel bladed onto a 520 cm 2 copper foil current collector and dried to form the electrode. Reproduced with permission. 29 Copyright 2013, Macmillan Publishers Ltd. 3 Types of NP‐Hydrogel Composites and Their Applications The innovative combination of nanoparticles and hydrogels create synergistic, unique and potentially useful properties that are not found in the individual components. Properties imparted to the composites depend on the type of nanoparticles incorporated, which in turn is determined by the proposed application of the designed composite. Different types of nanoparticle‐hydrogel composites and their associated properties and applications are described below. 3. 1 Metal NP‐Hydrogel Composites 3. 1. 1 Silver NP‐Hydrogel Composites Silver (Ag) NPs are known for their antimicrobial properties and have been widely used in dental fillings and, more recently in wound and burn dressings to prevent infections. 30 Ag‐NPs bind non‐specifically to bacterial membranes and other components, inducing structural changes that increase membrane permeability and mitochondrial dysfunction. 31 Controlled release of Ag‐NPs is necessary to sustain antimicrobial efficacy. As such, design of Ag NP‐hydrogel composites was expected to provide functional coatings for various applications ( Figure 8 a). 32 Furthermore, properties such as mechanical toughness, swelling ratio, stimuli responsiveness and bio‐compatibility/degradability would need to be investigated and optimized in such composites for effective application. Ag‐NPs have been incorporated into PAAm, 33 polyacrylic acid (PAA), 34 NIPAAm, 35 methyl methacrylate 36 and polyvinyl alcohol (PVA) based hydrogels. 37 These Ag NP‐hydrogel composites show promise as functional anti‐microbial coatings. Semi‐interpenetrating network (IPN) hydrogels offer an excellent alternative for applications requiring higher mechanical toughness. These materials also acted as templates for the synthesis of small (2–5nm) and uniformly distributed Ag‐NPs. 20 Efforts in recent years have shifted to utilizing naturally occurring materials such as chitosan, 38 carbohydrate polymers such as gum acacia and dextran 39 and gelatin 40 to produce bio‐compatible/degradable composite materials that have potential applications as implantable dressings. The controlled‐release of Ag‐NPs provides consistent protection for a period of time, without the need to remove the dressings. Tokarev et al. demonstrated that Ag‐NPs enhance the efficacy of surface plasmon resonance (SPR)‐based sensors. 41 Ag‐NPs incorporated into pH‐responsive hydrogels with the enzyme glucose oxidase function as glucose concentration sensors (Figure 8 b). The swelling‐shrinking transition of Ag‐NP filled hydrogels alter the inter‐particle distance and so affect the optical response of SPR‐based sensors, leading to sensitive glucose detection. 42 Additionally, Ag‐NPs have been incorporated into hydrogels to form electrically conducting hydrogels. The relationship between initial precursor Ag + ions concentration and swelling ratio of the hydrogel have a direct impact on conductivity. A higher concentration of Ag + ions resulted in better conductivity, but reduced swelling ratios, and vice versa. 19 High conductivity without affecting the swelling ratio could potentially be achieved with Ag nanowires (1D nano‐structures) rather than Ag NPs (0D nano‐structures). 43 Figure 8 a) Overview of potential medical applications for Ag NP‐hydrogel composites. Reproduced with permission. 37 Copyright 2013, Elsevier. b) Ag NPs incorporated into pH‐responsive hydrogel based glucose oxidase activity sensor. c) Shifts in silver nanoparticle absortion maxima (Δλ max ) as functions of glucose concentration for the plasmonic sensing device. d) Spectrophotometric glucose sensing. Change in absorption at 470 nm. Adapted with permission. 48 3. 1. 2 Gold NP‐Hydrogel Composites Stimuli‐responsive and switchable conductive Au‐NP‐hydrogel composites have been demonstrated by several groups. 25, 44 External stimuli such as temperature or pH cause a change in conductivity of the hydrogel due to the change in inter‐particle distance. Even though Au‐NP‐hydrogel composites have shown efficacy in SPR based sensors 45 and anti‐bacterial applications, 46 the high cost of gold has so far prevented wide‐scale adoption of Au‐NPs for such applications. Irradiation of light at the Au plasmonic peak induces localized heating within a temperature‐responsive gel matrix. 47 This phenomenon can be used for remote‐controlled drug delivery. If the temperature rises above the lower critical solution temperature (LCST) of the gel matrix, the gel structure collapses resulting in an on‐demand burst release of drugs as opposed to a diffusion‐controlled release. Shiotani and co‐workers have demonstrated this concept using a system ( Figure 9 a) of Au nanorods‐NIPAAm composite hydrogels and rhodamine‐based materials. 48 They reported a fast and reversible shrinking and re‐swelling of this composite hydrogel. The fast response was attributed to heat generation within the gel matrix, rather than external to the matrix. Cytotoxicity studies of these composite gels on cultured cells are underway, with clinical trials expected in the near future. 48 Different sized or structured Au‐NPs immobilized in a thermo‐responsive hydrogel matrix can also function as remote‐controlled microfluidic valves. The principle behind this application relies on the variation of thermal response with nanoparticle size. 49 The use of different excitation wavelengths renders opening of different valves when the correct amount of energy is delivered by matching the plasmonic resonant peaks of the NPs. On irradiation with 532nm light, Au‐colloids (3–10 nm diameter) with plasmonic peak at 532nm collapse the hydrogel network, opening the valve. Valves containing Au‐nanoshell (diameter 120nm, shell thickness 10nm) hydrogel composites remain unaffected. Opening of these valves containing Au‐nanoshells could be achieved upon illumination at 832nm (Figure 9 b). 50 Figure 9 a) Photo‐thermal effect of laser irradiated Au‐nano rods, causing localized collapse of hydrogel. White arrow indicates position of Au‐nano rods and laser irradiation. Reproduced with permission. 57 Copyright 2007, American Chemical Society. b) T‐junction of a microfluidic device with Au‐colloid hydrogel (right valve) and Au nano shell hydrogel (left valve) with 100 μM wide channels i) When the device is illuminated with 532nm green light the left valve opened ii) when the device is illuminated with 832nm IR light the right valve opened. iii) The absorption spectra of gold nano particles and gold nano shells. Adapted and reproduced with permission. 50 3. 1. 3 Other Metal NP‐Hydrogel Composites Although the majority of metal NP‐hydrogel composite studies involve Ag and Au‐NPs, several other metallic NPs also show promise in various fields including catalysis, magnetic components and environmental nanotechnology. Platinum (Pt) metal is well‐known as a hydrogenation catalyst. Pt‐NPs in a bola‐amphiphile hydrogel was found to be an efficient catalyst for the hydrogenation of p‐nitroaniline. 51 This procedure can be used for templated, in situ synthesis of nanoparticles leading to uniform distribution and high loadings. Magnetic NPs of cobalt (Co) or nickel (Ni) have also been incorporated into hydrogels to form soft, magnetic field driven actuators for muscle‐like applications. 52 Ni and Ni‐nickel oxide core‐shell NPs incorporated into PVA hydrogels were responsive to magnetic fields and used for separating and concentrating of chemical species from a mixture. 53 Copper (Cu) has been investigated as a cost effective alternative to Ag and Au‐NP anti‐microbial agents. Cometa and co‐workers have demonstrated an effective Cu‐NP‐poly(ethylene glycol diacrylate) hydrogel anti‐microbial coating, where the combined effect of the high surface/volume ratio of Cu‐NPs and charged quaternary ammonium salts led to potent bactericidal activity. 54 In situ synthesis of bi‐metallic NPs provides interesting nanocomposite hybrid materials. Fe‐Co bimetallic NPs, for example, were synthesized in a PAAm hydrogel network to form magnetic hydrogels for waste removal applications. 55 Poly(2‐acrylamido‐2‐methyl‐1‐propansulfonic) acid has been used as a template for the synthesis of Co:Ni bimetallic NPs, which can catalyse the hydrolysis of sodium borohydride (NaBH 4 ) for hydrogen production and are also magnetically responsive, to be sequestered when not required. 56 Zhang et al. recently used a poly(ethylene oxide propylphosphonamidate) (PEOPPA) hydrogel bearing multi‐amine groups to carry out in situ reduction with concomitant nanoparticle formation. Uniform immobilization of noble metal nanoparticles (Au, Ag, Pd, Pt and Ru) were obtained in the absence of any other reducing agents and stabilizers. 57 These nanoparticle‐infused hydrogels were used for the reduction of nitro aromatics in the presence of NaBH 4. This system showed excellent recyclability with retention of catalytic activity because aggregation induced deactivation was prevented by the hydrogel network. 3. 2 Metal Oxide NP‐Hydrogel Composites Metal oxide NPs hydrogel composites have been developed for their ferromagnetic and semi‐conducting properties. Iron oxide (FexOy) is a commonly synthesized ferromagnetic material that has recently been incorporated into hydrogels to form ferrogels. 58 Ferrogels made from poly(acrylamide‐co‐maleic acid), 59 pNIPAAm, 60 4‐vinylpyridine, 61 chitosan blends with PAA 62 and methacrylate co‐polymers 63 loaded with magnetite NPs are effective absorbents for toxic ions such as lead (Pb 2+ ), chromium (Cr 2+ ) and arsenic (As 5+ ). The embedded pollutants can then be collected magnetically. 64 Similarly, Ozay et al. ( Figure 10 a) used a poly(2‐acrylamido‐2‐methyl‐1‐propansulfonic acid‐co‐vinylimidazole) hydrogel loaded with Fe(II)oxide magnetite NPs to absorb and remove Cu 2+ ions from solution. 65 Ferrogels can also be magnetically driven actuators to mimic muscle movement. Caykara and co‐workers prepared Fe 3 O 4 NPs immobilized in poly(N‐tert‐butylacrylamide‐co‐acrylamide) hydrogels by in situ oxidation of iron precursor in an alkaline medium. 66 These ferrogels displayed actuation under magnetic field (Figure 10 b), suggesting that further research could lead to devices exhibiting human‐like actuation. Magnetite NPs embedded in thermo‐responsive hydrogels can be employed for drug delivery and microfluidic valve control in a similar fashion to that described above for light activated Ag and Au NP composites. 38 In this case, alternating magnetic fields (AMFs) result in a localized temperature change around the NP, stimulating the hydrogel matrix. 67 Titanium oxide and ZnO NPs, have been developed for their UV protective and photocatalytic properties. These materials have potential applications in skin care products and self‐cleaning surfaces. 68 Figure 10 a) Removal of Cu2+ ions in aqueous medium, i) Ferrogel in solution of Cu2+ ions, ii) Magnetic rod attracts ferrogel, iii) Removal of ferrogel is accompanied by the removal of Cu 2+ ions as well. Reproduced with permission. 65 Copyright 2010, Elsevier. b) Bending of ferrogel under varying magnetic field. i) No bending under no field. (ii–iv) Increasing bending observed as magnetic field is increased from 0. 5T to 1. 3T. Magnetic field strength is indicated in the pictures. Reproduced with permission. 66 3. 3 Non‐Metal NP‐Hydrogel Composites Non‐metal based NPs such as carbon‐based materials (graphene oxide, nanodots, nanotubes), quantum dots (CdTe) and Si have been used to create composite materials with unique properties and functions. Si‐NPs have been traditionally used as a support for catalysts or functional materials. Yang et al. incorporated Si‐NPs as cross‐linking centers in PAA based hydrogels that resulted in excellent mechanical strengthening ( Figure 11 a). 69 Si‐NP‐hydrogel composites have also been shown to have higher drug release efficiency when loaded with a drug such as doxorubicin. 70 Alvarez and co‐workers developed an antibiotic loaded (gentamicin) Si‐NP‐collagen composite hydrogel that exhibited prolonged antibacterial activity and increased mechanical strength. 71 Similarly, Lee and co‐workers used Si‐NPs‐collagen hydrogel composites for delivering nerve growth factors for neural tissue engineering. 72 Annaka and co‐workers developed a hydrophobically modified polyethylene glycol containing an hydrophilized Si‐NPs hydrogel composite for injectable intraocular applications. 73 In a similar approach Liu et. al, designed an injectable tissue adhesive using dopamine‐modified four‐armed poly(ethylene glycol and with a synthetic nanosilicate, Laponite. [[qv: 4a]] A subcutaneous implantation study of these materials in rats demonstrated improvement in mechanical and adhesive properties with enhanced cellular infiltration and minimum inflammatory response. Carbon nanotubes, 74 graphene oxide 75 and even melanin, 76 when placed in hydrogels act as NIR light absorbing materials useful for photo‐thermal drug delivery. These materials present synthetic alternatives to Au or Ag‐NP based remote‐controlled photo‐thermal systems. Semiconductor quantum dots CdSe/ZnS and CdSe/CdS have been incorporated into hydrogels to produce fluorescent hydrogels bio‐markers. Chang and co‐workers prepared highly fluorescent, robust cellulose‐QD hydrogels (Figure 11 b) 77 while Salcher and co‐workers made CdSe/CdS QDs/PNIPAAm hydrogel beads for cell labelling. 78 Figure 11 a) Left: Physical demonstration of extreme elasticity of Si‐NPs hydrogels by stretching. Right: Stress‐strain plots showing excellent toughness by Si‐NP hydrogels. SNP 0. 09 indicate gels made with 0. 09mg/mL surface treated silica nanoparticle, SNP0. 009 a indicate gels made with same amount of silica without surface treatment. CC1 is the hydrogel without silica nanoparticles. Reproduced with permission. 69 Copyright 2013, Royal Society of Chemistry. b) Demonstration of cellulose networks in the hydrogels protecting the CdSe/ZnS structure and preserving the quantum dots characteristics. Appearances of the Quantum dot (QD) ‐cellulose hydrogels under a 302 nm UV lamp (left), and Photoluminescence spectra of CdSe/ZnS (core/shell) QDs with average diameter 2. 8 nm (green), 3. 0 nm (yellowish‐green), 3. 2 nm (yellow) and 3. 6 nm (red) respectively in buffer solution. QD‐cellulose hydrogel emission peaks are matching with (right, bottom) with emission peaks of the free quantum dots in buffer. Adapted and reproduced with permission. 77 Copyright 2009, Royal Society of Chemistry. 3. 4 Polymeric NP‐Hydrogel Composites Polymeric NPs composed of micelles, 79 nano‐gels, 80 core‐shell particles, 81 dendrimers, 82 hyper‐branched polymers, 83 and liposomes 84 have been developed for a variety of applications. The inclusion of these particles in a hydrogel imparts multi‐functionality because the polymer particles themselves possess multiple functional groups. Many designs have biomedical applications including drug delivery and bio‐sensing. Zhong et al. enhanced the biological stability of collagen by the physical incorporation of poly amidoamine dendritic NPs. 85 Such physical incorporation resulted in enhanced mechanical properties due to the numerous interactions within the hydrogel network which, in turn, increased human conjunctival fibroblast proliferation. Zhang et al. overcame the limitation of poor drug loading and lack of control over the lower generation dendritic NPs by using generation 5, hyper‐branched polyamine ester nanoparticle in an hydrogel network. 86 These hydrogels allowed week‐long, controlled release of active ingredients which was not possible with hydrogels that did not possess NPs. The use of polymeric NPs for reinforcement of hydrogels is not restricted to biomedical applications. Polystyrene (PS), a material commonly used for packaging and storage, could be used as filler in hydrogels to impart mechanical strength. 87 Bait et al. developed an acrylamide and hydroxyethyl methacrylate hydrogel with PS‐NP fillers for superior elasticity in dermatological patches. 88 Thevenot et al. developed hybrid materials with differing elastic moduli by using alternating layers of PAAm hydrogel composites with and without PS‐NP fillers. Physical deformation of this bi‐layer gel produced an electrical potential, useful in developing soft tactile‐sensing devices ( Figure 12 ). 89 Polypyrrole (PPy) is a semi‐conducting conjugated polymer used in organic‐based optoelectronic devices. Luo et al. developed an agarose/alginate double network hydrogel incorporating PPy NPs for use as an infrared responsive releasing system, much like those employing reduced graphene oxide NPs. 38 Figure 12 a) Mechanical compression of bilayer gels with and without poly styrene nanoparticles (PSNPs) Gel 1: without PS NP (top) and Gel 2: with 26% PS NP (bottom part of the cylindrical gel) are attached together and compression is given from top (arrow). b) Time profile of deformation ΔL 1 /L and ΔL 2 /L (top) and electric potential ΔV generated (bottom) upon deformation. Reproduced with permission. 89 Copyright 2007, Royal Society of Chemistry. 4 Conclusions and Outlook We have reviewed nanoparticle‐hydrogel composites as a state‐of‐the‐art, versatile class of materials suitable for a wide range of applications. Synthetic methods and strategies and the unique synergistic properties of the composite that is absent in individual components, together with their applications, have been summarized. Nanoparticle‐hydrogel composites exhibit multi‐functional and stimuli responsive properties, making them ideal for “smart” materials, including i) antimicrobial gels/barriers/matrices; ii) photo thermally active hydrogels; iii) soft material catalysts; iv) environmental absorbents; v) drug delivery vehicles; vi) optical detection sensors and vii) soft actuators. Potential applications include i) safe, clinically implantable nanoparticle‐hydrogel composite systems for bio‐sensing and therapy; ii) environmental remediation systems for catalytic oxidation of toxins/removal of pollutants; iii) recyclable catalytic nanoparticle hydrogel composites for chemical synthesis and iv) composite hydrogel patches for cosmetic applications. We expect that the classification of synthetic approaches and applications of nanoparticle‐hydrogel composites presented in this review provide a better understanding of the system and enable the reader to design innovative combinations for new applications. Control of the covalent and supra molecular interactions by synthetic design and prediction of the resultant properties in the nanoparticle‐hydrogel composite are major areas to be developed in this emerging field. Such predictions upon experimental validation will form the foundation of design for the next generation nanoparticle‐hydrogel composites with optimal properties for a desired application. In the coming years, such design of nanoparticle‐hydrogel composites will not only result in advanced applications, but will also steer the fundamental understanding of material interactions, aiding computational prediction of properties of new composites from existing components.
|
10. 1002/advs. 201500026
| 2,015
|
Advanced Science
|
Virus Nanoparticles Mediated Osteogenic Differentiation of Bone Derived Mesenchymal Stem Cells
|
There are few methodologies that allow manipulating a biomaterial surface at nanometer scale, which controllably influence different cellular functions. In this study, virus nanoparticles with different structural features are selected to prepare 2D substrates with defined nanoscale topographies and the cellular responses are investigated. It is demonstrated that the viral nanoparticle based substrates could accelerate and enhance osteogenesis of bone derived mesenchymal stem cells as indicated by the upregulation of osteogenic markers, including bone morphogenetic protein‐2, osteocalcin, and osteopontin, at both gene and protein expression levels. Moreover, alkaline phosphatase activity and calcium mineralization, both indicators for a successful bone formation, are also increased in cells grown on these nanoscale possessed substrates. These discoveries and developments present a new paradigm for nanoscale engineering of a biomaterial surface.
|
1 Introduction It is well established that cell–material interactions regulate numerous cellular functions. 1 Biological processes such as adhesion, growth, differentiation, and apoptosis, are controlled by cell shape and cytoskeletal organization which is directed by cell–surface interactions. 2 Meanwhile, the surface chemistry and topography of materials play a very crucial role in altering cell behaviors at many stages of cell growth and development. [[qv: 1c]], 3 Although the dimensions of mammalian cells are in the order of a few micrometers, cellular sensing of the external environment and interaction with biomaterials occurs at the nanometer level. 4 Cell interactions with nanometric surfaces often result in a specific sequence of gene and protein regulations. These series of events initiate as early as the cell begins to sense the surrounding environment. Therefore, the understanding of various topographical cues that are responsible for cellular behaviors is a key to advance tissue engineering. In general, topographical cues can be classified as: 1) the roughness of the underlying surface, 2) the ligand‐display pattern and density, 3) the size and shape of the contact area for cell spreading, and 4) the geometry of topological features at a nanometer scale. 5 Reviewing the effect of individual cues is often complicated due to the difficulty in controlling and altering particular topographical features while preserving others. Micro/nanofabrication techniques are required to enable the recapitulation of topographical cues in the cell niche in a controllable and reproducible fashion. Examples of these technologies are mechanical roughening, 6 nano‐ and microindentation, and substrate‐templating using a well‐defined relief to impart topography with solvent‐casting, electrodeposition, chemical‐vapor depositions, or compression‐molding processes. 7 These engineered micro/nanoscale topographical cues mimic the micro/nanoscale features in the physiological environment, which can be used to demonstrate how individual cues or the combination of topographical cues affect a particular cellular response. However, all these methods suffer from the laborious process, the inability of predictably generating chemistry and topography in a simultaneous fashion, the requirement for high‐cost equipment, and the limited class of material can be used. [[qv: 2c]], 8 Certain virus particles, especially the plant viruses, have well characterized 3D structure and can be produced in high yield and purity. 9 The multivalent organization of the coat proteins make the viral particles powerful scaffolds for display of a variety of functional groups via chemical conjugation or genetic modification. [[qv: 9a]], 10 In the past two decades, using virus particles as building blocks, novel materials with unique structural features have been developed for a wide range of applications, including electronics, sensing, gene/drug delivery, bioimaging, and vaccine development. [[qv: 10c]], 11 However, so far there is no systematic investigation about how the nanoscale topographical cues of various plant virus particles coated substrates impact cell behaviors, specifically, osteogenesis of bone derived mesenchymal stem cells (BMSCs). From our unexpected, yet significant, observation that rodlike plant virus nanoparticle, tobacco mosaic virus (TMV) coated two dimensional (2D) substrate dramatically accelerates osteogenesis of BMSCs. The study suggested that the virus does not act as soluble inducer as supplementing cell culture media with TMV solution failed to mediate the differentiation. 12 We have hypothesized that shape of virus nanoparticle and/or nanoscale topography provided by surface structure of virus particle is necessary for the enhanced osteogenesis. Therefore, in this study, we generate a series of plant virus nanoparticles coated substrates with distinct morphology and nanotopography via electrostatic interaction. We applied these virus based scaffolds to investigate cellular responses to two types of the topographical cues: 1) the geometry of topological features by testing effects of three different viral particle shapes including rigid rod, spherical, and flexible fiber; and 2) the size and shape of the contact area for cell spreading at nanoscale level by utilizing viral particles with same shape but different in nanoscale features, constructed from different structure of coat protein that assembles around virus genomic material as shown in Figure 1. Our results show that some of these virus based scaffolds accelerate and enhance osteogenic differentiation of BMSCs. This finding presented here may provide a new route for enhancing the performance of orthopedic implants by regulating stem cell differentiation with nanotopography. Figure 1 A–I) Molecular models shows surface topography of plant viruses used in this study. A, B) Tobacco mosaic virus (TMV); C) potato virus X (PVX); D, E) turnip yellow mosaic virus (TYMV); F, G) turnip vein clearing virus (TVCV); H, I) cowpea mosaic virus (CPMV). Scale bar indicates 10 nm in (A), (C), (D), (F), and (H) and 5 nm in (B), (E), (G), and (I). The models were generated using Pymol ( www. pymol. org ) with coordinates obtained from RCSB protein data bank. 2 Results and Discussion 2. 1 Fabrication of Virus‐Coated Scaffolds by Layer‐by‐Layer Deposition Method We fabricated 2D virus based substrates from five plant viruses which can be categorized into three groups by morphology of the viral particles rod shape virus; TMV and turnip vein clearing virus (TVCV), filamentous virus; potato virus X (PVX), and spherical virus; turnip yellow mosaic virus (TYMV) and cowpea mosaic virus (CPMV). Not only are these plant viruses morphologically different, but also they are nanotopographically dissimilar as shown in Figure 1. Since all these viral particles have isoelectric pH less than 5. 5, overall surface charges on these particles are negative in neutral condition. Via an electrostatic interaction, negatively charged viral particles can be strongly adsorbed on 12‐well plates coated with poly‐ d ‐lysine (PDL), a positively charged biocompatible polymer. And the interaction helps to retain the viral particles on surface of the substrates. By depositing structurally and nanotopographically distinct viral particles on PDL coated substrate, we can readily construct an array of virus‐coated scaffolds with various topographies offered by the intrinsic morphology and micro/nanotopography of each viral particle. The presence of viral particles on PDL coated surface was confirmed by atomic force microscopy (AFM) ( Figure 2 ). The AFM images also show a nearly complete coverage of substrates by intact viral particles. The virus particles are randomly adsorbed on 12‐well plates coated with poly‐ d ‐lysine, however, some area of the virus coated substrates appeared to show direction of virus particles coating under AFM. This coating pattern results from the natural irregularity of the cell culture surface of 12‐well plate. To prevent the effect of plate pattern, same lot of 12‐well plates was used throughout this study. The virus substrates have been characterized in term of root mean square roughness from data collected from AFM images ( n = 4). There is no significant difference of microscale roughness across the virus coated substrates, created from deposition of numerous virus particles on the substrate surface, across these five virus substrates. However, different type of virus particle has dissimilar nanoscale topography provided by surface of each particle of virus as shown in Figure 1. Figure 2 Representative AFM images showing the coverage of PDL coated substrate with different virus nanoparticles indicate the viral particles, A) TMV; B) TVCV; C) PVX; D) TYMV; and E) CPMV, are mostly intact and fully cover the coating area. F) Root mean square roughness of different virus nanoparticles coated substrates by AFM analysis. Scale bars indicate 1. 25 μm in (A)–(C) and 0. 5 μm in (D) and (E). The data are expressed as mean ± s. d. ( n = 4) ns indicates nonsignificant and p > 0. 05 based on ANOVA. 2. 2 Viral Particles Coated Substrates Promote Osteogenesis To investigate the effect of surface topography on osteogenesis, we culture BMSCs on PDL coated substrate and the five virus‐based substrates and study the osteoblastic differentiation. BMSCs are isolated and cultured as reported in literature. The purity of the stem cells populations has been previously verified with several stem cells markers such as Cluster of Differentiation 73 and 90 (CD73 and CD90). 13 The difference in the expression of bone morphogenetic protein‐2 (BMP2) gene, an early osteogenic marker, [[qv: 12b]] among BMSCs cultured on PDL and virus substrates were recorded at 6 h after osteoinduction (Figure S1, Supporting Information). Moreover, after 7 d of induction, osteocalcin (BGLAP) and osteopontin (SPP1) genes expressions were higher compared to uninduced BMSCs ( Figure 3 ). These two genes are noncollagen genes actively involved during proliferation period. Osteocalcin is a specific marker for the osteoblast differentiation and mineralization, and is expressed exclusively during the postproliferative period and reaches its maximum expression during mineralization and accumulates in the mineralized bone. 14 Osteopontin is known to serve as a bridge between the cells and the hydroxyapatite through the arginine‐glycine‐aspartic acid (RGD) peptide and polyaspartate sequences present in it. It is one of the early markers of osteoblastic differentiation. 15 We observed significant changes in the expression of all three osteospecific genes in cells plated on the virus based substrates, except CPMV coated substrate, compared to cells grown on bare PDL substrate. Interestingly, in the case of spherical‐shaped viral particles, while TYMV coated substrates increased BMP2 gene expression by fourfold and dramatic increment of BGLAP and SPP1 were observed, there was no significant difference in these gene expressions between cells plated on PDL and CPMV substrates. Figure 3 The expression of osteogenic markers in BMSCs cultured on PDL and different virus nanoparticles coated substrates under osteogenic conditions. Quantitative real‐time PCR analysis (RT‐qPCR) showed upregulation of A) osteocalcin and B) osteopontin in cells grown on TMV, TVCV, PVX, and TYMV (but not on CPMV) coated substrates at 7 d after osteogenic induction. C) Immunohistochemical staining reveals that osteocalcin, a canonical osteogenic marker, is exclusively located in cell aggregates growing on TMV, TVCV, PVX, and TYMV substrates (not for CPMV coated substrate). Color representation: nucleus (blue), osteocalcin (red). Scale bar is 100 μm. The data were expressed as mean ± s. d. ( n = 3, * p ≤ 0. 05, ** p ≤ 0. 01 based on ANOVA). In consistence with gene expression data, immunofluorescence imaging of BMP2 (Figure S2, Supporting Information) and osteocalcin (Figure 3 C) revealed that the morphogens are localized in the cell aggregates on the four virus coated substrates. BMSCs cultured on TMV, TVCV, PVX, and TYMV develop greater cell nodules, a notable feature of BMSCs undergoing osteogenesis. In order to quantify the differences in the spatial distributions of cells on each substrates, we acquired the coordination of cells and applied nearest neighbor analysis. 16 The spatial distributions of BMSCs on TMV, TVCV, PVX, and TYMV substrates were similar to the theoretical “cluster” distribution, which indicates cells tend to cluster to form the cell nodules ( Figure 4 ). On the other hand, the spatial distribution of BMSCs on PDL and CPMV were similar to the “independent” distribution and shifted toward “regular” distribution. The data suggest that TMV, TVCV, PVX, and TYMV coated substrates are more favorable to the osteogenesis of BMSCs than PDL and CPMV substrates. Figure 4 Nearest neighbor analysis of BMSCs cultured on PDL and virus substrates. A–C, G–I) DAPI immunohistochemical staining and D–F, J–L) bright field microscopy images of BMSCs on A, D) PDL, B, E) TMV, C, F) TVCV, G, J) PVX, H, K) TYMV, and I, L) CPMV. M) Schematic diagrams of the nearest neighbor analysis. In this analysis the distribution of cells can range from independent (represented by a theoretical Poisson's distribution), to clustered, or regular. N) Plot of BMSCs spatial distribution on PDL control and virus substrates demonstrated cluster growth pattern which correlated to appearance of cells nodules on TMV, TVCV, PVX, and TYMV virus coated substrates. Scale bar is 200 μm. These cell clusters displayed robust positive staining for BMP2 in cell aggregates (Figure S2, Supporting Information). No fluorescence signal was detected in cells grown on PDL control and CPMV substrates. Similarly, immunohistochemical staining of osteocalcin at 14 d indicates that the canonical osteogenic marker was exclusively found in cells aggregates on TMV, TVCV, PVX, and TYMV substrates. In addition to the analysis of osteo‐specific markers, alkaline phosphatase (ALP) activity and calcium mineralization supported the osteogenic differentiation of cells on the four virus based scaffolds. ALP is an early marker of osteogenesis and its activity mediates matrix mineralization. 17 Cytochemical analysis of the osteogenesis process of BMSCs on PDL and virus coated substrates at day 4 and 7 after osteogenic induction suggested that cells on TMV, TVCV, PVX, and TYMV substrates had an increase in ALP activity at day 4, whereas CPMV substrates did not alter the enzyme activity when compared to PDL control. The enzyme activity drops to baseline at day 7 for cells on TMV and TVCV substrates ( Figure 5 A). It is possible that cells on these two virus substrates undergo differentiation and reach mineralization period earlier than cells on other substrates since alkaline phosphatase activity rises during cell proliferation and achieves maximum level as the culture progresses into mineralization stage. However, cellular level of ALP declines as mineralization progresses. 18 Additionally, cells on the four virus substrates at day 7 were positively stained by Alizarin red S which showed deep red color for calcium deposition in large cell nodules, whereas negatively stained was observed on PDL substrates (Figure 5 C). Cells on CPMV substrate only formed small nodules that were also stained with Alizarin red S. Quantification of dissolved Alizarin red S dye from cells nodules by UV–vis absorbance indicated that the mineralization of cells on TMV substrates doubled that of PDL, and PVX and TYMV substrates increased the mineralization by fourfold, while TVCV substrates slightly increased the mineralization of cells compared to PDL control substrates but not statistically significant (Figure 5 B). However, the calcium mineralization is an accumulation process, longer incubation time of cells on these substrates could increase the difference in calcium deposition between each substrate and may increase difference of the mineralization between cells on TVCV and PDL coated substrates. Cells on CPMV substrate have comparable calcium mineralization to cells on PDL control. The combined results from quantitative real‐time PCR analysis (RT‐qPCR), immunohistochemical staining, nearest neighbor analysis, enzyme activity, and calcium mineralization unambiguously indicate that TMV, TVCV, PVX, and TYMV substrates can accelerate and enhance osteogenesis of BMSCs. The accelerated osteogenic differentiation of BMSCs on TMV and TYMV substrates has been demonstrated before in our previous studies when BMSCs were cultured on the viruses coated APTES glass coverslips. 12, 19 In this study, we have confirmed that it is the topography created by deposition of virus nanoparticles on substrates, not underlying material, which mediates such differentiation, as we apply different backup material; poly‐ d‐ lysine coated tissue culture plate. We also expand the library of virus based substrates to include another morphology of virus nanoparticle; flexible fiber (PVX) as well as other types of virus nanoparticles with dissimilar nanoscale topography (TVCV and CPMV). Figure 5 Cytochemical analysis of the bone differentiation process of BMSCs on PDL and viruses coated substrates at 4 and 7 d after osteogenic induction. A) Alkaline phosphatase activity of cells cultured on different substrates. The data are expressed as mean ± s. d. ( n = 3, * p ≤ 0. 05, ** p ≤ 0. 01, **** p ≤ 0. 0001 based on ANOVA). B) Alizarin red staining of each sample at day 7. Cells on virus substrates are positively stained for calcium deposition, whereas negatively stained is observed on PDL substrates. The data are expressed as mean ± s. d. ( n = 3, ** p ≤ 0. 01, *** p ≤ 0. 001, **** p ≤ 0. 0001 based on ANOVA). C) Absorbance at 548 nm normalized to cell number to indicate a relative amount calcium deposit at day 7 stained by alizarin red solution. The mineralization of cells on TMV substrates doubles that of PDL, while PVX and TYMV substrates increase the mineralization by fourfold. TVCV substrates slightly increase the mineralization of cells compare to PDL control substrates. These evidences suggest an improvement in osteogenesis by virus coated substrates. 2. 3 Nanotopography of Viral Based Scaffolds Alters Cells Morphology and Induces Differentiation The majority of cells cultured on the four virus substrates have noticeably smaller size at 24 h after seeding compared to those on PDL and CPMV substrates. Previous study illustrated that cell shape and size are associated with adhesion strength of cells on a substrate. 20 Additionally, several reports showed that integrin‐mediated focal adhesion (FA) is an important regulator of osteogenesis. 21 It is hypothesized that too strong substrate binding may inhibit osteogenic differentiation. Mendonça et al. observed higher osteogenic differentiation of stem cells that attached looser on rough titanium disks than strongly attached cells on smooth substrate. 22 This could possibly be due to the limitation of cells movement or migration. Strength of cell adhesion and larger focal adhesion size are correlated to an increase in localization of vinculin. 23 Therefore, we investigated cell adhesion on virus substrates by using fluorescence imaging of vinculin, a protein of focal adhesion complexes (FAC), to analyze average focal adhesion size of cells grown on PDL and virus substrates for 24 h prior to osteoinduction. Vinculin signals were captured by fluorescence microscopy for size analysis by Slidebook 5 software. The data revealed the reduction in vinculin size of cells on TMV, TVCV, PVX, and TYMV but not CPMV substrates ( Figure 6 ). Figure 6 Immunochemical staining showing the difference in vinculin size of cells on PDL or virus coated substrates for 24 h. A) Immunofluorescence images of cells on different substrates at 24 h prior to osteoinduction (top panel). Color representation: nucleus (blue), vinculin (green), and phalloidin (red). The bottom panel demonstrates vinculin masking and selection of vinculins for size analysis. The selected vinculin spots are highlighted in blue. Scale bar is 50 μm. B) Average vinculin size of cells on different substrates. The data were expressed as mean ± s. d. ( n = 3, * represents p ≤ 0. 05 based on ANOVA). These results suggest that BMSCs attached to the four virus substrates weakly, whereas larger size of FACs dictates stronger cell–substrate adhesion in PDL and CPMV substrates. 24 The significantly smaller FA size for cells on the four virus substrates might increase cellular motility and facilitate the formation of cell aggregates within 6 h of osteoinduction. The larger FA size observed in CPMV sample, which did not improve osteogenic differentiation, might be due to the expression of vimentin binding ligand on CPMV coat proteins. 25 The vimentin cytoskeleton was shown to regulate focal contact size and help stabilize cell–matrix adhesion in endothelial cells. 26 Since major cytoskeletal component of mesenchymal cells is vimentin, the presence of vimentin binding ligand on CPMV substrate could supply additional adhesion points and consequently leads to higher adhesion strength of cells cultured on CPVM coated substrate, therefore mitigating cell migration and differentiation. Several reports previously described that elongated shapes and geometries that present features of subcellular concavity at the cell perimeter increase the cytoskeletal tension in mesenchymal stem cells (MSCs), thus promoting the preference for osteogenesis. 27 These similar geometries of BMSCs were also observed in our study. Representative actin and vinculin immunofluorescent heat maps of cells initially adhere on PDL and each virus coated substrates suggests that cells on TMV, TVCV, PVX, and TYMV were more elongated with higher actin stress fiber on the long axis of cells, and the majority of them had concave features that led to high cytoskeleton tension in the region. Furthermore, vinculin of cells that grew on these four substrates were highly localized at the protrusion area which was different from those of cells on PDL and CPMV coated substrates. The majority of cells on PDL and CPMV coated substrates were round in shape with evenly distributed actin filaments and vinculin around cell perimeter (Figure S3, Supporting Information). Moreover, overall morphology of cells on each virus substrates, which can be investigated from Figure 4 and Figure S2 (Supporting Information), reveals that cells on CPMV have more spread out shape compared to cells on other virus substrates. These data of morphology and immunofluorescence heat maps along with small FA size suggest that loose attachment of cells on unfriendly four virus, TMV, TVCV, PVX, and TYMV, coated substrates result in cytoskeleton tension, thereby enhancing osteogenic differentiation of BMSCs. Interestingly, data from this study suggest that the effect of nanoparticle morphology on differentiation is negligible. As observed from all experiments, osteogenic differentiation is comparable in cells cultured on substrates coated with different shape of virus. 3 Conclusion A series of assorted micro‐/nanoscale features possessed 2D peptide based scaffolds can be simply constructed from structurally distinct viral bionanoparticles by using fundamental electrostatic interaction. These virus based 2D scaffolds were used to investigate osteogenesis of BMSCs. The combined results from RT‐qPCR and immunostaining of BMP2 suggest an early osteogenesis of cells on TMV, TVCV, PVX, and TYMV coated substrates as early as 6 h after osteoinduction. Furthermore, the confirmation of the strong commitment in osteoinduction in longer term was evidenced by RT‐qPCR and immunostaining of osteocalcin and osteopontin, as well as enzyme activity, and calcium mineralization. These results suggest that topographies created by TMV, TVCV, PVX, and TYMV coated substrates stimulate and enhance osteogenic differentiation. The underlying mechanisms of the observation are proposed that the stress created by the unfavorable surface from the four viral nanoparticles causes the reduction in FA size, which in turn increases cell motility and facilitates the formation of cell aggregates. The unfavorable surface may also obstruct cell spreading therefore increased cytoskeleton tension which results in high aspect ratio or subcellular concavity at the cell perimeter, thus promoting osteogenesis. Further investigation about topography‐induced differentiation is necessary for a better understanding of how surface topography provided by viral particles affect cell–material adhesion complex and facilitate the differentiation. Additionally, a continued study can be done on the investigation of the alignment or patterning of virus particles on the cellular responses as the unique structure or morphology of virus particles make them feasible for hierarchical structure formation in both 2D and 3D substrates. 28 More importantly, it will be very interesting to study if our discovery can be extended to other synthetic substrates and employed in clinical tissue engineering applications. 4 Experimental Section Virus Purification from Infected Leaves : Purification of TMV, TVCV, TYMV, and CPMV were done by first, infected leaves were blended with three volumes of 0. 1 m potassium phosphate buffer pH 7. 0 and 0. 1% β‐mercaptoethanol. The mixture was filtered, and the filtrate was subjected to centrifugation to remove bulk plant material. The supernatant was collected and clarified by adding an equal volume of CHCl 3 /1‐butanol (v/v = 1:1). The aqueous layer was then collected followed by precipitation of virus with 4% PEG 8K and 0. 2 m NaCl. The pellet was centrifuged and resuspended in buffer before it was subjected to low speed centrifuge to remove PEG. The virus in supernatant was finally pelleted by ultracentrifugation and resuspended in buffer. For purification of PVX, infected leaves were blended with two volumes of 0. 1 m potassium phosphate buffer pH 8. 0, 10% ethanol, and 0. 1% β‐mercaptoethanol. The mixture was filtered, and the filtrate was subjected to centrifugation to remove bulk plant material. The supernatant was collected and clarified by adding 1% Triton X‐100. After centrifugation the supernatant was collected and processed by adding 4% PEG 8K and 0. 2 m NaCl to precipitate virus. The pellet was centrifuged, resuspended in buffer, and purified by sucrose gradient. Fabrication of Virus Based Scaffolds : 1 mg mL −1 TMV, TYMV, CPMV, 10 mg mL −1 TVCV, and 2. 67 mg mL −1 PVX in aqueous solution 0. 7 mL were dropped into 12‐well plates that were coated with poly‐ d ‐lysine using protocol suggested by Corning. The virus solutions were incubated with the PDL coated plate under sterile cells culture hood for overnight. Then the bottoms of each well were rinsed briefly with 18. 2 mΩ water before used for BMSCs culture. Surface Characterization of Virus Based Scaffolds by AFM : The surface morphology of virus based scaffolds was observed by AFM (Nanoscope IIIA MultiMode AFM (Veeco)). The bottoms of each 12‐well plate were cut out after virus coating and rinsed with 18. 2 mΩ water, then dried with a stream of nitrogen gas before mounting onto AFM sample holder for imaging in the tapping mode. BMSC Isolation and Expansion : Primary BMSCs were isolated from the bone marrow of young adult 80 g male Wister rats (Harlan Sprague‐Dawley Inc. ). The procedures were performed in accordance with the guideline for animal experimentation by the Institutional Animal Care and Use committee, School of Medicine, University of South Carolina. Cells were maintained in primary media (Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), penicillin (100 U mL −1 ), streptomycin (100 μg mL −1 ), and amphotericin B (250 ng mL −1 )), kept at 37 °C in a CO 2 incubator with 95% air/5% CO 2 and passaged no more than seven times after isolation. To induce osteogenesis, primary media were replaced with osteogenic media consisting of DMEM supplemented with 10% FBS, penicillin (100 U mL −1 ), streptomycin (100 μg mL −1 ), and amphotericin B (250 ng mL −1 ), 10 × 10 −3 m sodium β‐glycerolphosphate, l ‐ascorbic acid 2‐phosphate (50 μg mL −1 ), and 10 −8 m dexamethasone. Media were replenished every 3–4 d. RT‐qPCR Analysis : PDL and virus coated substrates were seeded with 4. 0 × 10 4 cells well −1 in primary media and allowed to attach overnight. The unseeded cells were used as a control to normalize the change in gene expression. The media were replaced to osteogenic media and cultured for 6 h, 4 d, 7 d, and 14 d. The cell cultures were terminated at these time points and total RNA was subsequently extracted using E. Z. N. A. RNA Isolation Kit, OMEGA. At least two separate experiments were conducted with each type of sample. The purity and quantity of the extracted RNA were analyzed using Thermo Scientific Nanodrop 2000c spectrophotometer and was reverse transcripted by qScript cDNA Supermix (Quanta Biosciences). RT‐qPCR (iQ5 real‐time PCR detection system Bio‐Rad Laboratories) was done by the method described as: 60 cycles of PCR (95 °C for 20 s, 58 °C for 15 s, and 72 °C for 15 s), after initial denaturation step of 5 min at 95 °C, by using 12. 5 μL of iQ5 SYBR Green I Supermix, 2 pmol μL −1 of each forward and reverse primers and 0. 5 μL cDNA templates in a final reaction volume of 25 μL. Glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) was used as the house keeping gene. Data collection was enabled at 72 °C in each cycle and C T (threshold cycle) values were calculated using the iQ5 optical system software version 2. 1. The expression levels of differentiated genes and undifferentiated genes were calculated using Pfaffl's method (M. W. Pfaffl, G. W. Horgan, and L. Dempfle, Relative expression software tool) for group‐wise comparison and statistical analysis of relative expression results in real‐time PCR, using GAPDH as the reference gene. Quantification of gene expression was based on the C T value of each sample which was calculated as the average of at least two replicate measurements for each sample analyzed. “Pairwise fixed reallocation randomization test” was performed on each sample and a value of p < 0. 05 was regarded as significant. The primers used for RT‐qPCR are shown in Figure S4 (Supporting Information). Three independent experiments were performed and analyzed for each gene expression study. ALP Activity : After 4 and 7 d of induction in the osteogenic media, the BMSCs seeded on PDL and virus coated substrates were determined as number of cells on each substrate by CellTiter Blue assay. Then the cells were fixed with 4% paraformaldehyde for 15 min at room temperature prior to analyze ALP activity by incubating the briefly fixed cells with 1‐Step p ‐nitrophenylphosphate solution (Thermo Scientific) for 15 min at room temperature. The solution was transferred to a new microfuge tube containing 250 μL of 2 n NaOH and the absorbance at 405 nm was measured. The measured ALP activity from each sample was normalized to the corresponding cell number. Three independent experiments were performed and analyzed for ALP activity. Alizarin Red Staining and Quantification : Calcium deposition on each substrate was visualized and quantified to confirm and compare osteogenic differentiation by Alizarin red staining. Fixed cell on day 7 were stained with 0. 1% Alizarin red solution (Sigma‐Aldrich) pH 4. 1–4. 5 for 30 min in the dark. The samples were washed with water (18. 2 MΩ) prior to imaging. To quantify the amount of dye on each substrate, 300 μL of 0. 1 n NaOH was added to each sample to extract the dye from the sample. The extracted dye solution measured the absorbance at 548 nm wavelength. The measured absorbance from each sample was normalized to the corresponding cell number from CellTiter Blue assay. Three independent experiments were performed and analyzed for Alizarin red staining and quantification. Immunofluorescence Assays and Image Analysis : For immunofluorescence assays and image analysis, PDL or viral particles coated glass coverslips were used as substrate for BMSCs culture. The substrates were seeded with 4. 0 × 10 5 cells sample −1. The cultures were terminated at 24 h after seeding to be used as vinculin immunostaining samples, 6 h after osteoinduction for BMP2 immunostaining analysis and 14 d after osteoinduction for osteocalcin immunostaining study. After termination, cells were fixed in 4% paraformaldehyde at room temperature for 30 min. Each of the samples was then permeabilized for 20 min by 0. 1% Triton‐X 100 for 15 min and blocked in 1. 5% bovine serum albumin (BSA, Sigma Aldrich) in PBS for 1 h at room temperature. After the blocking, the cells were incubated overnight with mouse monoclonal antibody targeting BMP2 (R&D Systems) at 1:100 dilution in blocking buffer or rabbit polyclonal antibody targeting osteocalcin (Santa Cruz Biotechnology) at 1:100 dilution in blocking buffer or mouse monoclonal antibody targeting vinculin (Neomarkers) at 1:200 dilution in blocking buffer. After overnight incubation, secondary goat antimouse antibody conjugated with fluorescein (Chemicon) was used at 1:400 dilution for 2 h at room temperature with BMP2 and vinculin samples. Secondary goat antirabbit antibody conjugated with Alexa Fluor 546 (Invitrogen) was used at 1:800 dilution for 2 h at room temperature with osteocalcin samples. Rhodamin phalloidin (1:100 in PBS) was used to stain filamentous actin in BMP2 and vinculin samples. Fluorescein phalloidin (1:500 in PBS) was used to stain filamentous actin in osteocalcin samples. Nuclei were stained with DAPI (4, 6‐diamidino‐2‐phenylindole, 100 ng mL −1 ). The samples were then mounted and sealed with clear nail polish before imaging. Images of the stained substrates were taken on an Olympus IX81 fluorescent microscope. SlideBook 5 was used to select and analyze immunofluorescence images of vinculin. After setting the threshold for masks, the criteria used to select vinculin spots to be analyzed were XY shape factor larger than 1. 5 and area size between 0. 5 and 1. 5 μm 2. The average size of vinculin for each image was calculated, followed by the calculation of average vinculin size of cells on PDL and virus substrates and the standard deviation from average values of three individual images which provide more than 500 vinculins for analysis per sample. Immunofluorescence heat maps of actin and vinculin were generated by ImageJ software. Color histogram was generated by measuring pixel intensity across the immunofluorescence heat maps of representative cells on each substrate. Spatial Distribution Analysis of BMSCs Cultured on PDL and Virus Coated Substrates : The spatial distribution of the cells on different substrates was analyzed by NIH ImageJ and R ( http://www. R‐project. org ) software packages. The fluorescence images of cell nuclei were primarily processed with ImageJ to be presented as particles, and their centroid coordinates were determined. These data were then imported into R for nearest‐neighbor analysis using the SpatStat module. The spatial distribution patterns of cells were identified for 70–90 cells on each substrate. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500069
| 2,015
|
Advanced Science
|
Multifunctional Hydrogels with Reversible 3D Ordered Macroporous Structures
|
Three‐dimensionally ordered macroporous (3DOM) hydrogels prepared by colloidal crystals templating display highly reversible shape memory properties, as confirmed by indirect electron microscopy imaging of their inverse replicas and direct nanoscale resolution X‐ray microscopy imaging of the hydrated hydrogels. Modifications of functional groups in the 3DOM hydrogels result in various materials with programmed properties for a wide range of applications.
|
Hydrogels are synthetic or natural polymer networks crosslinked either chemically or physically to maintain 3D structures that are able to absorb and retain large amounts of water. 1 Porous biocompatible hydrogels combine porosity with biocompatibility and play an important role in various biomedical applications, such as scaffold materials in tissue engineering, cell transplantation, and regenerative medicine. 2 The properties of porous hydrogels are substantially influenced by the porosity related parameters, including surface area, pore size, pore interconnectivity, interfacial properties, etc. 3 Therefore, rational design of hydrogels with desired structures and properties is an essential requirement for advanced technological applications. The properties can be additionally programmed by modification of functional porous hydrogels. Macroporous hydrogels prepared by colloidal crystal templating have attracted much attention in recent years due to their unique structural features, holding immense promise as responsive materials for detection, 4 scaffolds for tissue engineering, 5 and absorbents for CO 2 capture. 6 The colloidal crystal templating is a facile and effective method to construct 3D ordered macroporous (3DOM) structures with interconnected pores and well‐controlled pore size. 7 However, the porous structures of 3DOM hydrogels have not been definitively determined in many cases, which is largely due to the difficulty in morphological characterizations, and this has posed an obstacle to further precise studies. Herein, we analyze the reversible 3DOM porous structures in colloidal crystal templated macroporous hydrogels through two approaches: one procedure is electron microscopy imaging of their inverse replicas and the other is noninvasive and nondestructive nanoscale resolution X‐ray microscopy (nano‐XRM) imaging of the hydrated hydrogels. Thus, one focus of this work is the development of two novel methods for the characterization of 3DOM structures in hydrogels. This work also demonstrates for the first time the intrinsic shape memory properties of 3DOM hydrogels. The additional aim of this work is further modification of functional 3DOM hydrogels. The properties of such structurally tailored and engineered macromolecular ( stem ) gels can be programmed, depending on the applied chemistry. Thus, the functional hydrogels represent a robust platform for constructing well‐defined functional materials with preselected properties targeting a wide range of applications. The preparation of 3DOM hydrogels is schematically shown in Figure 1. Uniform poly(methyl methacrylate) (PMMA) colloidal spheres with average diameter of 286 nm were synthesized by surfactant free emulsion polymerization and assembled to face centered cubic crystalline structure by centrifugation (Supporting Information, Figure S1). An aqueous solution of a mixture of monomer (i. e. , poly(ethylene oxide) methacrylate (PEOMA)) and crosslinker (i. e. , poly(ethylene oxide) dimethacrylate (PEODMA)) was infiltrated into the voids between the stacked PMMA spheres by capillary force, which should reduce the defects generated in the porous structure compared to infiltration under partial vacuum. [[qv: 6b]] Thermal‐initiated free radical copolymerization (FRP) of the monomer and crosslinker formed a crosslinked network (Figure 1 a). The colloidal crystal template was then removed by washing with acetone to generate the desired porous structure. The colloidal crystal templating method for the preparation of macroporous hydrogels is as simple as other techniques, including freeze‐drying, porogenation, microemulsion formation, etc. The size of the macropores can be easily controlled by tuning the diameter of colloidal crystal templates by varying the synthetic conditions used for the preparation of the latex colloidal spheres. 8 Figure 1 Preparation and characterizations of 3DOM hydrogels. Scale bar: 2 μm. The structure of the resulting hydrogels was tentatively characterized by scanning electron microscopy (SEM); however, no pores were observed in the SEM images (Figure 1 b). Nevertheless, the brilliant colors originating from Bragg diffraction of the hydrogels, when they were placed in solvents, indicated the presence of an ordered macroporous structure (Figure 1 d–f). Thus, it was conceivable that the pores in the hydrogels collapsed during drying, prior to the SEM measurements. In order to prevent collapse of macropores, more rigid water‐soluble monomers (i. e. , methacrylates containing quaternary ammonium groups, see Supporting Information, Figure S2) were used, and the 3DOM hydrogels were synthesized by controlled radical polymerization (i. e. , atom transfer radical polymerization (ATRP), see Supporting Information, Figure S3), 9 but still no macropores were observed in the SEM images of the resulting hydrogels. It should be noted that the mechanical properties of the hydrogels can be influenced by the properties of both crosslinker and monomer as well as their ratio 10 (Supporting Information, Figure S4). This allows the macropores to be persevered in hydrogels that are sufficiently strong, i. e. , with very high content of rigid monomer/crosslinker (Supporting Information, Figure S5). Since collapse of macropores was still a major problem for soft hydrogels, new approaches were required to characterize the porous structure of soft 3DOM hydrogels. A facile method was developed to resolve this problem and indirectly characterize the 3DOM structure by SEM. The porous hydrogels were soaked in an acetone solution of a crosslinker (i. e. , divinylbenzene (DVB)), which was then thermally polymerized to form arrays of crosslinked DVB spheres inside the macropores of the hydrogels (Figure 1 ). The resulting composites were mechanically stable under vacuum because of the rigid crosslinked DVB networks, which were inverse replicas of the initial 3DOM structures. The SEM images of the section surfaces of the composites displayed ordered sphere arrays (Figure 1 c), revealing the shape of the pores in the original 3DOM structures in the swollen hydrogels. This method was also successfully applied to characterize the 3DOM hydrogels with various crosslinking densities (10%, 30%, 50%, 70%, 90%, and 100%) (Supporting Information, Figure S6) and with different compositions (Supporting Information, Figure S7). The macroporous structures of the 3DOM hydrogels were retained and highly reversible through repeated collapse and reformation cycles. The original 3DOM structures stayed intact as the 3DOM hydrogels were repeatedly dried under vacuum and then swollen by solvents for ten cycles. This was confirmed by the ordered sphere arrays in the SEM images of their inverse replicas (Supporting Information, Figure S6). This demonstrated the intrinsic reversible shape memory nature of the 3DOM structures in the hydrogels. In addition, a direct approach was employed to resolve the 3DOM structure in the presence of water by nano‐XRM in Zernike phase contrast mode. 11 The 3D nano‐XRM images ( Figure 2 and Supporting Information, Figures S8–S10) and movies (Supporting Information, Movies S1–S3) displayed orthogonal virtual slices through the reconstructed volume, the surface view of a segmented cropped volume, and a volume rendering of the pore phase, which had a maximum spatial resolution of 50 nm (16 nm voxels). 12 It revealed the 3D distributed porous structures with interconnected windows in the swollen samples. This is a nondestructive way to visualize the porous structures of swollen hydrogels in ambient conditions without requiring vacuum or pretreatment of samples. Figure 2 3D nano‐XRM images of trypsin immobilized 3DOM hydrogels soaked in water: A) orthogonal raw tomography slices through the reconstructed volume along with the surface view of a segmented cropped volume (the dark phase represents the pores), and B) volume rendering of the pore phase of the cropped volume. These well‐defined 3DOM hydrogels can serve as a versatile platform for the preparation of a variety of functional materials. The presence of surface functional groups on the pores of the 3DOM hydrogels, originating from the functional monomers, permitted to design materials with new properties by further chemical modifications of the accessible functionalities via grafting with functional organic compounds and polymers. The functional hydrogels can be considered as stem (structurally tailored and engineered macromolecular) gels (by analogy with stem cells) that can evolve into materials with final properties and function that can be programmed, depending on the chemistry applied for modification. By the esterification of the hydroxyl groups of the 3DOM hydrogels with dodecanoyl chloride, the surfaces of hydrogels modified with the long alkyl chains became hydrophobic. The alternation of the hydrophilicity/hydrophobicity was verified by water contact angle measurements. Water droplet was absorbed by the pristine hydrogels, while it stayed on the surface of the modified hydrogels ( Figure 3 A). A condensation reaction between the hydroxyl groups on the 3DOM hydrogels with rhodamine B afforded fluorescent hydrogels which could become fluorescent under UV light (Figure 3 B). The hydroxyl groups of the 3DOM hydrogels were also converted to carboxylic acid groups by reacting with succinic anhydride and then polyaniline was generated through in situ oxidative polymerization of infused aniline using ammonium persulfate (Supporting Information, Figures S11–S13). The resulting 3DOM hydrogel/polyaniline composites were electronically conductive, enabling them to connect a circuit and light a light‐emitting diode (LED) lamp (Figure 3 C). Figure 3 A) 3DOM hydrogel modified with long alkyl chains: the change of hydrophilicity/hydrophobicity was visualized by placing a drop of water on a) 3DOM hydrogel containing hydroxyl groups (immediate penetration of the water into the sample occurred and no drop is visible) and b) long alkyl chains modified 3DOM hydrogel (the water droplet stayed on top of the sample). B) 3DOM hydrogels modified with RhB: a comparison of 3DOM hydrogels containing hydroxyl groups before (left column) and after (right column) conjugation with RhB when exposed to white light (top row) or a UV source (bottom row). C) 3DOM hydrogels modified with polyaniline: photos of a LED circuit a) without and b) with connection of polyaniline modified 3DOM hydrogels. D) 3DOM hydrogels grafted with PNIPAM: change of hydrophilicity/hydrophobicity visualized by placing a drop of water on PNIPAM grafted 3DOM hydrogels at 25 and 45 °C (immediate penetration of the water into the sample occurred at 25 °C and no drop is visible; while the water droplet stayed on top of the sample at 45 °C). In addition, ATRP initiators were anchored onto the surface of the pores by reaction between hydroxyl groups of the 3DOM hydrogels and α‐bromoisobutyryl bromide. Poly( N ‐isopropylacrylamide) (PNIPAM) was then grafted from the pore walls by ATRP (Supporting Information, Figure S14) and the resulting copolymer showed temperature‐responsive surface properties due to the lower critical solution temperature (LCST) of PNIPAM (32 °C). As shown in Figure 3 D, the PNIPAM grafted 3DOM hydrogels absorbed the water droplet quickly at 25 °C, whereas they became hydrophobic and repelled the water droplet at 45 °C. The 3DOM hydrogels modified with bioactive species were useful for digestion or separation of bio(macro)molecules. For instance, trypsin was grafted onto the pores of the 3DOM hydrogels by the reaction of trypsin amino group with carboxylic acid groups from the hydrogels. The resulting 3DOM hydrogels with pore immobilized trypsin displayed bioactivity for tryptic digestion of bovine serum albumin (BSA) ( Figure 4 b) and N‐ α‐benzoyl‐ l ‐arginine p ‐nitroanilide (BAPNA) (Supporting Information, Figures S16 and S17). While the use of enzyme immobilized inverse opals as heterogeneous biocatalysis has been demonstrated in previous reports, 13 the present work allows fabrication of a novel column with trypsin‐immobilized packing (Figure 4 a), which displays high performance in online hydrolysis using N ‐α‐benzoyl‐ l ‐arginine ethyl ester hydrochloride (BAEE) as the substrate (Figure 4 c and Supporting Information, Figures S18–S26). The advantages are that the 3DOM structures can provide large surface area and reduced mass transfer resistance, making the enzyme‐immobilized 3DOM hydrogels efficient biocatalyst for column filtration/separation. 14 Figure 4 a) Schematic illustration of trypsin immobilized 3DOM hydrogels column. b) The digestion of BSA (0. 5 mg mL −1, 1 mL) by 3DOM hydrogel‐trypsin (2. 5 mg): UV–vis spectra of 25 μL of the BSA solution in 200 μL of bicinchoninic acid (BCA) protein assay reagent (bicinchoninic acid) from 0 to 4 h. The decrease of the absorbance at 562 nm originating from BCA/copper complex indicated the decrease of protein concentrations. c) Online monitoring of the absorption at 223 nm versus retention time in RP18 column for (1) 5 × 10 −3 m BAEE solution and (2) collected solutions from step 1 with 5 × 10 −3 m BAEE solution. The disappearance of the peak at ≈5. 5 min indicated the complete digestion of BAEE by trypsin immobilized 3DOM hydrogels column. The 3DOM hydrogels were also loaded with inorganic particles to form novel organic–inorganic composites. This is illustrated by reduction of AuCl 3 by NaBH 4 on the surface of 3DOM hydrogels modified with the carboxylic acid groups to form Au nanoparticles (NPs) with an average diameter of 7. 1 nm that were uniformly distributed on the pores. The distribution of the formed AuNPs can be observed in the transmission electron microscopy (TEM) images of a thin‐section of the sample ( Figure 5 b, c, and Supporting Information, Figure S27). The 3DOM hydrogel‐Au NPs composite displayed high catalyst activity as a heterogeneous catalyst for the reduction of 4‐nitrophenol (Figure 5 d and Supporting Information, Figure S28). Figure 5 Upper row: 3DOM hydrogels loaded with Au NPs: a) a cartoon representation of the overall structure, b, c) TEM images of a ≈100 nm thin‐section sample, and d) successive UV–vis spectra for the catalytic reduction of 4‐nitrophenol into 4‐aminophenol. Lower row: 3DOM hydrogels loaded with Fe 3 O 4 NPs: e) a cartoon representation of the overall structure, f, g) TEM images a ≈100 nm thin‐section sample, and h) photos of the sample in water in the absence (left) and presence (right) of a magnetic field. Another example of the organic–inorganic composite is a magnetic 3DOM hydrogel nanocomposite. By electrostatic self‐assembling, negatively charged magnetite (Fe 3 O 4 ) NPs with an average diameter of 3. 1 nm were decorated onto the positively charged pore surfaces of 3DOM hydrogels containing quaternary ammonium groups, affording magnetic 3DOM hydrogels (Figure 5 e–h and Supporting Information, Figures S29 and S30). In summary, 3DOM hydrogels were prepared by either conventional FRP or ATRP of hydrophilic functional monomers and crosslinkers in the presence of latex colloidal crystals as the template. This was confirmed by simple and effective visualization of the 3DOM structures through both indirect electron microscopy and direct nano‐XRM characterizations. The highly reversible 3DOM structure of the hydrogels was demonstrated by multiple drying/swelling cycles that indicated the shape‐memory nature of 3DOM hydrogels. Further modifications of functional 3DOM hydrogels with organic moieties, polymers, bio(macro)molecules, and inorganic particles generates materials with new properties (e. g. , hydrophobicity, fluorescence, conductivity, and stimuli‐responsivity) into the structured 3DOM hydrogels and result in novel composites for (bio)catalysis and separation applications. The unique 3DOM structures with well‐controlled pore size and excellent pore interconnectivity can also serve as models for systematic studies of the influence of pore size and pore interconnectivity on the properties of the resulting materials (such as fluid mechanics of the flow through pores, catalytic efficiency, etc. ). This work suggests that functional 3DOM hydrogels represent a versatile platform for a wide range of applications and there are even more multifunctional materials based on 3DOM gels to be explored. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500082
| 2,015
|
Advanced Science
|
Multistimuli Sensitive Behavior of Novel Bodipy‐Involved Pillararene‐Based Fluorescent Rotaxane and Its Supramolecular Gel
|
Fluorescent rotaxane BC12P5 is successfully constructed with 1, 4‐diethoxypillararene as wheel over a long alkyl axle with Bodipy chromophore as one stopper for the first time. NMR spectra clearly reveal its molecular shuttle nature triggered by multiple external stimuli including solvent polarity and temperature. In particular, the fluorescence nature introduced into rotaxane BC12P5 renders it a good sensor for the external stimuli. Nevertheless, the supramolecular gel successfully fabricated from this novel rotaxane system via self‐assembly in dimethyl sulfoxide (DMSO) also shows reversible gel–sol phase transition upon multiple external stimuli such as heating/cooling, shaking/resting, or the addition of different anions. Interestingly, exposure of the supramolecular gel film to HCl or ammonia vapor induces the change in the film fluorescence intensity, endowing this system with a potential application in gas detecting.
|
1 Introduction This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Inspired by naturally occurring biological motors such as the ATPase rotary motor and the kinesin or myosin linear motor systems, 1 chemists have tried to construct a variety of artificial molecular machines 2 including the unidirectional rotors, shuttles, scissors, and molecular muscles that can perform diverse molecular motions. 3 Rotaxanes, with a typical mechanically interlocked structure, have been widely employed as crucial precursor and building blocks for the fabrication of advanced supramolecular architectures 4 like molecular shuttles and switches with molecular motions upon certain external stimuli. 5 Thus far a series of macrocyclic compounds including crown ethers, cyclodextrins, and cucurbit[ n ]urils have been applied to construct the rotaxane‐based molecular motors. 6 Pillar[ n ]arenas, as the new member of functional macrocyclic family, have also been employed as good building block to create rotaxanes shortly after their first synthesis in 2008. 7 In 2011, Stoddart and co‐worker reported the first pillararene‐based rotaxane formed between a pillararene and N, N ′‐bis(3, 5‐di‐ tert ‐butylbenzyl)octane‐1, 8‐diamine, 8 which was followed by the construction of different species of rotaxanes 9 and novel rotaxanes. 10 However, pillararene‐based rotaxanes with multiple external stimuli responsiveness, especially those incorporating fluorogenic functionalities remain rarely explored. On the other hand, boron dipyrromethene (Bodipy) dyes have constituted one of the most important families of simple organic luminophores due to their special absorption and emission properties. 11 Their strong absorption and emission in the visible and near‐infrared range render them great application potential in chemosensors and probes, biological labels, laser dyes, photodynamic therapy agents, and a plethora of photonic devices. 12 Bodipy‐involved rotaxanes have, however, been rarely reported thus far, limited to several crown ether/cucurbit[ n ]uril‐based rotaxanes, 13 to the best of our knowledge. In the present paper, we describe the preparation and characterization of a new type of fluorescent rotaxane BC12P5 constructed on the basis of pillararene wheel and a dumbbell‐shaped axle with Bodipy chromophore as one stopper, Scheme 1, which appears to represent the first Bodipy‐involved pillararene‐based rotaxane. The pillararene wheel of this novel rotaxane BC12P5 was revealed to move over its dumbbell‐shaped alkyl axle under multiple external stimuli including the solvent polarity, temperature, and pH value on the basis of the NMR and fluorescent spectroscopic investigations. Nevertheless, rotaxane BC12P5 is able to self‐assemble into supramolecular gel in dimethyl sulfoxide (DMSO), which also shows reversible gel–sol phase transition upon multiple external stimuli like heating/cooling, shaking/resting, or the addition of different anions. Scheme 1 Schematic molecular structures of rotaxane BC12P5 under external stimuli. At the end of this section, it is noteworthy that in the past decade supramolecular gels formed by self‐assembly of organic molecules into entangled structures to immobilize the solvents have attracted extensive research interests due to their potential applications in chemosensors, optoelectronic devices, drug delivery, tissue engineering, biomaterials, and surface science. 14 The present Bodipy‐containing pillararene‐based fluorescent gel is therefore expected to find applications in related fields. 2 Results and Discussion 2. 1 Molecular Design and Synthesis Generally, rotaxane is constructed by wheel and axle components. In the present case, 1, 4‐diethoxypillararene (EtP5) is chosen as the wheel of the target rotaxane BC12P5 with 12‐(1H‐imidazol‐1‐yl)dodecanol ( 1 ) stopped by a carbamic unit and a meso‐chloro‐benzyl‐Bodipy unit ( 2 ) at both ends as an axle, Scheme 1. Interestingly, the Bodipy unit introduced as one stopper of the axle also provides the rotaxane BC12P5 with an effective fluorescence chromophore, enabling the detection of the responsiveness to external stimuli by fluorescence method. Both EtP5 and the semiblocked rod‐like Bodipy derivative Bodipy‐C12OH ( 3 ) were prepared according to the published procedures. [[qv: 7d]], [[qv: 12b]] Reaction of EtP5 with the semiblocked rod‐like component Bodipy derivative 3 in CHCl 3 led to the formation of pseudorotaxane structure, which then reacted with 1‐isocyanato‐3, 5‐dimethylbenzene in CHCl 3 afforded the target rotaxane BC12P5. Satisfactory elemental analysis result was obtained for the newly prepared rotaxane BC12P5 after repeated column chromatography followed by recrystallization. The matrix‐assisted laser desorption/ionization time of flight (MALDI‐TOF) mass spectrum displayed intense signal at m / z = 1627. 95, corresponding to the molecular ion [M–Cl] +. Nevertheless, the key intermediate and the target rotaxane BC12P5 were also characterized by 1 H and 13 C NMR spectroscopies, Figures S1–S6 (Supporting Information). 2. 2 NMR Characterization Figure 1 and Figure S7 (Supporting Information) show the 1 H NMR and 2D nuclear Overhauser enhancement spectroscopy (NOESY) of rotaxane BC12P5 in CDCl 3. For comparative study, the NMR spectra for the two components, namely the wheel and the axle Bodipy‐C12OH‐isocyanato ( 4 ), in CDCl 3 were also recorded and shown in Figure 1. Comparison in the NMR spectrum of these three species reveals that the 1 H NMR spectrum of rotaxane BC12P5 is not a simple superimposition of the spectra of pure compound 4 and EtP5 in the same deuterated solvent, indicating the effective interaction between the host wheel and the guest axle in the supramolecular rotaxane system. As can be seen, Figure 1 and Table S1 (Supporting Information), after being fabricated into rotaxane BC12P5, the signals of the methylene protons H 9, H 10, H 11, and H 12 on the axle (which are adjacent to the imidazolium unit in the axle) take obvious upfield shift from 1. 31, 1. 31, 1. 96, 4. 29 to 0. 39, −046, −1. 19, and 3. 98 ppm, respectively. This is also true for the imidazolium protons H 13 and H 15 with substantial upfield shift from 11. 47 and 7. 09 to 8. 39 and 6. 30 ppm, respectively. These results suggest the encapsulation of the imidazolium moiety and its adjacent methylene groups in the axle by the pillararene wheel in rotaxane BC12P5 in CDCl 3. Additional support for this point comes from the cross‐peaks between the aromatic protons H b of pillararene ring and the methylene protons H 9 and H 10 observed in the NOESY spectrum of rotaxane BC12P5, Figure S7 (Supporting Information). Figure 1 1 H NMR spectra of A) compound 4, B) rotaxane BC12P5, and C) EtP5 recorded in CDCl 3 at 25 °C. To reveal the solvent polarity effect on the rotaxane BC12P5 conformation, a more polar solvent, DMSO‐ d 6, was utilized for comparative NMR investigations. As shown in Figure S8 and Table S2 (Supporting Information), the signals of the methylene protons H 1, H 2, H 3, H 4, H 5, and H 6 that are adjacent to the carbamic stopper in the axle exhibit substantial upfield shift (Δ δ = −0. 38, −0. 84, −1. 67, −2. 08, −1. 91, and −1. 47 ppm, respectively) in the rotaxane system in comparison with those for pure compound 4, revealing the shielding effect of the host EtP5 cavity on these protons and in turn suggesting the methylene groups adjacent to the carbamic stopper in the axle threaded into the cavity of the pillararene ring in rotaxane BC12P5 in DMSO‐ d 6. This is further confirmed by the cross‐peaks between the signals of methylene protons H 3, H 4, H 5, and H 6 of the alkyl chain and the phenyl proton H b of the pillararene moiety observed in the 2D NOESY spectrum of rotaxane BC12P5 in DMSO‐ d 6, Figure S7 (Supporting Information). Obviously, different conformation was employed by rotaxane BC12P5 in DMSO‐ d 6 from that in CDCl 3 due to the difference in the solvent and rotaxane BC12P5 intermolecular interactions, suggesting the possible solvent polarity‐driven molecular shuttle nature of this system. 2. 3 Solvent Polarity‐Driven Molecular Shuttle As described above, 1 H NMR measurements indicate that the solvent polarity change might be able to induce the pillararene cavity to move on the alkyl chain in rotaxane BC12P5. As a result, systematic studies over the 1 H NMR spectra of rotaxane BC12P5 in a series of mixed solvents with different ratio of CDCl 3 /DMSO‐ d 6 were carried out. As shown in Figure 2 and Table S3 (Supporting Information), along with the decrease in the solvent polarity due to the stepwise addition of CDCl 3 into DMSO‐ d 6, the signals of the methylene protons H 9, H 10, H 11, and H 12 and the imidazolium protons H 13 in the axle of rotaxane BC12P5 experience substantial upfield shift. However, the signals of the methylene protons (that are adjacent to the carbamic stopper) such as H 2, H 3, H 4, H 5, and H 6 take obvious downfield shift, demonstrating the gradual movement of the pillararene moiety on the axle from the methylene groups adjacent to the carbamic stopper to those adjacent to the imidazolium unit along with the decrease in the solvent polarity, revealing the solvent polarity‐driven molecular shuttle nature of rotaxane BC12P5. Figure 2 Systematic change in the 1 H NMR spectrum of rotaxane BC12P5 along with the change in the ratio of CDCl 3 /DMSO‐ d 6 (v/v): A) CDCl 3, B) 1:1, C) 1:1. 5, D) 1:2, E) 1:2. 5, F) 1:3, G) 1:3. 5, H) 1:4, I) 1:5, J) 1:6, K) 1:7, L) 1:8, M) 1:9, N) 1:10, O) 1:20, and P) DMSO‐ d 6 recorded at 25 °C. 2. 4 Thermodriven Molecular Shuttle In order to try to investigate the molecular motion of rotaxane BC12P5 under temperature stimuli, the temperature‐dependent 1 H NMR spectra for the rotaxane system were recorded in DMSO‐ d 6. As exhibited in Figure S9 and Table S4 (Supporting Information), along with increasing the temperature, the signals of the methylene protons H 9, H 10, H 11, and H 12 (which are close to the Bodipy stopper) and the imidazolium protons H 13, H 14, and H 15 gradually move to upfield direction. For instance, the signal of proton H 12 shifts from 3. 46 to 2. 19 ppm along with the temperature change from 25 to 115 °C. In contrast, the signals of methylene protons H 2, H 3, H 4, H 5, and H 6 (which are adjacent to the carbamic stopper) gradually move to the downfield direction as the temperature increases as exemplified by the shift of proton H 2 signal from 0. 75 ppm at 25 °C to 1. 10 ppm at 115 °C, Figure S10 and Table S4 (Supporting Information). These results clearly reveal the temperature‐driven molecular shuttle nature of rotaxane BC12P5. Nevertheless, on the basis of the just above section and according to these NMR spectroscopic results, the pillararene ringrotaxane BC12P5 should locate on the methylene groups of the axle that are adjacent to the carbamic stopper at low temperature in DMSO‐ d 6. Along with increasing the temperature, the pillararene ring gradually slides to the imidazolium unit. 2. 5 Fluorescence Properties of rotaxane BC12P5 in Solution 2. 5. 1 Effects of Solvent Polarity on the Fluorescence Properties Due to the incorporation of the Bodipy fluorescent chromophore in the present rotaxane system, the fluorescence properties of this system were therefore studied following the solvent polarity change. As displayed in Figure S11 (Supporting Information), with the system concentration being fixed at 1 × 10 −5 m, the fluorescence intensity of rotaxane BC12P5 gradually gets decreased along with the addition of CHCl 3 into the solution of DMSO. In pure CHCl 3, a total decrease by the most of 23% in the fluorescence intensity was achieved in comparison with that in pure DMSO. In good contrast, the fluorescence property of compound 4 was also studied. As shown in Figure S11 (Supporting Information), the fluorescence intensity of 4 gradually gets decreased along with the addition of CHCl 3 into the solution of DMSO. In pure CHCl 3, the fluorescence intensity of compound 4 decreases for about 37% in comparison with that in pure DMSO. This is in line with that observed for rotaxane BC12P5. However, the decrease in the fluorescence intensity for compound 4 is larger than that of rotaxane BC12P5. This seems to indicate the relatively less effect of the solvent polarity‐driven molecular shuttle motion on the fluorescence intensity, suggesting the more effect of the solvent polarity on the fluorescence intensity. In line with previous investigation, 15 in the present case higher fluorescence intensity for rotaxane BC12P5 in polar solvent is achieved due mainly to the decrease in the nonradiative rate constant (which minimizes the nonradiative energy loss) with the help of the solvent polarity‐driven molecular shuttle motion. 2. 5. 2 Effect of Temperature on the Fluorescence Properties As can be easily expected, the fluorescence intensity of rotaxane BC12P5 at a fixed concentration of 1 × 10 −5 m in DMSO also takes systematic change along with the change in temperature, gradually decreased along with the temperature increase, by the most of 52% at 25 °C in comparison with that at 115 °C, Figure 3. This is also true for the reference compound 4. As shown in Figure S12 (Supporting Information), along with increasing the temperature, the fluorescence intensity of 4 gets gradually decreased in a similar manner to that of rotaxane BC12P5, indicating the weaker influence of the thermodriven molecular shuttle movement of rotaxane BC12P5 on the axle to the fluorescence intensity than that due to the consumption of more nonradiative energy at high temperature. 16 Figure 3 Systematic change in the fluorescence spectrum of rotaxane BC12P5 in DMSO (1 × 10 −5 mol L −1 ) along with the temperature change from 25 to 115 °C. 2. 5. 3 Effect of Acid/Base Change on the Fluorescence Properties Acid/base titration experiments were also carried out to reveal the sensor property of rotaxane BC12P5. As can be seen in Figure 4, along with the gradual addition of triethylamine (TEA) into the CHCl 3 solution of rotaxane BC12P5 from 0 to 30 μL, the fluorescence intensity at 516 nm undergoes a consecutive decrease of 38%. However, the fluorescence spectrum could completely recover upon addition of trifluoroacetic acid (TFA) into the above‐described CHCl 3 solution from 0 to 20 μL, indicating the influence of the pH on the fluorescence intensity. On the basis of previous research result, 17 the fluorescence‐quenching photoinduced electron transfer (PET) between the amine and Bodipy core at the extreme of high pH is able to occur because electron transfer may occur through space, resulting in the recovery of the fluorescence intensity of rotaxane BC12P5 due to the neutralization upon addition of TFA. As can be expected, the fluorescence intensity for 4 upon addition of TEA into the solution also gets decreased in quite a similar manner to that of rotaxane BC12P5, Figure S13 (Supporting Information). Nevertheless, the fluorescence intensity for this system also gets recovered after adding TFA. Figure 4 The fluorescence emission spectra of rotaxane BC12P5 (1 × 10 −5 mol L −1 ) in CHCl 3 upon addition of increasing amount (0, 5, 10, 15, 20, and 30 μL) of TEA (A) and then of increasing amount (0, 5, 10, 15, and 20 μL) of TFA at 25 °C (B). 2. 6 Preparation of rotaxane BC12P5 Supramolecular Gel Supramolecular gels constructed from low‐molecular‐weight molecules (LMWMs) simultaneously possessing both toughness and flexibility as gelators depending on reversible noncovalent interactions play important role in the development of soft material science. [[qv: 14d]] Due to the relatively rigid structure of the host component and the soft structure of the guest component usually employed by the rotaxanes and pseudorotaxanes, either rotaxanes or pseudorotaxanes with a mechanically interlocked structure have been used as suitable gelators to construct supramolecular gels. 18 In the present case, the π–π interactions between the phenylene moieties of neighboring pillararene rings of rotaxane BC12P5, as well as between the Bodipy–Bodipy stacking, Bodipy, and imidazolium moieties, with the help of the van der Waals forces between long alkyl chains in the neighboring rotaxane BC12P5 systems result in the formation of 1D supramolecular arrays, which subsequently self‐assemble into the cross‐linked network depending on the similar intermolecular interactions as mentioned above between neighboring rotaxane BC12P5 systems in different 1D arrays, Scheme 2. This 3D network then entraps the DMSO molecules with its porous structure, leading to the formation of a novel supramolecular gel containing Bodipy fluorescence chromophore depending mainly on the hydrogen bonding interaction between the solvent DMSO molecules and rotaxane BC12P5 systems. 19 The critical gelation concentration for rotaxane BC12P5/DMSO was about 11. 2 wt%. Figure 5 shows the scanning electron microscope (SEM) image of the gel formed from rotaxane BC12P5. The 3D network constructed from nanofibers with an interconnected porous structure observed clearly reveals the gel nature of this system. Scheme 2 Schematic representation of self‐assembling rotaxane BC12P5 into the supramolecular gel via 1D aggregate in DMSO. Figure 5 SEM image of the supramolecular gel of rotaxane BC12P5 formed in DMSO by drop casting on the copper grid. 2. 7 Multiple Stimuli‐Responsive Reversible Gel–Sol Transitions of the Supramolecular Gel Due to the sensitivity of the noncovalent interactions to external stimuli including solvent, temperature, pH, and mechanical stress, supramolecular gels usually exhibit stimuli responsiveness to the environment. As a consequence, multiple stimuli‐responsive behaviors of the present supramolecular gels were also investigated. Similar to other gels, a reversible gel–sol transition could be easily achieved by shaking (here through ultrasonic waves) or resting of this gel system. In addition, after adding CF 3 COOAg into the supramolecular gel system, the gel gradually collapses and finally becomes a solution after removing the AgCl precipitate. Upon addition of a little excess amount of tetrabutylammonium chloride (TBACl) into the solution, supramolecular gel is reformed. [[qv: 5j]], 20 Nevertheless, most probably associated with the temperature‐dependent nature of its building block, the supramolecular gel fabricated from rotaxane BC12P5 is also sensitive to temperature. As displayed in Figure 6, along with increasing the temperature, the supramolecular gel gradually becomes a solution with the gel‐to‐solution phase transition temperature ( T gel ) of about 338 K. Reversibly, along with decreasing the temperature, the solution formed over 338 K return to the gel phase. Figure 6 Photographs of reversible sol–gel transition upon cooling/heating under A) ambient light and B) illuminated at 365 nm. This gel also shows acid/base stimuli‐responsiveness. In comparison with rotaxane BC12P5 in DMSO solution, the supramolecular gel thin film prepared by coating onto quartz slide shows a redshifted broadened fluorescence at 534 nm with obviously weakened intensity due to the enhanced intermolecular interactions of rotaxane BC12P5 in the gel state, Figure S14 (Supporting Information). Interestingly, upon being exposed to HCl gas, the thin film gel fluorescence intensity gets significantly decreased, by the most of 50%, accompanied also by a visually clear fluorescence color change from yellow to purplish red under UV light, Figure S15 (Supporting Information). In contrast, when being exposed to the NH 3 gas, the fluorescence intensity of the supramolecular gel thin film gets increased by the most of 100% but without showing the visually color change behavior. These results seem to suggest the potential of this supramolecular gel as the acidic/basic gas sensor. 3 Conclusion In summary, the first Bodipy‐involved fluorescent rotaxane BC12P5 was designed and prepared. This novel rotaxane BC12P5 system exhibits molecular shuttle nature under multiple external stimuli including solvent polarity and temperature according to NMR spectra. In particular, the fluorescent nature introduced into rotaxane BC12P5 renders it a good sensor for these external stimuli. Nevertheless, the self‐assembled supramolecular gel formed from this rotaxane system with the help of DMSO also shows multiple external stimuli‐induced reversible gel–sol phase transition upon shaking/resting, heating/cooling, or the addition of different anions. In particular, exposure of this supramolecular gel film to the HCl gas leads to obvious decrease in the fluorescence intensity accompanied by a visually clear fluorescence color change under UV light, endowing the system with a application potential in acidic gas detecting. 4 Experimental Section General Remarks : All reagents were obtained from commercial sources without further purification. The compounds of 1, 2, 3, 4, and EtP5 were prepared according to the literature procedure. [[qv: 7d]], [[qv: 12b]] Measurements : NMR spectra were recorded on a Bruker DPX 400 spectrometer in CDCl 3 and DMSO‐ d 6. Electronic absorption spectra were recorded on a Hitachi U‐4100 spectrophotometer. Steady‐state fluorescence spectra were performed on an F4500 (Hitachi). MALDI‐TOF mass spectra was taken on a Bruker BIFLEX III ultra‐high resolution Fourier transform ion cyclotron resonance mass spectrometer with α‐cyano‐4‐hydroxycinnamic acid as matrix. Elemental analysis was performed on an Elementar Vavio El III. Preparation of 12‐(1H‐Imidazol‐1‐yl)dodecanol ( 1 ) : 1H‐imidazole (6. 80 g, 0. 10 mol), NaOH (4. 00 g, 0. 10 mol), and 12‐bromododecan‐1‐ol (2. 65 g, 0. 10 mol) in DMSO (20 mL) were stirred at 70 °C for 24 h. The solvent was poured into water. After filtration, the residue was dried by air to give 1 as a white solid with 80% yield. 1 H NMR (400 MHz, CDCl 3, 25 °C, δ): 7. 48 (s, 1H), 7. 07 (s, 1H), 6. 92 (s, 1H), 3. 94 (t, J = 12 Hz, 2H), 3. 66 (s, J = 12 Hz, 2H), 1. 77 (m, 4H), 1. 28 (m, 16H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 137. 21, 129. 52, 118. 90, 63. 15, 47. 18, 36. 06, 32. 96, 31. 19, 29. 93, 29. 65, 29. 56, 29. 54, 29. 51, 29. 49, 29. 13, 27. 36, 26. 64, 25. 87, 25. 69. MS m / z : [M+K] + calcd for C 15 H 28 N 2 O, 252. 39; found, 291. 75. Anal. calcd for C 15 H 28 N 2 O: C 71. 38, H 11. 18, N 11. 10; found: C 71. 29, H 11. 23, N 11. 21. Preparation of Meso‐chloro‐benzyl‐Bodipy ( 2 ) : 4‐(Chloromethyl)benzoyl chloride (3. 84 g, 20. 3 mmol) was added dropwise to a stirred solution of 2, 4‐dimethyl‐1H‐pyrrole (3. 86 g, 40. 6 mmol) in dichloromethane (200 mL) at room temperature under nitrogen, and the mixture was heated at 50 °C with stirring for 2 h. After vacuum evaporation of the solvent, toluene (150 mL), dichloromethane (15 mL), and triethylamine (13 mL) were added to the residual solid. The mixture was stirred at room temperature for 30 min under nitrogen and boron trifluoride diethyl etherate (18 mL) was then added. After heating at 50 °C for 1. 5 h, the solvent was removed under vacuum. The crude product was purified by silica gel column chromatography with hexane/ethyl acetate (5/1, v/v) eluent to give 2 with 45% yield. 1 H NMR (400 MHz, CDCl 3, 25 °C, δ): 7. 53 (d, J = 8 Hz, 2H), 7. 29 (d, J = 8 Hz, 2H), 5. 98 (s, 2H), 7. 03 (s, 4H), 4. 66 (s, 2H), 2. 55 (s, 6H), 1. 38 (s, 6H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 155. 84, 143. 17, 141. 09, 138. 77, 135. 28, 131. 49, 129. 38, 128. 61, 121. 49, 45. 73, 14. 73. MS m / z : [M + ] calcd for C 20 H 20 BClF 2 N 2, 372. 65; found, 372. 15. Anal. calcd for C 20 H 20 BClF 2 N 2 : C 64. 46, H 5. 41, N 7. 52; found: C 64. 38, H 5. 46, N 7. 64. Preparation of Bodipy‐C12OH ( 3 ) : 1 (0. 25 g, 1. 0 mmol) and 2 (0. 37 g, 1. 0 mmol) were refluxed in CH 3 CN (100 mL) for 7 d. After filtration and solvent evaporation, the crude product was precipitated by diethyl ether to yield compound 3 as a red solid with 67% yield. 1 H NMR (400 MHz, CDCl 3, 25 °C, δ): 11. 41 (s, 1H), 7. 64 (d, J = 8 Hz, 2H), 7. 37 (d, J = 8 Hz, 2H), 7. 18 (s, 1H), 7. 10 (s, 1H), 5. 98 (s, 2H), 5. 81 (s, 2H), 4. 32 (t, J = 16 Hz, 2H), 3. 65 (t, J = 16 Hz, 2H), 2. 55 (s, 6H), 1. 95 (m, 2H), 1. 57 (m, 2H), 1. 32 (m, 22H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 156. 14, 142. 81, 140. 25, 139. 05, 136. 64, 134. 42, 131. 31, 129. 94, 129. 50, 121. 76, 121. 65, 121. 12, 63. 03, 53. 17, 50. 59, 32. 90, 30. 29, 29. 49, 29. 38, 29. 34, 29. 30, 28. 93, 26. 34, 25. 81, 14. 74, 14. 60. MS m / z : [M–Cl] + calcd for C 35 H 48 BF 2 N 4 OCl, 589. 59; found, 589. 37. Anal. calcd for C 35 H 48 BF 2 N 4 OCl: C 67. 26, H 7. 74, N 8. 96; found: C 67. 33, H 7. 66, N 8. 89. Preparation of Bodipy‐C12OH‐isocyanato ( 4 ) : A mixture of 3 (0. 31 g, 0. 5 mmol), dibutyltindilaurate (one drop), and 1‐isocyanato‐3, 5‐dimethylbenzene (0. 22 g, 1. 5 mmol) in CHCl 3 (0. 3 mL) was stirred at −6 °C for 24 h. After filtration and solvent evaporation, the crude product was purified by flash column chromatography with (CH 2 Cl 2 /MeOH, 15:1, v/v) as eluent to yield compound 4 as a red solid with 76% yield. 1 H NMR (400 MHz, CDCl 3, 25 °C, δ): 11. 47 (s, 1H), 7. 64 (d, J = 8 Hz, 2H), 7. 37 (s, J = 8 Hz, 2H), 7. 17 (s, 1H), 7. 09 (s, 1H), 7. 01 (s, 2H), 6. 70 (s, 1H), 6. 55 (s, 1H), 5. 98 (s, 2H), 5. 81 (s, 2H), 4. 32 (t, J = 16 Hz, 2H), 4. 15 (t, J = 16 Hz, 2H), 2. 55 (s, 6H), 2. 28 (s, 6H), 1. 94 (m, 2H), 1. 62 (m, 2H), 1. 34 (m, 22H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 156. 16, 142. 80, 140. 20, 139. 31, 138. 89, 137. 97, 136. 69, 134. 32, 131. 30, 129. 91, 129. 52, 125. 24, 121. 65, 120. 92, 116. 59, 65. 41, 53. 18, 50. 59, 30. 30, 29. 55 29. 52, 29. 48, 29. 40, 29. 32, 29. 07, 29. 02, 26. 39, 25. 95, 21. 52, 14. 75, 14. 61. MS m / z : [M–Cl] + calcd for C 44 H 57 BF 2 N 5 O 2 Cl, 736. 76; found, 736. 57. Anal. calcd for C 44 H 57 BF 2 N 5 O 2 Cl: C 68. 44, H 7. 44, N 9. 07; found: C 68. 35, H 7. 53, N 9. 11. Preparation of EtP5 : To a solution of 1‐ethoxy‐4‐methoxybenzene (3. 35 g, 20. 0 mmol) and paraformaldehyde (0. 75 g, 25. 0 mmol) in 1, 2‐dichloroethane (300 mL), boron trifluoride diethyl etherate [BF 3 ·O(C 2 H 5 ) 2, 2. 52 mL, 20. 0 mmol] was added under nitrogen atmosphere at 25 °C. Then, the mixture was stirred for 4 h. The solution was washed by saturated sodium chloride solution and dried by anhydrous sodium sulfate. The solvent was removed and the residue was purified by flash column chromatography on silica gel with CH 2 Cl 2 as eluent, affording EtP5 as a white solid with 46% yield. 1 H NMR (400 MHz, CDCl 3, 25 °C, δ): 6. 81 (s, 10H), 3. 89 (m, 20H), 3. 77 (s, 10H), 1. 36 (m, 30H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 149. 83, 128. 54, 114. 70, 63. 67, 40. 24, 31. 71, 29. 71, 22. 77, 15. 35, 14. 24. MS m / z : [M + ] calcd for C 55 H 70 O 10, 891. 14; found, 890. 76. Anal. calcd for C 55 H 70 O 10 : C 74. 13, H 7. 92; found: C 74. 23, H 7. 86. Preparation of Rotaxane BC12P5 ( 5 ) : A mixture of 3 (62. 49 mg, 0. 10 mmol) and EtP5 (0. 37 g, 0. 40 mmol) was stirred in CHCl 3 (0. 40 mL) at −6 °C for 2 h. Then dibutyltindilaurate (one drop) and 1‐isocyanato‐3, 5‐dimethylbenzene (0. 20 g, 1. 3 mmol) were added. The mixture was further stirred for 3 h. The solvent was removed and the residue was purified by flash column chromatography on silica gel with (CH 2 Cl 2 /MeOH = 25/1, v/v) eluent to afford 5 as a red solid with 76% yield. 1 H NMR (400 MHz, DMSO‐ d 6, 25 °C, δ): 9. 12 (s, 1H), 7. 93 (s, 1H), 7. 60 (d, J = 8 Hz, 2H), 7. 54 (d, J = 8 Hz, 3H), 7. 15 (s, 2H), 6. 84 (s, 5H), 6. 79 (s 5H), 6. 65 (s, 1H), 6. 20 (s, 2H), 5. 63 (s, 2H), 3. 97 (m, 3H), 3. 86 (m, 16H), 3. 63 (m, 13H), 3. 49 (m, 2H), 2. 45 (s, 6H), 2. 24 (s, 6H), 1. 44–1. 35 (m, 36H), 0. 97 (m, 2H), 0. 74 (m, 2H), 0. 59 (m, 6H), 0. 27 (m, 2H), −0. 2 (m, 2H), −0. 4 (m, 2H), −0. 65 (m, 2H), −0. 81 (m, 2H); 13 C NMR (400 MHz, CDCl 3, 25 °C, δ): 156. 06, 150. 60, 149. 47, 142. 85, 138. 93, 136. 10, 135. 76, 133. 69, 131. 47, 130. 32, 130. 09, 129. 15, 125. 84, 125. 26, 122. 89, 121. 79, 121. 62, 116. 90, 116. 51, 114. 79, 66. 22, 65. 30, 63. 86, 52. 07, 48. 37, 31. 28, 30. 78, 30. 50, 30. 06, 29. 85, 29. 70, 29. 40, 28. 93, 28. 88, 27. 12, 26. 57, 25. 73, 21. 54, 15. 74, 15. 58, 14. 76, 14. 69. MS m / z : [M–Cl] + calcd for C 99 H 127 BF 2 N 5 O 12 Cl, 1627. 90; found, 1627. 95. Anal. calcd for C 99 H 127 BF 2 N 5 O 12 Cl: C 71. 49, H 7. 70, N 4. 21; found: C 71. 41, H 7. 78, N 4. 26. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500118
| 2,015
|
Advanced Science
|
Rational Design of Materials Interface for Efficient Capture of Circulating Tumor Cells
|
Originating from primary tumors and penetrating into blood circulation, circulating tumor cells (CTCs) play a vital role in understanding the biology of metastasis and have great potential for early cancer diagnosis, prognosis and personalized therapy. By exploiting the specific biophysical and biochemical properties of CTCs, various material interfaces have been developed for the capture and detection of CTCs from blood. However, due to the extremely low number of CTCs in peripheral blood, there exists a need to improve the efficiency and specificity of the CTC capture and detection. In this regard, a critical review of the numerous reports of advanced platforms for highly efficient and selective capture of CTCs, which have been spurred by recent advances in nanotechnology and microfabrication, is essential. This review gives an overview of unique biophysical and biochemical properties of CTCs, followed by a summary of the key material interfaces recently developed for improved CTC capture and detection, with focus on the use of microfluidics, nanostructured substrates, and miniaturized nuclear magnetic resonance‐based systems. Challenges and future perspectives in the design of material interfaces for capture and detection of CTCs in clinical applications are also discussed.
|
1 Introduction Metastasis, the spread of tumor cells from the primary tumor site to vital distant organs through the circulatory system, is directly responsible for most carcinoma‐related deaths in cancer patients. 1, 2, 3 Understanding the metastasis process and investigating the cause of metastasis will benefit the diagnosis and therapy of cancers, and have long been a focal point in the fight against malignant tumors. 4, 5, 6 More than a century ago, Ashworth found tumor cells in the blood of an individual with metastatic cancer and suggested that these circulating tumor cells (CTCs) could originate from several tumors present in the patient. 7 CTCs up to now have been found in patients with malignant tumors including lung, prostate, breast, colon and pancreatic cancers, but not in healthy individuals or patients with non‐malignant tumors. Hence, the relationship between the presence of CTCs and the development of metastases has been an important subject in tumor studies, and the number of CTCs is believed to be an important indicator of carcinoma progression and metastasis. 8, 9, 10 Therefore, CTC enumeration can be used as a novel approach for cancer prognosis in which the enumeration values have been demonstrated to correlate to overall survival of patients with metastatic cancer. For example, patients with metastatic breast and prostate cancer have a lower survival rate if their CTC count is more than 5 CTCs per 7. 5 mL of whole blood when using the CellSearch System. 11 Compared to routine clinical analysis by collecting disseminated tumor cells via surgical removal or tumor biopsy, CTC enumeration from peripheral blood as a “liquid biopsy” is more convenient and amenable in practical operation. 12 Furthermore, for patients undergoing cancer treatment, the decline of CTCs number is reported with the decrease of tumor size. 13, 14 Hence, in addition to serving as a prognostic marker of cancer metastasis, CTC enumeration can also be used as a novel non‐invasive method to assess the efficacy of cancer therapeutic treatment and realize personalized therapy. In recognition of the promising potential of CTCs in early cancer diagnosis and treatment, there is a growing interest in developing strategies for capture and detection of CTCs. 15, 16, 17, 18 However, due to the extreme rarity of CTCs, with only 1–100 CTCs present in 1 mL of peripheral blood, which usually contains about five billion normal blood cells, CTC capture and subsequent detection are particularly challenging. 19, 20 Currently, in term of the differences in biophysical and biochemical properties of CTCs as compared to normal blood cells, CTC capture strategies can be divided into two broad categories: biochemical methods and biophysical methods. Biochemical CTCs capture methods achieve selective CTCs isolation by affinity capture of unique biochemical markers expressed on surface of CTCs. For example, CellSearch system, a typical biochemical interface developed for CTC capture and isolation, is the first FDA‐approved system that processes 7. 5 mL blood and enriches CTCs by the antibodies of CTCs unique biochemical marker of epithelial cell adhesion molecule (EpCAM) conjugated on magnetic beads, followed by microscopic cell imaging. 21 Biophysical CTCs capture methods rely on differences in the physical properties of CTCs compared to normal blood cells such as cell size, deformability and density. Filtration and density gradient are two typical conventional biophysical methods for CTC capture and isolation. 22, 23 Specifically, filtration provides size‐based separation of CTCs on the premise that they are larger than normal leukocytes and red blood cells, while density gradient centrifugation utilizes differences in cell density to separate CTCs from blood. Although CTC capture and isolation have been successfully achieved by these systems, the low CTC‐capture yield and purity of these systems are matters of concern. 24, 25 Therefore, it is critical and urgent to develop some advanced material interfaces to achieve efficient capture and subsequent sensitive detection of rare CTCs for advancing biological and clinical cancer studies and applications. Recently, by exploiting the unique biophysical and biochemical properties of CTCs together with the development of nanotechnologies and advances in microfabrication and microfluidics, various exquisite material interfaces have been designed for outstanding capture and high‐sensitivity detection of rare CTCs ( Figure 1 ). In this review, we summarize recent representative works on the development of advanced material interfaces for CTC capture and detection. First, we will briefly introduce the known biophysical and biochemical properties of CTCs that can be employed for the design of these material interfaces. Subsequently, we will review the key advanced material interfaces, newly developed for efficient capture and detection of CTCs that can potentially revolutionize the future healthcare technology in cancer diagnosis and therapy, with focus on microfluidics, nanostructured substrates, and miniaturized nuclear magnetic resonance‐based systems. Lastly, we will present the challenges and future perspectives in the design of innovative materials interface for CTC capture and detection in clinical applications. Figure 1 By exploiting the unique biophysical and biochemical properties of CTCs, various exquisite platforms for CTCs capture and detection have been designed, which mainly involve the microfluidic‐, nanostructured substrates‐, and μNMR‐based systems. Some representative examples are exhibited here. Reproduced with permission. 59 Copyright 2010, American Chemical Society. Reproduced with permission. 66 Copyright 2010, National Academy of Sciences. Reproduced with permission. 84 Copyright 2012, Royal Society of Chemistry. Reproduced with permission. 94, 101, 122 2 Biophysical and Biochemical Properties of CTCs By presenting the possibility of being exploited to discriminate CTCs from normal blood cells, the biophysical and biomechanical properties of CTCs have gained much attention. 26, 27, 28 In this section, we first review historical and recent studies of the biophysical properties of CTCs including density, size and deformability. Next, we present studies of the unique biochemical properties of CTCs concentrating on the specific surface receptors that can be used for the selective capture and isolation of CTCs from peripheral blood samples. CTCs were first identified by Ashworth in 1869 when he microscopically inspected the blood from metastatic cancer patients. Due to their similarity to the metastatic cancer cells, CTCs identification were initially done by trained cytologists in term of elongated nuclei and fragmentation of the chromatin based on Papanicolaou's criteria for malignancy. 29, 30 The density of CTCs were then investigated. Seal et al. studied the specific gravity of CTCs and leukocytes in density gradient centrifugation, and concluded that the specific gravity of CTCs was bigger than leukocytes and the method of density gradient centrifugation appeared to be a potential way for CTCs capture and isolation. 31 In addition to the biophysical property of density, the tendency of CTCs to form clusters rather than individual cells was reported. 32, 33, 34 With the development of microscopy and fluorescent staining technologies, more insight into the biophysical properties of CTCs including size and deformability were provided recently. 35, 36 Several studies in breast cancer and lung adenocarcinoma noted that the size of CTCs was typically larger than blood cells, which has been an important criteria generally used for current CTCs capture and isolation. 37, 38 In addition to size, cell deformability is another important biophysical property of CTCs frequently exploited in current capture and label‐free isolation of CTCs. Cell deformability refers to the ability of cells to change shape under a given level of applied stress without rupturing, which is an important biophysical property for CTCs to make them survive and transfer in the stressful environment of blood stream. Cell deformability can be indicated by the nuclear‐cytoplasmic ratio (NC ratio: the ratio of nuclear area and cell area with the nuclear area subtracted), where cells having larger NC ratio are likely to be less deformable. Meng et al. compared the average NC ratio of CTCs from 36 breast cancer patients with leukocytes, and concluded that CTCs always had larger NC ratio than leukocytes, which was consistent with the cell deformability data in which CTCs was less deformable than leukocytes. 39 The decreased deformability of CTCs suggests that CTCs are stiffer than leukocytes, which provides an efficient way for CTCs capture and isolation from leukocytes. 40 In summary, all the differences of biophysical properties between CTCs and leukocytes in term of density, size, internal structures and deformability demonstrated in these historical and recent studies, could be exploited for the label‐free isolation of CTCs. In addition to the aforementioned biophysical properties, CTCs also express some unique biochemical markers that can be utilized for selective CTC capture and isolation. Among these biochemical markers, EpCAM and human epidermal growth factor receptor 2 (HER2) are two typical biochemical markers frequently used for the isolation and enrichment of CTCs. EpCAM is a transmembrane glycoprotein mediating Ca 2+ ‐independent cell‐cell adhesion in epithelia. 41, 42 EpCAM is found expressed on a great variety of human adenocarcinoma cells, but it is absent in blood cells. 43 Hence, EpCAM is one expressed CTC‐associated biomarkers known, and CTC isolation techniques based on EpCAM antibodies are widely used. 44, 45 The popular CellSearch system, which has been extensively used to capture and isolate CTCs from the blood of patients with cancers of the breast, prostate, and colon, employs a conjugation of EpCAM antibodies to ferrofluidic beads to enable the capture of CTCs through a magnetic field. 46, 47, 48 In addition to EpCAM, several studies have reported that HER2 is overexpressed in CTCs of both metastatic and early breast cancer patients, and clinical data has shown that the change of HER2 status from low level expression to high level also occurred along with breast cancer recurrence and progression. 49, 50 On the basis of this finding, HER2 is now considered to be a potential CTCs‐associated biomarker, and has also been widely used for CTC isolation and enrichment in clinical applications. 3 Microfluidics‐Based Material Interface for CTCs Capture Based on the differences in biophysical properties between CTCs and normal blood cells described above, label‐free strategy for direct capture and isolation of CTCs can be developed. As a powerful separation approach, microfluidic technique with small sample‐volume requirement, fast processing times, multiplexing capabilities and large surface area‐to‐volume ratios, offer a good option for label‐free CTCs capture and isolation. 51, 52 Recently, with the progress in nanobiotechnology and microfabrication, various microfluidic devices with rationally designed material interfaces have been developed for efficient CTCs capture, isolation and enrichment. 53, 54, 55 These exquisite microfluidic systems will be briefly introduced in this section, and their performances on CTCs capture and isolation are summarized in Table 1. Table 1 The performances of microfluidic‐based platforms on CTCs capture and isolation Microfluidic platform Capture yield or efficiency Detection rate or sensitivity Cell survival rate Reference Microcavity filter 80% – 98% 59 CTC‐chip – 99% – 17 Herringbone‐chip 91. 8% 93% 95% 66 Arc‐shaped trap 89% 100% – 80 Spiral microfluidics 80% 100% – 84 Immunomagnetic microfluidic chip 90% – – 69 John Wiley & Sons, Ltd. As a relatively straightforward technique with low cost, size‐based microfluidic filtration is one of the first approaches employed for CTCs capture and isolation, based on the fact that CTCs are larger than blood cells such as white blood cells (WBCs) and red blood cells (RBCs). 56 Several membrane‐based microfilters have been developed for size‐based microfluidic capture and isolation of CTCs from peripheral blood samples. 57, 58 Hosokawa et al. developed a microfluidic device equipped with a nickel microfilter for size‐based selective CTCs capture and isolation ( Figure 2 A). 59 In this microfluidic device, the nickel microfilter composed of 100×100 holes with the diameters between 8 and 11 μm was integrated between two PDMS funnels. Based on this microfluidic device, efficient isolation of CTCs from peripheral blood samples was obtained with high efficiency of greater than 80%. One advantage of this microfluidic device over conventional immunomagnetic CTCs separation platforms is its unique ability for efficient isolation of EpCAM‐negative CTCs. Moreover, approximately 98% of captured CTCs were found to be viable after fluorescent staining and washing processes. In another report, Vona et al. developed a polycarbonate microfilter‐based microfluidic device for size filtration of CTCs. The polycarbonate microfilter presented holes with diameter of 8 μm, and CTCs were efficiently trapped by this microflter while RBCs and WBCs with smaller sizes passed through it. Furthermore, this microfluidic system could run 12 samples in parallel and performed further identification and characterization of CTCs. In addition to above mentioned 2D microfilter‐based microfluidic devices, Zheng et al. recently fabricated a microfluidic system based on a microfilter with 3D pore structure for CTC isolation. 61 The 3D microfilter was composed of two 10 μm thick Parylene C membranes separated by a 6. 5 μm gap. Since the microholes fabricated on the top membrane were misaligned with that fabricated on the bottom membrane, large CTCs were efficiently trapped in this filter and finally, CTCs isolation from peripheral blood samples was successfully achieved based on this 3D microfilter‐based microfluidic platform. Moreover, a comparative study between the 2D and 3D microfilters was carried out, and it was found that the 2D microfilter might damage CTCs during the microfiltration process but the 3D microfilter could resolve this problem. Figure 2 Evolution of the CTCs capture and enrichment methods based on microfluidics‐based material interfaces. A) Microcavity filter. Reproduced with permission. 59 Copyright 2010, American Chemical Society. B) CTC‐chip. Reproduced with permission. 17 Copyright 2007, Macmillan Publishers Ltd. C) Herringbone‐chip. Reproduced with permission. 66 Copyright 2010, National Academy of Sciences. D) Arc‐shaped trap. Reproduced with permission. 80 Copyright 2010, Elsevier. E) Spiral microfluidic channel. Reproduced with permission. 84 Copyright 2012, Royal Society of Chemistry. Additionally, microfluidic devices with functionalized microchannels have been developed for CTCs capture and isolation. 62, 63, 64 Launiere et al. fabricated a microfluidic system with channels modified by alternating patterned biomimetic proteins (EpCAM antibody and E‐selectin) to increase target CTCs capture while reducing leukocyte's non‐specific adhesion by up to 82%. 62 In addition to microfluidic platforms with barely immunoaffinity‐based strategy for CTCs capture, microfluidic platforms with functionalized microstructure arrays have also been developed for improved CTCs capture and isolation. Sheng et al. reported an aptamer‐functionalized microfluidic device with micropillar array for enhanced CTCs capture. 65 This platform consisting of 59 000 micropillars, which could improve the interactions between CTCs and the aptamers, achieved efficient CTCs capture with efficiency of 95% from non‐processed whole blood samples. Nagrath et al. developed a unique microfluidic platform with EpCAM antibody‐coated microposts array for efficient and selective CTCs capture and isolation. 17 Based on the selective interactions between target CTCs and the EpCAM antibody‐coated microposts, viable CTCs isolation from peripheral whole blood samples was achieved by this platform under precisely controlled laminar flow conditions. It was found that this platform successfully identified CTCs from peripheral whole blood samples of patients with different cancers with high sensitivity of 99% (Figure 2 B). In addition to the functionalized micropost array‐based microfluidic platform, the same group also reported a herringbone‐based high‐throughput microfluidic mixing device (Herringbone‐Chip) for enhanced CTCs capture and isolation (Figure 2 C). The Herringbone‐Chip consisted eight microchannels with patterned herringbone structures designed to generate microvortices and provide passive mixing of blood cells to enhance the interactions between CTCs and EpCAM antibody‐coated chip surface. Consequently, CTCs capture with efficiency of 91. 8% was obtained based on this herringbone‐Chip in the blood samples prepared by spiking defined numbers of cancer cells into blood, as well as clinical blood samples from patients with metastatic disease, thereby indicating its great potential in clinical settings. 66 Liu et al. developed a microfluidic devices with EpCAM antibody‐functionalized deterministic lateral displacement (DLD) chamber composed of triangular micropost arrays for CTCs capture. 67 Based on the combination of microfluidic DLD array and high affinity‐based capture approach, efficient CTCs capture with 90% efficiency was obtained from spiked blood samples at low cell concentration (10 2 cells mL −1 ). Kamande et al. reported a modular microfluidic system containing three different functional regions by which isolation, enumeration, and phenotyping of CTCs could be finished in one device. 68 By combing the immunomagnetic separation strategy and microfluidic technique, Immunomagnetic‐based microfluidic systems have also been developed for efficient CTCs capture and isolation. 69, 70, 71, 72, 73, 74, 75, 76 Hoshino et al. developed an immunomagnetic‐based microfluidic device for CTCs capture. 69 In this work, blood samples were firstly labelled with magnetic nanoparticles functionalized by EpCAM antibodies, and target CTCs were then efficiently captured with high efficiency of 90% when the blood samples flowed through the microfluidic channel closely above arrayed magnets. Huang et al. also reported an immunomagnetic‐based microfluidic system for CTCs isolation from blood samples. 70 This microfluidic system was operated in a flip‐flop mode in order to reduce the stagnation and non‐specific adhesion of normal blood cells on microfluidic surface in the process of CTCs capture, and high capture efficiency of 90% was finally achieved based on this platform. The same group also developed a versatile immunomagnetic nanocarrier‐based microfluidic platform for capturing CTCs in whole blood. 71 In this work, CTCs were selectively targeted by EpCAM antibody‐functionalized magnetic nanocarriers and isolated from whole blood samples by magnetic force in a microfluidic chamber with capture efficiency greater than 90%. Chen et al. fabricated a graphite‐coated magnetic nanoparticles microarray chip for CTCs capture and isolation. 72 The graphite‐coated magnetic nanoparticles with good biocompatibility and stability were synthesized by using the chemical vapor deposition, and the graphite modification on its surface provided functional groups for subsequent antibody labelling to achieve specifically CTCs recognition. Based on this graphite‐coated magnetic nanoparticles microarray chip, efficient CTCs capture from spiked blood samples was successfully achieved even at very low cell concentrations. Similarly, Yu et al. developed a microfluidic system with micropillar array decorated with graphite oxide‐coated and antibody‐functionalized magnetic nanoparticles for CTCs capture and isolation. 73 Under magnetic field manipulation, the decoration of functionalized magnetic nanoparticles on the micropillars increased the interactions between target CTCs and micropillar surface, and successful CTCs capture from two spiked media was achieved with capture efficiency greater than 70% in culture medium and greater than 40% in blood sample. In another report, Issadore et al. developed a microfluidic chip‐based micro‐Hall detector to capture immunomagnetic nanoparticle‐tagged CTCs from whole blood sample with high efficiency and high‐throughput ability. 74 High deformability is a distinctive biomechanical property for cells circulating in the peripheral blood, especially for CTCs with larger size than normal blood cells in order to rapidly go through capillaries with small diameters of 6–8 μm and successfully metastasize. 77, 78 Atomic force microscopy (AFM)‐based single cell stiffness study for different cancer cells including lung, breast and pancreatic, have shown that malignancy increase cell deformability at the single cell level although CTC are still more stiffer than blood cells. 40 Therefore, in addition to size, the unique deformability of CTCs is also a factor that can be used for selective capture and isolation of CTCs. Based on the fact that CTCs are always larger and stiffer than normal blood cells, our group developed a microfluidic device equipped with an array of traps for CTCs capture and isolation from peripheral blood samples. 79, 80 For the structure of trap array, each trap was composed of three pillars with a diameter of 3–4 μm and was arranged in an arc shape with 5 μm distance between pillars. In the process of CTCs capture and isolation, small sized RBCs and WBCs with higher deformability could pass through the 5 μm gaps, while larger CTCs were stuck in the arc‐shaped traps, achieving highly efficient CTCs capture and isolation (Figure 2 D). Furthermore, a pre‐filter with 20 μm gap was mounted to prevent larger clumps and debris from clogging up the cell trap area. Highly isolated CTCs can be finally collected from this microfluidic‐based material interface for downstream applications, such as immunological staining and molecular analysis. In a straight microfluidic channel, fluid shear can generate lateral forces to cause transverse migration of particles. 81, 82 While in a spiral microfluidic channel, an inertial focusing of particle according to its size can be observed due to combination of shear induced life force and Dean drag force, which has been used for size separation of particles, giving some illumination for CTCs isolation by using a spiral microfluidic channel. Separation of CTCs in a spiral channel with rectangular cross‐section has been reported. 83 Recently, a novel spiral microfluidic device with trapezoidal cross‐section was developed for rapid and efficient label‐free isolation of CTCs from clinically blood samples, by utilizing the inherent Dean vortex flow and inertial lift forces present in the spiral microfluidic channel (Figure 2 E). 84 Compared to conventional spiral microfluidic devices with rectangular cross‐section, the position of Dean vortex core in spiral microchannel with trapezoidal cross‐section can be altered by which larger CTCs will focus and be collected at the inner channel wall outlet while smaller hematologic cells will focus and be removed at the outer wall outlet, thus achieving efficient CTCs isolation and enrichment. Based on this platform, high CTC capture efficiency of greater than 80% were successfully achieved within 8 min from both spiked cancer cells blood samples and clinical peripheral blood samples from patients with advanced stage metastatic breast and lung cancers, providing a powerful tool for CTCs capture and isolation. 4 Nanostructured Substrates‐Based Materials Interface for CTC Capture In tissue engineering and regenerative medicine, nanostructured substrates have been widely employed to mimic the natural extracellular matrix (ECM) and basement membrane. 85, 86, 87 These substrates can promote cell attachment due to enhanced local topographic interactions between nanostructures and nanoscale components of the cellular surface such as microvilli and filopodia, thereby assisting the capture and isolation of CTCs. 88, 89 Furthermore, nanostructured substrates can provide more surface area for immobilization of CTC affinity molecules. 90, 91 Hence, nanostructured substrates can be combined with the affinity interactions‐based CTC capture strategy, which can further improve CTC capture efficiency and emerge as a promising platform for isolation, and enrichment of CTCs. In this section, different types of nanostructured substrates‐based platforms for CTCs capture and isolation available will be briefly introduced including nanowires, nanopillars, nanodots, nanofibers, nanosheets, nanotubes and nanopores, and their performances on CTCs capture and isolation summarized in Table 2. Table 2 The performances of nanostructured substrates‐based platforms on CTCs capture and isolation Nanostructured substrate platform Capture yield or efficiency Detection rate or sensitivity Cell survival rate Reference Nanopillar 40% – 84–91% 92 Nanopillar‐micromixer 95% – – 94 Nanowire 67. 5% – – 93 Nanofiber 45% – – 97 Nanosheet 73% – – 96 Nanopore 80% 101 John Wiley & Sons, Ltd. Inspired by the surface components of cells, the nanowire and nanopillars‐based substrates have been designed and utilized to make use of the surface adhesion of the cells and aid the capture of CTCs in blood samples. For example, Wang et al. firstly utilized anti‐EpCAM‐coated Si nanopillars (SiNPs) substrates to identify and capture CTCs ( Figure 3 A). Using a wet chemical etching approach, densely packed nanopillars of 100–200 nm in diameter were prepared on silicon wafers; additionally, the length of these nanopillars could be easily controlled by altering the etching times. To test the cell capture efficiency of the SiNPs, cell suspension solution of MCF7 cells (an EpCAM‐positive cell line) was introduced for 1 h into the SiNPs and also flat silicon substrates. It was found that more cells were captured on SiNPs (45–65%) than on flat silicon substrates (4–14%), suggesting that nanopillars are responsible for enhanced cell capture. The performance of SiNPs on CTC capture and isolation was tested in the artifical CTCs blood samples prepared by spiking blood with different densities of tumor cells, and improved capture efficiency of CTCs (40%) was obtained by the SiNPs platforms comparted to some commercially available technologies. 92 Similar to SiNPs, quartz nanowires (QNWs) was also fabricatied and employed for CTCs capture and quantification in the spiked blood samples to investigate its potential in clinical use. 93 By increasing the contact frequency between cell and nanopillar substrate, even higher capture efficiency of CTCs could be further achieved. For instance, Wang et al. integrated SiNPs into a microfluidic device with serpentine chaotic micromixers, obtaining a nearly 100% capture efficiency (Figure 3 B). To test the performance of this integrated platform for CTC capture, a series of CTC samples was firstly prepared by spiking three kinds of solutions (whole blood, lysed blood, and PBS buffer) with cancer cell lines of MCF7, T24 and PC3, respectively. Under the optimal conditions of flow rate, more than 95% of capture efficiency of target cancer cells was found in all CTCs samples mentioned above by this integrated platform, providing an efficient way for isolation of CTCs and early diagnosis of cancer metastasis. 94 Figure 3 Efficient capture and enrichment of CTCs by using nanostructured substrates‐based material interface. A) Nanopillar substrates. Reproduced with permission. 92 B) Integrated chaotic micromixer‐nanopillar substrates. Reproduced with permission. 94 C) Nanodot. Reproduced with permission. 95 D) Nanofiber. Reproduced with permission. 97 E) Carbon nanotubes. Reproduced with permission. 99 F) Nanopore‐based 3D graphene foam. Reproduced with permission. 101 Nanodot and nanosheet substrates have also been demonstrated to show efficient CTCs capture ability. The dot size and density were controlled by the voltage applied and could be easily reproduced and tuned. Five tumor cell lines of interest were examined and they were either overexpressed with EpCAM antigens or without EpCAM antigens on their cell membranes. Although the aspect ratios of nanodots were small, the efficiency of specific cell capture by anti‐EpCAM conjugated to the nanodots, was enhanced by four to five times in comparison with smooth films (Figure 3 C). 95 The enhancement is most likely due to a synergistic effect from ligand‐receptor interaction, and nanostructure matching of tumor cells and nanodot substrate. In addition to nanodot, nanosheet substrates have also been employed for efficient CTCs capture and isolation. Yoon et al. fabricated a graphene oxide nanosheet substrate‐based device and used it for CTCs capture from blood samples. 96 After functionalized by CTCs‐selective antibodies, this nanosheet substate‐based device exhibited ability for CTCs capture with efficiency of 73% from blood samples of pancreatic, breast and lung cancer patients at low cell concentrations (3–5 cells mL −1 ). Inspired by ECM scaffolds, nanofiber‐based substrates have been fabricated and well developed for their efficient CTCs capture efficiency. Different materials such as TiO 2, and poly(lactic‐co‐glycolic acid) (PLGA) can be electrospun to form desired nanofibers with controllable diameters and lengths. Zhang et al. fabricated TiO 2 nanofibers of 100–300 nm diameter from a spun composite of titanium n‐butoxide and polyvinyl pyrrolidone (Figure 3 D). By coating anti‐EpCAM onto the surface of nanofibers, functionalized platform for CTCs capture was prepared. Using these nanofibers deposited substrates, cancer cells from artificial CTCs blood samples, as well as from whole blood samples of colorectal and gastric cancer patients were reliably captured. 97 In another report, Hou et al. developed a PLGA‐nanofiber embedded chip (PN‐nanovelcro chip) which not only captured CTCs with high efficiency, but also enabled highly specific isolation of single melanoma cell immobilized on the nanosubstrate. 98 The PN‐nanovelcro chip was composed of an overlaid PDMS chaotic mixter and a transparent PN‐nanovelcro substrate fabricated by electrospining PLGA nanofibers onto a commercial laser microdissection (LMD) slide and functionalized by a melanoma‐specific antibody. Based on the enhanced local interaction between cell and PLGA nanfibers, target melanoma cells were efficiently captured, and single cell isolation was subsequently isolated by using the highly accurate LMD technique. In order to specifically identify melanoma cells captured on the PLGA nanofibers, a four‐color immunocytochemistry method was also developed in the PN‐nanovelcro chip system. Nanotubes and nanopores have also been reported to have great potential for CTCs capture and isolation. For example, functionalized multiwalled carbon nanotubes (MWCNTs) films have been successfully used for K562 cells (leukemia cells) capture and electrochemical sensing (Figure 3 E). 99 They prepared the films by covalent coupling between ‐NH 2 groups in 3‐aminophenylboronic acid (APBA) and ‐COOH groups in acid‐oxidized MWCNTs. Due to the high affinity interacions between the boronic acid groups of APBA and the carbohydrate on cell surface, the K562 cells could be efficiently captured by the APBA‐functionalized MWCNTs films. Compared to bare APBA films, the functionalized MWCNTs one not only exhibited more boronic acid groups for K562 cell recognition, but also provided enhanced local cell‐MWCNTs interactions, which improve K562 cells' adhesion on its surface. Furthermore, the high electrical conductivity of MWCNTs maked the APBA‐MWCNTs film a good electrode for subsequent cell electrochemical sensing, presenting a promising way for efficient capture and highly sensitive electrochemical detection of CTCs. In another report, King et al. explored a method to more efficiently capture leukemic and epithelial cancer cells from flow by altering the nanoscale topography of the inner surface of P‐selectin‐coated microtubes. 100 In this work, halloysite nanotubes were naturally attached to the inner surface of microtubes to alter their nanoscal topography via a monolayer of poly‐L‐lysine. It was found that the capture efficiency of leukemic cells could be increased by halloysite nanotube coatings and mainly affected by halloysite content and selectin density, making the functionalized microtubes with nanoscale topography a promising platform for enhanced CTCs capture and isolation. In addition to nanotubes, nanopores also open up a new opportunity in CTCs capture and diagnosis. In a recent report, we reported a 3D hierarchical graphene platform that combines microporosity from reduced graphene oxide foam with anti‐EpCAM coated ZnO nanorod array (Figure 3 F). 101 The advantage of this novel composite structure stems from its high density of ZnO nanorods, which increases cell‐substrate contact frequency, as well as its microporosity, which lets through normal RBCs but specifically captures CTCs due to the introduction of EpCAM antibodies. When thickness of the foam reached 5 mm, the cell‐capture yield was more than 80%, indicating its potential CTCs capture capability for clinical blood samples. 5 Miniature Nuclear Magnetic Resonance System‐Based Materials Interface for CTCs Capture and Detection As a novel sensing technology, micro‐nuclear magnetic resonance ( μ NMR) exploits magnetic resonance technology to detect target labelled with immunospecific magnetic nanoparticles (MNPs), showing great potential in rapid and highly sensitive biodetection. 102 The typical MNPs used in μ NMR are superparamagnetic and have small size (tens of nm), which is different from the conventional magnetic nanoparticles used in immunoseparation. The mechanism of μ NMR‐based sensing technique is based on the phenomenon that MNP‐labeled targets exhibit faster relaxation of NMR signals due to local magnetic fields created by MNPs. 103 By systematically optimization of nanoagents, MNP‐target conjugation method, and NMR detectors, several exquisite μ NMR‐based platform have been developed for rapid and sensitive detection of biomolecules including nucleic acids, proteins, bacteria, and tumor cells. 104, 105, 106, 107, 108, 109 Compared to conventional biosensing methods, μ NMR‐based technique do not need sample purification procedures and can simultaneously achieve target capture and detection, gaining much attention in the fields of CTCs capture and detection. This section will briefly introduce recent developments of μ NMR‐based biosening systems and their potential applications in CTCs capture and detection. Based on the “ T2 ‐shortening” effect of MNPs in NMR measurements, the detection of CTCs labelled with MNP can be achieved. In NMR measurements, MNPs can produce local magnetic dipole fields with strong spatial dependence, and subsequent destroy the coherence in the spin‐spin relaxation of water protons. Therefore, target labelled with MNP will show shorter transverse relaxation time in NMR measurements, namely the phenomenon of “ T2 ‐shortening” effect, compared to target without MNPs label, making detection of CTCs possible ( Figure 4 A). 102 For μ NMR‐based CTCs detection system, engineering MNPs for high transverse relaxivity and efficient MNP labeling on cells are two important issues which need to be addressed for highly sensitive μ NMR sensing. 110 For the first issue, elemental iron (Fe) exhibiting the highest saturation magnetization and low magnetocrystalline anisotropy among ferromagnetic crystals, may be a good candidate of constituent material for MNPs and it is possible to synthesize superparamagnetic Fe‐MNPs with high transverse relaxivity. Recently Yoon et al. synthesized a new type of hybrid Fe‐MNP with high magnetic moments and transverse relaxivity. 111 This hybrid particle composed of an elemental Fe core and a protective ferrite shell, showed high transverse relaxivity and stable magnetic properties against oxidation. In addition to the synthesis of MNPs with high transverse relaxivity, strategies for efficient cell MNP‐labeling are also needed for μ NMR‐based CTC capture and detection. Recently, a novel labelling strategy for target‐MNPs constructs preparation called BOND (Bioorthogonal nanoparticle detection) was developed by Lee et al. 112 Based on the reaction between tetrazine (Tz) and trans‐cyclooctene (TCO), namely the Diels‐Alder cycloaddition, BOND can rapidly achieve the covalent binding of MNP to biological targets at room temperature without catalyst (Figure 4 B). BOND chemistry has been employed for cell MNPs labelling using a two‐step approach: cell labelling with TCO‐modified antibodies, and the subsequent covalent binding between cell‐antibodies‐TCO and Tz‐loaded MNPs. Since one antibody can be modified by multiple TCO tags without loss of its affinity, multiple attachment of Tz‐MNPs to cells can be subsequently achieved by using the antibodies as scaffolds. Therefore, compared to the method for cell‐MNPs preparation directly using MNP‐antibody conjugates, the two‐step BOND strategy can efficiently amplify MNP‐binding to cells, and then will amplify NMR signals and ultimately enhanced the detection sensitivity, thus showing great potential in μ NMR‐based CTC capture and detection. Figure 4 A) Principle of CTC detection based on μ NMR system. CTCs tagged with MNPs can accelerate the transverse relaxation of water protons. Compared to the non‐tagged samples (left), the NMR signal will decay faster in time domain (right), providing a sensing mechanism. Reproduced with permission. 102 Copyright 2008, Macmillan Publishers Ltd. B) Bioorthogonal nanoparticle detection (BOND). The method is based on the Diels–Alder cycloaddition between trans‐cyclooctene (TCO) and tetrazine (Tz). Cells are pre‐labeled with TCO‐antibodies and targeted with Tz‐MNPs. The antibody provides sites for multiple MNP binding. Reproduced with permission. 112 Copyright 2010, Macmillan Publishers Ltd. C) Typical example of the μ NMR system. This system consists of an array of microcoils for NMR detection, microfluidic channels for sample handling, embedded NMR electronics, and a permanent magnet. Reproduced with permission. 102 Copyright 2008, Macmillan Publishers Ltd. D) CTC deteciton performance comparison between μNMR and CellSearch system. Reproduced with permission. 114 Copyright 2012, Neoplasia Press. Till now, different models of μNMR devices with miniaturized system for CTCs capture and point‐of‐care detection have been developed. The miniaturization of μNMR system endows the following advantages: 1) improving detection sensitivity by reducing sample volumes and increasing the concentrations of targets; and 2) generating stronger radio‐frequency NMR magnetic fields due to the use of smaller magnets in miniaturized system. Figure 4 C shows a typical miniaturized μNMR device developed for CTCs capture and detection, which included four main parts: microcoils, microfluidic network, custom‐designed NMR electronic and a portable small permanent magnet. On the bottom of this μNMR device, eight planar microcoils with volume of 10 μL were arranged into a 2 × 4 array format to achieve parallel CTCs detection. A microfluidic network was implemented on the top of the microcoils to enable sample handling and distribution. In addition, to compensate for the inhomogeneity of magnetic field generated by the small permanent magnet, a NMR electronic was customarily designed and implemented to allow for spinecho measurement. Appreciating the rarity of CTCs in blood sample and the heterogeneity of CTCs surface biomarkers, quad‐μNMR platform with a quad biomarker “cocktail” (MUC‐1, EGFR, HER2, and EpCAM) for optimal signal and detection was developed for CTCs isolation and detection. 113 In this cocktail assay, CTCs were simultaneously targeted with TCO‐modified MUC‐1, EGFR, HER2, and EpCAM antibodies and subsequently incubated with Tz‐MNPs to get the quad NMR probes of CTC‐MNPs. For the performance of this quad‐μNMR platform, it was found that an average recovery rate of 38% across the various cell concentrations (200, 100, 50, and 25 spiked cells) tested was obtained, which was higher than the CellSearch system only with an average recovery rate of 9. 1%, ultimately leading to higher CTCs detection sensitivity ( Figure 5 D). Compared to CellSearch system, the quad‐μNMR platform demonstrated 400% fold higher CTCs detection sensitivity, showing great potential in isolation and detection of CTCs with low surface biomarkers expressing cell line, such as the MDA‐MB‐436 with known EMT behavior. Furthermore, CTCs isolation and detection from peripheral blood samples collected from 15 patients with ovarian cancer were successfully and sensitively achieved by using the quad‐μNMR platform, indicating its potential in clinical applications. 114 In summary, the Quad‐μNMR platform expands the range of CTCs isolation and detection conditions, making it not limited to the case of higher CTCs concentrations such as stage IV, progressive disease, or in patients not pursuing active therapy. Figure 5 Strategies for controllable CTC release from nanowire substrates. A) Enzymatic treatment. Reproduced with permission. 121 B) temperature stimulation. Reproduced with permission. 122 C) pH and glucose stimulation. Reproduced with permission. 123 Copyright 2013, American Chemical Society. 6 Approaches for CTC Detection and Identification Once CTCs are captured and enriched, subsequent detection and identification are needed to investigate their origin and genetic profile from which more valuable insight into the biology of metastasis can be obtained. 115, 116 In μNMR‐based platforms, CTCs detection can be easily achieved by analyzing the NMR signals of MNP‐labelled target tumor cells without prerequisite isolation and enrichment processes. Hence, this section focuses on the approaches of CTCs detection and identification used in microfluidic‐ and nanostructured substrates‐based platforms, which mainly involve immunological and molecular methods. Most CTCs immunological identification assays use different fluorescent dyes to simultaneously stain cytokeratins (positive marker for epithelial tumor cells) and leukocyte antigen CD45 (exclusion marker). The cell staining process is always carried out in situ in microfluidic‐ and nanostructured substrates‐based platforms. For example, in size filtration‐based microfluidic system with microcavity arrays, two fluorescent immunological probes of FITC‐labelled anti‐CD45 antibody and PE‐labelled anti‐EpCAM antibody were employed to detect and identify CTCs captured on the microcavity arrays. 59 Similarly, in a recent study, CTCs captured on the aptamer‐functionalized SiNWs substrates were distinguished from non‐specifically trapped WBCs by using a three‐color immunological method based on FITC‐labeled anti‐EpCAM, Cy5‐labeled anti‐CD45, and DAPI nuclear staining. 117 Most CTCs molecular identification methods use DNA testing techniques such as polymerase chain reaction (PCR) and restriction fragment length polymorphism (RFLP) to analyze the specific DNA or mRNA of CTCs enriched. 117, 118 PCR‐based analysis technique are the most widely used molecular method for CTCs detection and identification. For example, Devriese et al. used PCR‐based technique to analyze a panel of gene marker of CTCs including cytokeratin 7, cytokeratin 19, human epithelial glycoprotein and fibronectin 1 for selective identification and detection of CTCs in non‐small lung cancer, and achieved sensitivity of 46% and a specificity of 93% in 46 cancer patient. 119 In another study, Hoe et al. used PCR to do the single CTC genotyping for a key melanoma drug target mutation after capturing CTCs using a nanofiber‐embedded microchip. 98 For approaches of CTCs detection and identification, there are still two factors needed to be taken into consideration. Firstly, heterogeneity among CTCs is a problem that can not be ignored for CTCs detection and identification, which makes cell‐to‐cell variations occur in same cancers and only a very small fraction of CTCs that may eventually acquire the ability to seed the metastatic tumor. Hence, how to characterize molecular, phenotypic and functional difference of CTCs at the single‐cell level is a critical problem that need to be reviewed. Lee et al. reported a laser scanning cytometry‐based method for CTCs detection and identification, by which automated and rapid characterization of physical and functional cellular properties such as size, shape and signaling proteins of CTC at the single‐cell level was quantitatively achieved. 93 In another report, by combining the PLGA nanofiber substrate with the laser microdissection (LMD) technique, an exquisite platform for CTCs capture and detection was successfully developed by Tseng group. 98 In this work, based on an LMD microscope, captured CTCs could be cut out and harvested at the single‐cell level, making subsequent single‐cell molecular analysis possible. In addtion to the problem of CTCs heterogeneity, controllable release strategy of CTCs after capture are also needed for subsequent detection and identification of CTCs. For microfluidic CTCs capture platforms, magnetic‐based release strategy is the main method widely used in CTCs molecular identification process. 70, 71, 72, 73, 120 For example, in a recent study, Yu et al. developed a microfluidic platform with micropillar array decorated with magnetic nanoparticles for CTCs capture with efficiency of greater than 70% when the magnetic field was applied, and the captured CTCs could be released with high efficiency of 92. 9% upon the removal of applied magnetic field. 73 Moreover, it was found that 78% of the released CTCs was viable, laying a solid foundation for subsequent molecular analysis. For nanostructured substrates‐based CTCs capture platforms, different CTCs release strategies have been reported, which can be divided into three categories: enzymatic treatment, temperature, and pH and glucose dual stimulation. Detailed explanation for the three strategies will be given below. Enzymatic strategy for controllable CTC release after capture was firstly demonstrated for the SiNWs‐based CTCs capture platform by Shen et al. (Figure 5 A). In their work, by modified the SiNWs substrates with CTC selective DNA aptamers generated via Cell‐SELEX process, a new integrated SiNWs‐microfluidic chaotic mixture‐based CTCs capture platform was fabricated. This aptamers‐functionalized platform could not only achieve efficient CTCs isolation from blood with improved capture efficiency compared to the conventional EpCAM‐functionalized platform, but also realize controllable CTCs release after capture by nuclease treatment. 121 In another report, Hou et al. developed an exquisite platform with CTC capture and on‐demand release ability based on thermally responsive Poly( N ‐isopropylacrylamide) (PNIPAAm) brushes‐modified SiNWs substrate (Figure 5 B). This platform exhibited superior performances in capturing cancer cells with high efficiency at 37 °C, and releasing the captured cancer cells with great viability and retained functionality at 4 °C. 122 Recently, Jiang et al. developed a pH and glucose‐responsive strategy for CTCs release after capture based on poly(acrylamidophenylboronic acid) (polyAAPBA) brush‐grafted aligned SiNWs substrate (Figure 5 C). By precisely controlling pH and the glucose concentration in CTCs samples, reversible capture and release of CTCs could be successfully achieved with dual‐responsive performance. Specifically, the polyAAPBA‐grafted SiNWs substrate changed its state from cell‐adhesive to cell‐repulsive with the increase of pH from 6. 8 to 7. 8 in the presence of 70 mM glucose. Under the condition of pH 7. 8, the polyAAPBA‐grafted SiNWs substrate became glucose responsive, which could capture targeted cells in the absence of glucose and release them in presence of 70 mM glucose. The dual‐responsive capture and release of CTCs on this polyAAPBA‐grafted SiNWs substrate is noninvasive with higher cell viability of 95%. 123 7 Challenges and Future Perspectives In the previous sections, we described various advanced materials interface mainly based on microfluidics, nanostructured substrates, and micro‐nuclear magnetic resonance systems for CTCs capture and detection. Although promising results have been achieved by these interfaces in terms of capture efficiency and detection sensitivity, most of them still remain in the laboratory level and little of them unequivocally shows clinical validity and utility. 124 Heterogeneity among CTCs is a problem that cannot be ignored for CTCs isolation. CTCs always express variable biomarkers on their membrane, which affects their morphology and characteristics and makes cell‐to‐cell variations occur in same cancers, same patient or even within a single blood draw. 125, 126, 127 Therefore, efficient CTCs capture and isolation are challenging due to this heterogeneity of CTCs, indicating that not a single cell surface biomarker can confidently be used for total CTCs isolation. That is why current widely used EpCAM antibodies‐based CTCs capture systems do have concerned limitations in clinical applications. Similarly, the immunological capture methods predominantly used in nanostructured substrates and nuclear magnetic resonance‐based platforms do have similar concerns. By exploiting the inherent unique biophysical properties of CTCs, label‐free microfluidic strategies seems to have greater potential in clinical capture and isolation of CTCs compared to the currently utilized biomarker‐based immunological methods by which only a subset of CTCs expressing the selected surface markers are isolated. However, the label‐free microfluidic approach might also introduce false positives results in clinical CTCs capture and isolation by capturing cells that may not directly originate from the primary tumors. 128 In addition, captured CTCs by some label‐free microfluidic platforms are no longer intact after being subjected to shear forces, thus making subsequent CTCs identification and detection difficult. Beyond the issue of capturing and isolating CTCs, achieving a better understanding of the molecular characteristics of CTCs is also important from which new biomarkers for efficient CTCs capture and isolation can be discovered. However, traditional CTCs molecular analysis is always performed by using large ensembles on the order of 10 3 –10 6 cells, thereby only giving the average genotypic or phenotypic characteristics of the cell population. In addition, the way the isolated cells are cultured to expand CTC numbers is not recommended, given that cancer cells have a feature that modifies their characteristics to survive when the surrounding microenvironment is changed. 129 Single cell analysis has been widely used to explore cellular heterogeneity in gene and protein expressions responsive to environmental change and chemotherapeutic stimuli, providing an efficient way for CTCs heterogeneity study. 130, 131 Therefore, rationally designed platforms with highly efficient CTCs capture and single cell evaluation ability are expected to be promising tools for future CTCs study. Learning from the experiences of previous literatures reviewed above, single‐cell evaluation technique such as laser scanning cytometry and microfluidic systems with nanostructure arrays may be a good candidate for the expected objective, given that microfluidic techniques offer efficient label‐free CTCs separation while the existence of nanostructure arrays confine the CTCs migration and enhance the interactions between target CTCs and microfluidic surface. Furthermore, in order to achieve highly efficient CTCs capture, multi‐biomarkers of CTCs (e. g. , EpCAM, HER2, EGFR, and MUC‐1) can be patterned to the different regions of microfluidic system, by which CTCs with different biophysical and biochemical properties can be isolated. Moreover, with the help of single‐cell evaluation technique, captured CTCs can be further sensitively detected and characterized by immunological and molecular methods at the single‐cell level.
|
10. 1002/advs. 201500122
| 2,015
|
Advanced Science
|
Injectable Hydrogels for Cardiac Tissue Repair after Myocardial Infarction
|
Cardiac tissue damage due to myocardial infarction (MI) is one of the leading causes of mortality worldwide. The available treatments of MI include pharmaceutical therapy, medical device implants, and organ transplants, all of which have severe limitations including high invasiveness, scarcity of donor organs, thrombosis or stenosis of devices, immune rejection, and prolonged hospitalization time. Injectable hydrogels have emerged as a promising solution for in situ cardiac tissue repair in infarcted hearts after MI. In this review, an overview of various natural and synthetic hydrogels for potential application as injectable hydrogels in cardiac tissue repair and regeneration is presented. The review starts with brief discussions about the pathology of MI, its current clinical treatments and their limitations, and the emergence of injectable hydrogels as a potential solution for post MI cardiac regeneration. It then summarizes various hydrogels, their compositions, structures and properties for potential application in post MI cardiac repair, and recent advancements in the application of injectable hydrogels in treatment of MI. Finally, the current challenges associated with the clinical application of injectable hydrogels to MI and their potential solutions are discussed to help guide the future research on injectable hydrogels for translational therapeutic applications in regeneration of cardiac tissue after MI.
|
1 Introduction Heart diseases are among the major causes of death worldwide. The main difficulties in the treatment of cardiac injuries arise from the limited capability of cardiac tissue to regenerate itself, i. e. , the irreversibility of cardiac damages, the inability of cardiac tissues to tolerate ischemia, and the extremely limited period of viability of cardiac tissues postischemic injury. 1, 2 Until now, the golden standard for treatment of cardiac damage has been heart transplantation. However, there is a huge gap between the large number of people who are approved for heart transplants and the small number of available donors, creating a severe shortage and thus an extreme limitation on the efficiency of this method alone. For example, for every 77 people who receive a heart transplant, 20 people die due to shortage while 98 000 remain on waiting lists of at least 3 years in the US alone. 3 These statistics are even more grim in underdeveloped countries where the basic infrastructure required to achieve the timely linking of compatible donors to patients on waiting lists, and to perform the successful transplant itself, are lacking or nonexistent. As a result, there is a huge need for developing new solutions to repair or replace damaged cardiac tissues. Injectable hydrogels have emerged as a promising approach for cardiac repair in regenerative medicine. 4 The ability to rely on engineered or regenerated cardiac tissues, instead of a donated heart, would be a profound step forward in improving the rates of patient survival, as well as furthering the development of new treatments for myocardial infarction (MI), and many other life‐threatening maladies. Further importance of this approach arise from the fact that the body's natural response after MI is a process of fibrous remodeling that ultimately leads to scar formation instead of functional myocardium formation. 2 This scar formation prevents the heart from functioning properly and eventually leads to complete heart failure. Injectable hydrogels have the potential to deliver therapeutic agents, cells or engineered tissues locally to the damaged area of the heart in order to regenerate functional cardiac tissue and, therefore, provide a revolutionary treatment option for MI. They provide an ideal and novel delivery strategy that has the ability to overcome the drawbacks associated with current treatments. In recent years, numerous injectable hydrogels have been developed and many of them have been tried for application in cardiac repair after MI. A number of reviews have also been published on the subject. However, to date, the clinical application of injectable hydrogels in repair or regeneration of post‐MI‐heart has not been realized. In this article, we present a comprehensive overview of various injectable hydrogels for potential application in cardiac repair or regeneration. A brief overview of the pathology of MI has been presented in addition to the current clinical treatments and their limitations. The suitability of injectable hydrogels for cardiac regeneration has been discussed, along with other key important factors that need to be considered when designing the best hydrogel, the delivery method, and the therapeutics to be delivered. Finally, recent advancements in the field have been discussed followed by discussions on future perspectives for addressing the challenges hindering the widespread clinical application of injectable hydrogels in treatment of injured cardiac tissues. 2 Myocardial Infarction and Cardiac Tissue Damage MI can be caused by a number of cardiac pathologies such as hypertension, blocked coronary arteries, and valvular heart diseases, thereby leading to ischemic cardiac injury. 2 Death of cardiac cells, or myocardial necrosis, can occur as a result of insufficient blood flow which results in reduced supply of oxygen to the infarcted tissue. 5, 6 Infarction disrupts the collagen fiber connections between cardiomyocytes and weakens the extracellular matrix which results in thinning and dilation of the ventricular wall. Abnormal stresses result in the healthy myocardium 7 in an attempt to counterbalance the reduction in the cardiac function. 8 This causes changes in the shape, structure, and functionalities of the heart and stimulates the body to respond to these changes and to remodel the injured ventricles 2 leading to a cascade of consequences. Cytokines and growth factors are released to stimulate the immune systems in order to clean the infarcted region. 2, 5 Fibroblasts, endothelial cells, and stem/progenitor cells at the infarcted zone each play important roles in the formation of granulation tissue, which is replaced by the extracellular matrix to form scar tissue. The newly generated scar tissue lacks the contractile properties needed for the heart to pump blood efficiently, which subsequently leads to heart failure. 3 Available Treatments to Repair Cardiac Damage after MI: Motivation for Use of Injectable Hydrogels Three available treatments used for the remedy of cardiac tissue damage include pharmacological treatments, medical devices with interventional therapies including ventricular assist devices, and heart transplantation. 2, 9 Medications are used to alleviate pain, reduce the cardiac workload and cardiac demand, and defend the heart from internally stored toxic substances. Medical devices and surgical therapies are used for various purposes such as restoring blood flow to the heart or reducing the generated stress with metallic and drug eluting stents or bypass auto grafting. 8 These devices along with the interventional therapies optimize the functions of remaining viable cardiomyocytes, whereas transplantation replaces the damaged nonfunctioning tissues. The two techniques, namely, pharmacological treatments and medical devices with interventional therapies cannot sufficiently manage the progression of the disease and hence heart transplantation remains the only effective treatment 10 which replaces the infarcted heart with a healthy heart from a donor. 11 With an increasing number of patients, reduced number of donors and the immune complications that develop after heart replacement, transplantation remains an inadequate and inefficient technique. 10, 12 Therefore, new methods are needed to tackle the issue more effectively. Cardiac repair using tissue engineering and regenerative medicine appears as a promising solution for cardiac repair after MI. 6, 13 In both tissue engineering and regenerative medicine approaches, the repair of the cardiac tissue can be performed by transplantation of healthy living cells and/or suitable biomaterials into damaged cardiac environments. In regenerative medicine approach, the appropriate cell types are injected to the injured myocardium. These cells can be progenitor cells or stem cells that can be differentiated into the necessary cardiac cell types 5, 9 or even fully differentiated cardiac cells. In one of the approaches of the regenerative medicine, some biomaterials‐alone can be injected near the infarcted heart tissue to provide the mechanical support to the injured heart tissue. In tissue engineering, the desired cell types are grown in some natural or synthetic biomaterials or sometimes even without any biomaterials, for the in vitro formation of cardiac tissue that could be implanted around the damaged areas of heart. 3 By using these two techniques, different methodologies can be applied including cell delivery, cell‐encapsulated hydrogel delivery, hydrogels delivery with or without some biomolecules, and hydrogel delivery with both cells and biomolecules. 9, 14, 15, 16, 17, 18 While in case of cell delivery without hydrogel, the low yield of cell retention at the targeted site is a concern, in the case of cell encapsulated hydrogel delivery, the hydrogel provides a matrix of support for the cells to adhere to prior to being injected into the body. Therefore, injectable hydrogels can overcome the problems of the lack of cell retention observed when cells have been injected by themselves and therefore they can produce more effective cardiac regeneration. 19 In regards to delivering biomolecules or drugs, hydrogels can be used to encapsulate these substances, and be delivered locally to the targeted sites allowing to control the release kinetics and provide optimal sustained release of the molecules such as growth factors, chemokines and DNA plasmids that promote anti‐apoptosis, angiogenesis, and endogenous cell recruitment while preventing detrimental systemic effects. 7 4 Injectable Hydrogels Hydrogels are “water‐swollen polymer networks” 7 that have a high percentage of water content identical to human tissues. They can be injected as a liquid and be crosslinked to a gel phase using certain physical or chemical stimuli. Alternatively, they can be injected in a partially crosslinked gel form as well. The gel formation after injection allows introducing the material inside the body in a minimally invasive way and permits the addition of bioactive molecules before the injection. Thermo‐sensitive hydrogels are specifically designed to initiate gelation at body temperature. Therefore, once injected, the transition from liquid to gel can occur. 20, 21 Other methods of in situ cross‐linking can be photo cross‐linking, pH‐dependent crosslinking, or ionic crosslinking. Some hydrogels exhibit structures similar to that of the extracellular matrices, so once injected in the defected location, they can enhance the formation of a new extracellular matrix and improve integration within the body. 22, 23 The design parameters during the synthesis of hydrogels play an important role in determining their properties and behavior. These design parameters need to be chosen based on the key properties that the hydrogel must exhibit and, therefore, must be ultimately based on the designated application of the hydrogel. 4. 1 Design Parameters in Synthesis of Hydrogels In order to design the optimal hydrogel, there are several parameters that need to be considered including the physical, material, and biological characteristics. In other words, the main factors that are essential to consider in the development of hydrogels in tissue engineering are the physical and material properties, primarily related to hydrogel mechanics, and the biological properties, involved in cell adhesion, for instance. Ideally, injectable hydrogels should be biocompatible, biodegradable, and bioresorbable to prevent triggering an immune response. 20 The engineered tissue constructs, obtained using hydrogel, for cardiac implantation should be vigorous, contractible, and elastic with the ability to sustain periodic contraction and relaxation. They should also be well‐vascularized so that the cells encapsulated in them receive enough nutrients. 10, 24 While natural polymers generally present more biocompatible features than synthetic polymers, the latter has greater varieties of structure, composition, and properties particularly better mechanical properties, which suggests a need for careful consideration of design parameters for the synthesis of hydrogels for cardiac applications. 25, 26 While the cross‐linking density can be more easily controlled in covalently linked hydrogels, the formation of cross‐linking in ionic cross‐linking‐based hydrogel is controlled using multivalent counter‐ions. 27 For example, the cross‐linking of hydrogels in situ with mild temperature fluctuation in polymerization is notably adapted in orthopedic applications. 28 Maintaining an area for tissue growth, adhesion, and gene representation of cells is of paramount importance in the design of polymers. Scaffolds, cross‐linking types, density, and polymer‐chain rigidity are all important factors for consideration, while evaluating the mechanical properties of hydrogels. 29 The design of the controlled degradation rate of hydrogels should take both hydrolysis and enzymatic reactions into consideration. 30 Moreover, cell–hydrogel interaction contributes to adhesion which may be dependent on cell type, receptor–ligand interaction, and differentiation. 31 4. 2 Delivery Routes for Injectable Hydrogels Along with investigating the proper hydrogels, cells, biomolecules, and drugs to promote the greatest amount of cardiac repair and regenerations, the optimal delivery pathway must also be established. When considering which pathway is the best, it is important to consider which methods produce the best results and are clinically viable. The main objective of any route of administration of therapeutics is to have optimal amount of active drugs delivered to the target site creating least risk to the patient. An important feature of injectable hydrogel is that it can be directly injected intramyocardially to the site of interest which potentially allows a minimally invasive treatment procedure with shorter hospitalization time. Intramyocardial delivery involves injections of drugs or cells directly into the myocardium, usually into the left ventricle (LV) using an epicardial method or a catheter technique. While epicardial injections are reliable, this approach is invasive, limits access to the septum, risks puncturing the LV, and can lead to systemic embolization. The catheter‐based approach achieves better retention and avoids local toxicity effects but is also subject to the drawbacks of the epicardial method. Intravenous delivery is a low‐risk procedure and is minimally invasive but shows low cell retention and relies profoundly on cell homing. 32 Intraperitoneal injections are tough to perform and injection sites are difficult to monitor. 33 Disadvantages to nasal systemic drug delivery include variable amount of drug absorption, upper airway infections, and risk of long‐term damage to nasal epithelium. 34 The use of locally injectable hydrogels has clear advantages over these other methods. First, the direct placement of material to damaged tissue makes the treatment pinpointed. Also, no surgery is needed using trans‐endocardial methods. Rather, electrophysiological guidance can be used to direct the injections. While controlling the amount of injected material is difficult and cell mortality during injection is inevitable, new strategies have been explored to mitigate these shortcomings including vacuum stabilization of the treated area. 35 The feasibility of local injection methods coupled with the accurate and controlled delivery of therapies for damaged tissue makes this a desirable technique for MI treatment and requires further investigation. 4. 3 Types and Compositions of Injectable Hydrogels Injectable hydrogels can be made from a vast range of biomaterials that can be classified as either natural or synthetic in origin. Both natural and synthetic biomaterials present strengths and weaknesses in terms of biomedical applications that must be considered prior to the hydrogel synthesis. They can also be divided into two types depending on the type of cross‐linking: either a chemical cross linking type, linked by covalent bonds or a physical cross linking type, physically connected by combining polymeric chains and nanoparticles. 20 Some of these hydrogels both natural and synthetic types are discussed below. 4. 3. 1 Natural Polymer Based Hydrogels Naturally derived hydrogels are usually biocompatible and supportive of cellular activities. However, they have low mechanical strength, may induce an immune response, and are subject to batch‐to‐batch variations. Furthermore, structural modification is difficult due to their structural fragility and complexity, 36 ] Figure 1 a. Despite these drawbacks, naturally derived biomaterials continue to be a promising component of hydrogels due to their bioactivity and the fact that many of them are naturally present in the human body. This correlation can be extremely valuable because one of the major goals of hydrogel synthesis is to produce tissues that are analogous to the native tissues. The characteristics of specific natural biomaterials are discussed below as well as how these characteristics pertain to injectable hydrogels. Figure 1 Chemical structures of various natural and synthetic hydrogels that have been tried or have the potential for application as injectable hydrogels in cardiac tissue regeneration. a) Some natural hydrogels: i) collagen, 134 ii) gelatine, 135 iii) hyaluronic acid, 135 iv) alginate, 136 v) agarose, 137 vi) chitosan, 138 and vii) keratin. b) Some synthetic hydrogels: i) polyacrylic acid, ii) poly(ethylene oxide), iii) polyvinyl alcohol, iv) polyphosphazene, v) polypeptide chains. Collagen : Collagen is a natural polymer; present in the extracellular matrix of skin, bone, tendons, ligaments, and cartilage of mammalian tissues, Figure 1 a(i). It is composed of combinations of amino acid sequences that are biocompatible in terms of cell recognition and are amenable to enzymatic degradation in the presence of collagenases. The advantages of using collagen include its biocompatibility, strong cellular activities, and thermal reversibility. The disadvantages of using collagen for hydrogel synthesis include its low physical strength or inferior mechanical properties, high synthesis cost, and inherent immunogenic responses. In order to improve the physical properties of collagens, chemical cross‐linking using glutaraldehyde 37 or diphenylphosphorylazide 38 is recommended. Collagen has been widely used as a scaffold for 3D cell cultures and tissue engineering of artificial skin. The cell–collagen attachment mechanisms can be controlled through chemical modification by incorporating fibronectin, chondroitin sulfate, or low levels of hyaluronic acid into the collagen matrix. 39 Collagen has been successfully used as a natural polymer based hydrogel to repair MI in vivo. Dai et al. investigated the use of collagen hydrogels (named Zyderm from INAMED Corp which consisted of 95% collagen type I and 5% of collagen type III) into rat MI model. 40 Results showed improved stroke volume (SV), ejection fraction (EF), and wall thickness in collagen hydrogel group compared to the controls. The authors claimed that the high density and concentration of injected collagen provided the beneficial outcomes, suggesting that collagen hydrogels with proper optimization are a promising biomaterial for cardiac tissue repair. Gelatin : Gelatin, Figure 1 a(ii), is formed by decomposing the collagen triple‐helix structure into single strand molecules. Furthermore, the preparation of gelatin is distinguished by the postbreakage treatment of the collagen structure. Acidic treatment yields gelatin of type A, while alkaline treatment, characterized by high carboxylic content, yields gelatin of type B. 41 Altering the solution temperature leads to the formation of gelatin that is renowned for its high biocompatibility and simplicity. Vascularization in tissue engineering is promoted through gelatin gels that present growth factors for delivery. 41 Being a natural polymer and a derivative of collagen, gelatin also has a high potential for application in cardiac repair after MI. Hyaluronic Acid : Hyaluronate, a component of glycosaminoglycan, is formed into hydrogels by covalent cross‐linking with hydrazine derivatives 42 along with the radical polymerization of glycidyl methacrylate, 43 Figure 1 a(iii). Hyaluronidase, a primary enzyme found in cells and serum, plays roles in degradation of hyaluronate. The application of hyaluronate has been investigated in various tissue engineering applications such as wound healing and development of engineered skin tissues and intradermal implants. 44 One major disadvantage of hyaluronate, similar to other natural hydrogels, is its weak mechanical properties that hinder the scope of its applications. However, the properties can be improved or controlled by modifying the molecular structure and composition with various functionalization. In a recent study, a methacrylated hyaluronic acid (MeHA) macromere crosslinked with tetramethylethylenediamine (TEMED) and ammonium persulfate (APS) was applied in an MI model to ligate the descending and diagonal coronary artery. 45 The results showed significantly reduced infarct expansion and improved cardiac function. In another study, hyaluronic acid was crosslinked with PEG‐SH 2 after acrylation and applied as an injectable hydrogel at epicardial surface of infarcted site. 46 It was found that the modified hyaluronic acid hydrogel provided a significantly higher EF and SV index reduced infarct size, increased wall thickness, and better vessel formation compared to those of MI control group, suggesting that proper modification and optimization of hyaluronic acid hydrogels can offer promising solution for cardiac tissue repair after MI. Fibrin : Fibrin gels are formed at ambient temperatures by polymerization of fibrinogen with thrombin as a crosslinking agent. 47 Cell‐associated enzymatic activity promotes the degradation of fibrin during cell migration, the rate of which is controlled by apronitin, a proteinase inhibitor. 48 Matrix synthesis along with cell migration and proliferation is promoted with the use of fibrin gels in a mechanism yielding the incorporation of platelet‐derived growth factors along with the transforming growth factor I (TGF‐ I ). 49 Incorporation of the domain peptides (with a factor XIIIa substrate in one domain and a bioactive peptide containing RGD sequence in another domain) into fibrin gels, by the action of transglutaminase during the process of coagulation, constitutes the gateway for a variety of neurological applications of fibrin. 50 In multiple studies, injection of fibrin into the infarcted area of rat left ventricle resulted in increased cell transplant and survival, decreased infarct size, increased blood flow to ischemic myocardium, and improved cardiac function. 51, 52 Fibrin joins the ranks of hyaluronate with its mechanical properties limitations, reducing its scope of applications in the tissue‐engineering field. Natural wound healing and 3D scaffolds for tissue engineering are common applications of fibrin. Its properties in cell adhesion along with its capability to synthesize from the host's blood without triggering any inflammatory responses make fibrin highly attractive. Alginate : Alginate, also known as alginic acid or algin, hydrogels are anionic polysaccharides, extracted from the cell walls of a brown algae. They are primarily used in drug delivery for wound healing, due to their high biocompatibility, low cost, and simple gelation with Ca 2+, Mg 2+, Ba 2+, and Sr 2+, 53 Figure 1 a(iv). The transplantation of chondrocytes, hepatocytes, and islets of Langerhans has been used with alginate gel beads. 54 One main disadvantage of alginate hydrogel is its release of divalent ions to surrounding, resulting in limited long‐term stability. This mechanism can be counteracted with covalent cross‐linking using a variety of molecules which produce different cross‐linking densities. 30 Another solution involves the isolation of polyguluronate blocks from alginate and subsequent oxidation and covalent cross‐linking of these derivatives with adipic acid dihydrazide. 30 Oxidization of alginate leads to its degradation in aqueous media with temperature and pH being control factors. 55 The hydrophilicity of alginate constitutes a disadvantage in terms of protein absorption. However, ligand‐specific binding properties can be enhanced by modifying with lectin or other ligands. When covalently coupled with an RGD containing cell adhesion ligand, the adhesion and proliferation of differentiated phenotypes of skeletal muscle cells were significantly enhanced. 56 Alginate hydrogel has also been tested in cardiac applications for treatment after MI, due to its nonthrombogenic properties. 57, 58 In another study, a calcium crosslinked alginate hydrogel was tested in rat MI model and was found to be significantly effective in reducing left ventricular (LV) expansion and increasing the old‐infarcted heart wall thickness. 59 Similarly, alginate also showed promising results in large animal models such as a porcine model for treatment of MI. 60 Consequently, alginate has become the very first injectable material to enter clinical trials for the treatment of MI ( Ikara Holdings, Inc. IK‐5001 for the Prevention of Remodeling of the Ventricle and Congestive Heart Failure After Acute Myocardial Infarction. NCT01226563; 2010 ). Agarose : Agarose, Figure 1 a(v), like alginate, is a marine extracted algal polysaccharide except that it can form thermally reversible gels in a structure composed of double‐helices and junction zones consisting of multiple chain aggregations. 61 Modifying the physical structure of agarose by controlling the concentration can alter the pore size of the gel. By combining the properties of large pores and minimal stiffness at low concentrations cell migration and proliferation can be enhanced. 61 For an enhanced cell interaction mechanism, cell adhesion peptides (CDPGYIGSR) have also been covalently coupled with agarose. 62 In a recent study, agarose was used for encapsulation of cardiac stem cells (CSCs) and their delivery to an infarcted heart which resulted in enhanced cardiac repair and blood vessel formation. 63 Chitosan : Chitosan hydrogel is prepared by N ‐deacetylation of chitin and is characterized by high biocompatibility and low toxicity due to its structural similarity to natural glycosaminoglycans, Figure 1 a(vi). In addition, its solubility in an acidic environment, due to its high crystallinity, makes the application of chitosan highly effective. The hydrogel formation process of chitosan is performed by ionic 64 crosslinking or chemical cross‐linking with glutaraldehyde. Chitosan modification with sugar residues 65 and proteins such as collagen and albumin have yielded promising results in tissue‐engineering. 66 Chitosan‐based hydrogels can be used to protect the transplanted cells, promote angiogenesis, reduce infarct size, and improve cardiac function due to their ability to conjugate with various bioactive molecules. This can increase their bioactive functions without affecting their physical/chemical properties, such as gelation behaviors, swelling and degradation characteristics, network structure, and mechanical strength. 67 In a recent study, Lu et al. used chitosan with β‐glycerol phosphate and hydroxyethyl cellulose to synthesize a temperature‐sensitive hydrogel and applied in a rat MI model. 68 A significant improvement was found in infarct size, wall thickness, microvessel, end systolic diameter (ESD), and end diastolic diameter (EDD). Two other independent studies also reported similar results for the thermo‐responsive chitosan hydrogel in rat MI model, demonstrating its high effectiveness in cardiac tissue repair. 69, 70 Keratin : Keratin, Figure 1 a(vii), represents a broad category of fibrous structural proteins that make up human skin, hair, and nails. Monomerically, keratin forms intermediate filaments which subsequently form these macrostructures. Keratin has the ability to self‐assemble into fibrous scaffolds. 71 This creates an ideal matrix allowing for cellular permeation and proliferation. The wide availability of keratin makes it a desirable natural resource of biomaterials. More than 30 different cytokines and factors residual from hair morphogenesis might be useful in cardiac repair, augmenting myocyte cell viability, migration, and gene expression. 72 Three of these factors, namely, NGF, TGF‐ β1, and BMP4, were found to support angiogenesis. Also, keratin injections showed no evidence of inflammation, and keratin‐based scaffolds were found to regenerate nerve function and foster neuromuscular recovery. 71 Hydrogels made from keratin are notably biocompatible, have the proclivity to self‐assemble, and efficiently integrate into environments. Matrigel : Matrigel is a gelatinous protein mixture, secreted by Engelbreth–Holm–Swarm (EHS) mouse sarcoma cells, and commercially available under different trade names marketed by multiple companies. It is usually used as substrate coating to improve cell adhesion; however, it can also be used as an injectable hydrogel for cardiac repair. Kofidis et al. employed matrigel for delivery of embryonic stem cells to an infarcted myocardium, where the cells were encapsulated in the matrigel injected in the infarcted left ventricle. 73, 74 Their study proved that injectable matrigel may help in fixing the heart's shape, geometry, and functionality 74 after MI. Similarly, Zhang et al. showed that a mixture of collagen, matrigel, and cells as a source of cardiomyocytes offered the same functionalities once injected in vitro, as Kofidis and his colleagues demonstrated in their later research. 75 Furthermore, matrigel can inhibit cell death and enhance vascularization after being introduced to the infarcted region. 76 Decellularized ECM : In order to better mimic the composition of the native cellular microenvironment a number of groups have investigated the use of hydrogels from decellularized tissues or extracellular matrices (ECMs). ECM hydrogels replicate the native cellular microenvironment with specific characteristics based on the tissue from which they are extracted. They can provide polymeric hydrogels with native chemical and biophysical cues and therefore improve the attachment, survival and function of the transplanted cells. Besides, each tissue has a distinct composition of fibrous proteins, proteoglycans, and glycosaminoglycans that make up its ECM. 77 Recent studies have shown that ECM hydrogels made from decellularized cardiac tissues can help in repairing scarred myocardium after MI. Injection of ECM hydrogel from the myocardial matrix, into rats LV showed no embolization or ischemia, and no adverse effect on neighboring tissues or cardiac rhythm. 78 Additionally, when the interaction with human blood was evaluated, the decellularized matrix had no adverse effect on activated partial thromboplastin time or prothrombin time across different blood concentrations, verifying its hemocompatibility. 4. 3. 2 Synthetic Hydrogels Synthetic hydrogels are of particular interest in cardiac repair and other tissue engineering applications due to their strong mechanical properties and easily controllable features, 79 Figure 1 b. Here, we discuss various synthetic hydrogels which have been used in tissue engineering or regenerative medicine and have potential for application in cardiac repair. Poly(acrylic acid) Derivatives : Poly‐acrylic acid (PAA), Figure 1 b(i), has several derivatives that are used as biomaterials. Poly‐2‐hydroxyethyl methacrylate, also known as poly‐HEMA, is hydrolytically stable. The cross‐linking mediator can predict its permeability and hydrophilicity. The disadvantage of PAA‐derivatives is that they are not completely biodegradable. 4 A dextran‐modified poly‐HEMA has been developed that is found to be degradable by specific enzymes. Attaching the poly‐HEMA to oligo‐ d ‐lactide to form a gel phase of the poly‐HEMA without using lethal chemicals, have also been reported. 80 Another important derivative of PAA, worth mentioning, is poly‐ N ‐isopropyl‐acrylamide, or PNIPAAm, which is commonly used in tissue engineering for its temperature‐dependent biphasic behavior. It can change its phase from liquid to gel‐phase above its lower critical solution temperature (32 °C in water). Thus, it can be injected into the body as a prepolymer solution which can turn to a gel‐like phase at body temperature after injection, 4 which makes this polymer advantageous. The transition depends only on the body's temperature and is not based on a specific timing which makes them even more beneficial. On the other hand, these polymers are nondegradable and may be toxic, teratogenic, and/or carcinogenic. 4 In an earlier study, a biodegradable PNIPAAm hydrogel was injected in infarct heart and was found to result in improvement of heart function. 81 Similarly, in another report, PNIPAAm with a growth factor and antioxidants showed significant enhancement of MSC growth within the hydrogel providing a suitable microenvironment for heart cells to function. 82 Although the synthetic poly(acrylic acid) derivatives have promising potential for application in the treatment of cardiac tissue damage, their biocompatibility and inflammatory response may arise which are needed to be resolved for their extensive application in the future. Polyethylene Glycol or Polyethylene Oxide and Copolymers : Polyethylene glycol (PEG) or polyethylene oxide (PEO), Figure 1 b(ii), is a biocompatible and hydrophilic polymer which is also FDA approved for several biomedical applications. PEG is specifically used to prepare biological conjugates, 83 modify surfaces, 84 and stimulate cell membrane synthesis. 4 It does not cause immune responses and has low cell adhesion and protein binding. 85, 86 It is produced from the ionic polymerization of ethylene oxide and has end hydroxyl groups that facilitate the formation of PEG macromers which in turn contribute to its chain polymerization. The PEG macromers can be injected in the injured sites because of their low toxicity. PEG in its gel phase is usually used as a scaffold and it is coated to inhibit interactions between encapsulated cells and the polymer itself. 87, 88, 89, 90 The gel phase can be manufactured via UV photo‐polymerization of the polyethylene oxide with acrylate ends in the presence of R‐hydroxy acid. 91 When interaction with proteins is needed, the hydrogel can be reinforced with bioactive peptides 3 a property required for biomaterials used in tissue engineering and regenerative medicine. 92 In addition, copolymers of the PEG have been found to be useful in drug delivery. 93, 94 For example, PEG‐PPO‐PEG is an interesting triblock copolymer made of PEG and polypropylene oxide (PPO) which is a thermo‐sensitive polymer that can change to a gel phase at body temperature. 95 The drawback is that it is not biodegradable. To overcome this issue, a diblock (or triblock) of PEG and polylactic acid (PLA) was developed based on the knowledge that the PLA is biodegradable and harmless. 96 In an earlier study, PEG was used for understanding the cardiomyocytes–matrix interactions in a 3D microenvironment. The results showed an increase in viability and functionality of encapsulated cardiomyocytes. 97 In another study, the embryonic stem cells (ESCs)‐encapsulated PEG hydrogels demonstrated ESCs differentiation to the cardiomyocytes lineage and the cells showed cardiac‐like activity. 98 Wang et al. injected bone marrow‐derived MSC‐encapsulated PEG‐PCL‐PEG copolymer mixed with cyclodextrin in the MI region of a rabbit heart and observed formation of dense vessel network at the infarct site with significantly reduced cardiac infarction. 99 All of these studies demonstrated the superior activities of PEG copolymers in the treatment of MI. Polyvinyl Alcohol : Polyvinyl alcohol, Figure 1 b(iii), or PVA, is produced through the hydrolysis of polyvinyl acetate. The PVA hydrogel is developed using either physical or chemical cross‐linking. Chemical cross‐linking can potentially be harmful, so some recent studies have focused on replacing the chemical cross‐linking with photocrosslinking. The solubility and hydrophilicity of manufactured hydrogel depend on the molecular weight and the degree of hydrolysis. PVA can connect to multiple biological molecules. 3, 4 It can be used as a matrix due to its elasticity and role in improving the diffusion of mechanical signals. PVA is a neutral hydrogel and consequently its adhesion properties are relatively minimal but can be enhanced by mixing it with biological factors. PVA hydrogel has strong mechanical properties and a low rubbing coefficient. 3 All these excellent properties of PVA exhibit promising tissue engineering applications, especially cardiac tissue repair. Polyphosphazene : Polyphosphazene, Figure 1 b(iv), has attracted scientists' interest because of its biodegradability. By modifying its side‐chain organization the dynamics behind its degradation process can be regulated. The polyphosphazene polymer includes interchanging atoms of phosphorus and nitrogen and two side groups connected to each phosphorus atom. A transitional product, poly‐dichlorophosphazene, is formed before creating the final polymer. The gel phase of the end product, polyphosphazene, is a result of its hydrophilic backbone and its flexibility acquired following different substitution reactions. It can only be modified to create a thermo‐responsive hydrogel. From the polyphosphazenes, nonionic and ionic hydrogels can be prepared. The latter reacts with variation in pH or ionic strength and consequently they might be useful for protein drug delivery. 4 Polypeptides : Peptides, Figure 1 b(v) can form nanofiber hydrogels by means of charge interactions, hydrogen bonding, and other interactions. 100 The use of peptides is very beneficial since they can act as a scaffold that mimics a natural extracellular matrix (ECM). 101 Different types of peptide hydrogels can be formed. Thermo‐responsive peptide hydrogels have been made of elastin‐like proteins that contain tropoelastin. 102, 103 The ion‐induced cross‐linked peptide hydrogels change to a gel‐like structure based on interactions with a concentration of salt, which will increase the ion strength, specifically by minimizing the repulsion occurring in the positively charged side‐chain. 104 Some peptides can form pH‐responsive hydrogels. Two amino acids are essentially responsible for this activity: valine and lysine. In liquid phase, hydrophobic valine and hydrophilic lysine are on opposite sides of the peptide's secondary structure and this combination creates hydrogen bonds within the structure. As the pH decreases, the valine and adjacent lysine repel each other. As a result, the peptide structure unfolds and the hydrogel dissolves. Some peptides upon being attached to the fluorenyl‐methoxy‐carbonyl can form a gel‐like hydrogel with enhanced cell adhesion and proliferation properties. 79 4. 3. 3 Composite Hydrogels In addition to various natural and synthetic polymer‐based hydrogels, a number of hybrid and composite hydrogels have also been tried for application in cardiac repair. Some specific examples include ECM‐fibrin hydrogels, 105 alginate‐chitosan hydrogels, 106 and ECM‐polyethylene glycol hydrogels. 107 Due to the benefits of both fibrin and alginate as biomaterials for injectable hydrogels, a composite of these materials was used for cardiac repair after MI. Results showed that the expansion of the infarcted region of the left ventricle stopped upon injection of the hydrogel. Additionally, myocardial stiffness was relatively higher than those of control groups and fibrous collagen in the myocardium border region was increased. 108 Hence, the soluble collagen content in the infarcted zone was decreased, which accounts for the deceleration of damaged tissue enlargement. Thus, injection of fibrin‐alginate as well as other suitable composite hydrogels may provide an effective treatment to suppress the expansion of infarcted tissue. A recent strategy for achieving multiple functionalities in hydrogels is to incorporate various nanoparticles in them, thereby developing nanocomposite hydrogels. The nanoparticles to be incorporated can be polymeric, metallic, ceramic, inorganic, or carbon‐based as shown in Figure 2 a 109 while the strategies of incorporation can be physical entrapment, noncovalent immobilization, or covalent immobilization as shown in Figure 2 b. 110 The different strategies for the delivery of injectable composite hydrogels as well as other injectable hydrogels including acellular hydrogel alone, acellular hydrogel with some biomolecules, and the cell‐laden hydrogel delivery are shown in Figure 2 c. 76 Recent advances in the applications of these hydrogels in cardiac repair and regeneration with specific examples are discussed in the section below. Figure 2 Schematic representation of nanoparticle loaded injectable composite hydrogels, and delivery of various hydrogels including acellular hydrogels alone, acellular hydrogels with various biomolecules and hydrogels with various cells. a) A range of nanoparticles such as polymeric nanoparticles, metallic‐metal oxide based nanoparticles, inorganic nanoparticles, and carbon‐based nanomaterials can be incorporated in hydrogels to make nanocomposite hydrogels. Adapted with permission. 109 b) The common strategies for incorporation of nanoparticles in hydrogels: i) stabilization of inorganic or polymeric nanoparticles by nano–micro‐sized hydrogel particles, ii) noncovalently immobilized nanoparticles in a hydrogel matrix, and iii) covalently immobilized nanoparticles in hydrogel matrix. Adapted with permission. 110 c) Different strategies of injectable hydrogel delivery for treatment of MI. The injectable hydrogels can be delivered alone without any biomolecules or cells, with some biomolecules as carriers or with cells as a 3D matrix. Adapted with permission. 76 Copyright 2010, Royal Society of Chemistry. 5 Recent Advances in Application of Injectable Hydrogels in Cardiac Tissue Repair While numerous hydrogels have been synthesized for tissue engineering, only a few of them have been tested for cardiac applications. However, some of these hydrogels have shown strong promises in enhancing the repair of infarcted myocardium. Acellular injectable hydrogels without any therapeutic biomolecules were found to thicken the myocardial wall, thereby reducing abnormal stresses when injected directly after an infarction. 7, 13 Other reported improvements were enhanced cardiac function, reduced infarct size, and induced neovascularization, using injectable hydrogels as carriers for various biomolecules 111 and as carriers of cells. 13, 77, 112, 113, 114 In subsequent sections, we discuss the recent advances in applications of injectable hydrogels in cardiac tissue repair. 5. 1 Acellular Hydrogels for Treatment of MI As mentioned earlier, acellular hydrogels have exhibited promising results in cardiac repair as both a bulking agent to provide mechanical support to the infarcted heart when injected alone, and as a carrier for various biomolecules including growth factors, cytokines, and DNA plasmids. Several scientists have also focused on acellular biomimetic hydrogels without biomolecules in order to focus on materials that give the proper biological and chemical cues that mimic the native microenvironment. For example, in a recent study, an injectable ECM hydrogel derived from decellularized porcine myocardial tissue was used to reverse the negative remodeling process in infarcted myocardial tissue. 78 Upon injecting the hydrogel into infarcted pig hearts, the tissue self‐assembled into a porous scaffold, allowing cell infiltration. Histological characterization showed that the hearts that were injected with the ECM hydrogel developed a distinct layer of endocardium, while the control group exhibited a fibrillary layer and the saline‐injected control endocardium was moderately thickened, Figure 3 a–c. ECM hydrogel‐treated myocardial tissue exhibited 10% less collagen content compared to control groups, as well as the presence of cardiomyocytes as evidenced by cardiac troponin‐T staining, Figure 3 d. This reduction in fibrosis is essential for heart repair since fibrous tissue is noncontractile, which inhibits the heart's ability to pump blood properly. The tissue injected with the ECM hydrogel showed foci of neovascularization as well, whereas the control groups did not, Figure 3 g, h. This observation provided evidence of cardiac regeneration due to the fact that blood flow to the infarct zone provides oxygen essential for cardiomyocytes to grow. Figure 3 a–h) Enhancement of cardiac muscle and reduction of infarct fibrosis using myocardial matrix. Histological characterization of infarcted pig hearts: Masson's trichrome staining images are representative of six matrix‐injected pigs and four control animals. a) Matrix‐injected hearts had a distinct, thick endocardium (red stained muscle, indicated with an asterisk). b) Noninjected control animals had a loose fibrillar layer (blue) beneath the endothelium. c) In saline injected control animals, the endocardium was moderately thickened with minimal muscle (red). d) An adjacent tissue section for the matrix‐injected animal in a) stained for cardiac troponin‐T, indicating the presence of cardiomyocytes. e) Area of endocardial layer of muscle as a proportion of the infarct. f) Percentage of collagen content in the infarcts. Data are means ± SEM and were obtained from Masson's trichrome slides a–c). * P < 0. 05 (Student's t test). g, h) Matrix‐injected hearts contained foci of neovascularization in the area below the endocardium (g, arrows), but none of the saline or noninjected control hearts showed these areas of neovascularization h). Scale bar, 200 mm. i) Myocardial matrix is biocompatible and biodegradable. Representative histological sections of cell infiltration in matrix‐injected rat hearts injected with saline, PMM, or NDM at days 3, 14, and 28. Inflammation and multinucleated giant cells are present in the NDM groups at days 14 and 28 (arrows). The PMM (asterisk marked porous network) was completely degraded by 28 d. Scale bar, 200 mm. Reproduced with permission. 78 Copyright 2013, American Association for the Advancement of Science. Rodents were used to assess the biocompatibility and degradation properties of the myocardial ECM hydrogel, Figure 3 i. The ECM hydrogel exhibited complete degradation by 28 d postoperation. The group receiving nondecellularized porcine myocardial matrix (NDM) also showed similar degradation in addition to an immune response exhibited by inflammation and multinucleated giant cells. This further confirms that the ECM itself does not elicit an immune response, as long as the matrix is successfully decellularized. These positive results provide a base of support to move this ECM hydrogel toward clinical studies. The detailed study elicits how the components of ECM can be leveraged in engineering cardiac tissue. 5. 2 Injectable Cell‐Laden Hydrogels for Cardiac Repair Various injectable hydrogels, including tetronic‐fibrinogen (TF) and PEG‐fibrinogen (PF) conjugate hydrogels, have been used to enhance the efficacy of stem cell delivery in infarcted myocardium, 13, 112, 113, 114 Figure 4 a, b. In one study, a left thoracotomy was performed by antero‐lateral incision, followed by the injection of hydrogel precursor solution on four sides of the infarcted region. 13, 77 UV light was used to catalyze the crosslinking process. 13 Results showed that hydrogel treatment to the left ventricle (LV) led to an increase in wall thickness, which in turn increased the survival of viable cardiac tissue where infarction took place. 115, 116 Remodeling of the LV was revealed to some extent via echocardiography. The prevention of LV dilation and fractional shortening deterioration were observed in addition to minimal wall thinning. 23 Both cross‐linked PF and TE hydrogels with BMNCs showed increased neovascularization, leading to improvements in cardiac function. Injecting TF hydrogel and saline led to the same arteriole density, while PF 2% exceled the former by offering a greater restoration of heart function and neovascularization. 13, 117 Figure 4 Arterial staining and histological morphology of excised rat hearts treated with injected hydrogels 4 weeks post MI. Image (a) shows Alpha‐SMA positive vessels in the infarct and peri‐infarct zones, 4 weeks after MI. Panels (i) and (ii) show control (saline injection) peri‐infarct and infarct zone respectively, (iii) and (iv) show peri‐infarct and infarct treated with PF1%, (v) and (vi) show those treated with PF2% respectively, (vii) and (viii) treated with TF1%, respectively, and (ix) and (x) treated with TF2%, respectively. In image (b) 4 weeks after the MI the treated rat hearts were cut into equal transverse slices to show the gross appearance and histological morphology (i) in that of the untreated saline control group where saline injection was only used. The PF1% hydrogel injection in (ii), the PF2% hydrogel injection in (iii), the TF1% hydrogel injection in (iv), and the TF2% hydrogel injection in (v). Reproduced with permission. 13 Copyright 2014, Elsevier. Image (c) shows immune‐histochemical staining for the smooth muscles for the four groups 30 d after MI. Panel (i) shows the control group, (ii) shows hydrogel group, (iii) shows BMMNC group, and (iv) shows BMMNC + hydrogel group. Reproduced with permission. 124 Image (d) shows the staining with Masson trichrome of the infarcted wall forcollagen (green) and muscle (red) of the four groups. (i) PBS‐only group of the four groups; (ii) OPF‐only group; (iii) PBS + ESC group; and (iv) OPF + ESC group. Reproduced with permission. 19 Copyright 2014, Elsevier. Image (e) also shows staining with Masson's trichrome of the infarct area and shows similar level of tissue improvement with BADSCs treatment alone or Chitosan alone compared to the control group with Chitosan hydrogel delivery of BADSCs resulting in the greatest healing and infarct reduction. Reproduced with permission. 124 In another study, bone marrow derived stem cells (BMSCs) from rabbit tibia were encapsulated in Dex‐PCL‐HEMA/PNIPAAm hydrogels and injected into their infarcted myocardium after performing left thoracotomy and ligating the left coronary artery. One month later, the percent infarcted size was obtained from the ratio of infarcted wall to all the surface area of left ventricle (LV). 77, 118 The BMMNCs survived up to 3 d in hydrogel. Notably, the cell density in the animal group injected with cell‐laden hydrogel was much higher compared to that in the groups with either cells or hydrogels alone. After 48 h there was no significant difference in the results between the four groups on the level of LV end‐systolic diameter (LVESD), LV end diastolic diameter (LVEDD), and LV ejection fraction (LVEF). However, the echocardiography results after 30 d demonstrated that LVEF of the group that received BMMNCs encapsulated in the hydrogel was more significant than the other groups. In contrast, this group had lower levels in LVEDD and LVESD. Thus, the injected hydrogels enriched with encapsulated cells increased cell engraftment, where hydrogel could serve as the ECM, leading to an improved rate of cell retention and survival in the infarcted zone. [[qv: 77, 119, 120]] In a third study, embryonic stem cells (ESCs) extracted from mice were encapsulated into an oligo [poly(ethylene glycol) furmarate] (OPF) hydrogel and injected at the border of the infarcted region 1 week after inducing MI in mouse hearts. After 24 h, the hearts were excised, frozen and bisected into five different sections, followed by tracing Green Fluorescent Protein (GFP)‐labeled mouse ESCs via fluorescent microscope. After 4 weeks, immunochemical staining was used to determine graft size in the specimens. 19 The synergistic effect of OPF hydrogel and ESCs was evident by the fact that ESCs in OFP differentiated into cardiomyocytes, ESC count was higher, the ESCs were aggregated in the hydrogel, LV function for mice with ESC and OPF showed the most improvement, infarction size diminished the most, collagen deposition was minimal, no signs of prolonged inflammatory response were observed, as well as a significant increase in neovascularization was exhibited. Also, the OPF hydrogel completely degraded. 19, 121, 122, 123 In a recent study, cardiac progenitor cells (CPC) were found to exist in the form of a CD29 positive population. 124 Chitosan hydrogel was tested with incorporation of these cells as a method to repair damaged myocardium. CD29+ brown adipose derived stem cells (BADSCs) were obtained from rats, and then rats were subdivided into four groups containing 20 rats each. Each group received one of the following treatments: PBS only, chitosan hydrogel, BADSCs only, and BADSCs encapsulated in chitosan hydrogel. These treatment groups were put into place after the left coronary artery was ligated. 124 Lateral investigations were conducted to identify the cardiac differentiation of BADSCs in vitro, where they were treated with and without chitosan, Figure 4 e. As a result, the treated groups showed significant favorable results. In detail, the cross‐linked strained myofilaments were better organized than the control group. Analysis showed significant increase in the levels of collagen 1 when using chitosan treated with BADSCs compared to the control group. In addition, applying chitosan hydrogel led to enhancement in cell retention in the infarcted site, leading to survival of the grafted cell in ischemic myocardium. 124 Furthermore, chitosan + BADSCs had a synergistic effect as they led to a significant recovery of cardiac function—significant reduction took place in the left ventricle infracted zone, which led to a marked increase in the number of cells responsible for cardiomyocytes differentiation, as well as neovascular formation. 124 These studies clearly demonstrate how encapsulating cells in hydrogels can increase their therapeutic efficacy in terms of positive cardiac remodeling. The material properties of the hydrogel can be tailored specifically to provide the cell source with the correct chemical, mechanical, and biological signals to differentiate into the desired lineage, such as cardiomyocytes. Furthermore, hydrogels provide the cells with a mechanically stable, biocompatible environment to protect the cells from being washed away by the body's defense mechanisms. Therefore, these tunable properties give hydrogels great potential to be an integral part of the success of cell therapy in treating myocardial infarction. 5. 3 Carbon Nanotube‐Embedded GelMA Hydrogels for Cardiac Constructs In the case of engineering heart tissue, scaffolds do not usually exhibit the capability to conduct electrical current. However, the native tissue contains purkinje fibers that do in fact conduct electricity, allowing the heart to beat. A recent study employed carbon nanotubes (CNT) embedded GelMA hydrogel with improved mechanical and electrical properties to enhance the differentiation of the mesenchymal stem cells (MSCs) into cardiac cells, Figure 5. Another recent study showed a promising approach for high‐performance cardiac scaffold materials that can be created with the incorporation of CNTs into a photo‐cross‐linkable GelMA hydrogel. 125 The cardiac cells in CNT‐GelMA were observed to be in an elongated state and F‐actin fibers were demonstrated to be intact and more homogeneous in comparison with pristine GelMA. Immunostaining showed a significant increase in both sarcomeric α‐actinin and Troponin I on the CNT‐GelMA in comparison with the pristine GelMA. Also, sarcomere alignment, sarcomere interconnected structures, and troponin aggregation were observed on the CNT‐GelMA. Both CNT‐GelMA and pristine GelMA can generate synchronous beating activity. However, after the spontaneous beating rates were recorded over a 6 d period, the CNT‐GelMA showed a beating rate average that was both more stable and three times greater than the pristine GelMA. In order to study the cardiac beating, two model drugs known for disrupting the biological processes necessary for continued heart beats were introduced: heptanol and doxorubicin. In the case of heptanol, the gap‐junctional beating propagation is inhibited in the heart. However, in the CNT‐GelMA, synchronous beating was observed even after the drug was administered. In fact, it took 40–65 min for synchronous beating to stop in the CNT‐GelMA, as opposed to only 20 min in the pristine GelMA. This demonstrated that the CNT in the hydrogel assisted in maintaining the pathway that the heptanol disables and enhanced beating amplitudes and rates. 125 Figure 5 a–d) Structural, physical, and electrical properties of CNT‐GelMA hydrogels: a) schematic diagram illustrating the isolated heart conduction systems showing the purkinje fibers, which are located in the inner ventricular walls of the heart. b) Preparation procedure of fractal‐like CNT networks embedded in GelMA hydrogel. c) TEM image of GelMA‐coated CNTs. d) SEM images show porous surfaces of a 1 mg mL −1 CNT‐GelMA thin film. e, f) Adhesion, maturation, alignment, and phenotype of cardiac cells on CNT‐GelMA hydrogels. e) Confocal images of cardiomyocytes after culturing for 5 d on pristine GelMA and 1 mg mL −1 CNT‐GelMA revealed more uniform cell distribution and partial cell alignment on CNT‐GelMA. Higher magnification images showed well‐elongated cardiac cells and well‐developed F‐actin cross‐striations (bottom right, white arrows) on CNT‐GelMA but not on pristine GelMA (bottom left). f) Immunostaining of sarcomeric R‐actinin (green), nuclei (blue), and Cx‐43 (red) revealed that cardiac tissues (8 d culture) on (i) pristine GelMA and (ii) CNT GelMA were phenotypically different. Partial uniaxial sarcomere alignment and interconnected sarcomeric structure with robust intercellular junctions were observed on CNT‐GelMA. Immunostaining of Troponin I (green) and nuclei (blue) showed much less and more aggregated Troponin I presence on (iii) pristine GelMA than on (iv) CNT‐GelMA. g–i) Improved mechanical integrity and advanced electrophysiological functions of cardiac tissues on CNT‐GelMA. g) Spontaneous beating rates of cardiac tissues recorded from day 3 to day 9 on a daily basis. h) CNTs protected cardiac tissues against damages by heptanol. Plots of spontaneous beating amplitude over time (5 d culture) for 0–5 mg mL −1 CNTs in GelMA in response to 4 × 10 −3 m heptanol. i) Time lapse before sporadic beating and stop of beating induced by heptanol (* p < 0. 05). Reproduced with permission. 125 Copyright 2013, American Chemical Society. 5. 4 Injectable Hydrogel‐Based Gene Delivery Systems Gene therapy offers a unique opportunity to promote local healing of damaged myocardial tissue mainly by promoting new blood vessel formation and attenuating fibrosis. Some methods that are frequently used in delivering specific therapeutic genes using injectable hydrogels include upregulating or knocking down specific host genes or simply by overexpressing certain genes of therapeutic interest. One recent study aimed to improve the biofunctionality of injectable hydrogels by incorporating a gene delivery system using polyethylenimine (PEI) functionalized graphene oxide nanosheets (fGO) complexed with vascular endothelial growth factor‐165 (VEGF), all incorporated within methacrylated gelatin (GelMA), Figure 6 a. 126 The addition of the fGO increased mechanical strength, reinforced the physical network of the composite hydrogels, and established strong binding of fGO to plasmid DNA for eventual transfection to the host tissue. Biocompatibility of the formulated hydrogel, evaluated using quantitative PCR and ELISA analysis of proinflammatory tumor necrosis factor α, was confirmed as the GO‐GelMA nanocomplex did not induce any cytotoxic and inflammatory effects. Figure 6 Injectable hydrogels for gene delivery applications. a) Photocrosslinkable hydrogel for myocardial delivery of vascular endothelial growth factor (VEGF) carrying gene using cationic functionalized graphene oxide (fGO) nanoparticles. Schematic of stepwise formulation process for direct intramyocardial injection of damaged heart with acute myocardial infarction. Reproduced with permission. 126 Copyright 2014, American Chemical Society. b) Delivery of siRNA using biopolymer hydrogel—schematic of hydrogel formation for delivery of siRNA and subsequent inhibition of gene expression in incorporated and neighboring cells. Biomaterial solutions of alginate, photo alginate, or collagen are mixed with siRNA and GFP‐positive cells, and hydrogels are then formed by crosslinking, photo‐crosslinking, or thermos‐gelling, respectively. The siRNA diffuses through the hydrogel to affect incorporated cells, and it is also released from the hydrogel to locally affect surrounding cells that are part of the host tissue. Reproduced with permission. 127 Copyright 2009, American Chemical Society. c) Hydrogel for delivery of recombinant viruses using nanohybrid complexes. Controlled release of baculovirus (Bac) using CNT reinforced hydrogel. TEM images of (i) nonfunctionalized CNT with Bac, (ii) CNT functionalized Bac with arrow showing the baculovirus bound to CNT surface in magnified image (iii). Scale bar indicates 100 nm length. Self‐assembled nanocomplex of cationic PAA functionalized CNTs (f‐CNT) were hybridized with anionic Bac. (iii) Cumulative release kinetics of baculovirus from denatured collagen hydrogel (2. 5 mg mL −1 ) impregnated with Bac–CNT nanocomplex (v–viii–G) rMSCs were overlaid on the collagen hydrogel formulated with CNT (25 μg mL −1 ), BacMGFP, CNT/BacMGFP or functionalized CNT f‐CNT/BacMGFP. Abbreviations: TEM = transmission electron microscope; CNT = carbon nanotube; rMSCs = rat mesenchymal stem cells; PAA = poly (acrylic acid), MGFP = Monster Green Fluorescent Protein. Reproduced with permission. 128 Copyright 2014, Elsevier. d) Thermo‐responsive hydrogel for myocardial delivery of plasmid DNA. Thermosensitive sol‐to‐gel transition properties of biodegradable dextran‐poly(e‐caprolactone)‐2‐hydroxylethylmethacrylate‐poly( N ‐isopropylacryl amide) (Dex‐PCL‐HEMA/PNIPAAm) hydrogel at 1. 5 wt% concentration: (i) gel solution in fluidity at room temperature, (ii) the gel solution turned into gel at 37 °C, and (iii) sol–gel reversibility of gel solution at room temperature. (iv) Representative pictures of left ventricles from each group after Masson's Trichrome staining 30 d after treatments. Scale bar = 1 mm. Below images are the infarct size as percentages at 30 d. Infarct size is calculated as the ratio of infarcted to noninfarcted area of the left ventricle. Reproduced with permission. 129 Another study used hydrogels to overcome the drawbacks associated with the use of short interfering RNA (siRNA) to control gene expression, Figure 6 b. 127 The mechanism behind siRNA‐based therapy is that the mRNA that is complementary to the siRNA is degraded and therefore after transcription, the gene expression process is inhibited as needed. This study used calcium cross‐linked alginate, photo‐cross‐linked alginate, and collagen biodegradable hydrogels to encapsulate both the cells and the siRNA responsible for knocking down their gene of interest, GFP. The siRNA present in the hydrogel had the ability to affect the incorporated cells as well as locally affect the surrounding cells in the host tissue. The degradation rates of the polymers chosen in this study greatly affected the results in silencing the gene of interest. Due to the fact that the alginate degrades quickly, this study confirms that by using the degradation rates of different polymers, a sustained local release of siRNA can be delivered. When using these techniques, the parameters of the specific application need to be taken into consideration to determine the most beneficial timeline of delivery. Viral gene delivery systems have also been shown to be effective for myocardial regeneration therapy, Figure 6 c. In this case, hydrogels were used to deliver CNT hybridized baculoviruses to rat bone marrow stem cells. 128 These hydrogels were made up of a denatured collagen matrix containing MGFP gene carrying baculoviruses wrapped in single‐walled carbon nanotubes. The hydrogel was shown to have greater mechanical properties and a more sustained delivery of the recombinant baculovirus to the cells. The group containing the functionalized gel showed a much greater number of MGPF expressing cells compared to the control group, the baculovirus group and the nonfunctionalized hydrogel group using fluorescent microscopy. Last, in a recent study, short‐hairpin RNA (shRNA) of angiotensin converting enzyme was injected into rat myocardium postmyocardial infarction using a dextran‐poly(e‐caprolactone)‐2‐hydroxylethylmethacrylate‐poly( N ‐isopropylacrylamide) (Dex‐PCL‐HEMA/PNIPAAm) hydrogel, Figure 6 d. 129 This study used the RNA interference gene‐technique in order to silence the gene responsible for upregulation of angiotensin converting enzyme (ACE). This is important in cardiac repair and remodeling because upregulation of ACE leads to cell apoptosis and increasing infarct size, both of which ultimately lead to heart failure. This study proved that the hydrogel sustained an extended gene expression of the shRNA responsible for the silencing gene‐technique in vivo. Thirty days after intramyocardial injection of the ACE‐shRNA plasmid‐loaded hydrogel they observed decreased ACE expression, inhibited cell apoptosis, reduced infarct size, and improved cardiac function. These results were better than either the hydrogel or the ACE‐shRNA by itself. 6 Conclusions and Future Perspectives Injectable hydrogels offer huge potential for application in repair and regeneration of infarcted heart after MI. They can be used as carriers to deliver therapeutic drugs, biomolecules or cells to invoke a specific response in the infarcted heart tissues, or to form functional cardiac tissue constructs for replacement of infarcted cardiac tissues. In recent years, numerous hydrogels have been investigated for their application in cardiac repair and regeneration. Several of these hydrogels have shown great promises toward achieving cardiac tissue repair. To date, no hydrogel composition has shown both the necessary biological and mechanical properties for sufficient transition to clinical cardiac tissue repair. While naturally derived hydrogels from collagen, hyaluronic acid, chitosan, and ECM hydrogels have shown promise in the left ventricle modeling of the myocardium, synthetic hydrogels have been shown to offer better control over their properties such as degradation time, gelation time, and most importantly the mechanical stiffness of the hydrogel. One school of thought is that hydrogels can induce stem cell homing. However, not much is known about the exchange of signals that take part in the movement of stem cells to an injured myocardium posthydrogel treatment. The importance of hydrogels also manifests itself in their ability to deliver drugs and chemical signals throughout the body to the heart. In such cases, targeted delivery and long‐term controlled sustained release are the advantages. Cytokines, such as the stem‐cell factor (SCF), granulocyte SCF, or stromal‐cell‐derived factor (SDF), may enhance cell transplantation and infiltration. 130 When SCF was combined with G‐SCF, a 250‐fold increase in the number of circulating cells was made apparent in a model version. Cell microenvironments regulate and repair cellular fate and function. The ECM is a crucial element of the cell microenvironment that includes different chemical and physical cues. 131 Encapsulating cells in a suitable hydrogel before transplanting or injection also increases the chances for their survival, as it increases the cell‐ECM interactions. This is why much attention has now been focused on hydrogels obtained from ECM molecules. These biomaterials are now made to replicate ECM interactions, thus providing the best way to ensure the sustainability, survival and full function of the transplanted cells. 132 Besides, the hydrogels can be specifically designed to help in myogenic differentiation. 133 Despite a considerable amount of research having been done on repair of infarcted heart tissue using injectable hydrogels in small animals such as rats, mice, and rabbits, research on application of hydrogel therapies on large primates and humans is still lacking. Thus, more investigations are required before the injectable hydrogel therapies with cells or biomolecules for cardiac repair are implemented in humans. Challenges in this endeavor include ensuring the survival and integration of the delivered cells in the cardiac environment and their differentiation into the required myogenic phenotypes so that they can start performing like beating cardiac cells in minimal time from the time of injection. Another challenge is the use of chemical cross‐linking which can often be harmful for the cells. Photocrosslinking, ionic cross‐linking, or temperature or pH‐based crosslinking might be alternate options; however, application of these methods in situ is often difficult. Deficiency of needed cells, the lack of full integration with the host tissue, and the absence of electrical communication through gap junctions are also among challenges that can compromise the success of the injectable hydrogel‐based stem cell therapy or cardiac repair and require further research. Research has shown improvements in wall thickness, LV repair, and vascularization of the ischemic region with the use of injectable hydrogel based therapies. Studies performed so far in rodents and pigs give hope to the potential of hydrogel based therapies for cardiac repair in primates and humans.
|
10. 1002/advs. 201500125
| 2,015
|
Advanced Science
|
Simple 3D Printed Scaffold‐Removal Method for the Fabrication of Intricate Microfluidic Devices
|
An easy and cheap fabrication method for intricate polydimethylsiloxane microfluidic devices is presented. The acrylonitrile butadiene styrene scaffold‐removal method uses cheap, off‐the‐shelf materials and equipment for the fabrication of intricate microfluidic devices. The versatility of the method is proven by the fabrication of 3D multilayer, ship‐in‐a‐bottle, selective heating, sensing, and NMR microfluidic devices. The methodology is coined ESCARGOT: Embedded SCAffold RemovinG Open Technology.
|
Microfluidics 1, 2 is a continuously growing field, of great interest in chemistry, 3 physics, 4, 5 drug discovery, 6 biology, 7, 8 chemical biology, 9 biomedical research, 10 tissue engineering, 11 and most recently, organs‐on‐chip. 12 The small amounts of liquid required for experiments, the physics of fluids at the micro domain and the lab‐on‐chip approach make microfluidics one of the interdisciplinary field par excellence. 13 To date, polydimethylsiloxane (PDMS) is the most popular material in research laboratories for the fabrication of microfluidic devices. 14, 15, 16 It is relatively cheap and easy to manipulate, gas permeable and has a refractive index of 1. 4, close to the one of glass. For the manufacturing of microfluidic PDMS devices, generally a master is needed, usually obtained by clean‐room lithography of silicon wafers. Consecutively, PDMS is poured on the master, and after curing, the rubber must be carefully peeled off from the master and subsequently chemically bonded to another surface after activation with oxygen plasma or using chemical solutions. Notwithstanding the great potential, two main bottlenecks inhibit a more widespread use of PDMS devices. First, the PDMS fabrication method is considered too complex for many scientists without any experience in microfabrication. 9 Second, achieving a 3D (multilevel channels or a single channel with different sizes) using standard fabrication methods is rather complicated, as multiple layers of PDMS must be fabricated and then sealed together to create an internal 3D channel. 17 In recent years, sacrificial mold or fugitive ink is used for fabricating PDMS microfluidic devices. Although the use of sacrificial mold is a step forward in simplifying the fabrication of microfluidic devices, it still requires either harsh condition like the use of high temperatures for creating, 18 or removing, 19 a template, applying heavy swelling for pulling out the template, 20, 21 or the use of complex mold fabrication such as using chitosan 22 or isomalt printed with an heavily modified 3D printer and backfilled with epoxy resin. 23 Recently, 3D printing has been used either to print masters for soft lithography mold, 24 or to directly print microfluidic devices. 25, 26, 27 In the first case, although the mold is easily printed, it has the same limitation for creating multilayer and complex microfluidic devices than the standard clean room lithography. The limitation of 3D printing directly the microfluidic devices lies mainly in the material used and, so far only one example of 3D printed PDMS membrane is present in literature with the limitation of using PDMS mixed with colored photoresist, thus not pure PDMS and giving non transparent devices. 28 PDMS is usually preferred over other 3D printing plastics because of a) its gas permeability, useful in biology for keeping cells and bacteria alive for long time in the microfluidic chip; b) its elasticity, capable of making micro pumps and valves in the device and c) its simple chemical modification using well known silane chemistry, difficult thing to do on 3D printing plastics and d) its transparency. Moreover, embedding other functionalities as described in this research, is extremely hard or even impossible using a 3D printer for directly printing a microfluidic chip. Here we present an easy two‐step acrylonitrile butadiene styrene (ABS) scaffold‐removal method for achieving 3D, multilayer, intricate, micrometric channels in a single block of PDMS. We also show how, using the scaffold‐removal fabrication method, external components, such as heating elements, electronics or RF circuitry, can be embedded directly in microfluidic devices. A most striking example is the fabrication of a high‐resolution nuclear magnetic resonance device that provides molecular analysis of just microliter volumes. Using the ABS scaffold‐removal method, there is no need of lithography steps nor silicon masters, no need of bonding the PDMS on surfaces nor of repetitive procedures for obtaining multilevel channels, making the fabrication of microfluidic devices easy, low‐cost and opening up the field for a plethora of scientists working in different areas. We baptize this methodology ESCARGOT: Embedded SCAffold RemovinG Open Technology. In order to avoid the use of silicon masters and (clean room) lithography, and the subsequent bonding of PDMS to another surface or the complex fabrication of sacrificial molds, we propose the use of an off‐the‐shelf plastic polymer, used for 3D printing, as scaffold for creating micrometric sized channels. The scaffold plastic polymer can be inserted into liquid PDMS, and after curing of the latter, dissolved using a PDMS‐inert solvent, leaving an empty cavity inside the PDMS ( Figure 1, and Video S1, Supporting Information). With this method, any 3D channel structure, even extremely intricate ones, can be created in two easy steps, basically without knowledge of, or experience in, microfabrication or lithography. In addition to the simple fabrication method we also show how, using this method, it is easily possible to integrate external components such as UV‐LED, heating unit, ship‐in‐a‐bottle, and even a fully functional NMR microcoil. Figure 1 Schematic representation of the ABS scaffold‐removal fabrication method for manufacturing microfluidic devices. An ABS plastic scaffold is modeled, or 3D printed, in the desired shape (left). Consecutively, it is suspended in PDMS with or without the addition of external components and then the polymer is cured (middle). Finally, the scaffold is removed by immersion in acetone creating the microfluidic channels (right). One of the most common plastics used for the fused deposition modeling (FDM) 3D printing, is the cheap (less than 20¤ per kg) and commercially available ABS. 29 We extruded ABS plastic with the aid of a 500 μm nozzle giving filaments of approximately the same diameter. These scaffolds were then suspended into liquid PDMS and the latter was cured at 75 °C for 2 h, after which it was immersed in acetone for 12 h, dissolving the scaffolds. Acetone was the solvent of choice for dissolving ABS whilst its swelling ratio (S) for PDMS is as low as 1. 06. 30 A final flushing with acetone completely cleaned the inner channel, creating de facto a PDMS microfluidic device. Changing the nozzle of an off‐the‐shelf 3D printer to nozzle with smaller diameter of 400, 300, and 200 μm is easy and provides microfluidic channels of the same diameter (Figures S2 and S3, Supporting Information). It is not hard to imagine that, giving the rise of commercially available 3D printers, in the next year nozzles with diameters of about 100 μm or even smaller will hit the market. At those scales, roughness of few micrometers is non influential for the performance of the microfluidic chip, and this roughness is comparable to commercially available sandblasted glass microfluidic chips. Although the roughness due to the nozzle is in the order of few micrometers, the one coming from the layer by layer ABS deposition for more complex designs is much higher and it is dependent by the resolution of the printer, spanning from 100 to 10 μm. Many different 3D channels were readily created using the ABS scaffold‐removal method ( Figure 2, and Video S2, Supporting Information): spiral channels (Figure 2 a), multichannels with different geometries (Figure 2 b) and channels with compartments differing in size (Figure 2 c). As further proof of concept, a complex 3D multilevel scaffold based on the Hilbert curve 31 was designed and 3D‐printed utilizing ABS fuse deposition modeling. Also in this extreme case, with the single channel inside the PDMS having a length of 35 cm and containing 1. 4 cm 3 of ABS, it was still possible to remove the plastic with subsequent baths in dichloromethane and acetone (Figure 2 d). Figure 2 Various 3D multilayer PDMS microfluidic devices fabricated using the ABS scaffold‐removal method. a) Spiral microfluidic device; b) a microfluidic channel wrapped around another one; c) a single channel with different diameter and d) a microfluidic device fabricated using a 3D printed object used as scaffold. Diameter of the channels is 500 μm in (a) and (b), 500 μm and 90 μm in (c), and 2 mm in (d). Integrating external elements directly in the microfluidic device is desirable for lab‐on‐a‐chip approaches but difficult to achieve using standard PDMS fabrication methods. We incorporated stirring bars, electronic circuitry, heating elements, and radiofrequency (RF) components, illustrating the wide and versatile applicability of this method ( Figure 3, and Video S3, Supporting Information). Figure 3 Various electronic components embedded in microfluidic chips. a) a 390 nm UV‐LED illuminating a fluorescent dye in the channel; b) a resistance wire, used as selective heating unit; a thermochromic dye changes color only where the resistance wire is coiled around the channel; c) a 32 μm copper wire wrapped around a microfluidic channel is used as solenoidal microcoil allowing high‐resolution NMR spectroscopy on 2 μL sample volumes; the insert shows the 31 P‐coupled 19 F spectrum of a NaPF 6 solution in water, values in ppm; d) a fully functional Arduino microcontroller coupled with a color sensor embedded into the PDMS chip. The diameter of the channels in all the pictures is 500 μm. External components in a cavity bigger than the size of the channels guiding to it, so‐called “ship‐in‐a‐bottle, ” 32 can be included in the ABS polymer and subsequently released during the acetone treatment. In this way a small cylindrical 1 × 1 mm magnet was inserted in a microfluidic chamber (Video S3 and Figure S11, Supporting Information). As PDMS is nonconductive (resistivity 10 13 –10 15 Ω/cm) and acetone is a noncorrosive solvent, electronic components can also be embedded directly in the design. These can be simply inserted in the PDMS together with the microchannel scaffolds before curing it, then the acetone treatment removes only the scaffold leaving the electronics intact. In this way we inserted a 390 nm LED for the optical detection or electronic excitation of chemicals in the microfluidic channel (Figure 3 a). Another problem usually associated with microfluidic chip is the difficulty of heating only part of the channel inside the chip. Taking advantage of PDMS its low thermal conductivity (0. 15 W m −1 K) we envisioned a selective heating unit inside a microfluidic device. A 200 μm Nichrome resistance wire was loosely wrapped around the ABS polymer scaffold and inserted in PDMS. After the curing step and dissolving the ABS scaffold, a voltage of 1. 2 V with 0. 35 A sufficed for selectively heating a thermochromic dye above 27 °C only in the part of the channel surrounded by the resistance wire (Figure 3 b, and Video S3, Supporting Information). Temperatures can be varied and the 200 μm wire allows, for example, to boil water inside the channel (Video S3, Supporting Information). This simple and selective heating element embedded in the microfluidic chip can be of great value for designing chips to perform, e. g. , biological experiments such as PCR, sterilization inside the microchannels or for setting different temperatures for organ‐on‐chips or cell cultures. NMR spectroscopy is arguably one of the most powerful analytical tools available to the scientific community. However, NMR is a notoriously insensitive technique because of the unfavorable Boltzman distribution of the spin states, severely compromising analysis of mass‐ and volume‐limited samples, like in microfluidics. Approaches to solve sensitivity issues comprise most expensive and technologically demanding solutions as using extreme magnetic field strengths and complex NMR probe techniques. 33 Alternatively, down‐scaling the RF transceiver coils to match the size of the sample significantly increases the sensitivity, 34 although these small‐volume probes still require technological demanding fabrication methods. 35 We decided to exploit the concept of ABS scaffold‐removal to make a simple and cheap, yet most sensitive NMR sensor. A 32 μm copper wire was wrapped around a 500 μm ABS filament, resulting in a final channel encompassed by a solenoidal NMR microcoil (Figure 3 c), with a detection volume of only two μL; normal NMR tubes contain about 500 μL sample volume. This microfluidic device was integrated on a cylindrical aluminum probe insert and placed inside a 9. 4 T narrow‐bore superconducting NMR magnet. Tuning the resonance circuit to 376 MHz, straightforwardly allowed high‐resolution NMR spectra to be obtained (Figure 3 c insert). Line‐widths at half peak‐height were obtained of about 3 Hz and resolving heteronuclear spin–spin couplings, opening up the way to further optimization and applications (Figures S15 and S16, Supporting Information). The lines of the doublet observed in the 31 P‐coupled 19 F‐spectrum of a 1 m NaPF 6 sample correspond to an amount of one micromole of spins that is detected. This amount is in the order of the lower mM concentrations measurable in conventional NMR probes; further optimization of the probe coil is currently ongoing to allow for an increased concentration as well as mass sensitivity. In addition, we calculated that the material costs for fabricating this device is less than two euro. Thus, the microfluidic NMR‐device, in which sample container and transceiver coil are integrated, is cheaper than a standard NMR tube, and orders of magnitude cheaper than an NMR probe head worth several thousand euro. Last proof of the versatility of the ABS scaffold‐removal method was the embedding of a fully functional color sensor and a microcontroller directly in a microfluidic device (Figure 3 d). An Arduino micro and a color sensor were wired together and immersed in PDMS with an ABS scaffold. After curing the PDMS and removing the ABS polymer with acetone, the resulting microfluidic channel was right on top of the color sensor. Hooking up the Arduino to a computer revealed all the components of the microcontroller and the sensor to be working properly. Thus, even complex electronics can be easily embedded in PDMS microfluidic devices reducing the amounts of external components needed for using, screening and analyzing experiments in microfluidics devices. Although in its infancy, where roughness and channel sizes can still be improved by the ongoing 3D printing technology, the ABS scaffold‐removal method is simple and cheap compared to the current fabrication methods, yet powerful and versatile in creating 3D multilevel and intricate microfluidic channels in PDMS. Moreover additional elements like heating coils, RF circuitry or electronic components can be embedded directly, opening up new windows for the various fields of microfluidics and PDMS devices. Creating multilevel 3D microfluidic devices can be of benefit in many different fields, for example in fabricating complex vascular systems for organ on chips, or in handling spherical droplets in tubular channels of different sizes or in mixing liquids. Because of the multidisciplinary fields of applications we can envision, the simplicity of this method, coined ESCARGOT, is of great value for all scientists willing to work with microfluidic devices, regardless of their background. Experimental Section SYLGARD silicone elastomer 184 and SYLGARD silicone elastomer 184 curing agent were obtained by Dow Corning Corporation. A 3D SIMO pen was used for extruding 1. 7 mm ABS, plastic filament that was obtained from the same vendor. 3D print of Hilbert cube was ordered online and 3D printed by ridix. nl (Rotterdam, the Netherlands) using a Dimension SST 1200es printer and by 3dhubs. com using a Duplicator 4 printer. Acetone was obtained from Sigma‐Aldrich. Sparkfun RGB color sensor ADJD311 was bought from sparkfun. com, thermochromic dye from mindsetsonline. co. uk, Arduino micro and 200 μm Nichrome resistance wire from rs‐components. The ABS plastic filament was extruded through a 500 μm nozzle (3D SIMO pen) or a Craftbot 3D printer with 400, 300, and 200 μm nozzle and then modeled into the desired 3D shape with the help of a soldering iron set (100 °C) or printed with a fused deposition modeling 3D printer. The modeled ABS plastic was then immersed in a well mixed solution of 10:1 sylgard 184/sylgard 184 curing agent. The PDMS was then placed under vacuum for removing air bubbles and cured for 2 h at 75 °C, or overnight at room temperature. The PDMS was consecutively left for 12 h in acetone, after which the microchannels were cleaned with acetone and dried with a flow of compressed air. NMR spectroscopy was performed on an Oxford Instruments 9. 4 T superconducting magnet, equipped with a 14‐coils shimming set‐up, interfaced with a Varian Inova spectrometer. The ABS‐NMR device was a cylindrical shaped PDMS disk of about 4 cm in diameter and 1 cm height, positioned on top of an aluminum cylinder of a sacrificed NMR probe. The detection volume encompassed by the solenoid sums up to ≈2 μL. The two 32 μm ends of the copper solenoid wire were soldered to two connection leads with a variable (3–18 pF) capacitor in parallel, and wired to the tuning and matching circuit of the probe, respectively grounded, allowing the fine‐tuning of the resonance circuit to be done with the probe positioned inside the magnet. Experiments were run in non‐locked mode, typically using single‐scan acquisitions using the Vnmrj 2. 2D software. For the 19 F‐NMR experiments the coil was tuned to 376 MHz, and the acquired data were processed using Mnova (MestreLab, Santiago de Compostela, Spain). Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500149
| 2,015
|
Advanced Science
|
High‐Throughput Contact Flow Lithography
|
High‐throughput fabrication of graphically encoded hydrogel microparticles is achieved by combining flow contact lithography in a multichannel microfluidic device and a high capacity 25 mm LED UV source. Production rates of chemically homogeneous particles are improved by two orders of magnitude. Additionally, the custom‐built contact lithography instrument provides an affordable solution for patterning complex microstructures on surfaces.
|
Recent advances in fabrication techniques have created new opportunities for applications of polymer particles, beyond spherical particles. 1 Anisotropic polymer particles with precise shapes or heterogeneous chemistries, in particular anisotropic hydrogel particles, have demonstrated unique advantages in numerous fields. For drug delivery, tissue engineering, and diagnostic imaging, engineering nano and microparticles' shape is a way to tailor particle penetration and degradation properties. 2 In the field of biosensing, unique shape and graphical patterns of particles have brought new strategies for encoding complex particle libraries for multiplex sensing applications. 3 A common requirement to all these applications is the need for robust, affordable, and rapid techniques for particle fabrication. Conventional methods for the fabrication of micrometer‐sized hydrogel particles, such as dispersion, precipitation, and emulsion polymerization, are often limited to the production of polydisperse suspensions of spherical particles. 4 Similarly, droplet‐based microfluidic techniques enable high‐throughput polymer particle production but are usually restricted to spheres or spheroids. Contact photolithography and replica molding, already used to pattern polymeric structures on surfaces, have been successfully adapted to the production of nonspherical particles. Originally developed for the production of submicrometer features in the semiconductor industry, 5 photolithography techniques use light to transfer a pattern from a photomask to a photopolymerizable material. Shape‐coded hydrogel particles in the 50–1000 μm range were successfully patterned using contact photolithography, using a photomask placed in direct contact with a layer of monomer solution. 6 Replica molding, also known as imprint lithography, 7 is directly inspired from the soft lithography techniques developed for the fabrication of microfluidic devices. 5 Replica molding of particles consists of pouring a liquid monomer into a negative mold with the desired shape and dimensions, and photocrosslinking the material in the mold. Nevertheless, both techniques are static batch processes with limited throughputs and particle collection time and set‐up times in between runs often reduce the synthesis rates. The development of the flow‐photolithography technique enabled significant progress toward automation and scale‐up of microparticle synthesis using microfluidic channels. 8 Particles are synthesized inside a polydimethylsiloxane (PDMS) microfluidic channel filled with a photocurable monomer solution, using microscope‐based illumination and automated control of exposure to ultraviolet (UV) light. Where exposed to UV light, the monomer crosslinks and solidifies into a microparticle. Due to PDMS permeability to oxygen, oxygen is present at high concentration near the PDMS channel walls and locally inhibits the free‐radical polymerization. This inhibition creates a thin lubrication layer of uncured monomer (typically 2. 5 μm‐thick) at the top and bottom sides of the channel and results in free‐floating particles that can be transported through the channel with the stream of monomer. 9 Particles are collected in an outlet reservoir while the polymerization process is repeated inside the channel. The method was demonstrated on polyethylene glycol diacrylate (PEGDA) hydrogels, but is applicable to any free radical polymerization reaction. 9, 10 Several research groups successfully applied flow lithography to synthesize particles with complex graphical codes based on shapes, 11 1D‐barcodes, 12 or even 2D‐barcodes. 13 Recent studies also investigated 3D‐particle patterning. 14 The technique was initially proposed by Dendukuri et al. as continuous flow lithography (CFL), with sequential UV pulses sent through the photomask on a continuous flow of monomer. 15 This process was however limited in resolution at high flow rates, since the polymerizing particles moved significantly during exposure, resulting in blurred particles. In the next iteration of the technique, stop‐flow lithography (SFL), photopolymerization was performed in a stationary monomer, optimizing the patterning resolution. In addition, much higher flow rates could be used to flush particles out of the channel. As a result, both particle resolution (10–100 μm) and synthesis throughput (10 4 per hour) were enhanced compared to CFL. 8 While the conventional microscope‐based flow lithography brings multiple advantages, such as intense light power surface density through the objective, fine resolution, and control over focal adjustment, it critically limits the illumination area and significantly decreases the number of particles that can be synthesized in a single exposure. Typically, the homogenous illumination area with a 20× objective is less than 500 μm in diameter, which severely limits the number of particles per exposure and the particle synthesis rate. Moreover, the cost of the microscope instrument and objective hinder the possibility of using multiple parallel synthesis setups in terms of industrial scale up. To overcome the above limitations of CFL and SFL, we designed a novel bench‐top contact flow lithography system, with versatile lithography functions, and we successfully achieved particle synthesis at ultrahigh throughput. With our customized low cost contact photolithography system providing strong and homogeneous illumination across 23 mm and rationally designed microfluidic channels, we dramatically increased the particle synthesis rate by two orders of magnitude (>10 6 100 μm sized particles per hour) while maintaining excellent particle resolution and homogeneous physicochemical property of particles. Furthermore, the use of this cost‐efficient platform can be easily extended to a variety of photolithography applications. The investigated contact flow lithography station is composed of three major parts ( Figure 1 a, from bottom to top): an illumination unit triggering microparticle photopolymerization, a stage unit holding the microfluidic device, and an imaging unit (charged‐coupled device (CCD) based camera) enabling to align the mask with the microfluidic device. To build the illumination unit, a high power UV light‐emitting diode (LED) light source (365 nm, 700 mA) was collimated into a 25 mm beam using an aspheric condenser lens without diffuser and assembled to a precision XYZ‐rotation stage. A specially designed photomask adapter for 25 mm chrome masks was 3D‐printed and tightened to the UV illumination unit. The stage unit was fixed and fastened to a damped post to inhibit vibration from the solenoid valve during microparticle synthesis. Finally, the imaging unit was built into another XYZ linear translation stage for precision motion. In this manner, we are able to independently control the position of both the imaging unit and the illumination unit with regard to the stage unit. Before polymerization, the photomask is placed in contact with the UV transparent device to be patterned. Minimizing the distance between the mask and the microfluidic device decreases diffraction and aberration of the UV light. In the case of microscope‐based lithography, accurate positioning of the channel in the objective focal plane was critical for particle resolution and the objective depth of field was limiting the particle thickness. For the contact lithography instrument however, collimation of the UV LED light into a straight beam is the key to well‐resolved particles with straight edges. Figure 1 Contact lithography instrument. a) Schematic diagram of the contact lithography system. b) Photograph showing the extent of the polymerization area on glass slide (25 mm). c) Bright field microscopy image of various sized microstructure and d) fluorescence microscopy image of rhodamine‐B labeled microstructures with multiple shapes on a glass slide. In order to investigate the homogeneity of the illumination provided by the LED source, we polymerized an array of PEG microstructures with various shapes on a glass substrate across the entire beam of 25 mm diameter (Figure 1 b). The size and overall shape of polymerized structures appeared in excellent agreement with the mask pattern, with sharp edges and straight side walls. Structure quality decreased only in a 1 mm peripheral zone, where underpolymerization was observed (Figure S3, Supporting Information), resulting in a 23 mm effective area for reproducible patterning. Figure 1 c demonstrates successful patterning of structures of decreasing sizes, from 140 to 20 μm. Incorporating rhodamine‐B in the monomer solution and analyzing the fluorescent signal of the labeled microstructures ensured that the synthesized microstructures were not only physically but also chemically homogeneous, as shown in Figure 1 d (coefficient of variation (CV) = 6. 8%). The power density of the collimated LED beam was 125 mW cm −2, which is about three times lower than the value measured on our standard microscope‐based SFL system. Therefore, we increased UV exposure times from 70 to 200 ms to achieve similar chemical conversion for 100 μm sized particles. Flow lithography protocols were successfully developed for the contact flow lithography system, enabling the synthesis of chemically and physically homogeneous hydrogel particles with synthesis rates enhanced by two orders of magnitude. Figure 2 a describes the workflow for particle synthesis using contact flow lithography. The microfluidic device is secured on the fixed lithography stage (Figure S1, Supporting Information). The chrome photomask with the desired pattern is placed in the 3D‐printed mask holder and carefully elevated until in close proximity to the bottom of the microfluidic device, so that both the microchannel and the mask pattern can be observed simultaneously with the CCD camera. The position of the illumination unit is then adjusted to align the photomask with the microchannel (Figure S2, Supporting Information). Once aligned, the light source is elevated again until the chrome mask is in contact with the bottom of the microfluidic device. The device inlet is connected to a pressured monomer reservoir, and the outlet to a particle collection vial. The device is initially primed with monomer solution. From then, the stop‐flow lithography cycle can be decomposed in three steps. First, particles are polymerized with ≈200 ms UV exposure followed by a brief hold (≈250 ms) to ensure complete polymerization (Figure 2 b). Particles polymerize locally where the UV light reaches the monomer layer and the photomask pattern is transferred as a negative to the monomer layer. Second, the monomer flow is switched on again and particles are flushed out by flowing monomer solution for a few seconds through the microfluidic channel (Figure 2 c). Third, the flow of monomer is stopped for the next round of polymerization. Because of PDMS elasticity, a minimum response time is required to observe complete flow stoppage. 16 The pressure imposed at the inlet induces a deformation of the channel top wall. When this constraint is released, PDMS relaxation creates an opposite squeeze‐flow (Figure S3, Supporting Information), requiring additional seconds to reach a complete fluid stoppage and to ensure high patterning resolution. A compressed‐air flow control system and a solenoid valve control the pressure‐driven flow of monomer inside the device. Both the valve and the LED are computer‐controlled and synchronized, making the particle production a fully automated process. 17 Figure 2 Synthesis of microparticles in flow. a) Schematic diagram describing the workflow for stop‐flow contact lithography. b) Bright field microscopy image of an entire eight‐channel module filled with diamond‐shaped particles (5760 particles) after UV exposure. c) Sequential views of particles being flushed out of the channel ( W = 950 μm; L = 10 mm). Our contact flow lithography system provides a polymerization area ≈2000 fold larger than the microscope‐based system. To take advantage of this dramatically increased illumination area, we rationally redesigned the PDMS device to integrate multiple parallel synthesis channels. We tailored the channels layout and dimensions to maximize the particle synthesis rate. As shown in Equation (1), the particle synthesis rate depends not only on the number of particles polymerized per UV pulse ( n p ) but also on the duration of each cycle step: polymerizing ( t pol ), flushing particles out ( t flow ), and stopping the flow ( t stop ) (1) synthesis rate = n p t pol + t flow + t stop The microchannel dimensions (length L, width W, and height H) critically impact three of these parameters, namely, n p, t flow, and t stop. Theoretical analysis of a model single straight channel led to the scaling law given in Equation (2) (detailed analysis available in the Supporting Information), where μ represents the dynamic viscosity of the monomer fluid, E the elastic modulus of PDMS, and Δ P the pressure drop across the channel (2) synthesis rate ≈ 1 4 μ L E H 2 ( 1 Δ P W E + H 3 + 3 E Δ P W ) According to Equation (2), increasing the channel width or height leads to an increase in synthesis rate. We chose a channel height H of 50 μm in order to produce particles with 45 μm in height. The W / H aspect ratio was limited by fabrication constraints. Indeed, for W / H > 20 ( W > 1 mm), the top wall of the PDMS channel sags. PDMS delamination under high pressure imposes an additional practical limits on the pressure imposed at the inlet. With all other parameters fixed, Equation (2) shows that increasing the channel length tends to decrease the rate of particle synthesis. Indeed, longer channels increase both the particle flushing time and the flow stoppage time. Therefore, designs involving multiple short channels are preferable to a long serpentine channel. Contiguous parallel channels enable to maximize the coverage of the polymerization zone. As individual inlets and outlets would generate important dead space and excessive tubing, channels were grouped using with a splitting design. From a single inlet, the monomer flow was equally split into eight identical channels, using a design optimized through simulations (Figure S5, Supporting Information). Individual channel width was 950 μm and parallel channels were separated by 50 μm PDMS walls. With a channel length of 10 mm, two of these 8‐channel modules can be run side by side, covering a 16 × 10 mm polymerization zone. For microfluidic layouts with shorter channel length, the dead space occupied by the splitting flow modules upstream and downstream of the straight channels becomes too important relative to the effective particle polymerization module to be beneficial. Particles were successfully polymerized at high volume fractions in channels (>50%) without jamming, using high density mask patterns (Movie S1, Supporting Information). Typically features on the photomask were spaced from one another by at least 25 μm and from the channel wall by at least 50 μm. It should be noted though that, at such high particle density, the flow stoppage is critical, as a residual flow may cause particles to overlap during polymerization and clog the PDMS channel. As an example, with these dimensions, up to 720 diamond shaped‐particles (≈75 μm) fitted in a unit channel, leading to 11 520 particles polymerized per exposure in the 16‐channel device. At 8 psi, complete synthesis cycles were successfully run in 7. 5 s ( t pol 0. 5 s; t flow 4. 5 s; t stop 2. 5 s), leading 5. 6 × 10 6 particles h −1 (Movie S1, Supporting Information). This represents an increase in synthesis rate of two orders of magnitude compared to the microscope‐based stop‐flow lithography system. Although the extended dimensions of the microfluidic device require a polymerization cycle time ten‐fold longer than for microscope‐based flow lithography, the dramatic increase in the number of particles produced per UV pulse leads overall to a significant 100‐fold increase in synthesis rate. To assess the reproducibility of particle size, shape, and composition, a test panel of 12 shapes (≈75 μm) with distinctive aspect ratio and solidity was polymerized from a fluorescent monomer (Movie S2, Supporting Information). Table S1 (Supporting Information) summarizes the characteristics of the collected particles. All particles had sharp edges and straight side walls, and were flat ( Figure 3 ). The median particle thickness was 44. 8 ± 1. 5 μm (CV = 3. 3%), when the expected value was 45 μm (50 μm‐thick channel with top and bottom 2. 5 μm‐thick oxygen inhibition layers). Figure 3 Bright field microscopy image of PEGDA hydrogel particles with various shapes a) after UV exposure, b) after collection, c) magnified view (particles are 45 μm‐thick). d) Fluorescence microscopy image of rhodamine‐B labeled PEGDA hydrogel particles. For contact lithography, a 1:1 ratio between the photomask feature size and the particle size should be observed if the light source is perfectly collimated. The dimensions of the collected particles were in excellent agreement with mask feature size (from 98% to 106% for four shapes) and high reproducibility was demonstrated (CV = 3. 6%). Regarding particle composition, the variation of the median fluorescence intensity was found to be 7. 7% (across 108 particles). Furthermore, as such encoded hydrogel particles are typically engaged in biosensing experiments, [[qv: 13a]] we demonstrated the particle porosity and functionality by incorporating a biotinylated probe inside the gel material and validating the diffusion and capture of streptavidin–phycoerythrin molecules in the gel (data not shown). In addition to polyethylene glycol diacrylate aforementioned, the method can be extended to other photopolymerizable monomers as well, and was also demonstrated on polyurethane acrylate (Movie S3, Supporting Information). By revisiting our approach for lithography of particles, the design of our microfluidic device and the UV source, we managed to achieve very high synthesis rate and high reproducibility while working around clogging issues typical of particle suspensions at such high volume fractions. This manufacturing throughput combined with the low‐cost instrumentation paves the way for studies needing substantial numbers of particles; such as drug formulation, rheology, and 3D printing of custom suspensions. Besides the manufacturing of large quantities of particles per se for downstream applications, our system also offers a platform for fundamental studies of suspensions of complex microparticles with precise initial conditions. Indeed, it is possible to create large numbers of suspensions of arbitrarily shaped particles at high volume fractions in situ inside a microchannel, while precisely controlling the initial conditions, the material chemical and physical properties, the volume fraction, as well as the respective positioning of particles. Potential applications of interest include understanding of complex suspension dynamics, 18 particle trajectories in microfluidic devices, 19 particle jamming, or printing of unconventional materials. 20 For example, Figure 4 presents a study of jamming of particles in a microfluidic channel. An array of stiff polyurethane acrylate particles was photopolymerized inside a channel displaying a narrow constriction at its end. When flown through the constriction, particles progressively jam. The initial conditions of the suspension can easily be varied, in order to compare their respective influence on the suspension dynamics and the jamming event. Figure 4 Observation of a suspension of polyurethane acrylate particles jamming at a narrow channel constriction (channel height is 35 μm). An array of precisely shaped and positioned particles was polymerized in situ inside a microfluidic channel. Under flow (direction indicated by black arrow), the rigid free‐floating particles travel toward the channel outlet and progressively jam near the constriction. Finally, beyond particle synthesis, the contact flow lithography instrument offers versatile lithography applications and is in particular a cost‐efficient solution for patterning well‐resolved 10–1000 μm microstructures on surfaces, as previously mentioned (Figure 1 c, d). Patterned hydrogel microstructures have raised notable interest in the fields of biosensing, 21 cell culture, 22 and cell imaging (for example, neural stem cells 23 and spheroids 24. Various techniques were reported for fabricating hydrogel microwells in the 100–1000 μm range: replica molding of photocrosslinkable chitosan, 24 PEG/heparin multilayered structures, 25 or peptide‐based gels, 26 soft embossing of PEG gels, 23 as well as contact photolithography to pattern PEG‐based materials. 22 There is a need for techniques enabling to pattern multiple functional materials with high spatial control. For example, contact lithography methods were used to pattern multiple solutions containing cells and extracellular matrix components for tissue prototyping, 27 and to build complex millimetric hexagonal 3D tissue architectures with multiple cellular PEGDA hydrogels. 28 Arrays of hydrogel pads with variable stiffness polymerized in microfluidic channels 29 were used to study cell behavior. 30 The present contact lithography instrument enables to print well‐resolved microstructures with high reproducibility and flexibility, over a 23 mm circular area, with very short (<second) exposure times. As a proof‐of‐concept, a model experiment consisted of patterning thin monomer layers of controlled thickness (80–160 μm), sandwiched in between an acrylated glass slide and a PDMS‐coated slide. Figure 5 shows examples of resulting free‐standing structures polymerized on acrylated glass slides. High density arrays of wells with various shapes were successfully patterned (Figure S6, Supporting Information). Figure 5 c shows 5 μm cells seeded in 60 μm PEG wells. Two‐layered structures were achieved by two successive polymerization cycles on the same substrate. Figure 5 a shows a two‐layer microwell structure, with a functional bottom layer. A biotinylated bioprobe was immobilized in the bottom layer and could be selectively labeled with streptavidin–fluorophore conjugates later on. Figure 5 Complex multimaterial PEGDA structures patterned on surfaces (composite fluorescence images, 10× magnification). a) Two‐layered structure of 60 μm‐sized wells. A bottom sensing layer containing biotinylated oligonucleotides was grafted on the surface. Subsequently, a layer of hydrogel wells labeled with a red fluorophore was polymerized on top of the first layer. Later on, the hydrogel wells were incubated with a streptavidin‐AlexaFluor488 conjugate, which was captured by the biotin groups in the bottom layer. b) Nested designs. Posts (diameter: 20 μm, labeled with AlexaFluor488) and circles (diameter: 100 μm, labeled with AlexaFluor647) were sequentially polymerized on the surface. c) HSC‐3 cells labeled with calcein AM seeded in 50 μm diameter rhodamine‐labeled hydrogel wells. Furthermore, while larger UV sources have been used by others to pattern microstructures across wide surface areas using lithography, a major added‐value of this set‐up is the possibility to precisely align masks and substrates with the top view live‐imaging module and precise motion controls. Circumventing the need for an expensive aligner, our system provides an affordable solution for efficient and quick prototyping of features larger than 10 μm. It is possible to pattern hydrogel structures at precise locations in microchannels for example. Additionally, multiple masks can be used sequentially to print complex interlocked structures from different materials. Figure 5 b shows an example of such nested designs patterned with two different materials: 100 μm circular gels were polymerized around 20 μm posts (Figure S7, Supporting Information). In this report, we have presented a cost‐efficient and versatile approach for the fabrication of both free‐floating polymer microparticles in microchannels and polymer microstructures on surfaces. High‐resolution UV‐induced polymerization of microstructures (20–150 μm) was successfully achieved using contact lithography across a 23 mm circular area, without the need for an expensive optical objective. By rationally designing a multichannel microfluidic chip for contact flow‐lithography, we were able to increase the synthesis rate for chemically homogenous particles by a 100‐fold in comparison to our standard microscope‐based technique. Contrary to replica molding and static contact lithography, flow lithography can be easily automated and operated as a continuous process. Millions of graphically encoded particles can be synthesized within an hour, with high particle reproducibility and resolution while avoiding clogging issues typical of particle suspensions at such high volume fractions. Additionally, this process can be applied to the fabrication of bit‐coded particles with extruded holes as well. Finally, we believe that the reduced cost and portability represent a significant added value of the instrument. Indeed, there is a rising global interest in material science for solutions for scalable manufacturing and for equipment that enables accurate and efficient material fabrication at substantially lower overall cost. 31 An increasing number of innovative, flexible, and open‐source designs are being reported in the literature. 32 Built from scratch with an overall cost around $5000 (including the imaging unit), our system provides an efficient and versatile solution for particle and surface patterning, and easy prototyping, that can also easily be customized toward a specific application. In addition, the reduced cost, straightforward assembly, and small footprint enable parallelization of multiple polymerization stations and pave the path for particle production at industrial scale. Experimental Section Contact Flow Lithography Instrument : A detailed part list can be found in the Supporting Information. Microfluidic Device Fabrication : A microscope glass slide was spin‐coated with PDMS (200 μL, 3 min, 3000 rpm) and cured at 65 ºC for 30 min. The PDMS thick layer with channel imprints was fabricated through soft lithography on silicon wafers patterned with SU8. Inlet and outlet holes were punched using a 1. 5 mm biopsy punch before assembling top and bottom layers. The device was baked at 65 ºC for 1 h, rinsed with ethanol, and dried with argon before use. Contact Flow Lithography : A multichannel PDMS device was secured on the stage of the contact lithography instrument between slide holders. The PDMS device was connected to a pressured monomer reservoir and a collection vial using PTFE tubing (0. 75 mm ID). The monomer reservoir consisted of a 1. 5 mL Eppendorf microtube connected to a compressed air source (Tygon tubing 3/32 inch ID) using a 1. 5 mL small reservoir microfluidic kit (Elveflow, France). Typical monomer composition for particle synthesis was PEGDA700 (20%), PEG600 (40%), Darocur 1173 (5%), and Tris‐EDTA 3X buffer (35%). To fabricate fluorescent particles, rhodamine‐B acrylate (Laysan‐Bio, USA) was added to the monomer solution. A 25 mm square chrome photomask with the desired shape pattern (Front range, CO, USA) was placed on the mask holder on top on the LED source. The photomask was aligned with the microchannels using the top view camera and alignment marks, and then brought up in contact with the device. Particles were polymerized and collected through sequential exposure (250 ms, 125 mW cm −2 ), flow, and stoppage steps. Detailed descriptions of the flow and illumination control systems can be found elsewhere. 8, 12, 17 The collected particles were rinsed three times with aqueous buffer (Tris‐EDTA 1X buffer, 0. 05% Tween20) using a 1:10 volume ratio to remove nonpolymerized monomer. Particles were imaged using bright field and/or fluorescence microscopy. Particle dimensions and fluorescence intensity were analyzed using ImageJ software (NIH, USA). Patterning Microstructures on Surfaces (Posts, Wells) : Monomer solution was sandwiched between two thin glass slides, with double‐sided tape spacers (80 μm) used to control the height of the monomer layer. The bottom slide was acrylated beforehand to promote gel adhesion, whereas the top slide was spin‐coated glass with PDMS to prevent binding. Typical monomer composition for surface patterning synthesis was PEGDA700 (80%), Darocur 1173 (5%), Tris‐EDTA 3X buffer (15%). The device was placed on the instrument stage in contact with desired chrome photomask and exposed to UV light (typically 200 ms, 125 mW cm −2 ). Following exposure, the top slide was removed and the hydrogel layer was thoroughly rinsed with water to remove nonpolymerized monomer. When patterning multiple materials, the chamber was filled with the second monomer solution, sealed again with a PDMS‐coated slide, and aligned with a second photomask, before proceeding to the second polymerization run. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500393
| 2,016
|
Advanced Science
|
Sustained Delivery Growth Factors with Polyethyleneimine‐Modified Nanoparticles Promote Embryonic Stem Cells Differentiation and Liver Regeneration
|
Stem‐cell‐derived hepatocyte transplantation is considered as a potential method for the therapy of acute and chronic liver failure. However, the low efficiency of differentiation into mature and functional hepatocytes remains a major challenge for clinical applications. By using polyethyleneimine‐modified silica nanoparticles, this study develops a system for sustained delivery of growth factors, leading to induce hepatocyte‐like cells (iHeps) from mouse embryonic stem cells (mESCs) and improve the expression of endoderm and hepatocyte‐specific genes and proteins significantly, thus producing a higher population of functional hepatocytes in vitro. When transplanted into liver‐injured mice after four weeks, mESC‐derived definitive endoderm cells treated with this delivery system show higher integration efficiency into the host liver, differentiated into iHeps in vivo and significantly restored the injured liver. Therefore, these findings reveal the multiple advantages of functionalized nanoparticles to serve as efficient delivery platforms to promote stem cell differentiation in the regenerative medicine.
|
1 Introduction This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Hepatic failure including cirrhosis and fibrosis is a serious threat to human health, and recently embryonic stem cells (ESCs) have been recognized as promising therapeutic agents for the treatment of acute liver failure and advanced cirrhosis. 1, 2 Both murine and human ESCs were reported to differentiate into hepatocytes or hepatocyte‐like cells successfully. 3, 4, 5, 6, 7 As known, most of the current protocols used embryoid bodies (EBs) and simple adherent monolayer cultures with soluble growth factors and small chemicals to differentiate ESCs into hepatocytes; however, they are still subjected to inefficient differentiation and low purity of functional hepatocytes. 8, 9 The differentiation of stem cell could usually be affected by not only the internal factors, but also the exogenous growth factors, however, these substances tend to have short half‐lives and are needed to externally add to the culture media in order to maintain the effective level during the whole process. 10 While higher concentrations of added growth factors can induce cell toxicity. 11 Therefore, developing a stable and highly effective differentiation system with sustained delivery of growth factors is of the utmost importance. Recently, the application of biomaterials in the growth factors delivery system of stem cell has been reported. 12 The polymer‐based delivery systems have been developed to date, including poly(lactide‐ co ‐glycolide) (PLGA), polylactic acid (PLA), and polycaprolactone. PLGA particles were used as a delivery system to package growth factors within EBs to improve the differentiation of human embryonic stem cells into the vascular or osteogenic lineage. 13, 14 However, remaining organic solvents used in the fabrication of polymer nanoparticles may adversely affect sensitive protein bioactivity and the loading efficiency of larger biomolecules, such as growth factors. 15, 16 During the past decade, inorganic nanomaterial‐based carriers were widely used to deliver genes, drugs, and bioactive molecules due to their ability to overcome biological barriers and deliver larger biomolecules, leading to eliminate the loss of biological activity and likely enhance long‐term sustained release of bioactive molecules. Among various inorganic nanomaterials, silica nanoparticles have attracted significant attention as a superior multifunctional nanomaterial because of its biocompatibility, tunable pore volumes, unique interfacial features, and easily functionalized surfaces. 17, 18 In order to maximize loading capacity and improve the suspension stability or processability of delivery system, the surface of silica materials should be modified. One of the most attractive groups available for surface modification is the amino group by using amino silanes or cationic polymers such as polyethyleneimine (PEI). Positively charged PEI‐coated silica particles could dramatically enhance their dispersion stability and prevent the aggregation at physiological pH compared to corresponding bare materials, thus nowadays silica modified with PEI emerged as promising alternative delivery systems of biological molecules, such as nucleic acids, enzymes, and antibodies. 19, 20, 21 Chen et al. demonstrated that positively charged mesoporous silica nanoparticle (MSNs) delivering hepatocyte nuclear factor 3β plasmid DNA could promote the production of hepatocyte‐like cells from induced pluripotent stem cells. 22 MSNs modified with PEI were also reported to facilitate DNA and siRNA delivery because of the large number of amine groups on PEI, enhance particle uptake into cells through rapid endocytosis, and promote endosomal escape. 19, 23, 24, 25 Compared to unmodified silica particles, PEI as the polymer coating to the surface of MSNs not only enhances suspension stability of MSNs but also promotes bonding growth factors through electrostatic interactions and hydrogen bonding between amino groups of the MSNs and functional groups of the growth factors, 26, 27 which may improve growth factors loading, stabilize growth factors, and extend the time of cells exposure to the growth factors. In our study, we established an inorganic/polymer culture system for sustained promoting the differentiation of mouse embryonic stem cells (mESCs) into hepatocyte‐like cells by using silica nanoparticles modified with hyperbranched PEI as the carrier for exogenous growth factors ( Scheme 1 ). We found that PEI‐modified MSNs (PEI‐MSNs) could be internalized by the mESCs sustained delivery growth factors without impairing cell viability and efficiently deliver growth factors, such as Activin A, acidic fibroblast growth factor (aFGF) and hepatocyte growth factor (HGF), which were reported to play an important role in hepatic differentiation from the ESCs. 5, 28, 29 Furthermore, after three‐day treatment with growth factor (GF)‐PEI‐MSNs, predifferentiated definitive endoderm (DE) cells were sufficient to induce more robust differentiation of hepatocyte‐like cells in vivo upon transplantation into a mouse model of chronic liver injury. Moreover, the injured livers were more efficiently restored compared with the controls, which attributed to transplanted cells and the paracrine effect of predifferentiated cells. Thus, this treatment may offer a potential approach for regenerative medicine applications, especially in the treatment of liver diseases. Scheme 1 Schematic representation of polyethyleneimine (PEI)‐modified mesoporous silica nanoparticles (MSNs) for directed differentiation of mouse embryonic stem cells (mESCs) into hepatocyte‐like cells (iHeps) in vitro and in vivo. Growth factors were adsorbed on the PEI‐MSNs to form a GF‐PEI‐MSN complex. The mESCs induced by GF‐PEI‐MSN complexes exhibited significantly improved differentiation toward hepatocyte‐like cells with mature functions in vitro, and the induced cells reconstitute damaged hepatic tissues after transplantation in vivo. 2 Results and Discussion 2. 1 Characterization of MSNs and PEI‐Modified MSNs Silica nanoparticles were synthesized by using a previously described method, 30, 31 followed by modification with PEI to obtain amino‐functionalized MSNs. Transmission electron microscope (TEM) images of these MSNs ( Figure 1 A) exhibited well‐ordered hexagonal mesopores even after PEI modification. In addition, PEI‐modified MSNs appear a thin polymer film covering the whole surface of MSNs ((Figure 1 B). The average hydrodynamic size of these MSNs was ≈110 nm as measured by dynamic light scattering and increased to about 210 nm for the PEI‐MSNs in phosphate‐buffered saline (PBS) (Table S1, Supporting Information). With the large number of amine groups for PEI, the surface ζ potential of PEI‐coated MSNs changed from strongly negative to positive (Table S1, Supporting Information), which provides PEI‐MSNs with a platform for the delivery of growth factors. Figure 1 TEM images of A, B) mesoporous silica nanoparticles (MSNs) and PEI‐modified mesoporous silica nanoparticles (PEI‐MSNs), black arrow: PEI coating; C) FT‐IR spectrum; D) Thermogravimetric analysis (TGA) curves of the MSNs and PEI‐MSNs, and E) The cumulative release profiles of PEI‐MSNs loaded with Activin A, aFGF, and HGF. Data are presented as means ± SD (spontaneous differentiation (control)). The Fourier transform infrared (FT‐IR) spectrum of the PEI‐MSNs displayed absorption band at 1466 and 2930 cm −1, attributed to the stretching vibration of C=N and C—H, indicating the presence of amine surface functional groups on the surface of the particles, which could interact with the growth factors. In contrast, the MSNs did not show obvious absorption peak in these wave numbers (Figure 1 C). In addition, the amount of PEI in the MSNs was estimated by thermogravimetric analyses (TGA). MSNs and PEI‐MSNs showed a weight loss of 4. 2 and 15. 5 wt%, respectively (Figure 1 D), showing the adsorbed PEI content at ≈11. 3 wt%. As shown in Table S2 (Supporting Information), after functionalization with PEI, specific surface areas ( a s, Brunauer–Emmett–Teller(BET)) of MSNs significantly decreases, while the pore volumes and pore size for PEI‐MSNs is similar to that of MSNs, meaning that PEI molecules were successfully adsorbed onto the MSNs. Furthermore, obvious cytotoxicity of MSNs or PEI‐MSNs was not observed using a CCK‐8 assay, indicating that the synthesized nanoparticles could serve as excellent carriers (Figure S1, Supporting Information). 2. 2 Loading and Release Evaluation of Growth Factor on PEI‐MSNs To further improve the differentiation efficiency of every step of hepatic differentiation (mESCs to DE cells, DE cells to hepatoblasts and hepatoblasts to hepatocyte‐like cells), we investigated the delivery of key growth factors that regulate hepatogenesis, including Activin A, aFGF, and HGF. Activin A has been reported to induce definitive endoderm differentiation of ESCs; 10, 32 DE cells constitute the embryonic germ layer that produces hepatic cells; 33 aFGF and HGF are essential for liver development, with aFGF efficiently initiating hepatic differentiation of ESCs from the definitive endoderm and HGF promoting hepatic growth. 34, 35 The novel growth factors delivery system was prepared by mixing Activin A, aFGF, and HGF with positively charged PEI‐MSNs. After the growth factors were added, the hydrodynamic diameters of the growth‐factor‐loaded PEI‐MSN composites increased to larger sizes than before, while the ζ ‐potential of these complexes decreased slightly. This indicates that the growth factors were successfully attached to the positively charged PEI‐MSNs, possibly through electrostatic interactions and hydrogen bonding between amino groups of PEI and functional groups of the growth factors, which allowed the growth factors to be released more readily in the biological system as reported previously. 36 In addition, the loading efficiency of Activin A, aFGF and HGF on the PEI‐MSNs was ≈82. 23% ± 7. 98%, 81. 30% ± 0. 18%, and 77. 91% ± 4. 57%, respectively. The in vitro release profiles of Activin A, aFGF, and HGF from the PEI‐MSN in PBS are shown in Figure 1 E. Activin A release started with an initial burst followed by a slow release without an evident plateau, which is indicative of ongoing release thereafter. aFGF release pattern started with an initial burst followed by a decline phase which entered a steady‐state release. HGF release started with an initial burst and then was followed by a steady‐state release. Therefore our results show that PEI‐MSNs were considered as a novel system for long‐term delivery of the growth factors. 2. 3 PEI‐MSNs Loaded with Growth Factors Facilitate mESCs Differentiation into Hepatocyte‐Like Cells We investigated whether PEI‐MSNs loaded with growth factors could promote the differentiation of mESCs into hepatocyte‐like cells according to the protocol described in Figure 2 A, B. A quantitative polymerase chain reaction (qPCR (polymerase chain reaction)) analysis was performed at various times during the culture process to determine the degree to which the mESCs differentiated toward a hepatocyte‐like cell phenotype. As shown in Figure 2 C, the expression of pluripotent marker octamer‐binding transcription factor 4 (Oct4) 37 gradually decreased with time, and a significantly greater reduction in mESCs was induced by growth‐factor‐loaded PEI‐MSN complexes compared with other groups (mESCs induced by PEI‐MSNs, growth factors alone, or spontaneously differentiated mESCs), which indicated a more rapid loss of stemness during differentiation of the mESCs treated with GF‐PEI‐MSN complexes. The expression of DE transcription factor markers SRY‐box containing gene 17 (Sox1) 38 and forkhead box A2 (FoxA2) 39 increased within 3 d of the differentiation process and slowly diminished thereafter in all groups; however, the levels were significantly higher in the mESCs treated with GF‐PEI‐MSN complexes compared with other groups. The gene expression of hepatocyte‐related markers, such as alpha‐fetoprotein (AFP) 40 and albumin (ALB), 40, 41 was strongly upregulated in the mESCs containing GF‐PEI‐MSN complexes from day 8 compared with the other three groups. The mRNA expression of hepatic functional markers and hepatic metabolic enzyme genes, such as α‐1‐antitrypsin (AAT), 42 glucose‐6‐phosphatase (G6P) 43 and members of the cytochrome P450 subunit CYP7A1, 44 showed a time‐dependent upregulation from day 13 and were highly expressed at day 18 in the hepatocyte‐like cells that were induced by the mESCs treated with GF‐PEI‐MSN complexes. Figure 2 A) Schematic representation of the developed multistep differentiation procedure. B) Phase contrast images of differentiating cells at days 3, 8, 13, and 18. Scale bar: 200 μm C) Temporal expression patterns of genes by qPCR analysis during the induction of mESCs treated with GF‐PEI‐MSN complexes, PEI‐MSNs and growth factors alone or treatment without PEI‐MSNs and any growth factors. Data were calculated in relation to the expression of the housekeeping gene GAPDH (glyceraldehyde phosphate dehydrogenase) and used as an internal standard and are shown as the expression relative to that of the undifferentiated mESCs using the comparative CT Method (2 −CT ). Results represent the mean ± SD ( n = 3). To confirm the in vitro hepatic differentiation of the mESCs and expression of the endoderm‐specific markers Sox17 and FoxA2 at day 3 and the liver‐specific markers cytokeratin 18 (CK18), AFP, and ALB at day 18, immunofluorescence staining was used to detect these markers ( Figure 3 A and single channel images shown in Figure S2 in the Supporting Information). There was a greater upregulation of the expression of these markers by GF‐PEI‐MSN complexes in the induced cells compared with the other three groups, consistent with the results of the qPCR analyses. Figure 3 Immunofluorescence and flow cytometry analysis of the cellular stage‐specific protein expression in differentiating cells in vitro at days 3 and 18 induced by GF‐PEI‐MSN complexes, PEI‐MSNs, growth factors alone, and without treatment. A) Fluorescent images of definitive endoderm (DE) cell markers Sox17 (green) and FoxA2 (red) and liver‐specific markers CK18 (red), AFP (green), and ALB (green) acquired using an inverted fluorescence microscope. Nuclei were counterstained with DAPI (blue). Scale bar: 200 μm. B) Quantitative depiction of flow cytometric analysis for Sox17, FoxA2, AFP, and ALB expression in the four groups. All data are represented as the mean ± SD ( n = 3). * P < 0. 05, ** P < 0. 01, and *** P < 0. 001. To further quantify the expression levels associated with DE and hepatocyte‐like cells, the mESC‐derived cells were analyzed for the expression of Sox17 and FoxA2 at day 3 and expression of AFP and ALB at day 18 using flow cytometry analysis (Figure 3 B; Figure S3, Supporting Information). The percentages of Sox17, FoxA2, AFP and ALB positive cells in the spontaneously differentiated cells and directly differentiated cells by PEI‐MSNs, growth factors alone, and GF‐PEI‐MSN complexes were shown in Table 1. These results clearly demonstrated that the percentages of cells positive for Sox17, FoxA2, AFP, and ALB in the directly differentiated cells treated with GF‐PEI‐MSN complexes were significantly higher than in the other groups (Figure 3 B). These data demonstrate that more hepatic differentiation resulted from the treatment with GF‐PEI‐MSN complexes and the delivery of growth factors by PEI‐MSNs can provide an efficient platform for improving definitive hepatic differentiation in vitro. This may be attributed to increasing the growth factor concentration and bioavailability or extending time course for sustainable release in GF‐PEI‐MSN complexes group. Table 1 The percentages of positive cells in the differentiated cells of different treatment Positive cells name Control (without treatment) [%] PEI‐MSNs [%] Growth factors [%] GF‐PEI‐MSN complexes [%] Sox17 8. 97 ± 0. 90 11. 35 ± 2. 17 48. 40 ± 4. 36 72. 03 ± 0. 60 FoxA2 32. 53 ± 1. 88 30. 47 ± 2. 27 47. 33 ± 5. 98 73. 00 ± 1. 23 AFP 27. 10 ± 5. 35 30. 80 ± 4. 94 65. 73 ± 1. 79 93. 27 ± 1. 42 ALB 5. 01 ± 1. 04 5. 46 ± 0. 39 12. 00 ± 0. 28 23. 20 ± 2. 69 John Wiley & Sons, Ltd. 2. 4 PEI‐MSNs Loaded with Growth Factors Promote Functional Hepatic Maturation We further examined whether PEI‐MSNs loaded with growth factor complexes can promote hepatic differentiation and attain mature liver function in the treated mESCs. Firstly, we tested the ability of the differentiating cell populations to store glycogen, which is an important function of mature hepatocytes, at day 18 using periodic acid‐Schiff (PAS) staining. 45 A higher number of positive cells exhibited a pink to red‐purple cytoplasm in the GF‐PEI‐MSN complex treated group ( Figure 4 A), whereas a small number of positive cells was detected in the different controls and treatments. In addition, the uptake of indocyanine green (ICG) and Dil‐labeled acetylated low density lipoprotein (Dil‐ac‐LDL) was investigated. ICG is a nontoxic organic anion that is eliminated exclusively by mature hepatocytes, the uptake, and release of ICG can be used to identify differentiated hepatocytes in vitro. 46, 47 The differentiated cells displayed a pronounced capacity to take up ICG in the GF‐PEI‐MSN complex‐treated group (Figure 4 B) compared with those of other groups. We also examined the capacity for Dil‐ac‐LDL uptake, which is a critical hepatocyte function. 48 As shown in Figure 4 C, the positive immunofluorescent signals for Dil‐ac‐LDL uptake in the GF‐PEI‐MSN complex group increased significantly in the hepatic lineage differentiated cells at day 18 compared with those of the control and other treatments. These results demonstrated that PEI‐MSNs delivering growth factors can promote efficient hepatic differentiation of mESCs and significantly improve the maturation of hepatocyte‐like cells with hepatic functionality. Figure 4 Functional tests of mouse embryonic stem cell (mESC)‐derived hepatocyte‐like cells (iHeps) induced by GF‐PEI‐MSN complexes, PEI‐MSNs, growth factors alone, and without treatment. A) The functions of glycogen synthesis and storage were measured by periodic acid‐Schiff (PAS) assays on the mESC‐derived cells in the four groups. Glycogen storage is indicated by pink or dark red‐purple cytoplasms. B) The cellular uptake function of indocyanine green (ICG), and C) low‐density lipoprotein (LDL) in the different treatment groups was analyzed using a fluorescence microscope at the end of the differentiation process. Scale bar: 50 μm. 2. 5 Functional Evaluation of Transplantation of mESC‐Derived Definitive Endoderm Cells into Mice with CCl 4 Injury To further test the function of transplanted mESC‐derived DE cells induced by PEI‐MSNs loaded with growth factors in vivo, we used a mouse transplantation model for hepatic repopulation following carbon tetrachloride (CCl 4 )‐induced liver injury (Figure S4, Supporting Information). Studies have reported that transplanted cells have a selective advantage over host hepatocytes in injured livers. 49 Therefore, the four types of DE cells were derived from three‐day treatment with PEI‐MSNs, growth factors only and with GF‐PEI‐MSN complexes or without treatment and injected intrasplenically to mice with CCl 4 Injury. After 2 d of cell transplantation, the livers were harvested, and sections were examined by hematoxylin and eosin (H&E) staining. All of the livers were characterized by acute focal necrosis, ballooning necrosis, steatosis, and inflammatory cell infiltration in the treated groups compared with the control group (Figure S5, Supporting Information). After four weeks, all of the cell transplantations had ameliorated the effects of chronic liver injury and fibrosis. In addition, the cell transplanted livers in the GF‐PEI‐MSN complex group showed improved liver architecture that was similar to normal mice liver and lessened or absent fibrosis compared with the other groups ( Figure 5 A). Additionally, similar findings were also confirmed by enzyme‐linked immunosorbent assay (ELISA) analyses, which showed that serum alanine aminotransferase (ALT) (Figure 5 B) and aspartate aminotransferase (AST) (Figure 5 C) levels had decreased significantly with growth factors alone and the GF‐PEI‐MSN complex groups compared with the other groups. Moreover, the serum levels of ALT and AST returned to basal levels of normal mice or sham mice in the GF‐PEI‐MSN complex group. These findings support the conclusion that growth factors delivered by PEI‐MSNs improved the rescuing efficiency in CCl 4 ‐injured liver. Figure 5 Evaluation of transplantation of mESC‐derived DE cells by GF‐PEI‐MSN complexes, PEI‐MSNs, growth factors alone, and without treatment at four weeks in the recovery of the CCl 4 ‐injured mouse model. A) Hematoxylin and eosin (H&E) staining was performed on mouse liver sections in the different treatment groups. Scale bar: 50 μm. B) Serum ALT and C) AST levels in the different groups were measured by enzyme‐linked immunosorbent assays. All data are represented as the mean ± SD ( n = 3). * P < 0. 05, ** P < 0. 01, and *** P < 0. 001. Abbreviations: Normal, untreated normal mice; Sham, CCl 4 was not administered; CCl 4, CCl 4 group; PBS, PBS group. To date, numerous studies in rodents have shown that predifferentiated ESC transplantation can reverse acute fulminant hepatic failure. 6, 50, 51, 52, 53, 54 In our study, we employed day‐3‐induced DE cells as a source of isolated cells for transplantation into injured liver. The results showed DE cells were able to differentiate into hepatocyte‐like cells and attenuate AST and ALT levels after transplantation into injured liver, which confirmed the results of related reports. 55, 56, 57 To our knowledge, this is the first report demonstrating that PEI‐MSNs can be used as a delivery system for growth factors to improve DE cells from mESCs in vivo and significantly restore liver function. Although further studies will be necessary to determine the most suitable stages of mESCs differentiation procedure for transplantation, these results suggest the potential application of DE cells differentiated from mESCs for therapy of liver injuries or diseases. 2. 6 Engraftment and Derivation of Transplanted Cells Efficient cell engraftment and retention is critical for successful cell‐based therapy. To trace the homing of transplanted cells, liver sections were examined by in vivo bioimaging ( Figure 6 A, B) and immunofluorescence (Figure 6 C). Four weeks after transplantation, the red fluorescent signal of CM‐Dil‐labeled DE cells (chloromethylbenzamido(Cell Tracker CM‐DiI))was still detected in the mice livers (Figure 6 A). Furthermore, mESC‐derived DE cells treated with the GF‐PEI‐MSN complexes for 3 d exhibited a greater homing ability to detect in the injured liver compared with the other groups (Figure 6 B). These results were further confirmed by the CM‐Dil‐labeled cells detected in the recipient livers (Figure 6 C and single channel images shown in Figure S6 in the Supporting Information), which also revealed that more cells induced by GF‐PEI‐MSN complexes were engrafted into host livers. Figure 6 Cell detection of transplanted mESC‐derived DE cells from the different treatments in the CCl 4 injured mouse model. A) Cells were labeled with CM‐Dil prior to transplantation, and after four weeks, the transplanted cells were detected in the whole mouse host liver using an in vivo bioimaging system. B) Quantitative analysis of the transplanted cells from the different treatments shown in A. All data are represented as the mean ± SD ( n = 3), * P < 0. 05. C) Representative fluorescence images of the liver sections at four weeks after transplantation of CM‐Dil labeled cells. Inset, high magnification images of CM‐Dil labeled cells. Scale bar: 100 μm. Red, CM‐Dil; blue, DAPI. To evaluate hepatocyte differentiation of engrafted cells in the livers, ALB was used as a marker of mature hepatocytes, and the transplanted cells were evaluated by immunofluorescence staining. Although a small number of double CM‐Dil + /ALB + cells were detected in the livers of all recipients ( Figure 7 A), there was a higher percentage of double CM‐Dil + /ALB + cells (3. 62% ± 0. 19%) in the GF‐PEI‐MSN complex group compared with the growth factors (2. 74% ± 0. 22%), PEI‐MSNs (2. 22% ± 0. 23%), and control groups (1. 48% ± 0. 28%) (Figure 7 B). Moreover, under our conditions, teratoma formation (Figure S7, Supporting Information) was not observed in a series of grafts, although further studies using different animal models of liver diseases are needed to address the long‐term safety and efficacy of DE cells. These findings demonstrate that PEI‐MSNs carrying growth factors can act as an efficient platform to improve engraftment of transplanted DE cells to an injured liver and promote differentiation into hepatocyte‐like cells in vivo compared with controls and other treatments. Figure 7 Cell differentiation of transplanted mESC‐derived DE cells from the different treatments in vivo for hepatic repopulation. A) The transplanted livers were detected by fluorescence microscopy staining in the different groups. The exogenous origin of the differentiated hepatocyte‐like cells was confirmed by costaining for CM‐Dil (red) and ALB (green). The nuclei were counterstained with DAPI (blue). Scale bar: 100 μm. Inset and white arrowhead, high magnification images of differentiated cells. Representative field showing no CM‐Dil staining in a nontransplanted liver section (CCl 4 or PBS). B) Quantification of differentiated CM‐Dil+/ALB+ cells shown in A. All data are represented as the mean ± SD ( n = 3). * P < 0. 05 and ** P < 0. 01. Two primary mechanisms may explain the therapeutic effects of transplanted cells in injured liver. 58 First, ESCs could generate functional hepatocytes in vivo that are then engrafted efficiently within the host liver. 6 A study reported that 2. 5%–5% is necessary for reversing the liver injuries. 59 We also observed additional CM‐Dil + /ALB + cells in the GF‐PEI‐MSN complex group; however, low amounts of these cells could not significantly reverse the injured liver, which was mentioned by previous investigators. 60 One explanation for this low differentiation level may be that limited space is available for the donor cells to integrate into the CCl 4 ‐treated mouse, 61 and they do so with low efficiency. The second mechanism of promoting liver repair is indirect. Secretomes released from stem cells or their derivatives contribute to liver regeneration in response to acute damage. 62 Our study showed that a few engrafted cells expressed ALB in recipient livers, whereas these cells can improve liver function in a mouse disease model, which may also be related to the paracrine effect of predifferentiated cells. 3 Conclusions Stem‐cell‐derived hepatocyte transplantation is currently being evaluated as a potential method of providing metabolic support during acute and chronic liver failure, but is restricted due to inefficient differentiation and low purity of functional hepatocytes. In this study, we have successfully established a polymer‐modified nanoparticles‐based sustained delivery system for growth factors to direct stem cell differentiated into hepatocytes. Our findings show that this approach can help to overcome the limitations associated with current models and ensure efficient delivery of growth factors to improve mESC differentiation toward a hepatocyte‐like lineage with mature liver functions in vitro, including glycogen storage, indocyanine green and low‐density lipoprotein uptake. When transplanted into mice with liver injury, this system significantly repopulated the damaged liver, which was attributed to transplanted cells and the paracrine effect of predifferentiated cells. Therefore, MSNs with multifunctional surface properties are suitable delivery platforms for biomolecule delivery to induce specific differentiation and cell‐based therapies for treatment of hepatic disease and tissue engineering, which provides a powerful system not only for efficient differentiation of the stem cell, but also for developing therapeutic strategies in regenerative medicine. 4 Experimental Section Materials : Tetraethylorthosilicate (TEOS) and PEI (25 kD) were purchased from Sigma‐Aldrich (St. Louis, MO, USA). Cetyltrimethylammonium bromide (CTAB) was purchased from Alfa Aesar (Tianjin, China). All tissue culture materials were obtained from Gibco (Grand Island, NY, USA), and the other reagents (analytical grade) were purchased from Sigma‐Aldrich. All of the chemicals were of analytical grade and used without further purification. Animals : Experiments were performed using male ICR mice weighing 20–25 g and aged six to eight weeks (Laboratory Animal Center of Tongji University). All of the animal research procedures were approved by the Animal Experimentation Committee of Tongji University, and the animals were cared for in accordance with the Guidelines for Animal Experiments of Tongji University. Synthesis and Surface Modification of Mesoporous Silica Nanoparticles : The synthesis of MSNs was performed according to an earlier publication. 30, 31 Briefly, 300 mg CTAB was dissolved in a solution of 145 mL distilled water and 1. 05 mL sodium hydroxide (2. 0 m ). The solution was heated to 80 °C, and then TEOS (2. 5 mL) was added. The solution was vigorously stirred, and the reaction mixture was then stirred for 2 h. The resulting suspension was centrifuged and washed with water and ethanol several times. The precipitate was dried at 60 °C. To remove the surfactant template (CTAB), the samples were calcined for 5 h at 550 °C. PEI was coated onto the MSNs according to procedures described in an earlier publication. 19 MSNs (10 mg) were added to a solution containing 5 mg PEI (MW 25 kD) and 1 mL absolute ethanol. After the mixture was sonicated and stirred for 30 min, PEI‐coated MSNs were washed with ethanol and water. Characterization : The shapes and structures of the MSNs were characterized using a TEM (JEM 2011, JEOL, Japan). The particle sizes and ζ potentials were determined in PBS using a Malvern Zetasizer (Nano Series, Malvern Instruments Inc. , MA, USA). FT‐IR absorption spectra were recorded on a Nicolet Nexus 470 spectrometer. The total amount of PEI coating was determined by TGA (TGA7, PerkinElmer, USA). The MSN pore size, volume, and specific surface area were determined by nitrogen sorption measurements (Micromeritics TriStar 3000 analyzer, Micromeritics, USA). The surface areas and pore size distributions of the samples were calculated by the BET and Barrett−Joyner−Halenda (BJH) methods, respectively. Growth Factor Loading and Release : PEI‐coated MSNs were loaded with Activin A (R&D Systems, Minneapolis, MN, USA), aFGF (PeproTech, Rocky Hill, NJ, USA), and HGF (R&D Systems) by incubating 10 mg nanoparticles (sterilized under UV light) in a solution of 12 μg mL −1 Activin A, 5 mg sterilized nanoparticles in a solution of 1. 8 μg mL −1 aFGF, and 5 mg sterilized nanoparticles in a solution of 1. 3 μg mL −1 HGF, respectively, for 24 h at 4 °C in a thermomixer comfort (Eppendorf, Hamburg, Germany) with constant shaking at 1000 rpm. Following repetitive washings of the growth‐factor‐laden nanoparticles and removal of unloaded growth factors by centrifugation, the precipitates, Activin A‐PEI‐MSNs, aFGF‐PEI‐MSNs, and HGF‐PEI‐MSNs were redispersed in PBS for use in future experiments. The free growth factors present in the supernatant were determined using an ELISA according to the manufacturer's instructions. The growth factor loading efficiency of nanoparticles was calculated using the following equation Loading efficiency = ( total growth factor − free growth factor ) total growth factor In the release experiments, the growth‐factor‐laden nanoparticles prepared were placed in PBS (5 mL) and incubated under mild agitation at 37 °C. At predetermined time intervals, the particle suspension was centrifuged (at 10 000 rpm for 5 min) and the supernatant (4 mL) was removed and replaced by a new one. The amount of growth factors released was quantified by ELISA (R&D Systems). All analyses were conducted in duplicate. Mouse Embryonic Stem Cell Cultures : Undifferentiated mESC D3 cells were obtained from Professor Xiaoqing Zhang, School of Medicine, Tongji University. mESC D3 cells were maintained on irradiated mouse embryonic fibroblast feeder cells in 0. 1% gelatin‐coated dishes in Dulbecco's modified Eagle's medium (DMEM) supplemented with 15% (V/V) fetal bovine serum (FBS), 1% (V/V) non‐essential amino acids, 1 × 10 −3 m GlutaMAX, 0. 1 × 10 −3 m β‐mercaptoethanol (all from Gibco), and 1000 U mL −1 recombinant mouse leukemia inhibitory factor (LIF; Millipore, CA, USA). Media were changed every day. Hepatic Differentiation In Vitro : To evaluate the effect of GF‐PEI‐MSN complexes on hepatocyte differentiation, the mESCs were cultured according to previously published protocols with minor modifications as described in Figure 2 A. Before the initiation of cellular differentiation, the mESCs were dissociated into single cells at a seeding density of 1 × 10 5 cells mL −1 and cultured in 12‐well tissue culture plates (Corning, NY, USA) coated with 0. 1% gelatin (Sigma‐Aldrich) and without a feeder layer. To direct the differentiation of mESCs into hepatocytes, differentiation was induced by treating mESCs with differentiation medium that consisted of Glasgow minimum essential medium (GMEM, Gibco) supplemented with 2% FBS and Activin A‐PEI‐MSN complexes (100 μg mL −1 PEI‐MSNs and 100 ng mL −1 Activin A) for 3 d followed by treatment with differentiation medium consisting of GMEM supplemented with 10% FBS, 2. 5 × 10 −3 m sodium butyrate (Sigma‐Aldrich) and aFGF‐PEI‐MSN complexes (100 μg mL −1 PEI‐MSNs and 30 ng mL −1 ) for 5 d. The cells were further cultured in the maturation medium, which consisted of GMEM media supplemented with HGF‐PEI‐MSN complexes (100 μg mL −1 PEI‐MSNs and 20 ng mL −1 HGF) for 5 d and then followed by 5 d in 10 ng mL −1 oncostatin M (OSM; R&D Systems) plus 0. 1 × 10 −3 m dexamethasone (Dex; Sigma‐Aldrich). Experimental groups included mESCs with PEI‐MSNs loaded with Activin A, aFGF or HGF, whereas control groups included mESCs cells with or without an equal quantity PEI‐MSNs and growth factors only. RNA Extraction and Quantitative Reverse Transcriptase Polymerase Chain Reaction : Total RNA was extracted from cells on different differentiation days using RNAiso Plus (TaKaRa Bio Inc, Japan) and treated with Recombinant DNase I (RNase‐free) (TaKaRa Bio Inc, Japan) to remove genomic DNA contamination following the manufacturer's protocol. A total of 1 μg RNA was reverse transcribed into cDNA in a volume of 20 μL with M‐MLV ReverseTranscriptase (Promega, WI, USA) according to the manufacturer's instructions. qPCR analysis was performed on a BioRad iQ5 Real‐Time PCR System using the SYBR Green qPCR Master Mix (Bio‐Rad, California, USA). The PCR reaction consisted of 10 μL 2× SYBR Green PCR Master Mix, 1 μL 5 × 10 −6 m forward and reverse primers, 8 μL water, and 1. 0 μL template cDNA in a total volume of 20 μL. Conditions for PCR amplifications were as follows: 95 °C for 5 min, 40 thermal cycles of 95 °C for 30 s, 60 °C for 30 s, 72 °C for 30 s, and a final extension at 72 °C for 10 min. The specific primers used for the qPCR are listed in Table S3(Supporting Information). Each qPCR quantification experiment was performed in triplicate for each individual sample. The final results were reported as the relative expression normalized with the transcript level of the housekeeping gene glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) using the comparative CT Method (2 −CT ). Immunofluorescence Staining : Cells were fixed for 20 min at 4 °C in 4% paraformaldehyde and then washed three times in PBS. Cells were permeabilized in 0. 2% Triton X‐100 and then blocked with 10% goat serum in PBS for 20 min before incubation overnight at 4 °C with primary antibody diluted in 1% goat serum in PBS as follows: mouse anti‐Sox17 (1:100, R&D Systems), rabbit anti‐FoxA2 (1:500, Millipore), mouse anti‐CK18 (1:100, Abcam), goat anti‐AFP (1:50, Santa Cruz), and sheep anti‐ALB (1:100, Abcam). The cells were then washed three times in PBS and incubated with secondary antibodies, including rabbit anti‐mouse Alexa Fluor 488 conjugated (1:100), goat anti‐rabbit PE‐conjugated (1:200), rabbit anti‐mouse PE conjugated (1:200), rabbit anti‐goat fluorescein isothiocyanate conjugated (1:100), or rabbit anti‐sheep fluorescein isothiocyanate conjugated (1:100) (Jackson ImmunoResearch) for 1 h at room temperature. The cells were washed three times for 5 min each with PBS, counterstained with 4, 6‐diamidino‐2‐phenylindole (DAPI) (0. 25 μg mL −1, Molecular Probes) for 10 min, rinsed with PBS three times for 15 min and observed using an inverted fluorescence microscope (Leica DMI 4000B, Heerbrugg, Switzerland). Flow Cytometry : The cells were harvested by digestion with 0. 125% trypsin/ethylenediaminetetraacetic acid (EDTA), fixed with 4% paraformaldehyde for 30 min, and then permeabilized in staining buffer (PBS with 10% bovine serum albumin (BSA, Sigma‐Aldrich) and 0. 2% Triton X‐100) for 10 min. The cells were then incubated for 1 h with primary antibody against Sox17 (1:20, R&D Systems), FoxA2 (1:50, Millipore), AFP (1:200, Santa Cruz), or ALB (1:20, Abcam). After washing, the cells were incubated for 30 min with Alexa Fluor 488‐conjugated anti‐rabbit or anti‐mouse secondary antibody (Jackson ImmunoResearch) diluted to 1:100. Afterward, the cells were washed three times with PBS buffer. Finally, the cells were resuspended in 0. 5 mL ice‐cold PBS and analyzed using a Calibur flow cytometer (Becton Dickinson, CA, USA). Data were analyzed with the software FlowJo (Tree Star, version 7. 6). Periodic Acid‐Schiff Staining : The PAS staining system was purchased from Sigma‐Aldrich. Culture dishes containing cells were fixed in 4% paraformaldehyde, and the assay was performed according to the manufacturer's instructions. Indocyanine Green and Low‐Density Lipoprotein Uptake : For ICG (Sigma‐Aldrich) uptake assay, ICG was suspended in dimethyl sulfoxide (DMSO) (Sigma‐Aldrich) and a stock solution at 5 mg mL −1 and freshly diluted in culture medium to a final concentration of 1 mg mL −1. The cells were incubated in diluted ICG for 30 min at 37 °C followed by washing with PBS three times. The cells were returned to the culture medium, and the release of ICG was evaluated 6 h later. For the LDL uptake test, the differentiated cells were incubated in DMEM containing 10 μg mL −1 acetylated low‐density lipoprotein labeled with 1, 1′‐dioctadecyl‐3, 3, 3′3 ′‐tetramethylindocarbocyanine perchlorate (Dil‐Ac‐LDL, Invitrogen, USA) for 4 h at 37 °C. The cells were washed three times for 5 min each with PBS, counterstained with DAPI for 10 min, rinsed with PBS three times, and visualized using an inverted fluorescence microscope (Leica DMI 4000B). Animal Treatment and Cell Transplantation : To induce liver fibrosis, each mouse was administered intragastrically 1 mL kg −1 body weight 20% (V/V) CCl 4 (Chemical Reagent Company, Shanghai, China) dissolved in corn oil (Alfa Aesar, Tianjin, China) three times per week for four weeks. The day‐3 differentiated cells from different treatments were trypsinized at 37 °C with 0. 125% trypsin/EDTA and then resuspended in PBS. Prior to implantation, the cells were labeled with CM‐Dil (Life Technologies) according to the manufacturer's recommendations. Approximately 1 × 10 6 cells in 0. 1 mL suspension were injected intrasplenically into ICR mice ( n = 5). In the normal ( n = 5) groups and sham treatment ( n = 5) groups, CCl 4 was not administered. In the sham group, the mice were injected with the vehicle (corn oil) alone (three times per week for eight weeks). In the CCl 4 group, ICR mice were randomly divided into six groups: (1) mice injected with CCl 4 alone without cells; (2) mice injected with 0. 1 mL PBS (used as negative controls); (3) mice administered spontaneously differentiated mESCs cultured in the absence of PEI‐MSNs or any growth factors; (4) mice administered differentiated mESCs treated with PEI‐MSNs alone; (5) mice administered differentiated mESCs treated with growth factors alone; and (6) mice administered differentiated mESCs treated with GF‐PEI‐MSN complexes. Each experimental group contained five mice. Histological analysis of liver tissues was conducted by serial tissue section at two days and four weeks after cell transplantation. The blood at four weeks was harvested for further analysis. To observe the fate of transplanted cells in mice with liver injury, in vivo imaging was performed using a NightOWL imaging system and WinLight software (Berthold Technologies, Germany). Histology and Immunofluorescence : The livers were fixed in PBS containing 4% formaldehyde, embedded in paraffin, and sectioned into 8 μm sections. Sample sections were stained with H&E. For immunofluorescence staining, the livers were fixed and embedded in OCT compound (Tissue‐TEK, Sakura Finetek, CA, USA). After fixation, cryosections (8 μm) were incubated with ALB primary antibodies (1:100, Abcam) at 4 °C overnight, and the sections were then incubated with secondary antibodies at room temperature for 1 h. Finally, samples were counterstained with DAPI for 10 min. Sample sections were visualized and imaged using a fluorescence microscope. Aminotransferase Analysis : Blood samples were centrifuged at 3000 × g for 10 min to separate the serum, and they were then stored at −80 °C for subsequent analyses. ALT and AST were determined using commercial enzymatic kits (Jiancheng, Nanjing, China). Statistical Analysis : The results are given as the mean ± SD. An unpaired t ‐test or one‐way analysis of variance with Bonferroni post‐test was performed with the software GraphPad Prism 5. 0 (San Diego, CA, USA). The results were considered significant at P < 0. 05 (*), very significant at P < 0. 01 (**), and extremely significant at P < 0. 001 (***). Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201500413
| 2,016
|
Advanced Science
|
Two‐Dimensional Fluorinated Graphene: Synthesis, Structures, Properties and Applications
|
Fluorinated graphene, an up‐rising member of the graphene family, combines a two‐dimensional layer‐structure, a wide bandgap, and high stability and attracts significant attention because of its unique nanostructure and carbon–fluorine bonds. Here, we give an extensive review of recent progress on synthetic methods and C–F bonding; additionally, we present the optical, electrical and electronic properties of fluorinated graphene and its electrochemical/biological applications. Fluorinated graphene exhibits various types of C–F bonds (covalent, semi‐ionic, and ionic bonds), tunable F/C ratios, and different configurations controlled by synthetic methods including direct fluorination and exfoliation methods. The relationship between the types/amounts of C–F bonds and specific properties, such as opened bandgap, high thermal and chemical stability, dispersibility, semiconducting/insulating nature, magnetic, self‐lubricating and mechanical properties and thermal conductivity, is discussed comprehensively. By optimizing the C–F bonding character and F/C ratios, fluorinated graphene can be utilized for energy conversion and storage devices, bioapplications, electrochemical sensors and amphiphobicity. Based on current progress, we propose potential problems of fluorinated graphene as well as the future challenge on the synthetic methods and C‐F bonding character. This review will provide guidance for controlling C–F bonds, developing fluorine‐related effects and promoting the application of fluorinated graphene.
|
1 Introduction Since Andre Geim and Kostya Novoselov first isolated high‐quality few‐atom‐thick nanosheets (including single‐layer) from graphite, the ability to prepare graphene and its derivatives have triggered intense research in two‐dimensional nanomaterials all over the world. 1 Subsequently, graphene‐based materials receive much attention in nanotechnology because of their extraordinary properties, such as an ultrahigh theoretical specific surface area (2630 m 2 g −1 ), exceptional charge carrier mobility (200 000 cm 2 V −1 s −1 ), high thermal conductivity (≈5000 W m −1 K −1 ), high optical transmittance (≈97. 7%). 2 Despite these aforementioned superiorities, pristine graphene suffers from several shortcomings including structural defects, chemical inertness and a zero bandgap. Thus, many functionalization methods such as chemical bonding, loading or generating functional groups or free radicals on graphene (or its derivatives) have been utilized to improve structural integrity, surface activity and processability. 3 The functionalization not only inherits unique carbon conjugated structures but also brings about a promise to alter the graphene's properties including dispersion, orientation, interaction and electronic properties. 4 Graphene oxide (GO) 5 and halogenated graphene (CX m, X = F, Cl, Br, or I), [[qv: 4a]], 6 typical members of graphene derivatives including fluorographane 7 and thiofluorographene, 8 have thousands of oxygen functional groups or halogen atoms on carbon nanosheets with the transition of carbon atoms from sp 2 to sp 3 hybridization. The chemical modification endows graphene with many excellent properties, such as good dispersion in organic/water solvents, chemical activity on the surface via functional groups and tunable electronic properties, such as bandgap opening, charge transfer density and work functions. Despite numerous studies, GO and halogenated graphene continue to have several problems as follows: (1) GO has a variety of chemical bonds containing carboxyl, carbonyl, hydroxyl, lactone, and epoxide at graphene edge and basal‐plan. The amount and concentration of these functional groups are not controlled. (2) Halogenated graphene usually is a mixture of nanosheets with different degrees of substitution and different halogen‐carbon bonds. (3) GO and halogenated graphene show less chemical and thermal stability than graphene as a result of a great amount of defects or substituents on the surface, especially for brominated‐ and iodine‐doped graphene. The representative characteristics of halogenated graphene were shown in Table 1. 9 Table 1 The empirical characteristics of halogenated graphene 9 CF x CCl x CBr x CI x Value of x ≈0–1. 12 ≈0–0. 43 ≈0–0. 050 ≈0–0. 031 Synthesis methods Fluorination of graphene Exfoliation methods Chlorination of graphene Modification by bromine Doping by iodine Calculated bandgap when x = 1 ≈3. 1 eV ≈0. 9 eV Almost no bandgap — Stability at room temperature Stable unstable unstable unstable C–X in IR ≈1050–1250 cm −1 ≈790 cm −1 ≈600 cm −1 ≈745 cm −1 C–X in XPS ≈684–689 eV (F 1s) ≈200. 5 eV (Cl 2p 3/2 ) ≈202. 0 eV (Cl 2p 1/2 ) ≈71 eV (Br 3d) ≈184. 0 eV (Br 3p 3/2 ) ≈191. 0 eV (Br 3p 1/2 ) ≈619. 5 eV (I 3d 3/2 ) ≈631. 9 eV (I 3d 5/2 ) John Wiley & Sons, Ltd. Fluorinated graphene is regarded as the two‐dimensional basic structural element of fluorinated graphite first synthesized by Ruff et al. in 1934. 10 Subsequently, fluorinated graphite recevies much attention in self‐cleaning, solid lubricants, superhydrophobic coating, and the electrode of electrochemical cell because of its extremely low surface energy, good chemical and thermal stabilities, and high electromotive force (4. 57 V at 25 °C calculated by thermodynamic data) in lithium‐fluorinted graphite battery. 11 A typical method of preparing fluorinated graphite is the fluorination in fluorine‐containing atmosphere, and the F/C ratios and the C‐F bonds (covalent, semi‐ionic or ionic) usually depend on the fluorination conditions including the pressure, temperature and treatment time in fluorine‐containing atmosphere. Moreover, 2D fluorinated graphene with single or few layers can be obtained by exfoliating fluorinated graphite via mechanical or liquid phase exfoliation methods. High quality fluorinated graphene offers a great potential for modulating various properties by controlling the microstructures (layer, size and surface chemistry). Fluorinated graphene (CF x, x ≈ 0–1. 12), which is a stable and wide‐bandgap nanosheet in which a certain amount of C atoms is covalently bonded to F atoms, becomes a rising star of graphene derivatives because of its outstanding properties, such as a large negative magnetic resistance (a factor of 40 in a 9 T field), a wide optical bandgap (3. 8 eV) and a high room‐temperature resistance (>10 GΩ). [[qv: 4a]], 12 Fluorographene (fully fluorinated graphene, CF) is defined by Rahul R. Nair[[qv: 4a]] as a carbon monofluoride of graphene with the F/C ratio of 1. 0, which is also introduced or accepted by many groups. 13 Compared with other derivatives, fluorinated graphene shows many unique properties because of the formation of various types of C‐F bonds. First, because the F atoms has a higher electronegativity (4. 0) than C (2. 5), H (2. 2), and O atoms (3. 4), fluorinated graphene show great potential for using as an atomically thin insulator or a tunnel barrier based on the heterostructure. [[qv: 4a]] Second, because of the difference in electronegativity (1. 5) between C and F atoms, fluorinated graphene exhibits several C‐F bonding characters from ionic, semi‐ionic to covalent bonds controlled by the fluorination conditions. 14 The C‐F bonding character depends on the fluorination levels according to theoretical calculation. 15 Third, fluorinated graphene is regarded as an excellent cathode material for high‐energy lithium batteries because of its ability to electrochemically store and release high‐density energy (theoretical energy density of Li/CF 1. 0 is 2162 Wh kg −1 ). 16 Thus, a Li/CF x battery shows high energy densities, good chemical stability, a long‐term shelf life (>10 years) and minimal (<10%) self‐discharge. 17 In particular, the theoretical specific capacity of Li/CF x ( x = 1) is 865 mAh g −1 with an average discharge potential between 4. 5 and 5 V for a purely ionic C–F bond. 18 Fourth, covalent C‐F bonds show a high response to biological signals because of the high orientation and polarity of the C‐F bond. 19 Thus, fluorinated graphene can be developed for various biological applications, such as promoting neuro‐induction of stem cells[[qv: 19b]] or as a single multimodal material for magnetic resonance imaging. 20 Finally, C‐F bonds on the nanosheets greatly increase the hydrophobicity with an extreme low surface energy resulting in a super‐hydrohophic or amphiphilic film. 21 Recently, much progress has been made on the preparation and control of C‐F bonding characters, F/C ratios (the F/C ratio is defined as the mole ratio of fluorine to carbon) and configurations of fluorinated graphene. The brief roadmap of a synthetic strategy of fluorinated graphene is shown in Figure 1. However, a comprehensive review about the relationship between C‐F bonding character and various properties of fluorinated graphene has not yet been reported. In this review, we present the recent progress and advances on synthesis methods, C‐F bonding character, properties (bandgap, optical properties, stability, electronic conductivity, dispersibility, magnetic, tribological, mechanical (micromechanical) properties and thermal conductivity) and applications in energy conversion and storage devices, biological devices, quantum dots, supercapacitors and amphiphilic coating of fluorinated graphene. This review provides guidance for regulating a variety of properties and performances of fluorinated graphene based on designing and controlling its C‐F bonding character, F/C ratios and configuration. A strategy of structural design, potential problems and present/future challenges of fluorinated graphene are also proposed. Figure 1 Timeline showing recent synthetic methods regarding fluorinated graphene. Reproduced with permission. [[qv: 4a]] Reproduced with permission. [[qv: 12a]] Copyright 2010, American Chemnical Society. Reproduced with permission. [[qv: 24a]] Copyright 2014, Elsevier. Reproduced with permission. [[qv: 24b]] Copyright 2013, Royal Society of Chemistry. Reproduced with permission. [[qv: 25a]] Copyright 2012, American Chemical Society. Reproduced with permission. [[qv: 26a]] 2 Methods for Synthesizing Fluorinated Graphene The methods for synthesizing fluorinated graphene or fluorographene are mainly classified into two groups: fluorination and exfoliation methods. Fluorination mainly include direct gas‐fluorination, [[qv: 4a]], [[qv: 12a]], [[qv: 13b]], [[qv: 21a]], 22 plasma fluorination, 23 hydrothermal fluorination, 24 and photochemical/electrochemical synthesis. 25 Exfoliation methods includes sonochemical exfoliation, [[qv: 9a]], 26 modified Hummer's exfoliation, 20, [[qv: 21b]] and thermal exfoliation. 27 F/C ratios of fluorinated graphene prepared by different methods are summarized in Table 2. Table 2 Synthetic methods and preparation conditions for fluorinated graphene Synthetic method F/C ratios Reaction temperature Reaction time Ref Graphene‐based materials Fluorine agents/exfoliation solvents Direct gas‐fluorination Graphene membranes XeF 2 ≈0–1. 00 70 °C ≈1h–2 weeks [[qv: 4a]] Graphene films XeF 2 ≈0–0. 25 for single‐sided fluorination 30 °C ≈30–1200 s [[qv: 12a]] ≈0–1. 00 for double‐sided fluorination Graphene sheets XeF 2 ≈0–1. 00 350 °C ≈1–5 day [[qv: 13a]] GO F 2 ≈0–1. 02 From RT to 180 °C 20 min [[qv: 22a]] Highly ordered pyrolytic graphite F 2 ≈0. 70 600 °C 36–48 h [[qv: 22c]] Plasma fluorination CVD graphene SF 6 plasma — RT 6 s [[qv: 23a]] Epitaxial graphene SF 6 plasma ≈0. 10 RT 30 s [[qv: 23g]] GO CF 4 plasma 0. 17–0. 27 RT 5–20 s [[qv: 23k]] Graphene sheets CF 4 plasma — RT 3–20 min [[qv: 23i]] CVD graphene CF 4 plasma 0. 04–0. 18 RT to 200 °C 1–20 min [[qv: 23e]] Graphene Ar/F 2 plasma 0. 17 for 3 min RT 0. 5–30 min [[qv: 23h]] Hydrothermal fluorination GO dispersion HF (40 wt%) 0. 11–0. 48 150–180 °C 10–30 h [[qv: 24c]] GO films diethylaminosulfur trifluoride 0. 04–0. 05 0 °C or RT or 50 °C 17 h [[qv: 24a]] GO anhydrous BF 3 ‐etherate 0. 39 60 °C 24 h [[qv: 24b]] Photochemical synthesis CVD graphene CYTOP — RT — [[qv: 25a]] GO HF aqueous solution 0. 33 RT 48 h [[qv: 25c]] Electrochemical synthesis Graphite hydrofluoric acid aqueous solution (50 wt%) ≈0. 10 RT 6000 s [[qv: 25b]] Sonochemical exfoliation Fluorinated graphite sulffolane 1. 00 50 °C 1 h [[qv: 9a]] Fluorinated graphite 1‐butyl‐3‐methylimidazolium bromide 0. 25 or 0. 50 RT 3 h [[qv: 26a]] Fluorinated graphite NMP 0. 78–0. 31 RT 16–100 h [[qv: 26b]] Fluorinated graphite Chloroform or acetonitrile ≈0. 90 RT 6 h [[qv: 26c]] Fluorinated graphite cetyl‐trimethyl‐ammonium bromide and dopamine 0. 25 RT 10 min [[qv: 26d]] Modified Hummer's exfoliation Graphite fluorinated polymer H 2 SO 4 /H 3 PO 4 0. 07–0. 36 50 °C 2 h [[qv: 21b]] Thermal exfoliation Fluorinated graphite (CF 0. 57 ) — 0. 03–0. 40 480 °C 30 min 27 John Wiley & Sons, Ltd. 2. 1 Fluorination Methods 2. 1. 1 Direct Gas‐Fluorination Fluorographene was prepared by Nair et al. [[qv: 4a]] using XeF 2 gas to treat a graphene film at 70 °C ( Figure 2 a). The fluorination using XeF 2 gas is one of widely used technique to prepare fluorographene with different fluorinated structures becuase of mild and controllable process. The resultant fluorographene showed high thermal stability up to 400 °C even in an atmospheric environment. Subsequently, the fluorination process was investigated by Raman spectroscopy (Figure 2 b). An increase in the D band at 1350 cm −1 and a decreased 2D band at 2680 cm −1 indicated that the fluorination degree of graphene increased with a long XeF 2 treatment time. Fluorographene was obtained until all D, 2D, and G bands disappeared. [[qv: 4a]] Furthermore, fluorographene was also prepared by fluorinating graphene grown by chemical vapor deposition (CVD) on the Si substrate using XeF 2 gas at room temperature. [[qv: 12a]] Fluorographene showed a dominant stoichiometry of C 1. 0 F 1. 0 and a high F/C ratio of graphene film on both the front and back surface because of the effective etching on the Si substrate by XeF 2 gas. This effect was confirmed by fluorinated graphene (CF 0. 25 ) on Cu foils because the Cu substrate cannot be etched by XeF 2 (Figure 2 c). Raman spectra showed the D, D' and D+D' peak, while G peak was broadened by the exposure to XeF 2. This result indicated the introduction of a high degree of structural disorder in the fluorinated graphene. [[qv: 12a]] In addition, fluorographene was also synthesized by treating graphene sheets in XeF 2 at 350 °C for 1 and 5 days in an inert atmosphere. [[qv: 13a]] Despite a tunable F/C ratio, the large‐scale production of fluorographene is restricted by the high‐temperature fluorination and the expensive XeF 2. Figure 2 a) Various steps involved in the fluorination of grapheme (PMMA‐poly(methyl methacrylate)). b) Raman signatures of fluorinated grapheme (PMMA‐poly(methyl methacrylate)). Reproduced with permission. [[qv: 4a]] c) Optical changes of graphene upon single‐side fluorination. Reproduced with permission. [[qv: 12a]] Copyright 2010, American Chemical Society. d) Scheme for preparing fluorographene by direct‐heating fluorination of graphene‐oxide. Reproduced with permission. [[qv: 22a]] Copyright 2013, American Chemical Society. Fluorine gas (F 2 ) is another important fluorination agent to prepare fluorinated graphene because of its high reactivity. Fluorinated graphene with different F/C ratios was synthesized by Wang et al. [[qv: 22a]] GO was treated by F 2 at a low temperature (from room temperature to 180 °C) (Figure 2 d). The F/C ratios (0. 65, 0. 84, and 1. 02) of fluorinated graphene could be controlled by the concentration (2%, 5%, and 10%) of F 2 in a mixture of F 2 and N 2 gas. In addition, Cheng et al. [[qv: 22c]] prepared fluorinated graphene by the exfoliation of fluorinated highly oriented pyrolitic graphite (HOPG), which was treated using 1 atm F 2 at a high temperature of 600 °C. The resultant few‐layer fluorinated graphene showed a high F/C ratio (CF 0. 7 ). Interestingly, Sofer et al. [[qv: 22e]] also presented an easy and weighable method for the fluorination of Hummers GO and Staudenmaier GO in 20% F 2 /N 2 (v/v) at elevated temperatures and pressures. The high resolution XPS results indicated that the F/C ratio was 17. 8% and 5. 61% for Hummers GO and Staudenmaier GO, respectively. Despite the high activity, the fluorination of graphene using F 2 is limited by poor controllability of the C–F bonding characters (semi‐ionic or ionic bonds) and F/C ratios, special equipment requirements and environmental hazards (high toxicity and corrosion). Thus, many other fluorine‐containing agents such as SF 6, SF 4 or MoF 6 were also used for fluorinating graphene. Pumera et al. [[qv: 22d]] demonstrated the fluorination of GO using SF 6, SF 4 or MoF 6. The surface elemental composition showed that GO synthesized by the Hummers method were thermally fluorinated using SF 6, SF 4 and MoF 6 at 800 °C with different F/C ratios of 1. 92%, 0. 53%, and 0. 26%, respectively. Additionly, GO synthesized by Staudenmaier method were treated by microwave in SF 6 at 800 °C and 1000 °C and showed diffeirent F/C ratios of 4. 25% and 0. 49%, respectively. The results revealed that F/C ratios of GO could be tuned by different gaseous fluorine‐containing agents with the control of the temperature for the fluorination. The structural changes of GO also led to the changes of F/C ratios under the treatment of SF 6. [[qv: 22d]] Despite recent progress, the fluorination using XeF 2, SF 6, SF 4 or MoF 6 is still far from up‐scale industrial production. Thus, exploring a low‐toxic fluorine‐containing gas (or mixed gas) for mild, selective and high efficient fluorination is important for preparing various fluorianted graphene in the future. 2. 1. 2 Plasma Fluorination Compared with severe fluorination of fluorine‐based gas, plasma fluorination is considered to be an easy to control, mild and clean method for preparing fluorinated graphene. During plasma fluorination, the fluorine radicals generated by the plasma technique adsorb onto graphene and form different C‐F bonds. Recently, a variety of plasma sources, such as SF 6, [[qv: 23a, c, f, g, j]] CF 4, [[qv: 23b, d, e, i, k]] and F 2, [[qv: 23h]] have been used. Baraket et al. [[qv: 23a]] synthesized fluorinated graphene using electron‐beam generated plasmas in Ar/SF 6 ( Figure 3 a), and found that C‐F bonds in fluorinated graphene could be reduced to original C–C bonds by removing F atoms via annealing (500 °C). Sherpa et al. [[qv: 23g]] reported that the polarity of C‐F bonds, depending on the C‐F bonding characters (ionic, semi‐ionic, or covalent) between F and C atoms, could be induced in fluorinated epitaxial graphene using a SF 6 plasma‐treatment in a reactive ion etcher system. They found that work function of fluorinated graphene was controlled by the polarity of C‐F bonds as well as by the degree of fluorination. Recently, plasma fluorination of graphene using SF 6 plasma was also investigated by Yang et al. [[qv: 23j]] Interestingly, the fluorination of single‐layer graphene is much more feasible than multi‐layer because of large corrugations. Figure 3 a) Schematic diagram of the plasma processing systems in Ar/SF 6 mixtures. Reproduced with permission. [[qv: 23a]] Copyright 2010, American Institute of Physics. b) Schematic view of reactive ion etching system used for fluorination of grapheme in Ar/F 2 plasma. Reproduced with permission. [[qv: 23h]] Copyright 2012, American Institute of Physics. c) Schematic of the one‐step formation of a transistor by using a patterned buffer layer during the CF 4 plasma treatment. Reproduced with permission. [[qv: 23e]] d) The scheme showing a mechanism for fluorination by using CYTOP and laser irradiation. Reproduced with permission. [[qv: 25a]] Copyright 2012, American Chemical Society. Bon et al. [[qv: 23b]] reported the fluorination of GO, obtained from thermally exfoliated graphite oxide, by the treatment of CF 4 plasma. C‐F bonds in fluorinated graphene could be changed to C‐N bonds by reacting with butylamine (the nucleophilic reagent) at room temperature. [[qv: 23b]] Yu et al. [[qv: 23k]] also synthesized fluorinated reduced graphene oxide (RGO) using CF 4 plasma at room temperature, and the F/C ratios (F/C ≈ 0. 17–0. 27) were controlled by the plasma exposure time. Recently, Wang et al. [[qv: 23i]] reported the fluorination of CVD‐grown single‐layer graphene using CF 4 plasma. The results showed that F/C ratios of fluorinated graphene were tuned by the conditions of the plasma; however, the resultant fluorinated graphene and fluorographene consisted of a mixture of CF x (x ≈ 1–3), and the spatial distribution of F on graphene was highly inhomogeneous. [[qv: 23i]] K. I. Ho et al. [[qv: 23e]] presented a one‐step approach for the selective fluorination of graphene using CF 4 plasma in a plasma‐enhanced chemical vapor deposition (PECVD) system. During the fluorination, F‐radicals preferentially fluorinated graphene at a low temperature (<200 °C), while the defect was suppressed by screening out the effect of ion damage (Figure 3 c). When the fluorination time increases, D peak in pristine graphene is remarkably intensified and the G peak is broadened. Simultaneously, the D' peak originated from the intra‐valley resonance of Raman Scattering, is obvious. [[qv: 23e]] In addition to SF 6 and CF 4, F 2 is also used for plasma fluorination. Tahara et al. [[qv: 23h]] developed a highly controlled fluorination method of preparing fluorinated graphene utilizing fluorine radicals in Ar/F 2 plasma. To overcome ion attacks and facilitate the C=C addition reaction of graphene with fluorine radicals, graphene was placed on the other side of the Si substrate to avoid direct contact with Ar/F 2 plasma (Figure 3 b). High‐density plasma is important for fluorinating graphene with high F/C ratios because the fluoride‐containing ions (such as F −, CF 4 +, CF 3 + ) energies are lower than fluoride radicals. Desipte a simple and effective method, the plasma fluorination inevitably damages the carbon structure of graphene by severe ion bombardment at a relatively high temperature. [[qv: 25a]], 28 Furthermore, the production is limited because the preparation is highly restricted to plamas‐treated area and expensive equipment. And ion damage during the plasma treatment is inevitable. Thus, the up‐scale production of fluorinated graphene via plasma fluorination needs more developed technique and equipments. 2. 1. 3 Hydrothermal Fluorination Hydrothermal or solvothermal fluorination is another versatile method for preparing fluorinated graphene and fluorographene. The fluorination effect depends on fluorine precursors, such as hydrofluoric acid (HF), [[qv: 24c]] BF 3 ‐etherate[[qv: 24b]] and diethylaminosulfur trifluoride (DAST)[[qv: 24a, e]] and hexafluorophosphoric acid (HPF 6 ). [[qv: 24d]] Wang et al. [[qv: 24c]] presented a convenient method to fluorinate dispersed GO using HF through a simple hydrothermal process. Note that some oxygen‐containing groups were substituted by F atoms during the hydrothermal reaction. In addition, the F/C ratios were controlled by varying the temperature, times and HF concentration. Similarly, Gao et al. [[qv: 24a]] reported the solvothermal fluorination of GO films through converting the oxygen‐containing groups (mainly hydroxyl, epoxy, and carbonyl/carboxylic) to C‐F bonds by treating GO with DAST in chloroform at 50 °C. More recently, Samanta et al. [[qv: 24b]] prepared fluorinated RGO with fluorine coverage of 38 wt% using anhydrous BF 3 ‐etherate and alkyl thiol/alkyl amine on the gram scale. GO is an excellent nanosheet for the hydrothermal fluorianation because of many epoxide, hydroxyl, carboxylic and ketone functional groups on the surface The oxygen‐containing groups can be removed or substituted by the formation of C‐F bonds at high temperature using a suitable fluorination solvent. Thus, the hydrothermal fluorination shows great potential for fluoriated graphene with high F/C ratios. Unfortunately, the uniform distribution of C‐F bonds on fluorinated graphene by the hydrothermal fluorination has yet been reported. 2. 1. 4 Photochemical/Electrochemical Synthesis Lee et al. [[qv: 25a]] reported an environmentally friendly method of selectively fluorinating single‐side graphene using a solid fluoropolymer CYTOP (Cytop, CTL‐809) source and laser irradiation. The fluoropolymer CYTOP decomposed under laser irradiation on the surface of a single‐layer graphene film on a SiO 2 /Si substrate. Active fluorine radicals, generated by the decomposition of CYTOP, reacted with the sp 2 ‐hybridized carbon and formed C‐F bonds (Figure 3 d). Gong et al. [[qv: 25c]] prepared fluorinated RGO by employing UV irradiation on GO dispersion in HF at room temperature. The synthesis of oxy‐fluorinated graphene via an electrochemical method was demonstrated by Bruna et al. [[qv: 25b]] A graphite flake contacted a platinum wire as the working electrode was fluorinated in HF (50 wt%) as the electrolyte. Despite an environmentally friendly method, the F/C ratios of fluorinated graphene by photochemical fluorination are relatively low, and the special fluorination agents have yet to be developed. 2. 2 Exfoliation Methods 2. 2. 1 Sonochemical Exfoliation Sonochemical exfoliation of multilayer materials has been well researched because it is a versatile and nondestructive technique for preparing high‐quality two‐dimensional single‐ or few‐layer nanomaterials. Solution‐processed exfoliation has been employed for up‐scale production of two‐dimensional graphene and MoS 2. 29 Single‐layer fluorinated graphene and fluorographene were obtained by exfoliation from fluorinated graphite assisted by ultrasonication. To date, many intercalated molecules have been used to exfoliate fluorinated graphene including sulfolane, [[qv: 9a]] ionic liquids, [[qv: 26a]] surfactant, [[qv: 26d]] N‐methyl‐2‐pyrrolidone (NMP), [[qv: 26b]], 30 chloroform, [[qv: 26g]] 2‐isopropanol (IPA)[[qv: 26e]] and acetonitrile. [[qv: 26c]] The driving force of the intercalation can be evaluated by Gibbs free energy (Δ G ) of the intercalation compounding process triggered by the F atoms, which is defined in Equation (1) (1) Δ G = Δ H − T Δ S where Δ H and Δ S are the enthalpy and entropy for the intercalation of molecules or solvents respectively. Because of van der Waals attraction between two adjacent layers of fluorinated graphite, Δ H is generally expected to be positive; resulting in a small and positive ΔG, and thus the exfoliation is mainly affected by T Δ S. At high temperature and pressure, the increase in Δ S leads to a decrease in Δ G, which indicates increasing driving forces. Thus, compared with graphite, fluorinated graphene is easily exfoliated by the intercalation of molecules with relatively weak van der Waals attraction and a large interlayer space. High‐yield single‐ or few‐layer fluorinated graphene with a specific F/C ratio is obtained. Zborˇil et al. [[qv: 9a]] prepared fluorinated graphene (F/C = 1. 00) by a single‐step liquid‐phase exfoliation. In this process, fluorinated graphene was exfoliated from commercial fluorinated graphite suspended in sulfolane at a 135 W ultrasonic bath for 1 h at 50 °C. Chang et al. [[qv: 26a]] reported an effective and low‐cost exfoliation to obtain single and few‐layer fluorinated graphene (F/C = 0. 25 or 0. 50) in ionic liquid. In this method, ionic liquid (1‐butyl‐3‐methylimidazolium bromide) intercalated into the interlayer of commercial fluorinated graphite by mixing and incubating. After intercalation, black colloidal dispersion of fluorinated graphene was obtained by ultrasonication. Transmission electron microscopy (TEM) and atomic force microscopy (AFM) images revealed that two‐dimensional fluorinated graphene showed 1–5 layers with 2–10 μm in edge size. Among many organic solvents, NMP is considered to be an important intercalated molecule to exfoliate fluorinated graphene because of its dipole moment value of 4. 09 D ( Figure 4 a). [[qv: 26b]] According to Gong's studies, the intercalation of NMP into the interlayer of fluorinated graphite was accomplished by refluxing for 2 h, and the subsequent exfoliation was facilitated by ultrasonication for 100 h. Feng et al. [[qv: 26c]] reported a solvothermal exfoliation to prepare few‐layer (1–3) fluorographene with a high‐yield production of 15%. The semi‐ionic C‐F bonds of fluorographene exfoliated by chloroform (Figure 4 b) might be a result of hydrogen bonding during the intercalation. In addition, Wang et al. [[qv: 26d]] prepared fluorinated graphene by exfoliation of fluorinated graphite using a cationic surfactant of cetyl‐trimethyl‐ammonium bromide (CTAB) and dopamine (DA). This intercalation was carried out at room temperature in air (Figure 4 c). Zhu et al. [[qv: 26g]] demonstrated an easy method to synthesize fluorinated graphene nanosheets by means of a one‐pot sonochemical exfoliation of the commercially available graphite fluoride powders in chloroform under ambient conditions without any additional pretreatments, assistant reagents, or special protections (Figure 4 d). Figure 4 a) Schematic of the NMP intercalation and exfoliation fabrication processes for fluorographene dispersions. Reproduced with permission. [[qv: 26b]] Copyright 2012, Royal Society of Chemistry. b) Schematic of the preparation of fluorographene nanosheets by the solvothermal intercalation and exfoliation. Reproduced with permission. [[qv: 26c]] Copyright 2014, Royal Society of Chemistry. c) Schematic of the cooperative exfoliation process by PDA and CTAB to prepare fluorinated graphene sheets. Reproduced with permission. [[qv: 26d]] Copyright 2012, Royal Society of Chemistry. d) Illustration of the fabrication of fluorographene sheets via a chloroform‐mediated sonochemical exfoliation method. Reproduced with permission. [[qv: 26g]] Copyright 2013, Royal Society of Chemistry. Liquid‐phase exfoliation is a relatively simple method for high‐quality fluorinated graphene by optimizing the intercalation. The exfoliation at room temperature could reserve most original fluorine atoms. The chemicals for the intercalation and the reaction condition (time, temperature and pressure) are significantly important for the liquid‐phase exfoliation. In general, polar molecules are more effective for the intercalation than the non‐polar molecules. Besides, high temperature, long‐term and high pressure facilitate the exfoliation for single‐ or few‐layer fluorinated graphene after ultrasonication and separation. However, C‐F bonds of fluorinated graphene might be partially reduced during the high‐temperature exfoliation, and importantly, the numer of layers is hardly controlled because of the weak selectivity of the exfoliation. 2. 2. 2 Modified Hummer's Exfoliation Hummer's method was widely used to prepare GO by the intercalation and oxidation of bulk graphite. Recently, the modified Hummer's method attracted tremendous attention to the exfoliation of fluorinated graphite because of its convenient, easy‐control process. 20, [[qv: 21b]], 31 Pulickel M. Ajayan et al. [[qv: 21b]], [[qv: 31c]] developed a methodology to synthesize fluorinated GO using a modified Hummer's method. The magic‐angle spinning (MAS) 13C NMR results revealed that there were two types of fluorinated GO: partially fluorinated GO (FGO) and highly fluorinated GO (HFGO). FGO was hydrophilic similar to GO in hydrophilicity, while HFGO was relatively hydrophobic. Although the modified Hummer's method improves the dispersion of fluorinated GO in water or organic solvents by introducing numerous oxygen‐containing groups, this severe reaction inevitably partially destroys C‐F bonds of fluorinated graphene. 2. 2. 3 Thermal Exfoliation Fluorographene can also be exfoliated from fluorinated graphite by thermal exfoliation. Dubois et al. 27 prepared fluorographene by thermal exfoliation of fluorinated HOPG prepared using F 2. Fluorographene was obtained by fast elimination of interlaminar species of fluorinated HOPG with a sharp increase in temperature accompanied by the color changing from greyish to black. 27 3 Structures C‐F bonding character including C‐F bonds, F/C ratio, and configuration largely determines the chemical (electrochemical), electrical, electronic, optical, magnetic structures, stability and hydrophobicity of fluorinated graphene. Thus, the deep understanding of fluoro‐carbon structure is fundamental to control the properties and design the application of fluorinated graphene. In this section, we discusse the C‐F structural characteristics controlled by a variety of methods or technologies to offer a strategy for tuning C‐F bonds precisely and uniformly. 3. 1 C‐F Bond Chemical bonds are usually determined by the electronegativity between two bonding atoms. As a result, C‐F bonds vary from covalent bonds, through semi‐ionic bonds, to ionic bonds because of the extremely high electronegativity of fluorine. This feature results in a more electrostatic character in the covalent C‐F bond. [[qv: 14b]], 32 Sato et al. 33 experimentally confirmed the existence of semi‐ionic C‐F bonds in fluorine‐graphite intercalation compounds. Recently, Lee et al. [[qv: 14d]] synthesized fluorinated graphene with semi‐ionic bonds through a one‐step liquid fluorination using liquid ClF 3 as the fluorine agent. Moreover, semi‐ionic C‐F bonds in fluorine‐graphite intercalation compounds and fluorinated graphene were also reported based on theoretical calculations. 15, 34 However, the length of semi‐ionic and ionic C‐F bonds has never been experimentally determined. [[qv: 14a]], [[qv: 13b]], 34, 35 The semiempirical result is shown in Figure 5. Figure 5 The length of C‐F bons and the characteristic peaks of C‐F bond in C1s XPS. The fluorination of C–C bonds of graphene usually contains two competing reaction processes: (1) fluorine radicals react with graphene to form covalent C‐F bonds, in which the sp 3 ‐hybridized C atoms connect to F atoms and (2) fluorine radicals react with graphene to form semi‐ionic C‐F bonds, in which the sp 2 ‐hybridized C atoms connect to F atoms. C–F bonds change from ionic to semi‐ionic to covalent, accompanied by a decrease in F/C ratios by changing the fluorination conditions (e. g. , fluorination agents, temperature and time). [[qv: 14a]] Borini et al. [[qv: 25b]] reported oxy‐fluorianted graphene with semi‐ionic C‐F bonds through the electrochemical intercalation of graphite in hydrofluoric acid solution. Wang et al. [[qv: 19b]] found that C‐F bonds showed the partial transformation from semi‐ionic nature to covalence with the increasing F/C ratios of fluorinated grapehene, which was controlled by the exposure time in XeF 2 atmosphere. This transformation was also found in the liquid‐phase exfoliation with the appropriate solvent such as chloroform. Feng et al. [[qv: 26c]] found the partial transformation of covalent to semi‐ionic C‐F bonds in fluorinated graphene exfoliated by chloroform due to the formation of C–H…F hydrogen bonds between chloroform molecules and F atoms of fluorinated graphite. [[qv: 22a]], [[qv: 33b]], 36 Additionly, the low exfoliation temperature could contribute to the appearance of Csp 2 ‐F bonds. [[qv: 33b]] Previous studies indicated that the partial transformation between ionic (semi‐inoic) and covalent bonds could be caused by the interaction between C‐F bonds and other molecules or materials. Importantly, the natrue of C‐F bonds have a significant impact on the properties of fluorinated graphene, such as work function, [[qv: 23g]] reaction activity, 37 and electrochemical performance. 38 The presence and percentage of covalent, semi‐ionic or ionic C‐F bonds in fluorinated graphene are investigated by X‐ray photoelectron spectroscopy (XPS) and Fourier transform infrared spectroscopy (FTIR). According to previous studies, the characteristic peaks of semi‐ionic bonds between C and F atoms were observed at approximately 287–290 eV in the C1s XPS spectra (Figure 5 ), 685–688 eV in the F1s XPS spectra and 1050–1150 cm −1 in the FTIR spectra. [[qv: 19b]], [[qv: 22a]], [[qv: 25c]], [[qv: 26c, f]], 39 The characteristic peaks of C‐F bonds in fluorinated graphene are given in Table 3. Unfortunately, fluorinated graphene containing ionic C‐F bonds has seldom been reported. Table 3 Comparison of C‐F bonds in fluorinated graphene synthesized with different methods Method Raw materials Peak location in FTIR (cm −1 ) Peak location in XPS (eV) Ref. Graphene‐based materials Fluorine agents/exfoliation solvents Covalent C‐F Semi‐ionic C‐F Covalent C‐F Semi‐ionic C‐F Direct gas‐fluorination CVD graphene XeF 2 1211 1112 F1s 687. 5 F1s 685. 5 [[qv: 19b]] Direct gas‐fluorination GO F 2 1221 1150 C1s 289. 7 C1s 288. 0 [[qv: 22a]] Direct gas‐fluorination RGO F 2 /N 2 mixed gas 1212 1113 C1s 289. 3 C1s 288. 5 [[qv: 62b]] Photochemical synthesis GO HF solution 1212 1149 C1s 291. 2 C1s 290. 3 [[qv: 25c]] Hydrothermal fluorination Graphene ClF 3 1225 1107–1120 C1s 289. 4 C1s 288. 1 F1s 689. 6 F1s 686. 8 Sonochemical exfoliation Fluorographite NMP 1212 1084 C1s 290. 8 — [[qv: 26f]] Sonochemical exfoliation Fluorographite Cloroform 1216 1143 C1s 289. 9 C1s 288. 4 [[qv: 26c]] F1s 689. 0 F1s 688. 1 [[qv: 39a]] Thermal exfoliation FGO — — C1s 290. 3 C1s 288. 4 [[qv: 39b]] John Wiley & Sons, Ltd. 3. 2 F/C Ratio Precise control of the F/C ratio of fluorinated graphene is important for opening the bandgap, tuning electrical conductivity and optical transparency and understanding the structural transformation. Thus, beyond the aforementioned techniques in the second section, fluorination conditions, including the reaction temperature, the species of fluorination agents and catalysts, the type of carbon (e. g. , graphene, GO, and RGO), the treated side and the sonochemical time, are utilized to tune the F/C ratios of fluorinated graphene. [[qv: 23e, k]], [[qv: 24c]], [[qv: 26b]], 40 Yu et al. [[qv: 23k]] reported that F/C ratios (0. 17–0. 27) of fluorinated graphene were controlled by the time of CF 4 ‐plasma treatment. Similar results were also observed in a recent study by Kuan‐I Ho et al. [[qv: 23e]] In addition, Wang et al. [[qv: 24c]] presented an easy, low‐cost and efficient hydrothermal‐process to tune F/C ratios of fluorinated graphene. The contents of each C‐F‐containing group (such as C–CF, C–CF 2, and CF–CF 2, CF, CF 2, and CF 3 ) were dependent on the reaction temperature, time, and HF amount. An increase in the F/C ratios (from 0. 11 to 0. 48) was mainly attributed to the formation of the CF–CF 2 group. [[qv: 24c]] Interestingly, Robinson et al. found that fluorine saturation coverage differed when graphene films were fluorinated by XeF 2 on one or both sides. X‐ray photoelectron spectroscopy and Raman spectroscopy revealed that fluorine coverage saturates at 25% (C 4 F) for one‐side fluorination and at 100% (CF) for double‐side fluorination in XeF 2 at room temperature. [[qv: 12a]] Gong et al. [[qv: 26b]] also reported that the F/C ratios decreased with increasing ultrasonication time in NMP, which might be attributed to the increasing stretching vibration energy of C–F groups gained from the sonic power facilitating the departure of fluorine. [[qv: 26b]] 3. 3 Configuration Fluorinated graphene and fluorographene consisting of weakly bound stacked two‐dimensional carbon monofluorides are a basic building block of fluorinated graphite. [[qv: 4a]], [[qv: 12a]], 41 To gain insight into C‐F bonds, theoretical calculation on the configuration of fluorinated graphene is studied, such as chair, boat, stirrup, and twist‐boat configuration ( Figure 6 ). [[qv: 12b]] The chair configuration shows a two‐dimensional alternate layer of F atoms and C atoms on both sides, whereas in a boat configuration, F atoms alternate with C atoms in pairs. [[qv: 12b]], 42 In the stirrup configuration, each C atom is bonded to an F atom in the way that consecutive fluorine layers along a zigzag direction alternate with graphene layers, while the twist‐boat configuration derived from the boat configuration has a slight twist to F atoms connecting two unique C atoms. 43 Different configurations of fluorographene results in different properties including binding energy, chemical activity, stability, bandgap, Young's modulus and the lattice constant. [[qv: 12b]], [[qv: 43a]] For example, fluorinated graphene or fluorographene with the chair configuration has a lower theoretical binding energy than any other configuration, and the stirrup configuration is more stable than the boat and twist‐boat configurations. [[qv: 12b]], [[qv: 43a]], 44 Figure 6 Perspective views of optimized ordered configurations of fluorographe: a) chair, b) boat, c) stirrup, and d) twist‐boat, respectively. Light and dark grey spheres represent F and C atoms, respectively. The black box indicates the unit cell employed in the calculations. Reproduced with permission. [[qv: 12b]] The first‐principles density functional theory (DFT) calculation showed that fluorographene with the chair configuration had a direct bandgap of 3. 1 eV, 45 which is in good agreement with the experimental data, while the calculated data (7. 4 eV) based on the GW (where GW refers to the one‐particle Green's function with the dynamic screened Coulomb interaction) approximation (7. 4 eV) was twice as large as the experimental values. [[qv: 4a]], [[qv: 12b]], [[qv: 43a]] There are different bandgap values between DFT and GW because GW full account of the quasiparticles and their interaction with light in fluorographene included electron–hole (e–h) and electron–electron (e–e) interactions. [[qv: 12b]] The high Young's modulus E of the chair configuration was up to ≈228 N m −1, which was twice the experimental value (100 ± 30 N m −1 ). [[qv: 4a]] The difference between the calculation and experiment might be attributed to a large number of structural defects in fluorographene because a certain portion of C atoms was not bonded to F atoms but formed dangling bonds. [[qv: 43a]] However, compared with fluorinated carbon nanotubes, the energy difference among various configurations of fluorographene is very small, and this result indicates that fluorographene is unlikely to be a pure single‐crystal form in a chair, stirrup, boat, or twist‐boat configuration. 4 Properties Fluorinated graphene shows many excellent properites such as wide bandgap of 3. 1 eV, the highest theoretical specific capacity (865 mA h g −1 ), good thermal stability below 400 °C, distinct nonlinear feature and high hydrophobicity. In this section, we discuss a variety of properties including bandgap, absorption or luminescence, stability, electronic conductivity, dspersibility, magnetic properties, tribological properties, mechanical or micromechanical properties, and thermal conductivity. These properties are significantly important for the application of fluorinated graphene. 4. 1 Band Gap Graphene shows great potential for advanced electronic devices because of unique electronic properties, such as zero bandgap and high carrier mobility up to 200 000 cm 2 V −1 s −1. 41, 46 However, a zero band‐gap, specifically valence (π) and conduction band (π*) touching at a Dirac point, lowers achievable on‐off ratios for field emission transistors based on a graphene semiconductor. [[qv: 3b]], 47 Thus, opening the bandgap is crucial for the design and fabrication of high‐performance graphene‐based electronic devices. Theoretically, fluorographene shows a wide bandgap of 3. 1 eV because of the transformation from the trigonal sp 2 orbital to the tetragonal sp 3 orbital. [[qv: 4a]], [[qv: 9a]], 12, 48 This feature offers great potential for tuning the bandgap of fluorographene with different C‐F bonding characters. Robinson et al. [[qv: 12a]] prepared fluorinated graphene films (on one side) with fluorine coverage of 25% (C 4 F) using XeF 2. The calculation indicated that the bandgap of fluorinated graphene increases with an increasing F/C ratio because of the interaction between the p‐orbital of F and the π‐orbital of C. The formation of sp 3 bonds led to a large change in charge densities and scattering centers in the conduction band ( Figure 7 ). [[qv: 12a]] The band gap of C 4 F is 2. 93 eV according to the density of states calculations. Moreover, when graphene films were fluorinated on both sides, fluorographene (C 1. 0 F 1. 0 ) showed a large bandgap of 3. 07 eV. [[qv: 12a]] Liu et al. 48 investigated the bandgap of fluorinated graphene with different F/C ratios. The results indicated that the C‐F bonds in low‐fluorine‐coverage fluorinated graphene (CF 0. 031, CF 0. 056, and CF 0. 125 ) were polar covalent bonds because of the high electronegativity of F atoms, and thus, they exhibited a metallic behavior. This behavior could be changed by increasing F/C ratios. CF 0. 25 and fluorinated graphene (CF 1. 0 ) had wide bandgaps of 2. 92 eV and 3. 13 eV, respectively, according to the generalized gradient approximation (GGA) calculations. Interestingly, fluorinated graphene (CF 0. 5 ) in which C atoms bonded to F atoms on one side also exhibited metallic behavior ascribed to the exchange splitting of the dangling C‐p z orbital with a coupling with an impurity state induced by F atoms. The results indicate that the bandgap of fluorinated graphene is greatly influenced by F/C ratios. 48 Figure 7 a) Calculated binding energy per F atom compared to the F 2 gas state. b) Sketch of the calculated C 4 F configuration for the 25% coverage from (a). c) Calculated total density of states of single‐side fluorinated graphene for several fluorine coverages. Reproduced with permission. [[qv: 12a]] Copyright 2010, American Chemical Society. Based on the density‐functional GGA calculation, the bandgap can also be controlled by different configurations and layers of fluorinated graphene and fluorographene. Specifically, a chair configuration shows a bandgap of 3. 10 eV, while the bandgaps of the stirrup, boat, and twist‐boat configuration are 3. 58, 3. 28, and 3. 05 eV, respectively. [[qv: 12b]] In the chair configuration, F atoms are alternately distributed on the plane. One F atom locates above the carbon layer, while the other one is under the same layer. Thus, the chair configuration has more symmetry than the stirrup configuration. In addition, the stirrup configuration is more significant for the conduction state because charge density follows the chain characteristic. [[qv: 12b]] Li et al. 49 calculated the bandgap of C 4 F with different layers by means of DFT computation. The results implied that bi‐layer fluorinated graphene (C 4 F) had a much narrower indirect bandgap than that of monolayer fluorinated graphene. Additionally, the bandgap of C 4 F nanosheets was further decreased by increasing the number of stacked layers because the conversion from insulator to semiconductor based on the dipole‐dipole interaction between two C 4 F layers induce a subtle interlayer polarization. 49 4. 2 Optical Properties Fluorine‐substitution on carbon atom dramatically changes the optical properties of graphene including the absorption band, photoluminescence and transparency. Robinson et al. [[qv: 12a]] found that graphene film was optically transparent in the visible region after treatment by XeF 2. The absorption coefficient decreased after fluorination by SF 6 [[qv: 23c]] and CF 4 plasma, [[qv: 23e]] which was in agreement with other fluorinated carbon materials. [[qv: 14b]], 50 Recently, the absorption spectra of fluorinated graphene with different F/C ratios has been studied to appreciate the effect of fluorination on optical properties. [[qv: 4a]], [[qv: 24e]] Nair et al. [[qv: 4a]] investigated the optical transparency of fluorinated graphene by fluorinating in XeF 2 at 70 °C ( Figure 8 a). Graphene shows a peak at 4. 6 eV and an absorption edge at ≈2. 5 eV, which was in good agreement with a pronounced van Hove singularity, and was no longer linear above 2. 5 eV (Figure 8 a). [[qv: 4a]], 51 However, the absorption spectra were drastically changed after the fluorination. Compared with graphene, fluorinated graphene showed low‐intensity absorption with a weak and broad band in the range of 4. 0 to 5. 0 eV. It exhibited high transparency in the whole range because of the impurity scattering. [[qv: 4a]] Furthermore, fluorographene only absorbed light with energy >3. 0 eV (blue range) (Figure 8 a). This result indicated that fluorographene was nearly transparent in the range of visible light with the wide bandgap ≥3. 0 eV. [[qv: 4a]] Zhao et al. [[qv: 24e]] reported that fluorinated GO dispersed in CH 3 CN, synthesized by hydrothermal method with different reaction medium, exhibited two absorption peaks at approximately 220 nm and 250–350 nm, which were assigned to the π–π* transition of conjugated polyene‐type structures in the carbon nanosheets[[qv: 31b, c]], 52 and a couple of conjugated aromatic domains with different sizes, 30 respectively. Gong et al. [[qv: 39b]] found that the π–π* transition peak of GO red‐shifted from 230 to 260 nm after fluorination (Figure 8 b) because of an increase in the π‐electron concentration and structural ordering based on the restoration of sp 2 carbon and the possible rearrangement of atoms. [[qv: 26b]], 53 Figure 8 a) Changes in optical transparency of graphene due to fluorination. The absorption spectra of graphene (upper curve), partially fluorinated graphene (middle curve), and fluorographene (bottom curve). The solid curve is the absorption behavior expected for a 2D semiconductor with E g = 3 eV. Reproduced with permission. [[qv: 4a]] b) UV–vis absorption spectra of GO (dispersed in water) and FGO (dispersed in a mixture of ethanol and NMP) just after sonication. Reproduced with permission. [[qv: 39b]] Copyright 2014, Royal Society of Chemistry. c) NEXAFS spectra of pristine graphene and fluorographene with two different contents of fluorine. [[qv: 13a]] d) Room temperature photoluminescence emission of the pristine graphene and fluorographene dispersed in acetone using 290 nm (4. 275 eV) excitation. Reproduced with permission. [[qv: 13a]] Copyright 2011, American Chemical Society. Recently, the studies on the photoluminescence (PL) of fluorinated graphene have attracted attention because it not only yields insight into understanding electronic properties but is also crucial for advanced semiconductor devices and energy harvesting. Jeon et al. [[qv: 13a]] reported the room‐temperature PL spectra of graphene and fluorinated graphene dispersed in acetone using 290 nm (4. 275 eV) excitation (Figure 8 d). The results showed that fluorinated graphene (fluorination for 5 days) exhibited two emission peaks at approximately 3. 80 eV and 3. 65 eV indicating wide bandgaps, while no emission was obtained in graphene with zero bandgap. [[qv: 13a]], 54 Specifically, the peak at 3. 80 eV corresponded to the band‐to‐band recombination of a free electron and a hole, which was found in the bandgap of fluorinated graphene measured by near edge X‐ray absorption spectroscopy (NEXAFS) (Figure 8 c). [[qv: 13a]] The peak at 3. 65 eV was 156 meV (1260 cm −1 ) below the bandgap because of phonon‐assisted radiative recombination across the bandgap where the C–F vibration mode was excited when the electron‐hole pair recombined. Analogously, two accompanying peaks at 2. 88 eV and 2. 73 eV were also observed in low‐degree fluorinated graphene (fluorination for 1 day) (Figure 8 d). [[qv: 13a]] Based on unique PL, fluorographene can be developed for fabricating flexible near ultraviolet LEDs by optimizing quantum yield. Optical properties of ground‐state fluorinated graphene were also predicted by theoretical calculation based on DFT. 55 However, the calculation typically does not exactly match with experimental optical spectra because it does not take into account the interaction between two quasiparticles. 56 In this respect, the Bethe–Salpeter equation (GW‐BSE) represents a more precise method than DFT for calculating the direct transitions because it takes into account electron–electron (e–e) and electron–hole (e–h) interaction. [154–156] Samarakoon et al. [[qv: 12b]] reported the in‐plane absorption spectra of graphene and fluorographene calculated by GW‐BSE along with the random phase approximation (RPA) and GW‐RPA, respectively. RPA was regarded as the result of the DFT level. As shown in Figure 9, graphene had many notable peaks around 10–12 eV as a result of strong electron–hole correlations along with the appearance of bounded excitons in the ultraviolet region, opening the path toward an excitonic Bose‐Einstein condensate in graphene that was observed experimentally. [[qv: 12b]], 56, 58 This feature was also obtained for fluorographene. A distinctive peak around 9. 8 eV of fluorographene emerged in GW‐BSE that was evidently connected to strong electron‐hole coupling and was attributed to the transition from the near‐gap valence bands to the minimum conduction band. [[qv: 12b]] Theoretical calculation provides an insightful understanding of optical properties controlled by C‐F bonds, and results will promote more experimental studies on quantitative and qualitative descriptions of the optical properties of fluorinated graphene in the future. Figure 9 Calculated absorption spectra using RPA (dashed lines), GW‐RPA (dotted lines), and GW‐BSE (solid lines) for graphane (top panel) and flurographene (bottom panel), respectively. Reproduced with permission. [[qv: 12b]] 4. 3 Stability Compared with the instability of GO 59 and easily decomposed graphene, 60 fluorographene shows good chemical and thermal stability as a result of strong C‐F bonding energy. [[qv: 4a]] Raman spectroscopy ( Figure 10 a) was typically utilized to study the stability of fluorinated graphene because it can provide a wealth of information about the structures of graphene‐based materials. The stability of fluorinated graphene with different F/C ratios at high temperature has been recently reported in several studies. [[qv: 4a]], [[qv: 12a]], [[qv: 22c]] Fluorinated graphene with a low F/C ratio could be partially recovered to pristine graphene by a short annealing‐time at temperatures <400 °C, which was reflected by a continuous decrease in the D band. In contrast, fluorographene shows a high stability below 400 °C. [[qv: 4a]], [[qv: 12a]], [[qv: 22c]] The removal of both C and F atoms in fluorographene was observed by a prolonged annealing time at high temperature (≈450 °C). [[qv: 4a]] In addition, according to XPS data, fluorinated graphene, prepared using XeF 2 gas on SiO 2, Au, and Cu substrates, lost approximately 50–80% of the initial F/C ratios over 10 days until the F/C ratios were not changed. 61 The change in C‐F bonds by annealing at different temperatures was also demonstrated by an increase in electrical conductivity (Figure 10 b). [[qv: 4a]] No current could be detected when fluorographene was annealed T A below 200 °C. Fluorographene became weakly conductive, and the effective resistivity ρ = V/I decreased to ≈1 GΩ at 350 °C. The results indicated that the thermal stability and chemical inertness of fluorographene were similar to Teflon. [[qv: 4a]], [[qv: 13b]] Figure 10 a) Raman spectra of graphene fluorinated to various levels and then annealed at different T. A, B, C) Raman spectra for weakly, mode rately and highly fluorinated graphene, respectively. b) Changes in fluorographene's ρ induced by annealing and I–V characteristics for partially fluorinated graphene obtained by reduction at 350 °C. The curves from flattest to steepest were measured at T = 100, 150, 200, 250, and 300 K, respectively. Reproduced with permission. [[qv: 4a]] In addition, fluorographene showed a good chemical stability in many liquids such as water, acetone, and propanol, and under ambient conditions except for strong reductants. [[qv: 4a]], [[qv: 9a]] It was found that fluorinated graphene could be reduced by hydrazine, potassium iodide, ultraviolet irradiation and alkylamine compounds. [[qv: 9a]], [[qv: 12a]], 61, 62 Robinson et al. [[qv: 12a]] reported the low temperature chemical reduction of fluorographene by hydrazine with the process of 4CF n + n N 2 H 4 → 4C + 4 n HF + 2 n N 2. Radek et al. [[qv: 9a]] provided a pathway for defluorination using KI in DMF. In this process, fluorinated graphene transformed to metastable graphene iodide, which quickly decomposed to graphene and iodine at just 150 °C: CF+ KI → KF + [CI]; [CI] → C + 1/2I 2. Additionally, Lee and co‐workers[[qv: 14d]] reported that ionic C‐F bond was selectively reduced by acetone treatment at a low temperature with the equation 2C 2 F(semi‐ionic) + CH 3 C(O)CH 3 (l) → HF + 2C(s) + C 2 F(covalent) + CH 3 C(O)CH 2 (l). More recently, new fluorinated graphene derivatives were prepared by the covalent modification of fluorinated graphene. 7, 8, 63 Stine et al. [[qv: 63a]] fluorinated CVD‐grown graphene sheets followed by covalent modification with ethylenediamine. They found that the intensity of the F 1s peak was reduced by ≈90% whereas a large N 1s peak at 399. 5 eV appeared because of the removal of the fluorine. Urbanová et al. 8 synthesized thiofluorographene through the covalent functionalization (nucleophilic substitution). The thiofluorographene showed a small region where F atoms were substituted by –SH groups. Interestingly, the semiconducting properties of thiofluorographene could be potentially regulated by tuning the SH/F ratios. 4. 4 Electronic Conductivity Single‐layer graphene shows high electron mobility because of its sp 2 hybridized C atoms with a p z orbital forming a π ‐ conjugated bond. Fluorination is widely used to chemically tailor the electrical conducitivity because it enables the transition from metallic/semiconducting to an insulating nature controlled by different F/C ratios. Fluorographene, the thinnest two‐dimensional insulator, shows a distinct nonlinear feature of I–V curves. [[qv: 4a]], [[qv: 22c]], [[qv: 23i]] Different from C–C bonds of graphene, every C atom in fluorographene with sp 3 hybridization is bound to an F atom. Thus, fluorographene is an insulator because of the disappearance of π‐ conjugated bonds. Wang et al. [[qv: 23i]] investigated the typical I–V characteristics of fluorinated graphene prepared by CF 4 plasma. Fluorinated graphene showed a linear I–V curve with a short fluorination time (<10 min) with a resistance <10 MΩ because a low amount of F and sp 3 C atoms were regarded as defects in an sp 2 hybridized C network. Thus, the π‐conjugated network of graphene is preserved. However, this network was destroyed by a long fluorination‐time (>10 min) using CF 4 plasma. Fluorinated graphene underwent the transition from semiconductor to insulator, resulting in a nonlinear curve with a resistance >1 GΩ. Compared with graphene, the resistance of fluorinated graphene showed a sharp increase by more than 7 orders of magnitude (from 10 kΩ to >100 GΩ) when the fluorine coverage is a few tenths of a percent. [[qv: 23i]], 64 Consequently, fluorinated graphene with a low F/C ratio showed a semiconducting behavior with the sp 2 hybridized carbon network. 65 Moreover, the insulating properties of multi‐layer fluorographene with an extremely thin thickness (5 nm) could not be changed at temperatures <400 °C, and its dielectric constant and breakdown electric field (EBD) were ≈1. 2 and above 10 MV cm −1, respectively. 66 4. 5 Dispersibility Because of the presence of C‐F bonds, fluorinated graphene and fluorographene are highly hydrophobic and difficult to disperse or solubilize in most solvents because of its low surface free energy. [[qv: 21b]], 67 However, the dispersion of fluorographene in solvents is crucially important for the solution‐processed fabrication of devices or applications as precursors for electrodes and composites. [[qv: 26b]], [[qv: 26e]], 68 Previous studies reported that hydrophobic fluorographene could not be dispersed in ethanol because it has no free p z orbitals to form pseudohydrogen bonds with the hydroxy group of ethanol. [[qv: 13a]] The pseudohydrogen bond has been demonstrated to facilitate the dispersion of graphene with the π bond (abundant free p z orbitals) in ethanol. [[qv: 13a]], [[qv: 26b]], 69 Gong et al. [[qv: 26b]] studied the dispersibility of fluorinated graphene in a variety of organic solvents. It was found that fluorinated graphene showed much better dispersion in solvents with a large closed conjugated system formed by p z orbitals such as phenylethylene (PS), NMP, and THF than others with nonhybridized p z orbitals. [[qv: 26b]] Specifically, in a homogeneous solvent, a free p z orbital acted as an electron acceptor and formed pseudo‐hydrogen bonds with (C n F) x –F groups, thus resulting in an increase in dispersion of fluorinated graphene. [[qv: 13a]], [[qv: 26b]] Recently, the dispersion of fluorinated graphene in water was improved by fluorosurfactants in which perfluorinated units were adhered on the surface of fluorinated graphene and cationic or anionic units provided static repulsion. 70 Furthermore, fluorinated GO could be well dispersed in many organic solvents with nonhybridized p z orbitals, such as CH 3 CN, chloroform, and DMF. [[qv: 24e]] 4. 6 Magnetic Properties Graphene obtained by sonochemical exfoliation of high‐purity HOPG shows a strongly diamagnetic response and no sign of ferromagnetism over a wide range of temperature, T. 71 A weak sign of paramagnetism becomes noticeable only below 50 K attributed to the edge states and point defects. 71, 72 Interestingly, the introduction of F atoms in graphene causes a dramatic change in magnetic properties due to the presence of C‐F bonds. 20, 73 Nair et al. [[qv: 73a]] reported that the paramagnetism in the CF x samlpes with x increasing from 0. 1 to 1 was described by the Brillouin function. M = N g J μ B [ 2 J + 1 2 J ctnh ( 2 J + 1 2 J z ) − 1 2 J ctnh ( z 2 J ) ] where z = gJμ B H/k B T, g was the g ‐factor, J was the angular momentum number, N was the number of spins and k B was the Boltzmann constant. The number of spins N increased monotonically with x up to ≈0. 9, and then showed some decrease for fluorographene. The maximum M achieved by the fluorination of graphene was one order of magnitude higher than that achieved by irradiation. Unfortunately, the concentration of magnetic moments was only ≈0. 1% of the maximum hypothetically possible magnetism of one moment per carbon atom because F adatoms have a strongly towards clustering. [[qv: 73a]] Tang et al. [[qv: 73b]] reported that small F clusters that could be preferably formed around the vacancies in RGO produced a lot of magnetic edge adatoms. And such fluorinated RGO have a high magnetization of 0. 83 emu g −1, a high magnetic moment of 3. 187 × 10 −3 μ B per carbon atom and a high efficiency of 8. 68 × 10 −3 μ B per F adatom. Recently, fluorinated GO was used as an efficient magnetic resonance imaging (MRI) contrast agent by Ajayan et al. 20 Tang et al. 74 found that the uneven double‐side partially fluorinated graphene with the ripple structure become magnetic, whereas wrinkle structure showed nonmagnetic. And they also demonstrated that the magnetic moments could be significantly increased by external tensile strain. 74 4. 7 Tribological properties Generally, graphene shows good tribological performance due to its high chemical inertness, extreme strength, easy shear capability on its densely packed and atomically smooth surface. 75 Tribological properties of graphene are further improved by fluorination. 76 Fluorinated graphene is consider as one of important ultrathin solid lubricants or lubricant additive of lubricating oils, lubricating coatings and anti‐wear composites becuase of its low friction coefficient and high durability. Specfically, fluorine atoms bound on carbon structure enhances nanoscale friction and reduces the adhesion and the number of free electrons by developing few van der Waals contacts and wide band gaps. Thus, C‐F bonding structure endows fluorinated graphene with excellent tribological performance. [[qv: 76c, d]], 77 Carpick et al. [[qv: 76d]] systematically measured the friction between AFM tips and fluorinated graphene with different F/C ratios. This method was useful to illustrate the mechanism for the enhanced friction. They found that the enhanced friction was attributed to the significantly increased corrugation of the interfacial potential due to the highly localized negative charge concentrated at fluorine sites, consistent with the Prandtl–Tomlinson model. [[qv: 76d]] Park et al. reported that nanoscale friction on the fluorinated graphene was 6 times larger than that on pristine graphene, while the adhesion decreased somewhat becuase the attachment of F atom to the C atom enable the transition of graphene to the tetrahedral sp 3 configuration. [[qv: 76c]], [[qv: 77a]] Hou et al. 78 found that fluorinated graphene remarkably improved the reliability of the base oil and prolonged the friction time. Besides, trbological properties of fluorinated graphene are also controlled by its microstructures (the arrangement of F atoms, corrugation and the number of atomic layers), F/C ratios, surface chemistry (species on the surface of fluorinated graphene sheets). [[qv: 76c, d]], [[qv: 77b]], 78 Thus, many studies need to be presented to optimize trbological properties of fluoroinated‐graphene for ultrathin solid lubricant. 4. 8 Other properties 4. 8. 1 Mechanical or Micromechanical Properties Fluorination usually affects mechnical properities of graphene such as Young's modulus ( E ) and intrinsic strength ( σ ) because of the presence of C‐F bonds. Nair et al. measured the E and σ of fluorographene using AFM. [[qv: 4a]] Fluorographene exhibited a lower E (100 ± 30 N m −1 ) and a lower σ (≈15 N m −1 ) than of graphene ( E and σ of graphene are E = 340 ± 50 N m −1 and σ = 42 ± 4 N m −1, respectively). [[qv: 4a]], 79 They speculated that the decrease in E and σ arised from longer sp 3 hybridized C–C bonds in fluorographene than sp 3 hybridized C–C bonds in graphene. [[qv: 4a]] Interestingly, the elastic deformation σ / E of fluorgraphene showed litttle change in comparison of graphene because of the absence of structural defects during fluorination. [[qv: 4a]] The mechanism of controlling specifc mechanical properties including the strength, modules and deformation has yet been understood. 4. 8. 2 Thermal Conductivity Graphene exhibits superior thermal conductivity because of efficient phonon transfer in the 2D long‐range sp 2 carbon framework by lattice vibrations. 80 To date, many highly thermal conductive graphene film have been designed and prepared, and single‐ or few‐layer graphene is widely used as thermal condutive nanofiller in the polymer‐based composite to increase thermal conduction. [[qv: 80a, b, d]], 81 Recently, fluorinated graphene shows a great promise in combing heat dissipation and hydrophobic or self‐lubricating properties. Huang et al. calculated theoretical thermal conductivity of fluorinated graphene using non‐equilibrium molecular dynamic (NEMD) simulations. 82 Results showed that thermal conductivity of fluorinated graphene decreased during the fluorination, and it increased when the F/C ratio approached 1. 0. 82 They also found that thermal conductivity of fluorinated graphene was less sensitive to strain than of graphene. This result might be attributed to that the phonon become less sensitive to tensile strain after fluorination. 82, 83 Despite great interest, improving thermal conductivity (diffusity) by exploring the key C‐F structure is one of challenge for fluorinated graphene. 5 Applications 5. 1 Energy Conversion and Storage Devices Fluorinated carbon materials (CF x ) were first used as the cathode in lithium primary batteries by Watanabe et al. in 1972. 84 CF x was considered to be one of the ideal cathode materials for lithium primary batteries because of a variety of unique properties, such as high energy density, high average operating voltage, long shelf life, stable operation ability and wide operating temperature. Subsequently, Li/CF x batteries were first commercialized by Matsushita Electric Co. in Japan in 1975. 85 Importantly, Li/CF x batteries have the highest theoretical specific capacity (865 mA h g −1 ; x = 1) in primary battery systems. [[qv: 14c]], [[qv: 18c]] With an ultrathin two‐dimensional layer‐structure, fluorinated graphene and fluorographene are regarded as the most promising CF x to achieve the theoretical capacity because of their tunable F/C ratios and C‐F bonding characters, favorable diffusion kinetics of lithium ions and large specific surface area. Recently, many studies focused on the performance of Li/CF x batteries using fluorinated graphene or fluorographene as the cathode material. [[qv: 17b]], [[qv: 26c, e]], 86 Feng et al. reported that lithium primary batteries using fluorographene exhibited a remarkable discharge rate because of good Li + diffusion and charge mobility through nanosheets. [[qv: 26c]] Fluorographene exfoliated by chloroform with semi‐ionic F–C bonds showed a high specific capacity of 520 mA h g −1 and a voltage platform of 2. 18 V at a current density of 1 C, accompanied by a maximum power density of 4038 W kg −1 at 3 C, which was almost four times higher than that of fluorinated graphite ( Figure 11 a). [[qv: 26c]] Moreover, fluorographene showed an energy density of 1910 Wh/kg, which is higher than fluorinated carbon nanotubes (≈1600–1800 Wh/kg). 87 Recently, they also prepared nitrogen and fluorine co‐doped graphene with superior reversible specific discharge capacity (1075 mA h g −1 at 100 mA g −1 ), excellent rate capabilities (305 mA h g −1 at 5 A g −1 ), and outstanding cycling stability (capacity retention of ≈95% at 5 A g −1 after 2000 cycles) as the anode material for lithium ion batteries. 88 Such results was attributed to the increased disorder and defects as well as the electrically conductive graphitic N and semi‐ionic C–F bonds, and the highly wrinkled nanostructures caused by the co‐doping of N and F. 88 Zhan et al. presented a straightforward approach to fabricate self‐supporting fluorinated graphene nanosheets by liquid exfoliation of fluorinated graphite using IPA. [[qv: 26e]] Fluorinated graphene not only had abundant fluorine active sites for lithium storage but also facilitated the diffusion of lithium ions during charging and discharging. As a consequence, fluorinated graphene exhibited a high reversible capacity of 780 mAh g −1 at 50 mA g −1 and excellent cycle performance for 50 cycles (Figure 11 b). [[qv: 26e]] Rangasamy et al. fabricated a solid‐state Li/CF x battery with a solid electrolyte of Li 3 PS 4 that had dual functions: the inert electrolyte at the anode and the active CF x component at the cathode. [[qv: 17b]] The solid‐state Li/CF x battery exhibited excellent capacity, good rate performance and a stable potential profile with a capacity utilization of 1095 mAh g −1 beyond the theoretical capacity of a CF x cathode (when x = 1) (865 mAh g −1 ) (Figure 11 c). [[qv: 17b]] In recently, Jeon et al. 89 reported that edge‐selectively fluorinated graphene nanoplatelets (FGnPs), which prepared by mechanochemically driven reaction between fluorine gas (20 vol% in argon) and graphitic, demonstrated superb electrochemical performance with excellent stability/cycle life in lithium ion batteries. The FGnPs electrode showed an initial charge capacity of 650. 3 mAh g −1 at 0. 5 C and maintained a charge retention of 76. 6% after 500 cycles. 89 Meanwhile, the FGnPs based dye‐sensitized solar cells also displayed an outstanding performance (FF of 71. 5%, J sc of 14. 44 mA cm −2 and PCE of 10. 01%) because of the high electronegativity of F atom ( χ = 3. 98) and the strong C‐F covalent bonds (C–F, 488 kJ mol −1 ) at the edges. 89 Results indicate that edge‐selectively fluorinated graphene is one of excellent materials for energy conversion and storage devices. Figure 11 a) The galvanostatic discharge curves at different discharge rates. Reproduced with permission. [[qv: 26c]] Copyright 2014, Royal Society of Chemistry. b) Cycle performance of fluorinated graphene and fluorinated graphite electrodes at a current density of 50 mA g −1 and the rate and cycling performances of F‐graphene and F‐graphite electrodes obtained over a wide range of high current densities from 50 to 2000 mA g −1. Reproduced with permission. [[qv: 26e]] c) Discharge profile for the Li/LPS/CF x + C + LPS cell and illustrates cell capacity exceeding the theoretical maximum of 865 mAh g −1 for the CF x system. Reproduced with permission. [[qv: 17b]] Copyright 2014, American Chemical Society. d) Fluorinated graphene sheets as the scaffold for stem cell growth. Reproduced with permission. [[qv: 19b]] e) Magnetic properties of FGO. Reproduced with permission. 20 Xie et al. 90 first reported a prototype of Mg/fluorinated graphene battery with the capacity of 110 and 90 mAh g −1 at 10 or 50 mA g −1, respectively. They utilized the fast surface redox process to replace sluggish lattice migration to improve the kinetics of Mg batteries resulting in good reversibility and rate performance. High performance benefits from the surface reaction at accessible fluorinated functional groups of porous conductive frameworks. This proof‐of‐concept Mg/fluorinated graphene system bypasses the sluggish diffusion of multivalent cations into the host lattice and the structure distortion at the cathode. Vizintin et al. 91 used fluorinated RGO as an interlayer additive in lithium−sulfur (Li−S) batteries. Fluorinated RGO blocked the diffusion/migration of polysulfides from the porous positive electrode to the metallic lithium electrode and thus prevented the redox shuttle effect. The results showed that fluorinated RGO effectively improved the open circuit potential, cycling stability and capacitiy of Li–S batteries. 5. 2 Bioapplications Fluorinated graphene is of interest in many bioapplications because of its fascinating C‐F bonds that enable biological responses[[qv: 19b]], 92 and paramagnetic behavior. 20, [[qv: 73c]] Loh et al. [[qv: 19b]] used fluorinated graphene as the scaffold for the growth of mesenchymal stem cells (MSCs) (Figure 11 d). In their study, fluorinated graphene enhanced cell adhesion and proliferation of MSCs, exhibiting a neuro‐inductive effect viaspontaneous cell polarization. Fluorinated graphene films were highly supportive of the growth of MSCs, and C‐F bonds had significant effects on cell morphology and cytoskeletal and nuclear elongation of MSCs. [[qv: 19b]] Moreover, the introduction of C‐F bonds into GO caused a dramatic change in magnetic properties. Ajayan et al. 20 reported that fluorinated GO was an outstanding carbon‐based magnetic resonance imaging (MRI) contrast agent without magnetic nanoparticles (Figure 11 e). the results showed that fluorinated GO could be potentially developed for a theranostic material with multimodal imaging, including MRI, ultrasound and photoacoustics, as well as the potential to pack hydrophobic therapeutic agents along the hydrophilic fluorinated GO basal plane. 20, [[qv: 73c]] 5. 3 Fluorinated Graphene Quantum Dots Although fluorinated graphene is a semiconductor with a wide bandgap and shows UV‐fluorescence, [[qv: 12a]], [[qv: 13a]], 64 the bundling sheets and low fluoroescent intensity restrict the application in optoelectronic devices. [[qv: 13a]], 30, 93 Fluorinated graphene quantum dots (F‐GQDs) with the size <10 nm exhibits unique electronic and luminescent properties because of quantum confinement and edge effects. 94 Tang et al. demonstrated that F‐GQDs synthesized by cutting fluorinated graphene through hydrothermal method, exhibited bright blue photoluminescence and upconversion properties. 93 Gong et al. 95 developed a activating‐cutting strategy to obtain graphene fluoroxide QDs with the tunable size and controllable fluorine coverage. The graphene fluoroxide QDs with good solubility and stability in water, display stable blue luminescence in hostile environment. This feature shows a great potential for the fabrication of advanced optical nano‐devices. [[qv: 95a]] Sun et al. 96 developed a new top‐down method to simultaneously synthesize F‐GQDs and GQDs by combining a microwave‐assisted technique with the hydrothermal treatment. F‐GQDs showed excellent photo‐ and pH stability in long‐term and real‐time cellular imaging. Results open a gate to the application of fluorinated graphene in environmental engineering, solar cells, biological probes, bioimaging and energy technology. To date, the photoluminescence controlled by the defect and C‐F bonding character is still unclear. 5. 4 Other Applications Fluorinated graphene can also be used for applications in supercapacitors, [[qv: 24e]] electrochemistry 97 and amphiphobicity[[qv: 21b]], 98 applications based on unique properties controlled by F/C ratios and a two‐dimensional layer‐structure. Zhao et al. [[qv: 24e]] prepared solid supercapacitors using fluorinated graphene as an electrode material. Cyclic voltammetry measurements showed that fluorinated graphene prepared in dichloromethane exhibited the highest specific capacitance at 106. 6 F g −1, which was much better than GO. The results were also confirmed by charge/discharge curves. [[qv: 24e]] Pumera et al. 97 studied the electrochemical properties of fluorographite with three different F/C ratios of 0. 33, 0. 47, and 0. 75. The results revealed that the heterogeneous electron transfer was accelerated by increasing F/C ratios, and the fluorographite with the F/C ratio of 0. 75 showed the fastest rate of electron transfer ( k obs 0 ) at 2. 69 × 10 −3 cm s −1 and 4. 37 × 10 −3 cm s −1 in [Fe(CN) 6 ] 4–/3− and Eu 2+/3+ redox probes, respectively. And the overpotentials of ascorbic acid and uric acid oxidations decrease with the increasing F/C ratios. And the fluorographite with F/C ratio of 0. 75 provided a response to uric acid at 18. 46 μA mM −1, which more sensitive than that to ascorbic acid (2. 15 μA mM −1 ). 97 C‐F bonds drastically reduce the surface energy of graphene, resulting in a change in wetting behavior. Mathkar et al. [[qv: 21b]] reported an amphiphobic coating of fluorinated GO with a low surface tension of 59 dyn cm −1, synthesized by oxidizing the basal plane of fluorinated graphite. This method allows for unique, accessible, carbon‐based amphiphobic coatings. [[qv: 21b]] 6 Conclusion and Outlook In this review, we have given an overview of synthetic methods, structures and properties of fluorinated graphene that can be utilized for applications in high‐energy storage, unique biological response and magnetic resonance imaging, fluorinated graphene quantum dots, supercapacitors, electrochemistry and amphiphobicity. We have emphasized the importance and significance of controlling C‐F bonding characters, F/C ratios and configurations of fluorinated graphene, fluorographene and F‐GQDs by fluorination (gas or liquid phase) or exfoliation. The selective fluorination enables graphene with different two‐dimensional configurations for various properties including wide bandgap, blue luminescence, excellent electrochemistry, high stability and self‐lubricating. For example, an increase in the F/C ratio enlarges the bandgap of fluorographene, while a low F/C ratio usually ensures charge transport based on π‐conjugated structures. The covalent C‐F bonds in gas fluorination are crucial for thermal and chemical stability, while semi‐ionic and ionic bonds endow fluorinated graphene with a higher discharge potential for lithium batteries. Moreover, thermal conductivity, magnetic properties and luminescence of fluorinated graphene are not well developed because of a complicated fluoro‐carbon structure. Although significant progress has been made, additional challenge for uniform up‐scale synthesis/production, target‐oriented fluorination, the homogeneity of C‐F bonding character, solution processability and removal of other fluorides in applications must be addressed. It is very difficult to tune C‐F bonding precisely at the specifc microstructure and/or chemical structure because of the strong fluorination and the complicated chemical and microstructures (layer, size, defects, configuration) of graphene. A versatile, low‐cost and safe method of fluorinating graphene has not yet been found, resulting in the limitation of a wide range of use in commercial applications. Furthermore, because the mechanism of the formation of different C‐F bonding character is still unclear, fluorinated graphene is often composed of a mixture of various types of covalent, semi‐ionic and/or ionic C‐F bonds with different ratios. As a result, fluorinated graphene with different structures (layer, size, C‐F bonds) needs to be separated and/or purified before the use in advanced electronic devices. Thus, chemical methods or strategies for selectively fluorinating graphene are of paramount importance. To date, there is a huge number of opportunities and challenges for designing and synthesizing fluorinated graphene with different structures such as core–shell, nanoporous spheres, nanocages and topologically nontrivial assemblies. The investigation of diffenent fluorinated graphene will put insightful understanding of their properties. On the basis of controlling the selectivity of plasma‐treatment, gas‐ and/or mild fluorniation, the change in F bonding nature, bandgap or electronic interaction of fluorinated graphene with the increasing F/C ratios and/or the specific C‐F bonds will be investigated. This result will further illustrate the effect of C‐F bonding character and configuration on thermal conductivity, self‐lubricating and optical properties. In addition to controllability and uniformity, multifunctional fluorinated graphene will be an interesting subject of intense and fruitful research in the future. A systematic study of thermal conductivity, magnetic property and luminscence will provide more special application in fluorinated graphene and F‐GQDs as well as other fluorinated carbon materials. Besides, the electrochemical properties of fluorinated graphene will be further improved to maxmize the energy density and power density by optimizing the microstructure, C‐F bonding character, the interfacial wettability and cooperation with additives. By the integration of chemical groups, polymer chains and/or functional nanoparticles, fluorinated graphene and its composites will show great potential for secondary batteries (e. g. , Li, Na and Li–S battery), super‐insulating materials, light emitting diodes (LED) and display materials. Based on much effort focused on the controllability of structures and optimized properties, in the future, fluorinated graphene will exhibit excellent performance in flexible nanoelectronics, energy conversion/storage, special protective coatings, and tissue engineering.
|
10. 1002/advs. 201600058
| 2,016
|
Advanced Science
|
Biologically Inspired Smart Release System Based on 3D Bioprinted Perfused Scaffold for Vascularized Tissue Regeneration
|
A critical challenge to the development of large‐scale artificial tissue grafts for defect reconstruction is vascularization of the tissue construct. As an emerging tissue/organ manufacturing technique, 3D bioprinting offers great precision in controlling the internal architecture of a scaffold with preferable mechanical strength and printing complicated microstructures comparable to native tissue. However, current bioprinting techniques still exhibit difficulty in achieving biomimetic nano resolution and cooperating with bioactive spatiotemporal signals. In this study, a comprehensive design of engineered vascularized bone construct is presented for the first time by integrating biomimetic 3D bioprinted fluid perfused microstructure with biologically inspired smart release nanocoating, which is regarded as an aspiring concept combining engineering, biological, and material science. In this biologically inspired design, angiogenesis and osteogenesis are successively induced through a matrix metalloprotease 2 regulative mechanism by delivering dual growth factors with sequential release in spatiotemporal coordination. Availability of this system is evaluated in dynamic culture condition, which is similar to fluid surrounding in vivo, as an alternative animal model study. Results, particularly from co‐cultured dynamically samples demonstrate excellent bioactivity and vascularized bone forming potential of nanocoating modified 3D bioprinted scaffolds for human bone marrow mesenchymal stem cells and human umbilical vein endothelial cells.
|
1 Introduction Vascularization of large‐scale artificial bone tissue grafts is the most critical challenge for various large bone defect reconstruction. Traditional tissue engineering has been focusing on combining osteoconductive scaffolds, osteoinductive growth factors, and osteogenic precursor cells to repair and regenerate bone. However, nutrient and waste exchange between individual cells and capillary vessels in bone is limited to distances of 100–300 μm. Therefore, construction of vascularized bone grafts plays a vital role in regenerating and remodeling bone tissue. 1 The blood vessels in native bone are critical for transport of oxygen and nutrients to maintain skeletal tissue functions. Failed vascularization in implanted grafts results in necrosis of osteoblast in the interior and poor integration between neo and host tissues. 2 So far an ideal vascularized bone construct has not been produced, despite a great deal of research and effort. The main reasons are the deficiencies of perfused vascular structure in hierarchical bone scaffold design, and the effectively targeted stimulation of multiple functional signals. Currently, the strongly desired characteristics of advanced tissue scaffolds in the field involve both biomimetic properties in structure and the ability to regulate the cell behavior. Hence, an ideal vascularized bone scaffold that can integrate structure with functionality should be designed to regulate osteogenesis and angiogenesis. The engineering techniques that mimic the critical aspects of natural healing and growth cascade, is widely utilized to artificially augment the proliferation and differentiation of the recruited or implanted cells via the integration of growth factors and cytokines that providing suitable biochemical and physicochemical factors for tissue regeneration. Therefore, combining the design of a 3D biomimetic fluid perfused scaffold and an effective growth factor delivery method is regarded as a highly promising technique for vascularized bone regeneration research, especially for eventual clinical applications. 3, 4, 5 Regarding 3D scaffold fabrication techniques, phase separation, freeze drying, porogen leaching, and electrospinning may offer limited control over scaffold geometry, pore characteristics and internal channel architecture. All the deficiencies significantly decrease nutrient transportation, cell migration and survival. 6 Compared with traditional manufacturing technology, 3D bioprinting can provide the ability to construct multiple hierarchical and multi‐scale bone‐like scaffolds with controlled macro shape, porosity and microstructure, thus allowing for patient‐specific fabrication and customized clinical application. 7, 8 3D bioprinting with fused deposition modeling (FDM) has been one of most effective way to make macro‐scale bone implants with high mechanical strength which also contain microstructures with controllable features. However, the potentially high temperature used to process most common materials for this technique makes it difficult to incorporate bioactive components into scaffolds or include bioactive growth factors delivery. 6 In addition, current 3D bioprinting techniques (including FDM) exhibit difficulty in achieving biomimetic nano resolution for regulating cellular events. 8, 9 Therefore, the surface modification or other post fabrication technologies are no doubt promising choices to improve biocompatibility and functionality of 3D bioprinted scaffolds. Within the complex cascade of biological events, growth factors are well known to play a crucial role in regulating cellular behaviors and transferring signals between cells and their extracellular matrix (ECM) to stimulate endogenous repair and regeneration mechanisms, thereby leading to an accelerated functional restoration of damaged or defective tissues. The growth factors that are administered in their native form and without any protection are susceptible to biodegradation and can be rapidly eliminated from the blood circulation, resulting in insufficient amounts at targeted site for a worse therapeutic effect. Although direct adsorption, layer‐by‐layer (LbL) technology, multiphase loading, particulate‐based delivery, hydrogel‐based delivery, and their combination application as well as some intelligent delivery systems have been developed over the past decade, targeted transport and sustained release of growth factors with time‐ and dose‐dependent profiles still have little achievement. 3, 10, 11, 12 Incorporating smart stimuli‐responsive elements into growth factor delivery system is one highly innovative strategy to obtain specific release triggered by external stimuli. Drug or gene delivery in response to pH, temperature, magnetic, ultrasound, irradiation and electric stimuli has shown great promise, however, the delivery of growth factors via external triggers for tissue engineering remains limited to their intrinsic characteristics, including deactivation by exogenous stimuli due to poor protein stability, and poor encapsulation or release effects due to relatively large size. 11, 13 More importantly, few systems have addressed the cooperative biological signaling events of cells as a function of the changes in their dynamic microenvironment. The state of the art concept toward the delivery of dual or multiple growth factors is not only to make more efforts for developing sophisticated delivery platforms, but to explore a biologically inspired system that dynamically release multiple cues to regenerate complex tissues and more closely reproduce the evolving microenvironment that occurs in natural ECM. Therefore, with development of tissue engineering technology, new scaffold manufacturing technique and smart growth factor delivery approaches are strongly desired to develop forward to comprehensive engineering design and biologically inspired responsive induction. Consequently to vascularized bone regeneration, current synergistic therapy also lacks a biologically active control mechanism for responsive multiple growth factor delivery to induce angiogenesis and osteogenesis in spatiotemporal coordination. 12 Hence, there is a strong requirement for a vascularized bone scaffold that can integrate biomimetic structure with functionality to intelligently regulate osteogenesis and angiogenesis. In this study, we implemented an integrated set of manufacturing processes for the first time which combines biomimetic 3D structure design with post fabrication functionalization ( Figure 1 a, b). Research activities included: (1) bioprinting a 3D fluid perfused microstructure vascularized bone scaffold via computer‐aided design (CAD) and (2) fabricating a biologically inspired smart release nanocoating on the surface of the bioprinted scaffold to coordinate spatiotemporal angiogenic and osteogenic growth factor delivery. This engineered vascularized bone constructs were cultured in dynamic fluid surrounding which may provide an alternative to sacrificed animal experiment, to evaluate the availability of biologically inspired smart release system for improved vascularized bone regeneration. Figure 1 a) Schematic illustration of the fabrication process of nanocoating modified 3D bioprinted scaffolds. According to the native bone structure, the biomimetic perfused scaffold combining bone support and vascular channels was designed and printed by FDM printer. Then surface modification process was performed to obtain a bioactive vascularized bone construct. b) Schematic representation of sequential adsorption and biologically inspired release of growth factors in the nanocoating film. The rhBMP‐2 was adsorbed in first 15 dual‐layers and then rhVEGF was adsorbed in the top 5 dual‐layers together with genipin crosslinking reaction. When MSCs and HUVECs were co‐cultured in dynamic fluid, the secretion of MMP2 by HUVECs could trigger the release of growth factors. After 4 weeks of culture, the vascularized bone structure would be formed in vitro. 2 Results 2. 1 Biomimetic Engineered Complex Tissue Scaffold Bioprinting and Post Fabrication Through the optimization of the engineering design, we have successfully obtained a 3D bioprinted vascularized bone construct with a unique integration of fully interconnected microvascular network within a microstructured bone matrix. Within this vascularized bone model, “square pore shaped” scaffolds were composed of stacked units with a 200 μm line distance and a 250 μm layer height to form a porous cylinder. In order to mimic the arrangement of blood vessels in native bone, a series of interconnected horizontal and vertical channels (500 μm) were designed as shown in Figure 2 a. The microvascular design of the constructs can possess similar flow characteristics to native blood vessels under pulsatile arterial flow as demonstrated in our recent study. 14 3D models were printed using polylactic acid (PLA) on a FDM printer. Afterward, a novel and simply implemented surface modification strategy was employed to provide a nanoscale surface feature and immobilize bioactive cues onto the biomimic 3D scaffolds. Gelatin (Gel) and polylysine (PLL) with sequential adsorption of dual growth factors (recombinant human bone morphogenetic protein, rhBMP‐2 and recombinant human vascular endothelial growth factor, rhVEGF), were assembled layer by layer on the 3D scaffold via electrostatic interaction to form (Gel/PLL) 20 multilayer nanocoatings. The multilayer coating was then crosslinked by genipin (GnP) to form interpenetrating polymer networks (IPN) [(Gel/PLL) 20 ] GnP. Since human umbilical vein endothelial cells (HUVECs) express matrix metalloprotease 2 (MMP2) which is a type of gelatinase with the capacity to degrade gelatin to short peptide chains. 15 With the progression of vascular development and subsequent MMP2 accumulation, the crosslinked networks could be cleaved to release the growth factors. Therefore, a comprehensive design of engineered vascularized bone scaffold was presented for the first time which integrated biomimetic 3D printed structures with organic self‐modulatory mechanisms. Compared with traditional growth factor release system, this design can not only inherit all superiorities from LbL adsorption, but also be endowed with a particular desirable ability of biologically inspired release. Figure 2 a) Microstructural characterization of 3D bioprinted perfused scaffold based on CAD design by SEM. The red circle shows 500 μm vascular channels and the blue square shows 200 μm pores of bone scaffold. The scale bars indicate 200 μm. b) Images of different scaffolds, including PLA, bioactive nanocoating (Gel/PLL) 20 modified PLA (BC), Gnp crosslinked bioactive nanocoating [(Gel/PLL) 20 ]GnP modified PLA (cBC), bioactive nanocoating with growth factors (BCG), and Gnp crosslinked bioactive nanocoating with growth factors (cBCG). c) Red auto‐fluorescent image of cBC or cBCG. d) ATR‐FTIR spectra of different scaffolds. e) Surface morphologies of the different coating modified scaffolds, untreated PLA served as a control. A nanoscale islet‐like feature uniformly distributed over the surface and the adsorption of protein increased the roughness, whereas the crosslinking process weakened these changes. f) Mechanical properties of 3D bioprinted scaffolds. After the post fabrication modifying process, 3D bioprinted scaffolds maintained native bone‐like mechanical strength. 2. 2 Biologically Inspired Smart Release Coating Fabrication and Characterization The interactions between cells and biomaterials are mainly dependent on the physicochemical characteristics of the biomaterials' surfaces. 16 It is expected that this nanocoating could improve surface properties and provide a special functional domain for the 3D bioprinted scaffold to promote cell–substrate interaction. 17 Gel, a negatively charged biopolymer, consists of highly bioactive polypeptides that are derived from collagen. Numerous RGD (arginylglycylaspartic acid) integrins and other functional recognition sequences within gelatin are beneficial for cell attachment, migration, proliferation, and differentiation. Positively charged PLL is widely used to promote cell adhesion via enhancing electrostatic interaction with negatively charged ions of the cell membrane. In this study, the 20 dual‐layer assembly were designed to optimize the contribution of bioactive components and improve the loading of growth factors. The crosslinking process may also further stabilize the multilayer coating and avoid burst release of growth factors. After layer by layer assembly, the remaining amino groups from Gel and PLL contributed to the crosslinking reaction with GnP. GnP as a crosslinking agent in this reaction is an enzymatic product of geniposide isolated from the fruit of the gardenia plant and is reacted with free amino groups to form blue pigments. 18 Hence, a blue coating was observed on the surface of scaffold after the crosslinking reaction. Moreover, the coated scaffold exhibited a strong red fluorescence due to the intrinsic red fluorescence of GnP, illustrating that it could maintain this unique property for diagnostic imaging (Figure 2 b, c). ATR‐IR spectroscopy clearly confirmed the successful preparation of a GnP crosslinked nanocoating and effective loading of growth factors on the surface of 3D bioprinted scaffold (Figure 2 d). Plain PLA scaffolds exhibited a hydrophobic surface with an average contact angle of 75°. Through the surface modification, there was a distinct increase in hydrophillicity (contact angle about 50°) for the nanocoating (Figure S1, Supporting Information). Morphology analysis revealed a nanoscale islet‐like feature uniformly distributed over the surface and the adsorption of growth factors increased the surface roughness when compared with smooth and featureless PLA (Figure 2 e). The crosslinking process made these features more homogeneous and compact, which further increased interface stiffness. The assembly of the bioactive components not only affected the hydrophilicity of the substrate, but also changed the surface morphology, which would in turn influence the cell behaviors on the scaffolds. After the post fabrication modification, the 3D bioprinted scaffolds maintained excellent mechanical properties (Figure 2 f). They possessed a native bone‐like mechanical strength, with a compress modulus of about 0. 4 GPa and a yield stress higher than 15 MPa. This could provide a desired support for bone regeneration. Therefore, surface modification may be one of the most direct and effective strategies to improve the biocompatibility of scaffolds and modulate cellular events without causing a significant change to the intrinsic mechanical and microstructure properties of designed synthetic grafts. We also found these bioprinted PLA scaffolds exhibited unobvious degradation behavior in PBS or esterase solution during 4 weeks of culture, thus it can provide a stable surrounding for our nanocoating release system further to promoting tissue regeneration (Figure S2, Supporting Information). rhBMP‐2 is an osteogenic growth factor used extensively in both ectopic and orthotopic sites for bone generation. rhVEGF is an angiogenic factor critical for both intramembranous and endochondral bone formation. Dual application of rhBMP‐2 and rhVEGF has been regarded as one of the most efficient system for effective vascularized bone formation. 5, 19, 20 However, traditional delivery techniques have exhibited an unfavorable therapeutic effect. A burst release and low sustained doses of growth factors have a limited effect for the long term bone regeneration, while an excess of rhBMP‐2 may lead to undesirable incidences of hematoma, ectopic bone formation, and osteoclast induced osteolysis. 21 Additionally, excess amounts of rhVEGF can actually inhibit osteogenesis, associating with severe vascular leakage and hypotension. 22 Therefore, the amount and timing of rhBMP‐2 and rhVEGF delivery is critical to enhance bone formation and localized vascularization simultaneously. 20, 23 The smart nanocoating used in our design can not only be utilized to immobilize bioactive components onto biomaterial surfaces, but also to control the growth factors quantity and sequential release. More importantly, our system could control the release of growth factors to regulate cell behaviors through organic self‐modulatory mechanisms during vascularized bone formation. Wherein, the nanocoating was fabricated according to the protocol, with rhBMP‐2 being adsorbed in the first 15 dual‐layers and then rhVEGF layers being adsorbed in the top 5 dual‐layers. 24 We anticipated that rhVEGF would be initially released from the top layers to stimulate the formation of blood vessels, followed by rhBMP‐2 release for initiating osteogenic differentiation. The crosslinking process could further stabilize the growth factors in the nanocoating and prevent their rapid clearance. Figure 3 a illustrates organic self‐modulatory mechanism in our system, which is characteristic of sequential release from crosslinked multilayer films with a representation of the proposed film architecture by biologically inspired manner instead of the simple surface erosion. When using BSA as a model protein and MMP2 as a cleaved trigger to study the release profile (Figure 3 b), we found the protein could be sustained released from nanocoating at a minimal dose level over a prolonged time period of several weeks. The crosslinking process also greatly improved the loading stability of BSA in the nanocoating. Moreover, MMP2 could sensitively trigger the fracture of IPN to release protein. The goal being the creation of an effective release mechanism performed successfully on the cBCG scaffold. Instead of uncontrollable diffusion process by surface erosion in traditional LbL system, the controlled nature of localized release from our 3D scaffold surfaces can eventually enables much lower doses of growth factors to be effective for tissue regeneration. Figure 3 a) Illustration of the proposed assembly (i) and release (ii, iii) process of multilayer films without (A) and with (B) crosslinking, as well as our biologically inspired system (C), where the BMP2 (green spheres) and VEGF (red spheres) are loaded into films composed of PLL (blue) and MMP trigger‐cleavable Gel (red). Compared with traditional LbL film adsorption, crosslinking retain their stable immobilization and sequential release without highly inter‐diffusion. Moreover, surface erosion contributes film degradation where the therapeutic agent is released throughout the film, whereas biologically inspired system exhibits a controllable release behavior. The release profiles reflect the effect of crosslinking, and biologically inspired on kinetics of drug release (iv). b) Protein release profiles of nanocoating with BSA within 2 weeks. The cBCG could sustainably release up to 4 weeks (not shown in here). MMP2 was thought to trigger the cleavage of gelatin chain to controllably release growth factors. c) Confocal fluorescence images of hMSCs and HUVECs co‐culture on various scaffolds in a static culture condition for 5 days. hMSCs were labeled with cell tracker green, and HUVECs were stained with cell tracker red, respectively. The scale bars indicate 200 μm. The cBCG scaffold was also imaged as 3D scanning structure. d) Fluorescent images of hMSCs and HUVECs on the 3D bioprinted scaffolds with F‐actin (red) and nucleus (blue) staining in a static culture condition for 3 days. The hMSCs exhibited a well distributed spread on scaffold surface, while the HUVECs formed an aggregative microvascular networks. The scale bars indicate 100 μm. 2. 3 Human Mesenchymal Stem Cells (hMSCs) and HUVECs Co‐Culture on 3D Bioprinted Scaffold Some studies have indicated a positive effect of implanting biomaterial constructs co‐cultured with mesenchymal and vascular cells, where the development of vascularized tissues both in vitro and in vivo was enabled. 25 Therefore, co‐culturing hMSCs with HUVECs in our study was conducted to generate the vascularized bone tissue. The cellular organization of co‐culturing hMSCs and HUVECs on the scaffolds in a static culture condition was investigated after 5 days. Images of green labeled hMSCs and red labeled HUVECs showed that hMSCs homogeneously distributed on the surface of scaffolds. Meanwhile, HUVECs were inclined to aggregate and migrate to form line patterns on the scaffolds (Figure 3 c). In addition, both hMSCs and HUVECs on the nanocoating modified 3D bioprinted scaffolds exhibited excellent adhesion and proliferation, compared with an unmodified PLA control (Figures S3 and S4, Supporting Information). F‐actin staining showed that on the nanocoating, hMSCs spread well and maintained a spindle morphology, whereas HUVECs preferred to grow in lines and form highly aligned network structures (Figure 3 d). 2. 4 In Vitro Engineered Vascularized Bone Construction on Dynamic Culture Condition In order to generate a functional vasculature prior to osteogenic induction, we developed a two‐step culture protocol ( Figure 4 a). hMSCs and HUVECs were co‐cultured in endothelial growth media (EGM) for a week to induce the formation of vascular networks, and then incubated in osteoinductive media (OM)/EGM (1:1) for 3 weeks to induce bone formation. Moreover, to mimic the unique flow characteristics of the native vascularized bone microenvironment, a dynamic culture was conducted to investigate vascularized bone formation (Figure 4 b). The biomimetic‐engineered strategy was adopted in our customized flow fluid device as an alternative method of animal studies. Such conditions, when combined with our highly perfused scaffold, are beneficial to the formation of microvascular structures. 26 The immunofluorescence images of MSCs and HUVECs co‐culture showed faster and higher CD31 expression on our cBCG scaffold within 4 weeks induction, suggesting an ongoing process of perivascular coverage of capillaries induced from sustainable VEGF release (Figure 4 c and Figure S5, Supporting Information). On one hand, the fluid shear stress was performed on our perfused scaffold to accelerate microvascular formation through mimicking fluid surrounding in vivo; on the other hand, sustainable release of VEGF further promoted partial MSCs endothelialization and angiogenesis. Figure 4 a) Schematic illustration of experimental approaches. hMSCs and HUVECs were seeded in EGM and MSCGM at 1:1 ratio on scaffolds for first day. Then the vascular differentiation was induced for 1 week in EGM. At last, the OM/EGM (1:1) was supplied to induce osteogenic differentiation for another 3 weeks. b) Schematic diagram of dynamic culture in a custom‐designed flow bioreactor system. The system composes of four parts, which are perfused chamber, flow controller, nutrient controller, and gas controller. When culture medium flowed through constructs, the cells seeded on the scaffolds would be subject to fluid shear stress by mimicking fluid surrounding in vivo. c) Immunofluorescence staining of the vascularization marked with CD31 antibody for 2 and 4 weeks in a dynamic culture condition. The scale bars indicate 100 μm. To verify the self‐modulatory release ability of our scaffolds in the presence of HUVECs, a monoculture of hMSCs was conducted as a control in vascularized bone differentiation study. After 4 weeks of culture, the maturation of bone and vascular tissue on the scaffolds was assessed using immunofluorescence staining of the osteogenic differentiation marker osteopontin (OPN, red) and angiogenic specific marker von Willebrand factor (vWf, green), respectively ( Figure 5 a). In previous studies, hMSCs have been reported to differentiate into endothelial cells in the presence of rhVEGF, and hMSCs possessed the potential to directly form vascularized bone. 27 The hMSC monoculture displayed some evidence of vascular formation, however, the HUVECs inducted from MSCs showed a limited positive effect on the growth factor release triggered by the MMP2. In contrast, a marked maturation on vascularized bone was observed in the co‐culture system. This was a reasonable and expected result since a high density of endothelial cells in the co‐culture system shortened vascularization time. Figure 5 a) Immunofluorescence staining of the vascularized bone formation in the dynamic co‐culture condition. The fluorescence images for anti‐vWF (green) and OPN (red) showed that the cBCG scaffold possessed more vascular‐like network and osteogenesis than other control groups. The scale bars indicate 100 μm. b, c) Quantification of ALP activity. d, e) Total collagen synthesis. f, g) Quantification of calcium deposition content on different scaffolds comparing the dynamic co‐culture with dynamic monoculture. All data showed that the cBCG scaffold enhanced the osteogenic differentiation. As discussed, sequential adsorption allowed for rhVEGF release firstly from the nanocoating for inducing vascular formation. Then rhBMP‐2 was released to upregulate osteogenic differentiation. The nanocoating scaffold adsorbed with duel growth factors exhibited a higher expression for specific differentiation markers relative to other control groups. It is postulated that the crosslinked nanocoating modified scaffold released growth factors through a MMP2 regulative mechanism instead of diffusion effect. MMP2 secreted by HUVECs would act as on–off switch for the growth factor release, as the activation of the release system depends on the MMP2 expression to cleave the IPN. The nanocoating modified scaffolds with hMSCs and HUVECs co‐culture not only possessed excellent bone forming potential, but also exhibited well‐developed and aggregative microvascular networks. As a structure's innovative design of 3D bioprinted scaffolds, the microchannel networks present in our scaffolds are beneficial to the integration of neovascular formations into native vasculature in the implantation site. This would enable the formation of a circular and stable network, which is a preceding step to creating mature blood vessels in engineered new bone. hMSC osteogenic differentiation on various scaffolds was evaluated quantitatively by measuring alkaline phosphatase (ALP) activity (an early osteogenic differentiation marker), determining total collagen expression (which is main component for bone ECM), staining for bone mineralization, and quantifying calcium content. A rapid increase and high expression in the ALP activity in a short period was observed in all growth factor loading groups (Figure 5 b, c). In our design, the number of hMSCs in co‐culture group was one half of that in monoculture group. However, the two groups exhibited similar results on osteogenic differentiation. Compared to the hMSC monoculture, the ALP activity of hMSCs on the cBCG scaffold in the co‐culture system which triggered‐release the rhBMP‐2 with sustained low dose modality by HUVECs exhibited more rapidly increase with prolonged expression. Due to the initial burst release, the ALP activity on the non‐crosslinking nanocoating did not show any significantly sustained improvement. Therefore, the rhBMP‐2 could be well stabilized in the crosslinked networks and efficiently controlled release achieved with a prolonged time in the co‐culture system. The synthesis of total collagen was also evaluated to verify these characteristics and effects (Figure 5 d, e). Compared with the control groups, hMSCs on the BCG and cBCG scaffolds expressed significantly higher collagen by rhBMP‐2 release. In addition, the collagen content of hMSCs on the cBCG scaffold in the co‐culture system was significantly higher than that of non‐crosslinking group or monoculture groups. For the longer differentiation period, this controlled release behavior was desired to produce beneficial effect over the duration of the experiment, avoiding a rapid clearance of growth factor. Mineralization is ultimately the most important indicator of hMSC osteogenic differentiation, thus the calcium deposition on all scaffolds was investigated after 4 weeks of culture (Figure S6, Supporting Information). Compared with bare PLA scaffold, all nanocoating modified scaffolds showed a positive effect of mineralization. These results could be attributed to the charged surface which serves as a binding site for calcium ions or acidic phospholipids and as nucleation sites for mineralization. In addition, the crosslinked nanocoating (cBC and cBCG) presented an improved calcium deposition when compared with the non‐crosslinked nanocoating, suggesting that the increased surface stiffness could be beneficial to overall mineralization. This phenomenon was also observed in other papers, the matrix stiffness at the cell‐implant interface resulted in the greatest enhancement of the osteogenic differentiation. 28 A larger area of continuous Alizarin red staining was observed in those groups adsorbed with rhBMP‐2. In particular, the intensity of the staining and the size of the deposit were greatest on the cBCG scaffold in the co‐culture system. Similar to the staining results, calcium content analysis further confirmed those phenomena (Figure 5 f, g). The differentiation results demonstrated that the cBCG scaffold could provide a biomimetic bone‐like structure and regulate the release of growth factors for extended time periods to promote vascularized bone formation. We also studied the osteogenic differentiation of hMSC on the scaffolds in static culture conditions (Figure S7, Supporting Information). Compared with BCG scaffolds, our cBCG scaffolds provided stable performance in the dynamic fluid environment similar to in the static culture. Therefore all results indicated that, via the crosslinking process, the cBCG scaffold would theoretically have excellent properties on for efficient and enhanced, yet regulated vascularized bone formation in vivo. 3 Discussions For engineered tissue regeneration, the hierarchical and complicated tissue structure is difficult to precisely fabricate through traditional manufacturing technique of scaffolds. Although 3D bioprinting as an advanced manufacturing technology can precisely fabricate the internal macro‐architecture and complicated microstructures of scaffolds, current bioprinting techniques are still difficult to obtain nanoscale feature and directly cooperating with bioactive signals with controllable manner. Except for advanced structural design, determining the roles that growth factors play in tissue repair and regeneration is as important as designing, developing and applying suitable formulations that release them with spatiotemporal control. As previously reported, LbL assembly provided a simple and effective strategy to modify and functionalize scaffolds. Additionally, the sequential adsorption of multiple growth factors could exhibit release successively to promote tissue regeneration with time dependent kinetics. However, highly inter‐diffusion of polyelectrolyte layers resulted in undesirable leakage of growth factors without sequential release, and driving force of surface erosion made growth factors passively release with negative effects. In view of addressing all this drawbacks, we proposed a state of the art stimuli release manner, “biologically inspired release profile”, which depended on the coordinated interactions with cells or a local cellular microenvironment for triggering changes of delivery systems and thereby leading to controlled release of growth factors. In this study, we demonstrated that integrating a biologically inspired smart release nanocoating strategy with biomimetic 3D bioprinted fluid perfused microstructure could create a highly innovative vascularized bone construct with nano to micro features and self‐modulatory angiogenic and osteogenic growth factor delivery. In virtue of the precise microstructure of scaffold by 3D bioprinting, this bioactive nanocoating might perform a targeted immobilization of growth factor via proposed assembly protocol. Moreover, biologically inspired release system addressed the cooperative biological signaling events of cells as a function of the changes in their dynamic microenvironment. In this biologically inspired design, angiogenesis and osteogenesis were successively induced through a MMP2 regulative mechanism by delivering dual growth factors with sequential release in spatiotemporal coordination. Therein, crosslinking process greatly improved the loading stability of growth factors in the nanocoating without inter‐diffusion. Availability of this system was evaluated in dynamic culture condition, which was similar to fluid surrounding in vivo, as an alternative animal model study. When culture medium flowed through constructs, the cell seeded on the scaffolds would be subject to fluid shear stress by mimicking fluid surrounding in vivo. Our results demonstrated good bioactivity and vascularized bone forming potential of nanocoating modified 3D bioprinted scaffolds. The ability of such a strategy to intelligently regulate rhBMP‐2/rhVEGF release has great potential for improving vascularized bone regeneration and avoiding undesired harmful side effects in clinical applications. 4 Conclusion Although various 3D fabricated scaffolds, surface modification methods and growth factor delivery strategies have been investigated in biomedical application, integrating engineered perfused design of scaffolds and biologically inspired release system is yet to be explored in the manner of biomimetic hierarchical architecture and dynamic biological signaling events. This study makes use of a modular approach to generate bioactive nanocoating on perfused 3D bioprinted scaffold that release growth factors through MMP regulative mechanism, and demonstrates their stimuli‐responsive profiles toward improving vascularized bone regeneration. These results present a highly innovative release mechanism for growth factor delivery by biologically inspired process, which may not only benefit vascularized bone regeneration, but also extend to improving any complex vascularized tissue or organ regenerations. 5 Experimental Section Biomimetic Scaffold Design and 3D Bioprinting : The biomimetic scaffold was designed and printed based on previously reported method. Within this vascularized bone model, the “square pore shaped” scaffolds were composed of stacked units with a 200 μm line distance and a 250 μm layer height to form a porous cylinder. In order to mimic the arrangement of blood vessels in native bone, a series of interconnected horizontal and vertical channels were designed as shown in Figure 1. The diameter of vascular channels was nearly 2. 5 times greater than the pore size in the bone regions of the scaffold. The vascular tubes were long interconnected channels, while the pores of bone region were closely arrayed layer by layer to form regular networks. 3D models were printed into scaffolds layer by layer from PLA on an FDM printer. Additionally, representative CAD models of the scaffolds were used to analyze for surface area, volume, and pore density. The theoretical parameters of scaffold structure were calculated, including the wall thickness (≈200 μm), pore size (≈200 μm), porosity (≈50%), channel size (≈500 μm), and surface area/volume ratio (≈30). The 200 μm is regarded as an ideal pore size for the bone scaffolds and the larger channel may provide a biomimetic fluid environment and vascular invasion spaces in vivo. Crosslinked LbL Assembly Film Construction : For the construction of bioactive nanocoating modified 3D scaffold, the biocomponents were fabricated onto 3D bioprinted scaffold surfaces via electrostatic assembly. Briefly, aminolyzed PLA scaffolds were obtained by immersion in PEI solution (5. 0 mg mL −1 ) for 12 h. Then, polyanion (gelatin, Gel) solution and polycation (polylysine, PLL) solution (2. 0 mg mL −1 ) were alternatively assembled onto the scaffolds via 30 min immersions each, followed by three rinses with PBS buffer, until the desired (Gel/PLL) 20 architectures were obtained. During the assembly process, rhBMP‐2 and rhVEGF (0. 5 mg mL −1 ) were adsorbed into the coatings. The rhBMP‐2 was adsorbed in the first 15 dual‐layers and then the rhVEGF was adsorbed in the top 5 dual‐layers. For the preparation of the IPN, GnP in PBS (0. 50%, w/v) was used to crosslink the amino groups of polyelectrolytes. The LbL‐coated scaffold was immersed into GnP solution for 48 h at room temperature, and finally rinsed with PBS. ATR‐FTIR spectroscopy measurements were performed with a Perkin Elmer Spectrum BX system, to detect nanocoating structural changes. The degradation behavior of all scaffolds was studied in PBS and esterase solution for 4 weeks. 3D Scaffold Mechanical and Morphological Characterization : The mechanical properties of all scaffolds were tested using MTS criterion universal testing system equipped with a 50 k N load cell (MTS Corporation, US), according to international organization for standardization (ISO) and American society for testing and materials (ASTM). The scaffolds were compressed at a strain rate of 2 mm min −1 to a maximum strain of 20%. The slope of the linear elastic region of stress–strain curve was calculated to obtain the compressive modulus. The compressive strength was obtained corresponding to the stress value at the yield point. The morphology and surface topography of scaffolds were studied using a Zeiss SigmaVP scanning electron microscope (SEM). All scaffolds were coated with a roughly 10 nm thick gold layer and imaged using 5 kV electron beam. MMP Triggered Controllable Release : Release studies of nanocoating modified 3D scaffolds were performed using bovine serum albumin (BSA) as protein model by incubation in PBS (pH 7. 4) at 37 °C. Relative quantification of protein released from the nanocoating was determined using micro BCA protein assay kit (Thermo scientific). The BSA (1. 0 mg mL −1 ) was absorbed into the LbL coating in the assembly process, and matrix metalloproteinase 2 (MMP‐2, 50 ng μL −1 ) was used to cleave the crosslinked nanocoating in the release study. The release media was withdrawn at fixed time intervals and replaced with fresh buffer. The sample solutions were monitored using UV–vis spectrophotometry at 562 nm to determine BSA concentration. The calibration curve was plotted using standard protein solutions with known concentrations of proteins. hMSCs and HUVECs Co‐Culture : hMSCs (Texas A&M Health Science Center, Institute for Regenerative) were cultured in mesenchymal stem cell growth media (MSCGM) consisting of alpha minimum essential media, 20% fetal bovine serum (FBS), 1% l ‐glutamine, 1% penicillin/streptomycin. HUVECs (Life Technologies) were cultured in EGM consisting of Medium 200 and low serum growth supplement (LSGS). For osteogenic differentiation studies, hMSCs were cultured in OM (MSCGM supplemented with 10 × 10 −9 m dexamethasone, 50 μg mL −1 l ‐ascorbate acid and 10 × 10 −3 m β‐glycerophosphate (Sigma)). All experiments were performed with hMSCs and HUVECs of six cell passages or less. According to the previous study, a 1:1 ratio was optimally chosen in co‐culture studies as it provided robust and stable vascular networks while enabling bone formation. hMSCs and HUVECs (2 × 10 5 cells mL −1 ) were incubated with CMFDA and CMTMR (10 × 10 −6 m Molecular Probes, CellTracker Dye, life technologies) for 30 min at 37 °C, respectively. The cells were mixed in a 1:1 ratio and then cultured on the scaffolds in a static condition for 5 d. The cell location or arrangement on the 3D bioprinted scaffolds in co‐culture system was imaged with a Zeiss 710 laser scanning confocal microscope. Cell Adhesion and Proliferation : To study the effect of nanocoatings on hMSC and HUVEC attachment, the cells (2 × 10 5 cell mL −1 ) were seeded on various scaffolds for 4 h. The samples were assessed by the 3‐(4, 5‐dimethylthiazol‐2‐yl)‐2, 5‐diphenyltetrazolium bromide (MTT) assay. Briefly, MTT solution (0. 5 mg mL −1 ) was added in the plate and then incubated for 4 h. After the media was removed, isopropanol/HCl solution (1 m ) was added to dissolve the formazan crystals. The optical density (OD) was measured at 490 nm by photometric plate reader (Thermo Scientific). The cell proliferation was conducted 1, 3, and 5 d. Samples were seeded with 1 × 10 5 cell mL −1 and counted at each time point using the same MTT assay described above. To investigate the effect of surface features on the hMSC and HUVEC phenotype and spreading, the organization of actin filaments of adherent cells cultured on our constructs was evaluated after 3 d culture in the static condition. The cells' cytoskeleton was identified with double staining of actin staining (red) using Texas Red‐phalloidin and nuclei staining (blue) using 4, 6‐diamidino‐2‐phenylindole dihydrochloride (DAPI) (Invitrogen). Cells were fixed in 10% formalin for 15 min, permeabilized in 0. 1% Triton X‐100, and blocked with 1% BSA. Cells were then incubated with phalloidin for 20 min and DAPI for 3 min. Samples were observed and imaged using a Zeiss 710 confocal microscope. In Vitro Vascularized Bone Grafts on Dynamic Culture Condition : To induce vascularized bone formation, hMSCs and HUVECs (5 × 10 5 cell mL −1 ) was seeded onto scaffolds, and divided into three culture condition groups including static co‐culture, dynamic co‐culture, and dynamic hMSCs monoculture. A flow bioreactor system was utilized for incubating cells on 3D bioprinted scaffolds to study vascularized bone formation in a dynamic culture. The system consisted of a digital peristaltic pump (Masterflex, Cole‐Parmer), a fluid reservoir with culture medium, and a port for gas exchange with 5% CO 2 /95% air. Efficient transfer of nutrients and oxygen is facilitated by the convective forces provided by unidirectional creep flow through the scaffolds. The optimal culture condition is to utilize EGM for 1 w and then a mixed media composed by EGM and OM at 1:1 ratio for 3 weeks. At predesigned time points, cells were digested in lysed buffer via freezing at −80 °C and thawing at 37 °C. The lysate was collected to test ALP activity and collagen secretion. The ALP activity was determined for 7 and 14 d using ALP assay kit (Bioassay Systems) after the initiation of MSC osteogenic differentiation. ALP substrate was added to the digested suspension in the dark for 30 min, and then the absorbance was read at 405 nm. Measurements were compared to p ‐nitrophenol standards and normalized to total cell protein. The total collagen content was measured via Sirius red method. The suspension was dried, and then incubated in Sirius red solution (0. 1% Sirius red in picric acid) for 1 h. After washed in 5% acetic acid, the precipitate was dissolved in 0. 1 m NaOH for 30 min. The OD was measured at 550 nm and the measurements were compared to collagen standards. After cultured in OM for 3 w, alizarin red S (ARS) staining was used to assay calcium deposition or mineralization nodules on the scaffolds. The cells were fixed with 10% formalin for 10 min, then incubated with ARS stain solution (2% ARS, pH 4. 2) for 30 min. After washed in distilled water three times, the ARS stained scaffolds were imaged. In addition, a calcium detection kit (Pointe Scientific) was used to quantify the calcium deposition. The calcium deposition was dissolved in 0. 6 m HCl, and reacted with dye reagent. Samples were read at 570 nm wavelength, and the contents were calculated with CaCl 2 standards. For immunofluorescence staining, the cells were fixed with 10% formalin for 15 min, permeabilized in 0. 1% Triton X‐100 for 10 min and blocked in 10% BSA for 30 min. Then cells were incubated with primary antibodies at 4 °C overnight. The following primary antibodies were used for staining: goat polyclonal anti‐vWF antibodies (Santa Cruz Biotechnology) and mouse monoclonal anti‐OPN antibodies (Santa Cruz Biotechnology). After incubation with primary antibodies, donkey anti‐goat IgG‐FITC (Santa Cruz Biotechnology) and chicken anti‐mouse IgG‐TR (Santa Cruz Biotechnology) as secondary antibodies were added and incubated 1 h, respectively. Fluorescence images were observed using a confocal microscope. For immunostaining of vascular network, scaffolds were fixed in 10% formalin for 10 min, and permeabilized with Triton X‐100 (0. 1%) in PBS for 10 min. After blocked with BSA for 1 h, the samples were incubated with primary antibodies (Anti‐CD31 antibody, abcam) overnight. The scaffolds were stained with chicken anti‐mouse IgG‐TR secondary antibodies (Santa Cruz Biotechnology) overnight. Finally, the hydrogels were stained with DAPI, and imaged using a confocal microscope. Statistical Analysis : The data are presented as the mean ± standard deviation (SD). A one‐way analysis of variance (ANOVA) with Student's t ‐test was used to verify statistically significant differences among groups, with p < 0. 05 being statistically significant ( * p < 0. 05; ** p < 0. 01; *** p < 0. 001). Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201600160
| 2,016
|
Advanced Science
|
In Vivo Long‐Term Biodistribution, Excretion, and Toxicology of PEGylated Transition‐Metal Dichalcogenides MS
|
With unique 2D structures and intriguing physicochemical properties, various types of transition metal dichalcogenides (TMDCs) have attracted much attention in many fields including nanomedicine. Hence, it is of great importance to carefully study the in vivo biodistribution, excretion, and toxicology profiles of different TMDCs, and hopefully to identify the most promising type of TMDCs with low toxicity and fast excretion for further biomedical applications. Herein, the in vivo behaviors of three representative TMDCs including molybdenum dichalcogenides (MoS 2 ), tungsten dichalcogenides (WS 2 ), and titanium dichalcogenides (TiS 2 ) nanosheets are systematically investigated. Without showing significant in vitro cytotoxicity, all the three types of polyethylene glycol (PEG) functionalized TMDCs show dominate accumulation in reticuloendothelial systems (RES) such as liver and spleen after intravenous injection. In marked contrast to WS 2 ‐PEG and TiS 2 ‐PEG, which show high levels in the organs for months, MoS 2 ‐PEG can be degraded and then excreted almost completely within one month. Further degradation experiments indicate that the distinctive in vivo excretion behaviors of TDMCs can be attributed to their different chemical properties. This work suggests that MoS 2, among various TMDCs, may be particularly interesting for further biomedical applications owning to its low toxicity, capability of biodegradation, and rapid excretion.
|
1 Introduction In the past few years, 2D nanomaterials have received tremendous attention and been applied in various fields due to their fantastic physical and chemical properties. 1, 2, 3, 4, 5, 6, 7, 8 As an analogue of 2D graphene, transition metal dichalcogenides (TMDCs), generally described as the formula MX 2, in which M is the transition metal from groups 4–10 of the periodic table and X is a chalcogen (S, Se, or Te), have attracted great interests due to their rich electronic, optical, mechanical, and chemical properties. 4, 9, 10, 11, 12, 13 In particular, different combinations of transition metals and chalcogens, as well as their various arrangements in the 2D crystals, would lead to a substantial range of properties, making TMDC materials interesting for applications in biological systems. For example, many biosensing platforms have been fabricated based on TMDCs for detection of biological molecules. 14, 15 The high near infrared (NIR) absorbance of selected 2D TMDCs, such as molybdenum dichalcogenides (MoS 2 ), tungsten dichalcogenides (WS 2 ), and titanium dichalcogenides (TiS 2 ), have made them ideal agents for photothermal therapy. 16, 17, 18, 19 The extraordinary surface‐area‐to‐mass ratio of single layered 2D TMDCs affords them high drug loading capacity useful for drug delivery and combination cancer therapy. 20, 21 TMDCs and their doped or composite nanostructures have also shown great promises as novel contrast agents for multimodal biomedical imaging. 22, 23, 24, 25 Additionally, TMDCs with interesting mechanical properties have been explored in the area of tissue engineering. 26 Considering the great potential of TMDCs in biomedical applications, it is therefore critically important to systematically evaluate their long‐term in vivo behaviors. In the past few years, there have been a number of reports studying the toxicity of TMDCs nanomaterials in vitro and in vivo. 27, 28, 29, 30, 31 At the in vitro level, it was found that the cytotoxicity of MoS 2, WS 2, and WSe 2 nanosheets appeared to be low to different cell lines. 15, 18, 32, 33 On the other hand, in our previous reports about the use of polyethylene glycol (PEG) functionalized TMDCs for in vivo imaging and therapy applications, we have uncovered that PEGylated MoS 2 and WS 2 showed no appreciable acute toxicity to the treated mice at our tested dose. 16, 20, 23, 34 Similar findings have also been reported by several other groups regarding the in vivo toxicity of TMDCs with biocompatible surface coatings. 21, 30, 35 However, the detailed in vivo long‐term degradation and excretion behaviors of various TMDCs remain to be investigated to our best knowledge. In addition, it would be very interesting and important to find out which type of TMDCs has the fastest excretion and least long‐term retention, and thus could be the most promising one for further in vivo biomedical applications. In this work, we have synthesized three representative TMDCs (MoS 2, WS 2, and TiS 2 ) by a high‐temperature solution‐phase method and then broke them into small nanosheets under ultra‐sonication. After functionalization with PEG, these three types of PEGylated TMDCs nanosheets showed great physiological stability and quite low in vitro cytotoxicity. Healthy mice were intravenously ( i. v. ) injected with MS 2 ‐PEG (M = Mo, W, Ti) and then sacrificed at different time points with major organs collected. All the three types of MS 2 ‐PEG nanosheets showed dominant accumulation in reticuloendothelial systems (RES) such as liver and spleen. Interestingly, while high levels of W and Ti were detected in RES organs of mice post injection of WS 2 ‐PEG and TiS 2 ‐PEG, respectively, even after 30 d, we found that MoS 2 ‐PEG could be excreted from the body almost completely within 30 d by urine and feces, likely owing to the oxidization of Mo IV S 2 into water‐soluble Mo VI ‐oxide species (e. g. , MoO 4 2− ). The blood analysis and histological examination of those mice showed no obvious in vivo toxicity of all three types of PEGylated TMDCs at our tested dose, even for WS 2 ‐PEG and TiS 2 ‐PEG with long‐term retention in mouse RES organs. Our work suggests that MoS 2 could be a promising 2D nanomaterial for biomedical applications due to its biodegradability and relatively fast excretion. 2 Results and Discussion 2. 1 Synthesis and Surface Modification of MoS 2, WS 2, and TiS 2 Nanosheets TMDCs (MoS 2, WS 2, and TiS 2 ) nanoflakes were synthesized by a high‐temperature solution‐phase method, TEM images (Figure S1, Supporting Information) have showed the morphologies of MoS 2, WS 2, and TiS 2 nanoflakes. Then the synthesized MoS 2, WS 2, and TiS 2 were dissolved in 1‐methyl‐2‐pyrrolidinone (NMP) and broken into small and single‐layer nanosheets under ultrasonication 36 ( Figure 1 a). In brief, the metal precursors (MoCl 5, WCl 6, and TiCl 4 ) were first reacted with oleylamine (OM) to obtain M‐OM complexes (M = Mo, W, and Ti). Upon injection of the sulfur solution by dissolving sulfur powder in OM, the solution color immediately turned into black or deep brown, suggesting the rapid formation of TMDC nanoflakes. Transmission electron microscope (TEM) image (Figure S1, Supporting Information) indicated that the synthesized TMDCs exhibited flake‐like structures. Under ultrasonication in NMP solution, all TMDCs nanoflakes were broken into smaller nanosheets, which became soluble in water. The phase analysis of the as‐prepared nanosheets was also determined by power X‐ray diffraction (XRD) (Figure S2, Supporting Information). All peaks in these spectra matched well with the reported results. 16, 17, 20 Figure 1 Synthesis and characterization of PEGylated MS 2 (M = Mo, W, Ti) nanosheets. a) A scheme of MS 2 TMDC nanosheets synthesis process. b) TEM images of PEGylated MoS 2, WS 2, and TiS 2 nanosheeets. c) DLS size distribution of PEGylated MoS 2, WS 2, and TiS 2 nanosheets in water. d) UV–vis‐NIR absorbance spectra of PEGylated MosS 2, WS 2, and TiS 2 nanosheets with the concentration of 0. 02 mg mL −1. After ultrasonication in NMP, the as‐synthesized MoS 2 /WS 2 /TiS 2 nanosheets became water‐soluble, we have tested the dynamic light scattering (DLS) size distribution of these three nanosheets in water (Figure S3, Supporting Information). Since the as‐made MoS 2 /WS 2 /TiS 2 presented good water‐solubility but bad stability in physiological solutions in the presence of salts. We then chose lipoic acid conjugated 5 kDa PEG (LA‐PEG 5k ) to modify MoS 2 /WS 2 /TiS 2 nanosheets through M‐S (M = Mo, W, Ti) bond. Thermogravimetric analysis showed the weight percentages of MoS 2, WS 2, and TiS 2 in the PEGylated samples were determined to be 25. 99%, 59. 72%, and 46. 92% (Figure S4, Supporting Information), respectively. After PEGylation, the hydrodynamic diameters of MoS 2 ‐PEG, WS 2 ‐PEG, and TiS 2 ‐PEG were determined by DLS to be ≈91, ≈72, and ≈102 nm (Figure 1 c), respectively. We also have collected the statistics of TEM measured sizes about PEGylated MoS 2, WS 2, and TiS 2 (Figure S5, Supporting Information). Through the TEM images we could find that the diameters of PEGylated MoS 2 /WS 2 /TiS 2 were smaller compared with MoS 2 /WS 2 /TiS 2 nanosheets, which could be attributed to the functionalization process. MS 2 were synthesized by a high‐temperature solution‐phase method and then broken in NMP under ultrasonication to obtain small nanosheets. After that, the small nanosheets were functionalized by LA‐PEG under ultrasonication in the water phase. The modification process would further broke the nanosheets, which lead smaller size compared with MS 2 under TEM. All PEGylated TMDC nanosheets were rather stable in various physiological solutions including phosphate buffered saline (PBS), RPMI‐1640 medium, and fetal bovine serum (Figure S6, Supporting Information). UV–vis‐NIR spectra of PEGylated TMDC nanosheets all showed strong wide‐band NIR absorbance (Figure 1 d), with the weight extinction coefficients of MoS 2 ‐PEG, WS 2 ‐PEG, and TiS 2 ‐PEG measured to be 26. 36, 23. 42, and 22. 04 Lg −1 cm −1, respectively, at 808 nm. 2. 2 In Vitro Cytotoxicity Study for PEGylated TMDCs We then carried out a number of different assays to study the in vitro cytotoxicity of various MS 2 ‐PEG (M = Mo, W, and Ti) samples. First, the standard 3‐(4, 5‐dimethylthiazol‐2‐yl)‐2, 5‐diphenyltetrazolium bromide (MTT) assay was performed on mouse macrophage Raw 264. 7, human renal epithelial cell 293T and mouse breast cancer 4T1 cell lines (Figure S7, Supporting Information). After incubation with different concentrations of PEGylated MoS 2, WS 2, and TiS 2 for 24 h, we did not find any serious cytotoxicity for these three kinds of MS 2 ‐PEG nanosheets to both two cell lines even at high concentrations (e. g. , 200 μg mL −1, in terms of MS 2 weight concentrations) ( Figure 2 a, b). Next, the lactate dehydrogenase (LDH) leakage assay was performed to test the integrity of cell membrane after cells were exposed to PEGylated TMDCs for 24 h (Figure 2 c). 37 It was found that the all these three MS 2 ‐PEG (M = Mo, W, and Ti) showed little damage to cells even at high concentrations. When cells are under stimulation by foreign damage signals, the oxidative stress inside cells may be increased to activate the cellular defense system. 38 Herein, we used a dihydroethidine (DHE) probe to detect the intracellular reactive oxygen species (ROS, e. g. , peroxide and superoxide) in cells after incubation with MS 2 ‐PEG (Figure 2 d). Compared with the control group, the experimental groups also showed no noticeable increase of intracellular ROS generation after treatment by MS 2 ‐PEG. All these data put together indicated that our PEGylated MoS 2, WS 2, and TiS 2 nanosheets exhibited no appreciable in vitro cytotoxicity within our tested dose range. Figure 2 Cytotoxicity of PEGylated MS 2 (M = Mo, W, Ti). a, b) Relative cell viability of Raw 264. 7 cell line (a) and 293T cell line (b) after incubation with PEGylated MoS 2, WS 2, and TiS 2 at different MS 2 weight concentrations for 24 h. c, d) Relative LDH release level (c) and relative ROS values (d) of 4T1 cell line after incubation with PEGylated MoS 2, WS 2, and TiS 2 nanosheets at different concentrations for 24 h, respectively. 2. 3 In Vivo Biodistribution and Clearance Behaviors of PEGylated TMDCs The biodistribution and clearance behaviors of nanomaterials are highly important for their potential use in biomedicine. However, to our best knowledge, systematic studies to investigate the long‐term in vivo behaviors of various TMDCs in parallel are still missing to date. In this work, healthy Balb/c mice were i. v. injected with PEGylated MoS 2, WS 2, and TiS 2 (10 mg kg −1, in terms of TMDC weight concentrations). Their major organs including heart (H), liver (L), spleen (Sp), lung (Lu), kidney (K), stomach (St), intestine (I), skin (Sk), muscle (M), and bone (B) were collected at different time points post injection and then split into two halves, one for biodistribution study and the other for histology examination. For biodistribution measurement, the organs were weighted and digested by aqua regia. The contents of Mo, W, and Ti were measured by inductively coupled plasma‐atomic emission spectrometry (ICP‐AES) ( Figure 3 a–c). It was found that all three types of PEGylated TMDCs accumulated mostly in RES organs such as liver and spleen at 1 d post injection ( p. i ), as the results of phagocytose by the Kupffer cells and spleen macrophages in these two organs. At later time points, however, we found that the metabolic rate of MoS 2 ‐PEG was much faster than that of WS 2 ‐PEG and TiS 2 ‐PEG. After 30 d, the detected Mo levels decreased sharply to 0. 54 ± 0. 069 % ID g −1 in the liver (from 67. 74 ±10. 26% ID g −1 at day 1) and 0. 017 ± 0. 015% ID g −1 in the spleen (from 77. 70 ±14. 91% ID g −1 at day 1), with most of Mo excreted from the body (Figure 3 a). Different from MoS 2 ‐PEG, the other two types of TMDC nanosheets showed much slower excretion, with rather high levels of W and Ti levels detected in the liver and spleen after 30 d (Figure 3 b, c). Figure 3 d showed the Mo, W, and Ti levels in the liver at various time points. It was found that MoS 2 ‐PEG showed remarkably faster excretion from this organ compared to the other two types of PEGylated TMDCs. Unexpectedly, the W content in the liver increased slightly over time in mice injected with WS 2 ‐PEG, likely owing to the gradual translocation of WS 2 into liver from spleen, in which the W content significantly dropped, as well as the rather slow excretion of WS 2 from the liver. Figure 3 In vivo biodistribution and excretion of PEGylated MS 2 (M = Mo, W, Ti). a–c) Biodistribution of MoS 2 ‐PEG (a), WS 2 ‐PEG (b), and TiS 2 ‐PEG (c) in different organs after i. v injection (H: heart; L: liver; Sp: spleen; Lu: lung; K: kidney; St: stomach; I: intestine: Sk: skin; M: muscle; and B: bone). d) The clearance effect of liver after i. v injection of PEGylated MoS 2, WS 2, and TiS 2 at different time points. e, f) Urinary excretion (e) and fecal excretion (f) of PEGylated MoS 2, WS 2, and TiS 2 at different time points. Error bars in the above data were based on standard deviations of four mice per group. In order to further investigate the clearance pathway, Balb/c mice after i. v injection with PEGylated TMDCs were housed in metabolic cages to collect their urine and feces, which were then digested by aqua regia to measure metal ion contents using ICP‐AES. Interestingly, high concentrations of Mo were detected in both urine and feces from MoS 2 ‐PEG injected mice, while quite low levels of W and Ti were found in the urine and feces of mice injected with the other two types of PEGylated TMDCs (Figure 3 e, f). Those results further confirmed that MoS 2 ‐PEG, unlike the other two types of TMDCs, could be excreted effectively via both renal and fecal pathways. A serial of experiments were next carried out to find out the mechanism of different metabolism behaviors for these three TMDCs. Since the phosphate buffer saline (PBS) is the most frequently used physiological buffer, we have chosen PBS buffer for the in vitro degradation experiments. PEGylated MoS 2, WS 2, and TiS 2 were dissolved in PBS at room temperature for three months. Compared to WS 2 ‐PEG which only showed slightly decreased UV–vis‐NIR absorbance after storage for three months, the optical absorbance of both MoS 2 ‐PEG and TiS 2 ‐PEG samples decreased remarkably ( Figure 4 a–c). Notably, while a transparent colorless solution was left for MoS 2 ‐PEG after incubation for three months, white precipitate was observed in the TiS 2 ‐PEG sample (inset of Figure 4 a–c). We then used X‐ray photoelectron spectroscopy (XPS) to analyze different TMDC samples dissolved in PBS for three month. It could be found that most of Mo element in MoS 2 nanosheets was oxidized to the high valence state (Mo VI ) (Figure 4 d), suggesting the oxidization of Mo IV S 2 with dark brown color into colorless water‐soluble Mo VI ‐oxide species (e. g. MoO 4 2− ). Different from MoS 2, only a part of high valence state of W (W VI ) was observed for the WS 2 ‐PEG sample after three month storage in PBS (Figure 4 e), suggesting the formation of W IV S 2 /W VI O 3 compounds after such incomplete oxidation. Such a difference between MoS 2 ‐PEG and WS 2 ‐PEG could probably be due to the better chemical stability of W IV and stronger chemical band of W‐S in WS 2 compared to the counterparts in MoS 2. 39 As for the TiS 2 ‐PEG sample, the insoluble white precipitate seen in the bottom of Ependor tube was found to be TiO 2 based on both XPS (Figure 4 f ) and XRD data ( Figure S8, Supporting Information). TEM images of these three samples also showed that while no large nanoparticles were found in the degradation product of MoS 2 ‐PEG, lots of aggregates were formed for another two samples (WS 2 ‐PEG and TiS 2 ‐PEG, Figure 4 g–i). Figure 4 Mechanism of the clearance of the MS 2 ‐PEG nanosheets (M = Mo, W, and Ti). a–c) UV–vis‐NIR spectra of MS 2 ‐PEG ( I ) before and after three months ( II ) standing in PBS at the concentration of 0. 02 mg mL −1, (a) MoS 2 ‐PEG, (b) WS 2 ‐PEG, and (c) TiS 2 ‐PEG. Inset: Photos of the MS 2 ‐PEG (M = Mo, W, and Ti) samples before ( I ) and after three months ( II ) standing in PBS solution. d–f) XPS spectra of the Mo VI+ (d), W VI+ and W IV+ after oxidation (e), and Ti IV of TiO 2 nanoparticles from TiS 2 ‐PEG after three months standing in PBS (f). g–i) TEM images of the PEGylated MoS 2 (g), WS 2 (h), and TiS 2 (i) nanosheets after three months standing. j) A scheme showing the different pathways of the clearance of MS 2 ‐PEG nanosheets (M = Mo, W, and Ti). Based on the above observations, we conclude that the different in vivo excretion behaviors of three types of TMDCs should be due to their distinctive chemical properties (Figure 4 j). WS 2 shows relatively high stability in the physiological environment, and can hardly be degraded. Therefore WS 2 ‐PEG after i. v. injection would retain in RES organs for a long time without rapid excretion. TiS 2 is not stable and could be gradually oxidized into water‐insoluble TiO 2 aggregates, which however also could not be easily excreted from the mouse body. In marked contrast, MoS 2 within the physiological environment could be oxidized and transformed into water‐soluble Mo VI ‐oxide species (e. g. , MoO 4 2− ), which are then readily excreted from the mouse body via both renal and fecal pathways. 2. 4 In Vivo Toxicology Study of PEGylated MoS 2, WS 2, and TiS 2 At last, we carefully looked into whether PEGylated TMDCs would exert any toxic effect to mice, by both hematology assay and histology examination. For the same female Balb/c mice used for biodistribution studies, their blood was collected at 1, 7, 30, 60 d post i. v injection of MS 2 ‐PEG (M = Mo, W, and Ti). Various serum biochemical parameters including alkaline phosphatase (ALP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and urea nitrogen (BUN) were measured ( Figure 5 a–d). After 1 d treatment with MoS 2 ‐PEG, all the serum biochemical parameters were close to control group except AST, which increased at 1 d p. i. but dropped 7 d later. The biochemical parameters for the WS 2 ‐PEG treated group showed no obvious difference compared with untreated mice. As for the TiS 2 ‐PEG treatment group, the ALT and ALP activities in plasma were increased compared to the control group. However, all these parameters decreased into the control levels later, demonstrating no irreversible injury to the liver of mice induced by TiS 2 ‐PEG. Notably, all those variations were still within the reference ranges for healthy Balb/c mice. Figure 5 Blood biochemistry and hematology data of Balb/c mice treated with MS 2 ‐PEG nanosheets (M = Mo, W, and Ti). The data were collected at different time point after i. v injection, error bars were based on standard deviations of four mice per group. a) Alkaline phosphatase (ALP), b) alanine aminotransferase (ALT), c) aspartate aminotransferase (AST), and d) urea nitrogen (BUN) levels in the blood at different time point. e) Red blood cells (RBC), f) white blood cells (WBC), g) mean corpuscular volume (MCV), h) hemoglobin (HGB), i) mean corpuscular hemoglobin concentration (MCHC), j) mean corpuscular hemoglobin (MCH), k) platelet (PLT), and l) hematocrit (HCT) levels in the blood at different time point. Gray areas in the figures show the normal reference ranges of hematology data of male Balb/c mice. m ) H&E staining images of liver and spleen from mice treated with MS 2 ‐PEG nanosheets (M = Mo, W, and Ti). The organs were collected from the mice sacrificed at 30 d post injection. The control group was obtained from the untreated mice. For the blood routine examination, red blood cells (RBC), white blood cells (WBC), hemoglobin (HGB), mean corpuscular volume (MCV), mean corpuscular hemoglobin concentration (MCHC), hematocrit (HCT), mean corpuscular hemoglobin (MCH), and platelet (PLT) counts were measured (Figure 5 e–l). All the parameters tested in the treated groups during the monitoring period were within the reference normal ranges, 40 except some variations in platelet counts. The decrease in the platelet count 1 d after treatment may be due to the absorption of TMDCs nanosheets on blood cells, similar to the behavior of some other nanomaterials previously reported. 41, 42 Later on, the platelet count increased rapidly to the normal range. While the first halves of organs were used for biodistribution studies, the other halves of major organs from mice injected with different PEGylated TMDCs were sliced for hematoxylin and eosin (H&E) staining and histological examination. No obvious sign of abnormality, such as inflammation, was noticed in all examined major organs, including liver and spleen with domination accumulation of those nanomaterials (Figure 5 m ). Our results suggest that all of the three PEGylated TMDCs nanosheets have no significant toxicity to mice within 30 d. In particularly, considering the almost complete clearance of MoS 2 ‐PEG within 30 d, it is reasonably to predict that MoS 2 ‐PEG would not cause further long‐term toxicity to the treated animals in a reasonable dose range. 3 Conclusions In summary, we have systemically studied the in vivo biodistribution, excretion, and toxicology profiles of three respective types of TMDCs with surface PEGylation. The biodistribution results indicated abundant accumulation of those nanosheets in RES organs after i. v injection. Notably, after 30 d, MoS 2 ‐PEG could be excreted almost completely, while large amounts of injected W or Ti were still retained in the mouse RES organs for those injected with WS 2 ‐PEG or TiS 2 ‐PEG, respectively. Further degradation experiments uncovered that MoS 2 could be oxidized into water‐soluble Mo VI ‐oxide species (e. g. , MoO 4 2− ) to allow its rapid clearance from the body. In contrast, WS 2 with much higher chemical stability was more difficult to be oxidized and thus showed much longer in vivo retention, while TiS 2 could be oxidized to into water‐insoluble TiO 2 aggregates that were also hard to be excreted. Moreover, further histological and blood analysis showed no obvious long‐term toxicity of these three types of TMDC nanosheets at our tested dose. This work is the first time to carefully compare the in vivo biodistribution, degradation, and excretion behaviors of different TMDCs side‐by‐side. Our results suggest that MoS 2, among various TMDC nanomaterials, may be particularly promising for further biomedical applications owning to its biodegradability and relatively rapid excretion. 4 Experiment Section Materials : All the chemical reagents, unless specified, were purchased and used without further purification. Tungsten (VI) chloride (WCl 6 ), molybdenum (V) chloride (MoCl 5 ), titanium tetrachloride (TiCl 4 ), oleylamine (OM), 1‐octadecene (ODE), and NMP were purchased from Sigma‐Aldrich. Sulfur powder (S) was obtained from Sinopharm Chemical Reagent Co. , Ltd. (China). Lipoic acid was purchased from Sigma‐Aldrich and mPEG‐NH 2 (MW = 5K) were obtained from Biomatrik Co. , Ltd (Jiaxing, China). Deionized water used in our experiments was obtained by using a Milli‐Q water system. Synthesis of MoS 2, WS 2, and TiS 2 Nanosheets Synthesis : First, MoS 2, WS 2, and TiS 2 nanoflakes were synthesized according the previous report with slight modification. 19, 36 1 mmol MoCl 5 /WCl 6 TiCl 4 /was dissolved in 15 mL oleylamine (OM) and 10 mL 1‐octadecene (ODE) in a three‐necked flask under magnetic stirring, then heated to 150 °C with the protection of nitrogen atmosphere and kept for 20 min to remove the low‐boiling‐point impurities. Then the temperature was increased to 300 °C slowly. 2. 5 mmol sulfur powder dissolved in 4 mL OM was injected to the reaction solution rapidly. The solution temperature was kept at 300 °C for another 30 min and then cooled down to the room temperature. The product was dissolved in cyclohexane and washed with ethyl alcohol/cyclohexane (V:V = 1:1) for three times. After that, as‐synthesized MoS 2, WS 2, and TiS 2 nanoflakes were dissolved in NMP solution, and sonicated for 5 h using a water‐bath ultra‐sonicator. The nanosheets were precipitated by centrifugation with ethyl alcohol and then dissolved in DI water. Functionalization of MoS 2, WS 2, and TiS 2 Nanosheets : LA‐PEG was synthesized following a literature procedure. 43 20 mg LA‐PEG was added to 10 mL aqueous solution of MoS 2, WS 2, or TiS 2 (1 mg mL −1 ) under magnetic stirring at room temperature overnight. Then the solution was centrifuged 4000 rpm for 20 min by Amicon filters (Millipore) (MWCO = 10 kDa) and washed with DI water to remove excess polymer, obtaining PEGylated MoS 2, WS 2, and TiS 2. Characterization : The TEM images were taken by a FEI Tecnai F20 TEM. XRD measurement was performed by a PANalytical X‐ray diffractometer at Cuka radiation (λ = 0. 15406 nm). UV–vis‐NIR spectra of PEGylated MoS 2, WS 2, and TiS 2 were acquired by a PerkinElmer Lambda 750 UV–vis‐NIR spectrophotometer. The sizes of PEGylated MoS 2, WS 2, and TiS 2 were measured by dynamic light scattering (MALVERN ZEN3690). The absolute Mo/W/Ti contents as well as MS 2 weight concentrations in different MS 2 ‐PEG samples were measured by ICP‐AES (Vista Mpx 700‐ES). In Vitro Cytotoxicity Assay : The mouse macrophage RAW 264. 7 cells, murine breast cancer 4T1 cells, and human renal epithelial cell 293T were chosen to evaluate the cytotoxicity of PEGylated MoS 2, WS 2, and TiS 2 nanosheets. Both cells were originally obtained from American Type Culture Collection (ATCC) and cultured at 37 °C within 5% CO 2 in RPMI‐1640 cell medium supplemented with 10% fetal bovine serum (FBS). For the MTT assay, the cells were seeded in 96‐well plates and incubated with PEGylated MoS 2, WS 2, and TiS 2 at different concentrations. 24 h later, the cells were washed with fresh medium and preceded to the standard MTT assay or LDH assay following vendors' protocols. For ROS detection, cells after treatment with PEGylated MoS 2, WS 2, and TiS 2 for 24 h were incubated with dihydroethidine (DHE) probe for 40 min. Then cells were collected, washed with PBS twice, and preceded to flow cytometry analysis (FACSCalibur, BD). Biodistribution Study : Female Balb/c mice (20 g ± 2 g) were bought from Nanjing Peng Sheng Biological Technology Co. Ltd and used under protocols approved by Soochow University Laboratory Animal Center. Balb/c mice were i. v injected with PEGylated MoS 2, WS 2, and TiS 2 (200 μL per mouse, 1 mg mL −1, dosage = 10 mg kg −1, in terms of MS 2 weight concentrations). Control group were i. v injected with 200 μl PBS. Four mice were used per group. The mice were sacrificed with major organs including heart, liver, spleen, lung, kidney, stomach, intestine, skin, muscle, bone collected at day 1, day 7, day 14, day 30, and day 60 p. i. Those organs were added with 10 mL aqua regia (HCl:HNO 3 :HClO 4 = 3:1:1) and heated to 200 °C for 2 h. After being cooled down to room temperature, each sample was diluted to 10 mL by DI water and passed through a 0. 22 μm filter to remove any undigested tissue. The amount of Mo, W, and Ti were measured by ICP‐AES (Vista Mpx 700‐ES). Excretion Study : Female Balb/c mice were i. v injected with PEGylated MoS 2, WS 2, and TiS 2 (dosage = 10 mg kg −1 ) with four mice per group. Each mouse was placed in metabolism cage. Their urine and feces were collected at regular time points. The metabolism cages were washed after each round of collection. The collected urine and feces were digested with aqua regia for ICP‐AES measurement of Mo, W, and Ti contents (Vista Mpx 700‐ES). Toxicology Study : Female Balb/c mice were i. v injected with PEGylated MoS 2, WS 2, and TiS 2 (dosage = 10 mg kg −1, in terms of MS 2 weight concentrations) with four mice per group. At 1, 7, and 30 d p. i. , blood from those mice were drawn through their orbital venous plexus. For the blood chemistry analysis, alkaline phosphatase (ALP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and urea nitrogen (BUN) were measured. For the complete blood panel test, red blood cells, white blood cells, hemoglobin, mean corpuscular volume, mean corpuscular hemoglobin concentration, mean corpuscular hemoglobin, hematocrit, and platelet count were measured. All the assays were tested in Shanghai Research Center for Biomodel Organism. For histology study, heart, liver, spleen, lung, kidney collected from those mice were fixed in 10% neutral buffered formalin, processed routinely into paraffin, sectioned into 8 μm thickness slices, and then stained with hematoxylin & eosin (H&E). The slices were observed with a digital microscope (Leica QWin). Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201600285
| 2,017
|
Advanced Science
|
Self‐Assembled Bifunctional Peptide as Effective Drug Delivery Vector with Powerful Antitumor Activity
|
E‐cadherin/catenin complex is crucial for cancer cell migration and invasion. The histidine‐alanine‐valine (HAV) sequence has been shown to inhibit a variety of cadherin‐based functions. In this study, by fusing HAV and the classical tumor‐targeting Arg‐Gly‐Asp (RGD) motif and Asn‐Gly‐Arg (NGR) motif to the apoptosis‐inducing peptide sequence‐AVPIAQK, a bifunctional peptide has been constructed with enhanced tumor targeting and apoptosis effects. This peptide is further processed as a nanoscale vector to encapsulate the hydrophobic drug docetaxel (DOC). Bioimaging analysis shows that peptide nanoparticles can penetrate into xenograft tumor cells with a significantly long retention in tumors and high tumor targeting specificity. In vivo, DOC/peptide NPs are substantially more effective at inhibiting tumor growth and prolonging survival compared with DOC control. Together, the findings of this study suggest that DOC/peptide NPs may have promising applications in pulmonary carcinoma therapy.
|
1 Introduction Lung cancer is the leading cause of cancer death among males in many parts of the world, and in more developed countries, it has surpassed breast cancer among females. 1, 2 Advanced non‐small‐cell lung cancer (NSCLC) is responsible for most of the death. Docetaxel (DOC), erlotinib, and pemetrexed are the clinically approved second‐line therapies for NSCLC. 3, 4 DOC is a semisynthetic product derived from the needles of the European yew, Taxus baccata. The antineoplastic activities of DOC were played by promoting assembly of tubulin into microtubules, and rendering the microtubules resistant to depolymerization. 5, 6 In response to the increasing challenge of lung cancer worldwide, polymeric nanoparticles (NPs), polymeric micelles, and liposomes have been drawn significant attention to the delivery of DOC to lung cancer cells via a passive targeting mechanism. 7, 8 Although many strategies have been adopted to treat lung cancer, the low target specificity for cancer cells is the most common and serious problem. NPs have the capability of delivering high doses of therapeutic compounds to tumor cells and have attracted increasing attention in drug delivery and cancer therapy. 9, 10, 11 Encapsulation of hydrophobic drugs into NPs can make the drug dispersible in aqueous solutions, which can be used for intravenous applications. The nanoscale and hydrophilic interface of the NPs can prolong their circulation time and enhance the cellular uptake. 12 Recently, vaccines, peptides, proteins, small molecules, and other biomimetic carriers have been reported for various therapeutic payloads delivery. 13, 14, 15 Among these, small molecules or peptides have been developed for active targeting of cytotoxic agents to tumor cells, 16, 17 with the advantage of tissue‐specific targeting, thus off‐target effects could be minimized. 18, 19, 20 These vectors with unique structures are able to self‐assemble into a variety of nanostructures including nanocircles, nanotubes, and spherical NPs, which can enhance the gene or drug loading (DL) capability for their high‐aspect‐ratio structures. 21, 22 Moreover, taking the peptides for example, it is easy to endow it with targeting ability through simple modifications, such as conjugated with specific sequences and mediated by avidin–biotin interaction. 16, 23 RGD (Arg‐Gly‐Asp), a tripeptide motif, is an integrin‐recognition motif found in many ligands, allowing for homing of drugs to tumor blood vessels. 24, 25 RGD‐containing peptides have been widely used in tumor‐targeting research. 25 Peptides containing the Asn‐Gly‐Arg (NGR) motif which can recognize aminopeptidase N (CD13), a membrane‐bound enzyme associated with angiogenic tumor vessels, can be used for delivering various anti‐tumor agents to the tumor vasculature. 26 Based on this, RGD‐ and NGR‐containing peptides can actively target to tumor blood vessels by binding to integrin and CD13 receptors. Neuronal (N)‐cadherin, a cell adhesion molecule, can mediate cell–cell adhesion via a calcium‐dependent mechanism. Researchers have shown that inappropriate expression of N‐cadherin by tumor cells may promote motility and invasion in carcinoma cells. 27, 28 Histidine‐alanine‐valine (HAV) motif, is a highly conserved sequence at classical cadherin homophilic binding site. It has been confirmed that linear or cyclic peptides containing the HAV sequence of N‐cadherin can disrupt cadherin‐mediated cell adhesion, and hence inhibit cell aggregation, compaction, and neurite outgrowth. 29, 30 In addition, AVPIAQK, Smac/DIABLO (smacN7), is an inhibitor of apoptosis protein antagonist and a promoter of caspase activation. 31 In this study, we have designed a bifunctional peptide HAVRNGRRGDGGAVPIAQK (HRK‐19), conjugated with tumor targeting peptide sequences HAV and NGRRGD, as well as apoptosis‐inducing peptide sequence AVPIAQK. We further prepared self‐assembled NPs using the bifunctional peptide and DOC for potential application in drug delivery systems. The DOC loaded peptide NPs (DOC/peptide NPs) were prepared by directly dissolving in aqueous buffer with DOC and the bifunctional peptide which was pretreated with hexafluoro‐2‐propanol to achieve molecular level mixing. Antitumor effects of the DOC/peptide NPs were investigated both in vitro and in vivo. Molecular‐modeling study was employed to investigate the interactions of DOC and the bifunctional peptide. The cellular uptake and cell apoptosis of the DOC/peptide NPs were also examined. The potential mechanism of the DOC/peptide NPs inhibiting tumor growth in mice is illustrated in the Figure 1 : 1) Targeting tumor‐enriched‐cadherin by HAV sequence; 2) Targeting tumor‐enriched‐integrin and ‐CD13 by NGRRGD sequence; 3) Transportation of NPs into tumor cells through endocytosis; 4) Inducing antitumor effects by apoptosis‐inducing peptide AVPIAQK and DOC. Our findings indicated that the DOC/peptide NPs showed significant antitumor activity both in vitro and in vivo, and may present a promising DOC formulation for cancer chemotherapy. Figure 1 Schematic representation of mechanisms of the DOC/peptide NPs inhibiting tumor growth of mice. 2 Results 2. 1 Tumor‐Targeting Peptide Design and Self‐Assembly Peptide Nanoparticles Preparation The mass spectrum of the HRK‐19 peptide was shown in Figure S1 of the Supporting Information. And it can be seen the mass of HRK‐19 peptide was 1959. 80 g mol −1, which was consistent with the theoretical mass (1959. 20 g mol −1 ). The HRK‐19 peptide had ≥95% purity, as confirmed by analytical reversed‐phase (RP)‐high performance liquid chromatography (HPLC) (Figure S2, Supporting Information). As shown in Figure S3 of the Supporting Information, the HRK‐19 peptide contains no apparent secondary structure such as helix and Beta turn. As shown HRK‐19 is an amphiphilic peptide with its N‐terminal hydrophilic and C‐terminal hydrophobic (Figure S4, Supporting Information). To prepare a tumor‐targeting peptide, in this study, Asn‐Gly‐Arg‐Arg‐Gly‐Asp (NGRRGD) and three amino acids (HAV) were conjugated with apoptosis‐inducing peptide sequence AVPIAQK, forming bifunctional peptide (HRK‐19 peptide, HAVRNGRRGDGGAVPIAQK). As shown schematically in Figure 2 A, self‐assembly of the DOC/peptide NPs was prepared by the film dispersion method. The amino acid sequence and molecular model of the peptide was shown in Figure 2 B. The DL, encapsulation efficiency (EE), particle size and polydisperse index (PDI) of a series of DOC/peptide NPs were showed in Table S1 of the Supporting Information. It can be seen that the DL, and particle size increased with increase of drug/copolymer ratio in feed, whereas, the EE decreased accordingly. In consideration of DL and stability, the sample of S2 was chose for further applications and was characterized in detail. The DL and EE of DOC were 4. 52 ± 0. 18% and 90. 50 ± 3. 68%. According to the particle size distribution spectrum shown in Figure 2 C, the average particle size and PDI of the DOC/peptide NPs were 62. 41 ± 0. 83 nm, 0. 125 ± 0. 021, and the zeta potential was 3. 07 ± 0. 13 mV (Figure S5, Supporting Information). In addition, the prepared NPs were monodisperse and had a very narrow particle size distribution. Transmission electron microscopy (TEM) imaging was carried out to identify the possible nanostructures self‐assembled from the prepared bifunctional peptide. And it can be seen that the peptide NPs observed by TEM (Figure 2 D) was in good agreement with the results of particle size analysis. The nanostructure of DOC/peptide NPs observed by particle size analysis, as well as TEM imaging, demonstrated that the prepared DOC/peptide NPs were spherical NPs, and could be well‐dispersed in aqueous solution. Figure 2 E showed the appearance of Dulbecco's phosphate‐buffered saline (DPBS), and clear solution with Tyndall effect of DOC/peptide NPs in DPBS (right) could be observed, which revealed that DOC/peptide NPs are stable and could be well‐dispersed in DPBS. Figure 2 F presented the X‐ray diffraction (XRD) patterns of DOC, pure peptide, and DOC/peptide NPs. Pure DOC is crystalline, with characteristic peaks at 2θ = 8. 10°, 9. 32°, 11. 38°, 12. 60°, 13. 94°, 16. 94°, and 20. 42°. Pure peptide is noncrystalline. When comparing XRD diagrams of DOC, the absence of specific diffraction peaks in the DOC/peptide NPs diagram indicated that DOC was relatively completely encapsulated. In vitro release profile of DOC from DOC/peptide NPs in PBS solution at pH 7. 4 was presented in Figure 2 G. DOC was rapidly released and reached its peak release of 85. 37 ± 1. 83% of the total drug within 24 h. In comparison, DOC was released from the DOC/peptide NPs over an extended period. Figure 2 Self‐assembly characterization of the DOC/peptide NPs. A) Schematic illustration of the self‐assembly of DOC/peptide NPs. B) Amino acid sequence and molecular model of the peptide investigated. The carbon atoms on the imidazole ring of histidine residue are colored with green. C) Particle size distribution of DOC/peptide NPs. D) TEM image of DOC/peptide NPs in Dulbecco's phosphate‐buffered saline (DPBS), the scale bar represents 100 nm. E) Photograph of DPBS (a), and DOC/peptide NPs in DPBS (b). F) XRD patterns of docetaxel, peptide, and DOC/peptide NPs. G) In vitro release behavior of DOC from DOC/peptide NPs. 2. 2 Interaction between DOC and the HRK‐19 Peptide The interaction between two molecules of HRK‐19 peptide and DOC was investigated with the method of molecular dynamics simulation. At first, DOC was merged randomly to the simulated peptides in the workspace of Hyperchem so as to get the initial structure of a complex. Then two stages of molecular dynamics simulations were performed in the environment of water and solvation effect was considered implicitly. In the heating stage, complex was heated from 0 to 300 K within 100 ps. Subsequently, the complex was running at 300 K for 100 ps. Results of molecular dynamics simulation were presented in Figure 3. It was noted that both the HRK‐19 peptide and DOC were trying to adjusting their conformations and the position until a favorable interaction mode was reached. It can be seen from the overall shape of the HRK‐19 peptide and from the relative positions of the carbon atoms on the imidazole ring of peptides that the conformation of the peptide changed a lot. After 200 ps molecular dynamics simulation, a binding site between the peptide and the DOC can be formed on the surface of the HRK‐19 peptide due to continuous interaction (Figure 3 D). Interaction energy between two molecules of peptide and DOC, which was calculated at molecular mechanic level with CHARMM27 force field, 32 is −38. 378 kcal mol −1. At the initial state (Figure 3 A), the interaction energy between two molecules of peptide and DOC should be close to 0 kcal mol −1, and the interaction energy at final state (Figure 3 D) was −38. 378 kcal mol −1. This showed strong interaction between HRK‐19 peptide and DOC. Figure 3 Interaction modes between peptides and DOC revealed by molecular dynamics simulation in water environment. A) The initial conformation of complex composed of two molecules of peptides and docetaxel; Conformations (B)–(D) are corresponding to snapshots of the complex collected at 88. 000, 104. 250, and 200 ps, respectively. Peptides are represented with stick. The carbon atoms on the imidazole ring of peptides are colored with green. Docetaxel is shown ball‐and‐stick style and the carbon atoms of its terminal methyl group are colored with green. 2. 3 Cellular Uptake To investigate their ability to transport hydrophobic molecules into cancer cells of the DOC/peptide NPs, cellular uptake studies of drug loaded peptide NPs were performed on A549 cells. Coumarin‐6 (C6) was chosen as the model dye loaded into peptide NPs because of its strong fluorescence and low solubility in water, using the same procedure as for DOC loading. 33 Figure 4 A showed the time‐dependent cellular uptake of free C6, and C6/peptide NPs into A549 cancer cells, evidenced by the stronger green fluorescence in the cells with longer incubation time. In free C6 group, a very dim fluorescence was observed in the cytosol of several cells after 2 h, and fluorescence intensity was increased after incubation for 4 h. In contrast, stronger fluorescence intensity can be found for C6/peptide NPs group than free C6 after 0. 5 h. C6/peptide NPs could rapidly accumulate in the A549 cells in 2 h, and a more bright green fluorescence was observed after 4 h. These results were further confirmed quantitatively by flow cytometry (FCM) measurements (Figure 4 B), again revealing a consistent increase in fluorescence intensity when the incubation time changed from 0. 5 to 4 h. And it also showed that fluorescence intensity of cells treated with C6/peptide NPs is much stronger than that in free C6 group at the same incubation time. Figure 4 Cellular uptake of free coumarin, coumarin/peptide by A549 cells. A) Confocal images of cells treated with free coumarin and coumarin loaded NPs at 100 ng for 0. 5, 2, and 4 h. B) Mean fluorescence intensity of cells after treatment for 0. 5, 2, and 4 h. Cells incubated in media only were used as control. 2. 4 The Effects of DOC on Microtubule Formation Agents targeting microtubules are often used in cancer therapy. The taxanes act to stabilize microtubules and dampen microtubule dynamics to prevent the normal formation of mitotic spindles. 34 Both paclitaxel and DOC can bind to β‐tubulin in assembled tubulin, thereby reducing depolymerization and leading to cell death. 35 Therefore, we examined the effects of DOC on microtubule formation and cell apoptosis on A549 cells. Figure 5 showed the microtubule formation examined by fluorescence microscopy. Microtubule formation was assayed by the tubulin stain (red) and the nuclear was revealed by DAPI (4′, 6‐diamidino‐2‐phenylindole) stain (blue). As shown in Figure 5, in the absence of DOC, cells go through the cell cycle with the formation of normal tubulin spindle in (Figure 5, Control). However, at 10 × 10 −9 m, DOC begins to arrest A549 cells by interfering with the assembling/disassembling microtubule spindle. Figure 5 The effects of DOC on the microtubule formation on the A549 cell lines. A549 cells were seeded on coverslip and treated with docetaxel of indicated concentrations for 4 h. The microtubule was determined by fluorescence microscopy. The red indicates the α‐tubulin and the blue is the DAPI stain for DNA. Scale bar: 20 µm. 2. 5 Cell Apoptosis Assay Given the effectiveness of transporting hydrophobic drug‐DOC into A549 cancer cells by the bifunctional peptide NPs, we carried out the cytotoxicity experiments to further evaluate the anticancer activity of DOC/peptide NPs on A549 by Flow cytometry analysis. Flow cytometry analysis was performed to monitor the percentage of cell apoptosis after the cells being stained by Annexin V‐Fluorescein isothiocyanate/propidium iodide (Annexin V‐FITC/PI) apoptosis detection kit, and A549 cells incubated in media only were used as control ( Figure 6 A). The percentage of apoptosis of A549 cells (early and late apoptosis) incubated for 24 h with DOC/peptide containing DOC 5 × 10 −9 and 10 × 10 −9 m were 11. 56 ± 0. 91% and 20. 66 ± 1. 89%, respectively. For free DOC, they were 8. 87 ± 0. 79% and 15. 73 ± 0. 56%, respectively. Obviously, DOC‐loaded bifunctional peptide NPs induced significant cell apoptosis. In short, these results indicated that the bifunctional peptide NPs could facilitate the inhibition of tumor proliferation and induction of tumor cell apoptosis, as was consistent with the results concerning cellular internalization described above, leading to stronger anti‐tumor efficacy. Figure 6 Cell apoptosis assay in vitro and N ‐cadherin immunoreactivity in vivo. A) Ratio of apoptotic and necrotic cells treated for 24 h with various DOC formulations. The asterisks “*”and “**” on graph obtained by Student's t ‐test indicate significant differences at P < 0. 05 and P < 0. 01, respectively. B) Photomicrograph shows strong membrane positivity (the red) of N ‐cadherin in the tumor cells: a) normal saline group, b) blank peptide group, c) Taxotere group, and d) DOC/peptide NPs group. 2. 6 In Vivo Therapeutic Efficacy Subcutaneous A549 model and pulmonary metastatic A549 model were used to compare the antitumor activity of DOC/peptide NPs with that of Taxotere, blank peptide, and normal saline (NS). Figure 6 B showed the strong membrane positivity of N‐cadherin in the tumor cells. Cells treated with the free peptide can induce partial loss of cadherin labeling. And treatment with DOC/peptide NPs showed more loss of cadherin labeling than free peptide. As shown in Figure 7 A, fast tumor growth curves were obtained in the control groups (NS and blank peptide), suggesting that the A549 tumor growth was not inhibited by blank peptide. Compared with the NS groups, DOC/peptide NPs was effective in inhibiting tumor growth (Figure 7 A–D). As shown in Figure 7 B, Taxotere treatment induced a decrease in the body weight of the mice. According to Figure 7 D, tumor weight in DOC/peptide group (0. 035 ± 0. 005 g) is significantly lower than that in Taxotere (0. 3325 ± 0. 060g, P < 0. 001), blank peptide (1. 01 ± 0. 159 g, P < 0. 001), or NS group (1. 08 ± 0. 17 g, P < 0. 001). This further confirmed that DOC/peptide exhibited better therapeutic efficacy than Taxotere in A549 tumor‐bearing mouse model (Figure 7 D). Figure 7 DOC/peptide NPs inhibited growth in subcutaneous A549 model. A) Tumor development curve. Balb/c nude mice were subcutaneously with 100 µL of A549 cells on day 0. On day 7, the mice were randomly assigned to four groups that received intravenous normal saline (NS), blank peptide, Taxotere, or DOC/peptide NPs. B) The body weights of different groups. C) Representative photos of tumors in each treatment group. D) The tumor weights of the mice measured on the indicated days. E) CD31 immunofluorescent staining of tumors. F) Ki67 immunohistochemical staining of tumors. The asterisk “**” on graph obtained by Student's t ‐test indicates significant differences at P < 0. 01. In panels (D) and (E), the four groups were a) NS group, b) blank peptide group, c) Taxotere group, and d) DOC/peptide NPs group, respectively. Angiogenesis is a process vital to the continued development of a tumor mass. This process has been the subject of intense research due to its role in cancer development. 36 Sections of tumors from mice in each group were stained for CD31 immunofluorescence to determine the microvessel density (MVD) as a measurement of tumor angiogenesis. The DOC/peptide NPs treatment resulted in dramatic inhibition of angiogenesis in the tumors (Figure 7 E). The MVD in the DOC/peptide NPs group was 10. 67 ± 1. 53, which was dramatically lower than that in the Taxotere group (47. 33 ± 4. 04, P < 0. 01), blank peptide group (88. 00 ± 8. 54, P < 0. 01), and NS (92. 67 ± 7. 23, P < 0. 01) group. The results implied that antiangiogenesis may be another mechanism of inhibiting cancer by the DOC/peptide NPs in vivo. Moreover, the activities of various drug formulations on proliferation of tumor cells were analyzed by immune‐histochemical staining for Ki‐67, a proliferation marker (Figure 7 F). Compared with NS group (73. 50 ± 4. 11%, P < 0. 01), blank peptide group (72. 00 ± 9. 24%, P < 0. 01), and Taxotere group (53. 00 ± 7. 42%, P < 0. 01), the percentage of Ki‐67‐positive cells in the DOC/peptide NPs‐treated group (13. 67 ± 4. 11%) was significantly lower. In pulmonary metastatic model ( Figure 8 ), the body weight of the Taxotere treated animals was significantly lower than that of other groups (Figure 8 A). As shown in Figure 8 B, the median survival in DOC/peptide group (52 days) is significantly longer compared with Taxotere (45 d, P < 0. 05), blank peptide (35 d, P < 0. 05), and NS (30 d, P < 0. 05) group. The weight of lungs (Figure 8 D) in DOC/peptide NPs group (0. 31 ± 0. 05 g) was dramatically decreased compared with that in blank peptide (0. 74 ± 0. 13 g, P < 0. 001), or NS group (0. 71 ± 0. 14 g, P < 0. 001). However, the weight of lungs in Taxotere group (0. 23 ± 0. 07) was lower than that of DOC/peptide NPs group, that was because of the toxic effect of Taxotere on the mice. The H&E staining of lungs showed the anticancer effect of Taxotere group was worse than that of DOC/peptide NPs group. There has obvious tumor in the lungs of the Taxotere group (Figure 8 E). Figure 8 DOC/peptide NPs inhibited growth and metastasis in pulmonary metastatic A549 model. A) The body weights of the mice measured on the indicated days. Balb/c nude mice were injected intravenously with 100 µL of cell suspension containing 1 × 10 7 A549 cells on day 0. On day 5, tumor bearing mice were assigned randomly into four groups that received intravenous normal saline (NS), blank peptide, Taxotere, or DOC/peptide NPs. B) Kaplan–Meier survival curve of mice in each group. C) Representative photos of lungs in each treatment group. D) Weight of lungs in each group. E) H&E staining of lungs in each treatment group: a) NS group, b) blank peptide group, c) Taxotere group, and d) DOC/peptide NPs group. The asterisk “*” on graph obtained by Student's t ‐test indicates significant differences at P < 0. 05. 2. 7 Evaluation of Tumor Targeting and Penetrating Efficiencies of Drug/Peptide NPs in Vivo Figure 9 A showed the real time distribution and tumor accumulation of physiological saline (NS), free 1, 1′‐dioctadecyl‐3, 3, 3′, 3′‐tetramethyl indotricarbocyanine (DiD), DiD‐loaded bifunctional peptide NPs (DiD/peptide NPs), in the presence of excess free bifunctional peptide (DiD/peptide NPs) at 1, 4, 8, and 24 h after iv injection. After 1 h, high DiD fluorescence was observed in mice, especially in the tumor of mice treated with DiD/peptide NPs (Figure 9 A). As time elapsed, the fluorescence accumulated in the tumors of DiD/peptide NPs mice, while the average free DiD signal decreased significantly (Figure 9 A). This observation demonstrated that the active targeting group exhibited slower clearance, that is, it showed the long circulation effect of bifunctional peptide NPs systems, which was consistent with cellular uptake results. Figure 9 Evaluation of tumor targeting and penetrating efficiencies of drug/peptide NPs in vivo. Mice with forelimb tumors derived from A549 cells were given tail vein injections of NPs loaded with the fluorescent dye DiD (100 µg kg −1 ). Normal saline was served as the control. A) In vivo images of mice after treatment with free DiD or DiD/peptide NPs for 1, 4, 8, and 24 h. B) Frozen sections of tumors removed 1 or 4 h after treatment with free DiD or DiD/peptide NPs were stained with DAPI to label nuclei. To further investigate the accumulation and penetration of DiD/peptide NPs, nude mice bearing tumors from A549 cells were injected with free DiD, DiD/peptide NPs, at a DiD concentration of 100 µg kg −1. After treatment for 1 or 4 h, tumors were excised for examination. As shown in Figure 9 B, the fluorescence of tumor cells treated with DiD/peptide NPs was higher than free DiD at different time, which was congruent with the in vivo imaging data. 3 Discussion In this study, a type of amphiphilic peptide with tumor‐targeting NGRRGD sequence and HAV sequence and tumor apoptosis‐inducing AVPIAQK sequence were designed and synthesized. The proof‐of‐principle experiments were conducted to test the cancer targeting value of the bifunctional peptide and the therapeutic value of the self‐assembly of DOC/peptide NPs. The results showed that the DOC/peptide NPs were effective in suppressing growth of A549 NSCLC in vitro and in vivo. DOC, a chemotherapeutic agent with a large spectrum of antitumor activity, was approved for second‐line treatment of advanced NSCLC. 37, 38 Due to the poor water solubility, nonionic surfactants—Cremophor EL and Tween 80—were often used for their commercial formulations. The side effects with these nonionic surfactants include hypersensitivity reactions and peripheral neuropathy in human. 39 Many strategies have been used to improve its water solubility, among these, nanotechnology shows promising application in drug delivery system. The common NPs' delivery of cytotoxic cargoes to tumor tissues can be achieved by passive targeting, which makes the use of the inherent size of NPs and exploits the enhanced permeability and retention (EPR) effect. 40, 41, 42 However, the presence of the complex in vivo environment limits the EPR effect, thereby resulting in the poor efficacy in drug delivery. A feasible approach for overcoming this issue is to modify these carriers with different targeting ligands that facilitate the translocation across the endothelium of the tumor vasculature. 43 NGRRGD can mediate tumor‐homing by binding to α v β 3 and CD13 receptors, which are tumor angiogenesis biomarkers up‐regulated on various cells in tumors. 44 HAV is a sequence resides in the first extracellular repeat of the classical cadherins. Studies have shown that peptides mimetic the HAV motif in N‐cadherin are sufficient to inhibit cell–cell adhesion and neurite outgrowth. The self‐assembled peptides have been widely investigated because of their good biocompatibility and used for cancer treatment, tissue engineering, and wound healing. 45, 46, 47, 48 In this study, a bifunctional peptide conjugated with tumor targeting peptide sequences‐NGRRGD and HAV, and apoptosis‐inducing peptide sequence‐AVPIAQK was prepared. Then, we used this bifunctional peptide and a film dispersion method to obtain DOC nanoparticles (DOC/peptide NPs) with a very narrow particle size distribution, thus leading to remarkable improvement in solubility of the DOC in DPBS solution. In addition, the results of particle size distribution, combined with nanostructure of the DOC/peptide NPs observed by TEM, suggested that the prepared NPs were stable and could be well‐dispersed in aqueous solutions. In comparison with free DOC, a much slower release behavior of DOC/peptide NPs can be seen. Cellular uptake of peptide NPs was evaluated using confocal microscopy and FCM, respectively. The results indicated that the improved cytotoxicity of peptide NPs was attributed to their enhanced uptake by cells. Sugahara et al. showed that co‐administration of iRGD with cancer drugs was slightly more effective than the conjugated drug at inhibiting tumor growth and accumulation. 49 And Zhao et al. also showed that combined application of α‐helical peptides and antitumor drug doxorubicin at low concentrations was significantly more effective than either drug alone against HeLa tumors. 50 So in this study, we think coadministration of the bifunctional peptide loaded with DOC appears to have more effective anticancer activity and great clinical potential. The main mechanism of action of DOC against cancer cells is to stabilize microtubules and prevent depolymerization stabilized microtubules that could further lead to cell apoptosis. In cytotoxicity tests, we found that cells treated with free DOC or DOC/peptide NPs exhibited similar concentration‐dependent cell apoptosis and blank peptide had a very low toxicity on tumor cells. These experiments suggest a similar antitumor activity between DOC/peptide NPs and free DOC. Furthermore, in vivo animal experiments showed that the DOC/peptide NPs efficiently inhibited both growth and metastasis of tumors, and prolonged the survival of tumor‐bearing mice in pulmonary metastatic A549 model, all results further confirmed that DOC/peptide NPs exhibited better therapeutic efficacy than commercial formulation Taxotere under the same experiment conditions. Particles with a small size (<200 nm) can easily extravasate to extravascular tumor site due to the EPR effect. 51 On the other hand, the bifunctional peptide used in this study can improve its tumor tissue‐specific targeting due to tumor‐targeting sequences—NGRRGD and HAV. Therefore, DOC/peptide NPs may enhance drug accumulation in tumor tissues, and decrease drug extravasation from normal vessels into normal tissues. To confirm above hypothesis, we examined in vivo tumor targeting of peptide NPs with DiD as a fluorescent probe, and we found that DiD/peptide NPs accumulated at higher concentrations in the tumor than free DiD. Moreover, the in vivo antitumor study further demonstrated the superior antitumor activity of DOC/peptide NPs. Results of CD31 staining of tumor tissues suggested improved antiangiogenesis effect of DOC/peptide NPs. 4 Conclusions We report here the use of the bifunctional peptide as a molecular building unit to construct a nanoscale vector to encapsulate the hydrophobic drug DOC. After DOC was encapsulated into peptide NPs, cellular uptake and in vitro cytotoxicity effect of DOC were increased compared with free DOC. Besides, a sustained in vitro release behavior was observed in DOC/peptide NPs group. Bioimaging analysis showed that DOC/peptide NPs could penetrate into xenograft tumor cells with a significantly long retention in tumors and high tumor targeting specificity. In vivo xenograft studies showed that DOC/peptide NPs were more effective in inhibiting tumor growth and prolonged survival. Thus, the DOC/peptide NPs prepared in this work showed improved antitumor activity in vitro and in vivo, and may have potential applications as an intravenous therapy in lung cancer therapy. 5 Experimental Section Materials : The HRK‐19 peptide (HAVRNGRRGDGGAVPIAQK, theoretical mass = 1959. 20 g mol −1, 99. 5% purified powder) was synthesized commercially using solid phase synthesis methods by the Shanghai Bootech BioScience & Technology Co. , Ltd. , Shanghai, China. Peptide stock solutions were prepared by dissolving the peptide powders in DPBS to a concentration of 5 mg mL −1, mixing, sonicating for 40 s, filtering, and then storing at 4 °C. For the TEM measurement, the peptide stock solution was diluted to a concentration of 0. 25 mg mL −1 (100 × 10 −6 m ). DOC was purchased from Sichuan Xieli Pharmaceutical Co. , Ltd. , (Chengdu, China). Taxotere was commercially available from Sanofi (Paris, France). Annexin V‐FITC Apoptosis Detection Kit was obtained from KeyGen Biotech (Nanjing, China), Ki‐67 Rabbit Monoclonal Antibody was obtained from Abcam (Massachusetts, USA). DAPI, C6, and 3‐(4, 5‐dimethyl‐2‐thiazolyl)‐2, 5‐diphenyl‐2H‐tetrazolium bromide (methyl thiazolyl tetrazolium) were obtained from Sigma‐Aldrich (Saint Louis, USA). DiD was purchased from Biotium (Hayward, CA). Dulbecco's modified Eagle's medium and fetal bovine serum (FBS) were purchased from HyClone (Logan, USA). All the animals were provided with standard laboratory chow and tap water ad libitum and all the procedures were performed according to the protocol approved by the Institutional Animal Care and Treatment Committee of Sichuan University (Chengdu, People's Republic of China). All mice were treated humanely throughout the experimental period. HRK‐19 Peptide Analysis : RP‐HPLC of the peptide was performed on an Agilent 1100 liquid Chromatograph with a vydac C18 column (4. 6 × 250 mm 2, 5 mm). Samples performed with 0. 1% trifluoroacetic acid (TFA) in aqueous solution/0. 1% TFA in ACN (gradient elution from 80:20 to 10:90, v/v) over 20 min, and then run using gradient elution from 10:90 to 0:100 (v/v) for another 5 min, followed by a plateau to complete 30 min. The mass spectrometric analysis was carried out using the LCQ Deca XP mass spectrometer of Thermo Finnigan (San Jose, CA, USA) in the ESI (+) (electrospray ionization) mode. The peptide solution was introduced into the ESI source by direct injection using a Thermo syringe pump at a flow rate of 0. 5–1 µL min −1. Spray voltage was maintained at ≈5. 0 kV, the capillary voltage was about 15 V, and the capillary temperature was set at 250 °C, N 2 was used as a nebulizing gas. Secondary structure analysis was performed by using GOR IV online tools ( https://npsa‐prabi. ibcp. fr/cgi‐bin/npsa_automat. pl?page = npsa_gor4. html ). And hydrophobicity/hydrophilicity analysis was performed by using ProtScale online tools ( http://web. expasy. org/protscale/ ) based on Kyte & Doolittle. Preparation and Characterization of Self‐Assembly Peptide Nanoparticles : The drug encapsulation experiments were performed as follows: DOC and the appropriate bifunctional peptide were dissolved in 400 µL of hexafluoro‐2‐propanol (HFIP) in 5 mL glass vials and sonicated for 1 min to mix, before the HFIP was removed by rotary evaporation. The vials were then left to stand in the fume hood for at least 8 h at room temperature with the cap removed to allow any trace amount of HFIP to evaporate. Then 200 µL of 1 × DPBS was added to each vial and vortexed for 30 s. The solutions were aged for 8 h and then centrifuged (2000 g, 5 min) to remove any precipitated DOC, and the supernatant was carefully collected and analyzed using HPLC with the following conditions: Varian ProStar model 325 HPLC (Agilent Technologies, Santa Clara, CA, USA) equipped with an Agilent Zorbax‐C18 column (5 µm, 4. 6 × 150 mm); the flow rate was 1 mL min −1, with the mobile phase held at 35% A (MeCN with 0. 1%TFA) and 65% B (0. 1%TFA aqueous solution) for 5 min and start to gradient to 70% A at a period of 25 min, then gradient back to the initial conditions in 1 min and held for 4 min; the monitored wavelength was 237 nm. The drug EE was calculated as the percentage of DOC recovered from the supernatant to the DOC added. The DL capacity was calculated from the percentage of recovered DOC to the sum of recovered DOC and added bifunctional peptide. The particle size distribution, PDI, and zeta potential of prepared DOC/peptide NPs were determined by Malvern Nano‐ZS 90 laser particle size analyzer at 25 °C. All results were the mean of three test runs, and all data were expressed as the mean ± standard deviation (SD). The morphological characteristics of the self‐assembly peptide NPs were examined by transmission electron microscope (TEM, H‐6009IV, Hitachi, Japan). The NPs were diluted with distilled water and placed on a copper grid covered with nitrocellulose. Samples were negatively stained with phosphotungstic acid and dried at room temperature. The 3D Structures of DOC/Peptide from Computer Simulations : At first, the structure of the DOC molecule was built by using Marvin Sketch ( http://www. chemaxon. com ) and optimized at a molecular mechanical level using the MMFF94 method. 52 Then, it was further optimized at semi‐empirical level using the AM1 method with the Fletcher‐Reeves algorithm by employing Hyperchem software (HyperChem Professional 8. 0, Hypercube, Inc. , Gainesville, FL, USA). In order to understand in detail the interaction between the HRK‐19 peptide and DOC, 200 ps molecular dynamics simulation was performed on complex composed DOC and two molecules of the peptide. In the process of simulation, CHARMM27 was chosen as the force field and solvation effect was considered implicitly. 32 Cell Culture, Cellular Uptake, and Cell Apoptosis Assay : The human NSCLC cell line A549 were purchased from American Type Culture Collection and cultured with RPMI‐1640 (GIBCO, US), supplemented with 10% (V/V) FBS (GIBCO, US), 100 unit mL −1 penicillin, and 100 µg mL −1 streptomycin. Cellular uptake of C6 as a model drug loaded in self‐assembly peptide NPs was measured by confocal microscopy and flow cytometry analysis. A549 cells at log phase were seeded onto a 24‐well plate at 1 × 10 5 cells per well and cultured in 1 mL of medium. After 24 h, the media was removed, and cells were exposed to serum‐free medium containing free C6, or C6/peptide NPs at a final concentration of 10 µg mL −1. After incubation for 0. 5, 2, and 4 h, the media were removed and carefully washed with PBS. For observation by confocal microscopy, cells were fixed with cold acetone, washed again with PBS, stained with DAPI, and imaged using a fluorescence microscope (×400) (Leica DM2500, Germany). For the expression and localization of α‐tubulin, cells were incubated with the antibodies against α‐tubulin (Abcam, USA) followed by incubation with TRITC‐conjugated secondary antibodies. The images were obtained using a confocal microscope (DM6000 CS, Leica, Germany). To investigate the apoptotic effect of the DOC/peptide NPs, flow cytometry was performed using an ESP Elite flow cytometer (BeckmaneCoulter, Miami, FL, USA) with FITC‐conjugated Annexin V/propidium iodide (PI, BD PharMingen) staining as per the manufacturer's instructions. A549 cells cultured in 6‐well plates were treated with DOC/peptide NPs, free DOC, and blank peptide. Medium without treatment reagents was added as control. Both early apoptotic (Annexin V‐positive, PI‐negative) and late apoptotic (Annexin V‐positive and PI‐positive) cells were measured. In Vivo Tumor Targeting and Penetrating Efficiencies : For the in vivo animal experiments, tumor‐bearing female nude BALB/c mice were injected intravenously with free DiD or an equivalent amount of DiD (100 µg kg −1 ) loaded in the drug/peptide NPs. After i. v. injection of the drug, optical fluorescence imaging was performed by positioning each mouse on an animal plate in the vivo imaging system (Quick View 3000) with excitation and emission wavelengths of 645 and 715 nm, respectively. The images were obtained 1, 4, 8, and 24 h after injection of the test drugs. For the in vivo cellular uptake studies, BALB/c nude mice bearing subcutaneous tumors from A549 cells were injected with free DiD and DiD/peptide NPs. After 1 and 4 h, the tumors were isolated and embedded in optimal cutting temperature compound, and then cut into 5 µm slices. Sections were stained by DAPI and observed. In Vivo Tumor Model and Treat Plant : Antitumor activity of DOC/peptide NPs was investigated in pulmonary metastatic A549 model and subcutaneous A549 model. Eight‐week‐old female BALB/c mice nude were obtained from the Animal Center Laboratory of Beijing HFK Bioscience Co. , Ltd. In subcutaneous A549 model, the mice were injected subcutaneously with 100 µL of an A549 cell suspension (1 × 10 7 ) into the right flank. After the tumor mean diameter reached ≈6 mm, the tumor‐bearing mice were randomly assigned to four groups that received NS (control), blank peptide, commercial formulation Taxotere (Sanofi, Paris, France) (5 mg kg −1 body weight), or DOC/peptide NPs (5 mg kg −1 body weight) by intravenous injection into the tail three times for 3 weeks. When the mice in the control group began to die, all mice were sacrificed by cervical vertebra dislocation, and the tumors were immediately harvested and measured. To further investigate the antitumor activity of DOC/peptide NPs in subcutaneous A549 model, survival times of the mice were observed (10 mice per group). In pulmonary metastatic A549 model, BALB/c nude mice were injected intravenously with 100 mL of cell suspension containing 2 × 10 5 A549 cells on day 0. On day 5, tumor bearing mice were assigned randomly into four groups (12 mice per group). Mice were injected intravenously every 3 d for 2 weeks with 100 µL of NS (control), blank peptide, commercial formulation Taxotere (Sanofi, Paris, France) (5 mg kg −1 body weight), or DOC/peptide NPs (5 mg kg −1 body weight), respectively. For tumor growth inhibition assay (6 mice per group), mice were scarified by cervical verte bradislocation on day 25. On day 26, mice in NS group began to die. Lungs in each group were weighted, and tumor nodules in each lung were numbered. To further study the antitumor activity of DOC/peptide in pulmonary metastatic A549 model, survival times of the mice were observed (6 mice per group). Immunohistochemistry : Tumor tissue sections were prepared as described above for Ki‐67 staining using the labeled streptavidin‐biotin method. 53 The primary antibody and secondary antibody were rat antimouse monoclonal antibody Ki‐67 (Gene Tech) and biotinylated goat antirat immunoglobulin (BD Biosciences Pharmingen), respectively. To quantify Ki‐67 expression, the Ki‐67 labeling index (Ki‐67 LI) was calculated as number of Ki‐67 positive cells/total number of cells counted under×400 magnification in five randomly selected areas in each tumor sample by two independent investigators in a blinded fashion. For the CD31 assay, the tumors were stored at −80 °C to examine microvessel expression, then frozen sections of tumors were cut at 8–10 µm thickness using a cryostat (Leica Microsystems, CH), fixed in acetone, and washed with PBS. After permeabilization (Triton X‐100 (Sigma‐Aldrich, DE) 0. 1% (v/v) in PBS) and blocking (5% (w/v) bovine serum albumin (BSA) in PBS), the primary antibody (rat anti‐CD‐31 (1:50), BD Pharmingen, USA) was applied for 24 h at 4 °C, and followed by incubation with an FITC‐conjugated second antibody (Abcam, USA). Finally, sections were incubated with DAPI (Invitrogen, BE) (50 ng mL −1 ) for 5 min to visualize the cell nuclei. MVD was calculated as the average number of small CD31‐positive vessels in a high‐power (×400) field using a fluorescence microscope (×400) (Leica DM2500, Germany). The immunofluorescent analysis of N‐cadherin was evaluated as above. Briefly, the primary antibody (monoclonal antihuman N‐cadherin (1:200), Abcam, USA) was applied for 1 h at 37 °C, and followed by incubation with Alexa Fluor 647‐conjugated second antibody (Abcam, USA). Finally, sections were incubated with DAPI (Invitrogen, BE) (50 ng mL −1 ) for 5 min to visualize the cell nuclei. Statistical Analysis : Statistical analyses were performed using SPSS for Windows version 15. 0 (IBM Corporation, Armonk, NY, USA). Data were presented as means ± SD/SEM. Statistical analysis was performed using two‐tailed Student's t ‐test, and P ‐value <0. 05 was considered as significant difference. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file.
|
10. 1002/advs. 201600410
| 2,017
|
Advanced Science
|
Bioinspired Mechano‐Sensitive Macroporous Ceramic Sponge for Logical Drug and Cell Delivery
|
On‐demand, ultrahigh precision delivery of molecules and cells assisted by scaffold is a pivotal theme in the field of controlled release, but it remains extremely challenging for ceramic‐based macroporous scaffolds that are prevalently used in regenerative medicine. Sea sponges (Phylum Porifera), whose bodies possess hierarchical pores or channels and organic/inorganic composite structures, can delicately control water intake/circulation and therefore achieve high precision mass transportation of food, oxygen, and wastes. Inspired by leuconoid sponge, in this study, the authors design and fabricate a biomimetic macroporous ceramic composite sponge (CCS) for high precision logic delivery of molecules and cells regulated by mechanical stimulus. The CCS reveals unique on‐demand AND logic release behaviors in response to dual‐gates of moisture and pressure (or strain) and, more importantly, 1 cm 3 volume of CCS achieves unprecedentedly delivery precision of ≈100 ng per cycle for hydrophobic or hydrophilic molecules and ≈1400 cells per cycle for fibroblasts, respectively.
|
On‐demand delivery of molecules and cells is a pivotal theme in the field of controlled release. While most on‐demand delivery systems developed so far have placed much emphasis on achieving the release of cargos at controllable rates or by various release gates, some recently reported systems have achieved ultrahigh precision of release amount at nanogram regime. 1, 2, 3 For biomedical applications like those in bone tissue engineering, ultrahigh precision release is highly desirable for the delivery of therapeutic cargos (such as bone morphogenetic proteins, parathyroid hormone, miRNA, etc. ), for which accurate administration is imperative and the consequence of overdose is serious. Although porous scaffolds are highly efficient carriers for loading and transportation of biological entities, such as therapeutic molecules and cells, 4, 5, 6, 7 it is very challenging to achieve highly precise, on‐command delivery of molecules and cells by porous scaffolds since they usually rely on passive release mechanisms (molecular diffusion, scaffold degradation, etc. ). 8, 9, 10, 11 Recent advances in porous scaffolds for high‐precision controlled release focus on developing active delivery systems that, to a large extent, achieve the delivery of biological cargos in response to external stimuli, such as temperature, pH, and enzyme. 12, 13, 14, 15, 16 Owing to their responsive abilities to external stimuli, the active release systems possess considerable flexibility and versatility on the kinetics and amount of release. Since most of the aforementioned active delivery systems are designed to operate under static conditions, their efficacy has a high likelihood to be compromised if implanted in the mechanically dynamic environment of target tissues in vivo. For mechanically different scenarios, such as trabecular bone (compressive modulus E c > 50 MPa), cartilage ( E c = 0. 1–2 MPa), and cardiac tissue ( E c = 0. 01–0. 02 MPa), 17 the ability of delivery systems actively responding to such mechanically dynamic environment is highly desirable. Starting from the exemplary work by Mooney and co‐workers, mechanoresponsive materials triggered by mechanical stimuli (such as compression, tension, and shear) have demonstrated promising capacity of delivering cargos in a controlled and active manner. 18 Nevertheless, there are only a few studies of active porous systems relying on mechanical stimuli for delivery. Due to soaring demands for drug or cell delivery in force‐related biological systems like musculoskeletal and circulation systems, the new strategy of mechano‐sensitive delivery systems, which harness mechanical energy from host tissues to deliver wanted molecules and cells, have emerged. 19 Recent examples include Fe 3 O 4 /alginate composite scaffolds delivering cells and drugs in response to variable magnetic forces 2 and bacterial mechano‐sensitive channels acting like a dual‐control gate by cell membrane tension and MscL charges to modulate the delivery of bioactive molecules into live cells. 20 These strategies, however, are not applicable in achieving ultrahigh release precision in ceramic‐based porous scaffolds due to the high rigidity, limited biodegradability, and large interconnected pores of ceramic scaffolds. Furthermore, delicate on‐demand release behaviors (including gated release, repeated release, or logic release) have yet to be realized in porous ceramic scaffolds possessing macropores (e. g. , pore size >500 µm) and high porosity (e. g. , >80%). Inspiration from sea sponges (Phylum Porifera), which delicately masters water circulation through their bodies for food, oxygen, and waste exchanges, 21 provides novel insights into the design and construct of mechano‐active porous ceramic structures for ultrahigh precision delivery. Leuconoid sponges have the most complex body structures among all types of sea sponge, consisting of a hierarchical channel system where spherical choanocyte chambers are connected by incurrent canals and apopyles (water exiting pores). 22 The choanocytes, which are unique collar cells with flagella lining the internal chamber, take in water or expel it while capturing tiny food particles (nutrients, bacterial cells, etc. ) in the interstices between choanocytes ( Figure 1 a). 23 Besides cells, the sponge has a skeleton composed of mesohyl (mainly collagen) and are structurally reinforced by calcite or silica spicules and/or spongin fibers, 24 resembling essentially a ceramic‐polymer composite. This resilient composite, together with the hierarchical porous structure, renders the highly efficient water flow system for accurate transportation and exchange of materials from the nanometer scale to tens of micrometer (e. g. , food, nutrient, waste particles, and gas molecules). Figure 1 Leuconoid sponge‐inspired CCS with a hierarchically porous structure. a) Hierarchically porous structure of leuconoid sponge, 1‐choanocyte chamber, 2‐apopyle, 3‐canal and 4‐interstice. Arrow heads in the SEM image on the right indicate food particles and bacterial cells. b) Schematic diagram of the fabrication of CCS and molecule‐ or cell‐loaded CCS by a novel in situ foaming method. c) Photograph of CCS after being cut into different geometries (white) and BPB‐loaded CCS (blue). d) Three‐hierarchy porosity of CCS: spherical major pores in CCS, e) interconnected pores connecting adjacent major pores, and f) micrometer‐sized minor pores on the walls of major pores. g) Length scales of the hierarchical pores in leuconoid sponge and CCS. SEM image in (a) is reproduced with permission. 23 Copyright 2006, The Marine Biological Laboratory. Here, we report a mechanically regulated ultrahigh precision logic delivery system realized by bioinspired macroporous ceramic composite sponges (CCS). The CCS imitates the body plan and skeletal composition of leuconoid sponge to possess a three‐level hierarchy of porosity and high resilience, turning conventional rigid porous ceramics into mechano‐active sponges capable of programmable and repeatable deliveries of molecules and cells. We selected synthetic biomineral hydroxyapatite (HA) and natural cornstarch as CCS components and developed a novel foaming method to construct hierarchical structures, mimicking the leuconoid sponge. The resultant CCS is highly resilient and exhibits unique gate‐logic release behaviors with ultrahigh precision in response to mechanical stimuli. The mechanically regulated AND logic delivery system by a 1 cm 3 CCS reaches an unprecedentedly delivery precision of ≈100 ng per cycle for hydrophobic or hydrophilic molecules and ≈1400 cells per cycle for fibroblasts. In addition, the compressive modulus of CCS changes significantly at different moisture contents, granting it flexibility to comply with the mechanical environment of host tissues or extracellular matrix supporting regular physiology of different types of cells. 25, 26, 27 The adjustable mechanical properties, logic controlled delivery behavior, and ultrahigh release precision render CCS an effective and mechanically smart scaffold for the applications in controlling cell fate, 28 repairing tissue damage, 29 and regenerating tissues. 30 We developed a low‐temperature foaming technique that efficiently fabricates the sea sponge‐inspired hierarchical pore structures in ceramic‐based scaffold with feasibility to load molecules and cells in situ (Figure 1 b and Figure S1, Supporting Information). In order to imitate the composite nature of sea sponge's skeleton and aim at tissue regeneration applications, bioactive HA and gelatinized cornstarch were selected as bioceramic and organic phases, respectively. The cornstarch also acts as a foaming agent for the stabilization of the porous structure. This foaming technique is applicable at relatively low temperatures ranged from ambient temperature to 90 °C, depending on the gelatinization temperatures of starch and the requirement of cargo (i. e. , drug, protein, cell, etc. ). The ceramic–starch foam can be readily cut or machined into any shape and form robust self‐standing CCS after air‐drying or freeze‐drying (Figure 1 c). Dehydrated CCS generally has a porosity >85% (Figures S2 and S3, Supporting Information) and comprises of a three‐hierarchy pore structure with pore sizes varied from the millimeter to micrometer range (Figure 1 d–f), which highly mimics the porous structure of the leuconoid sponge. The first hierarchy is comprised of spherical major pores with a narrow size distribution from ≈500 µm to ≈1 mm (Figure 1 d), matching the dimension of the leucon's choanocyte chamber (Figure 1 g). The sizes of major pores in this range are also desirable for bone ingrowth and regeneration. 31, 32 The second hierarchy consists of interconnected pores with mainly open circular windows of 50–100 µm in diameter on the walls of the major pores (Figure 1 e), similar to the sizes of incurrent canals and apopyles in leucon. The third hierarchy corresponds to a large amount of micrometer‐sized pores (<10 µm) uniformly residing on the walls of the major spherical pores (Figure 1 f). Such minor pores resemble the interstices between choanocytes and, like the sponges using such interstices to trap food particles and bacterial cells, can serve as reservoirs for loading molecules and cells (Figure 1 g). Due to the structural and phase combination of hydrophilic HA particles and water‐absorbable starch, CCS exhibits moisture‐dependent mechanical properties that are attractive for on‐demand delivery and regenerative medicine. CCS with various moisture contents shows a viscoelastic stress–strain response ( Figure 2 a). Most notably, air‐dried CCS with 83 wt% HA content and 87% porosity (moisture content <5%) is found to have remarkable compressive strength (σ c = 1. 20 ± 0. 01 MPa) and modulus ( E c = 57. 1 ± 1. 3 MPa), which are much greater than the mechanical properties reported in porous systems of poly( d, Llactide‐co‐glycolide)/nano HA ( E c = 4. 56 ± 0. 3 MPa), 33 poly( l ‐lactic acid)/HA ( E c = 10. 87 ± 3. 20 MPa), 34 and HA coated with polycaprolactone (σ c = 0. 57 ± 0. 09 MPa). 35 The compressive strength of CCS with low moisture content (e. g. , <10%) is also comparable to strengths of demineralized and deproteinized trabecular bovine femur (DMB and DPB, Figure 2 a), 36 which are clinically used for bone substitution and regeneration. Figure 2 Moisture‐dependent mechanical properties of CCS. a) Compressive stress–strain relations of CCS (ceramic content = 83 wt%) with different moisture contents (<5, 10, and 35 wt%), comparing to that of demineralized trabecular bovine femur (DMB) and deproteinized trabecular bovine femur (DPB). 36 b) Moisture‐dependent compressive modulus ( E c ) of CCS in comparison to that of different tissues. c) Compression–decompression loop of CCS with 45% moisture content subjected to different strains. d) Resilience versus recovery time relation of CCS with 45% moisture content subjected to different compressive cycles and strains (Resilience = [( H − h )/ H ] × 100%). e) SEM images showing the microstructure of CCS (ceramic content = 86 wt%) before and after 150 compression cycles (max. strain = 10%). The red lines outline pore morphology and the numbers indicate large ceramic particles, both of which remain unchanged after compression. The high mechanical strengths of CCS are probably attributed to the reinforcing effect of starch molecules on ceramic particles. A close‐up examination of CCS pore walls reveals the structural interlocks between ceramic particles and starch molecules (Figure 1 f). Gelatinized starch molecules form a 3D network grasping and binding HA particles that consist of ≈80 wt% of CCS, which is a mimicry of sea sponge's skeleton of mesohyl matrix embedded with mineral spicules and/or spongin fibers. The interlocking effects between ceramic particles and starch networks result in a strong cohesion between starch molecules and neighboring ceramic particles via a “bridging mechanism” (arrows in Figure 1 f), ensuring the high robustness of CCS even at high ceramic content and porosity up to 86 wt% and 88%, respectively (Figure S3, Supporting Information). Interestingly, the same CCS can be sintered at 1250 °C for 2 h to obtain a porous ceramic scaffold with a porosity of 75% but its compressive strength and modulus drop to 0. 18 and 11. 6 MPa, respectively (Figure S3, Supporting Information). This decrease indicates that gelatinized starch in CCS actually reinforces the porous structure of HA rather than deteriorates its mechanical properties, agreeing to what has reported in the non‐porous HA/starch system. 37 Static mechanical properties of CCS can be adjusted dramatically as the moisture in the sponge changes (Figure 2 a, b) and its compressive modulus decreases exponentially to the range of kilopascal when the moisture increases to 35 wt%. This adjustability enables CCS to match the compliance of target tissues, which has been a persistent drawback of ceramic‐based 3D scaffolds in the applications of controlled release and tissue regeneration. 38 CCS can match the compliances of trabecular bone ( E c > 50 MPa), cartilage ( E c = 0. 1–2 MPa), and heart tissue ( E c = 0. 01–0. 02 MPa) at moisture contents of 3, 15–35, and 45 wt%, respectively (Figure 2 b). Given that internal moisture (water) contents in trabecular bone, cartilage, and heart muscle are ≈10, 60–80, and 70 wt%, respectively, 17 CCS implanted to these target tissues are expected to equalize its moisture content with surrounding tissues and eventually render its stiffness compatible with that of the host tissue. More importantly, like a real sea sponge, CCS with elevated moisture contents (usually >10 wt%) becomes remarkably resilient at different amounts of deformation. Figure 2 c demonstrates the recoverable loading–unloading loops of CCS measured at the 11th compression cycle of different strains (i. e. , CCS experienced ten compression cycles before the measurement), revealing a reversible elasticity at 1% strain while maintaining a full recover ability up to 10% strain. The compliance of CCS, however, is preserved at different compressive strains (≈60 kPa in Figure 2 c) and is much greater than the well‐known hydrogels 39 or ferrogel 40 reported for drug or cell delivery. Resilience of CCS remains at 99. 5% and 98% after 300 cycles of compression at 3% and 10% strains, respectively (Figure 2 d). The remarkable resilience of CCS is also reflected by its ultrafast recovery when the compression is removed (Figure 2 d and Movie S1, Supporting Information), taking only a few hundred milliseconds to recover even after 300 compression cycles of 10% strain. Direct observation before and after 150 compression cycles suggests CCS preserves its porous structure, with positions of pore ceramic particles remaining identical (Figure 2 e), demonstrating a high robustness and stability for mechanically modulated applications. The great resilience and adjustable compliance of CCS are predominantly rooted in high viscoelasticity and pliability of the gelatinized starch chain network which can hydrate or dehydrate reversibly. Abundant hydroxyl groups in both HA and starch can form a large quantity of hydrogen bonds in the presence of water, which plasticizes the starch molecules to achieve a substantial interfacial stability between ceramic and starch network, restricting the movement and rotation of ceramic particles and eventually maintaining the integrity of CCS during cyclic deformation. Similar to sea sponge's delicate transportation of multi‐scale food, gas, and waste particles in its hierarchical body system, CCS also allows accurate releases of various cargos from molecules to proteins to cells by mechanical modulation. Mechanically modulated cargo release by CCS exhibits a binary‐stage pattern in deionized water (DI water) (equivalent to ≈66 wt% moisture content). Model molecules of bromophenol blue (BPB) release only 0. 27–0. 35 µg cm −3 from CCS at compressive strains <2% for 20 consecutive compression cycles, but the release rapidly quintuples to and saturates at 1. 47–1. 58 µg cm −3 when the strain rises above 3% ( Figure 3 a and Movie S2, Supporting Information). This binary‐stage release resembles a mechanically modulated digital gate whose low and high release states correspond to 0 and 1, respectively, with a threshold at 3% compressive strain. Figure 3 Mechano‐active CCS for AND logic release. a) Cumulative release of BPB from CCS at different strains. CCS has a moisture content of 66% and is subjected to ten compression cycles at each strain. b) Simplified cylindrical pore model for the analysis of water expelled from pores in CCS. Before compression, the pore has a circular cross section with a diameter of 2 r and the hydrophilic liquid has a contact angle of θ. After compression, the deformed cylindrical pore has an elliptical cross section with two axes of (a) and (b). c) Estimated counter pressure (∆ P, blue triangles) and compressive strains (ε repel, red circles) as a function of pore size. d) Cumulative release of BPB from CCS at different moisture contents. CCS is subjected to ten compression cycles at a pressure of 0. 5 kPa. e) 3D diagram of AND logic release in CCS, where the vertical axis is the cumulative release of BPB and the horizontal axes are moisture content and pressure, respectively. CCS is subjected to ten compression cycles at each point. f) Logic map showing True (1) and False (0) relations between dual inputs of A) pressure/strain and B) moisture content, and C) output of release. Red zone represents C = 0 (False) and green zone represents C = 1 (True). Dot lines represent the thresholds of logic gates A and B. The hierarchical pore structure of CCS allows liquids to flow into the interior of the sponge and reach the large surface areas of pore walls where cargos (e. g. , drugs or cells) are preserved. The high resilience of CCS, on the other hand, allows the formation of strong water convection and shear force in the structure when compressed, 2 assisting the release of cargos. Theoretically, the cargo release capability enabled by fluid convection should be a function of the counter pressure applied to squeeze the fluid out of the pores against capillary pressure. Implied by the Young‐Laplace equation for a simplified, cylindrical pore structure (Figure 3 b), this counter pressure (Δ P ) is inversely proportional to the pore size for a specific liquid and was estimated to be on the order of kilopascal for water in the pores with sizes from 10–1000 µm (Figure 3 c; see the Supporting Information for calculation). The counter pressure presumably originates from the shear of pores under external compression and the strain required for completely repelling the water (ε repel ) can be estimated (Figure 3 c; see the Supporting Information for calculation). Assuming average sizes of 50–100 µm for the interconnect pores in CCS (also canals and apopyles in sea sponge), the minimum of ε repel is calculated to be 2. 9%, which agrees with the threshold of 3% in the strain gate observed in experiment (Figure 3 a). In addition, theoretical predictions of ε repel (0. 2%–17%) in a range of pore sizes are close to the values used in the experiments (3%–20%) to achieve satisfactory delivery of drugs or cells. Besides the pressure/strain gate, CCS has another binary gate of moisture which restricts the release down to <0. 25 µg cm −3 of CCS (20 cycles cumulative) at water contents <40 wt% while allowing up to a ten‐fold increase at water contents >45 wt% (Figure 3 d and Movie S3, Supporting Information). The binary gates of strain (or pressure) and moisture together construct an AND logic gate, by which the cargo is only released when true inputs from both binary gates are met (Figure 3 e). The logic map of dual‐controlled release is shown in Figure 3 f, revealing all AND logic conditions for CCS to deliver or retain cargos, which, interestingly, also reveals an interrelated true‐false boundary between strain (or pressure) and moisture gates. For example, the true input of the pressure gate required less than half of the pressure (1 kPa vs. 0. 5 kPa) when the moisture threshold increased from 26 to ≈60 wt%. Based on this logic map, cargos released by CCS can be accurately controlled by selecting appropriate strain (pressure) and water content. Therefore, at stable water contents, the amount and intermittence of release can be readily tailored by the strain/pressure, realizing high‐precision temporal, spatial and even programmable controls on drug or cell delivery. Ultrahigh precision logic delivery of molecules and cells are achieved by CCS under mechanical stimuli. CCS loaded with water‐insoluble BPB (payload 700 µg cm −3 ) and subjected to compression (state 1) and decompression (state 0) cycles (“on/off” status) of 3% strain exhibits linear cumulative release proportional to the cycle numbers ( Figure 4 a). More importantly, the release amount modulated by 3% compressive strain is highly repeatable in every cycle and reaches a remarkable precision of ≈70 ng per cycle cm −3 of CCS (Figure 4 b). Meanwhile, background release without mechanical stimuli (“off” status) is less than 10 ng per cycle cm −3. Besides hydrophobic molecules, ultrahigh precision is maintained in water‐soluble hydrophilic molecules of bovine serum albumin (BSA), reaching a precision of 150–200 ng per cycle per cm 3 CCS (Table S1, Supporting Information). To our best knowledge, this on‐demand, nanogram precision delivery of molecules by porous systems is reported for the first time and the release precision is much higher compared to other macroporous system reported recently (Table S2, Supporting Information). 41, 42, 43 This mechanically induced ultrahigh release precision of ≈10 ng mL −1 for proteins is more superior than an electrically controlled release by a hydrogel system reported recently, which releases lysozyme protein at a precision of 2000 ng mL −1 per cycle for a hydrogel layer much less than 1 cm 3 (≈1. 4 mm 3 ). 44 Figure 4 Mechano‐active CCS for cyclic delivery of molecules and cells. a) Cumulative release of BPB from CCS subjected to 300 consecutive compression cycles (blue boxes) and the release from CCS without compression (red circles, tested at the same time periods). The second y ‐axis denotes the percentage of released BPB compared to the total amount loaded in CCS. b) The average amount of BPB per cycle released from a unit volume of CCS during compression (“ON”, blue boxes) and the release from CCS without compression (“OFF”, red circles). CCS is tested for six consecutive sets and each set contains 50 compression cycles. Compressive strain and moisture content are 3% and 66%, respectively, for all the CCS tested. c) Cumulative number of cells released from CCS subjected to 80 consecutive compression cycles. The second y ‐axis denotes the percentage of released cells compared to the total amount loaded in CCS. d) The average number of cells per cycle released from a unit volume of CCS. CCS is tested for four consecutive sets and each set contains 20 compression cycles. Compressive strain and moisture content are 20% and 66%, respectively, for all the CCS tested. e) Fluorescence images of cells released from CCS I) without compression, II) after the first and III) the fourth compression sets, and IV) viable cells remaining in the CCS after 80 consecutive compression cycles. In the fluorescence images, live cells are stained green and dead cells are stained red. CCS, infiltrated with fibroblasts (payload 5 × 10 5 cells cm −3 ) and allowed cell adhesion for 4 h, demonstrates a similar linear cumulative release pattern to the BPB release and a repeatable on‐off release for the compression cycles of 20% strain (Figure 4 c, d). The precision of cell release per cycle reaches 1400 cells per cm 3 CCS, tested in 10 mL of culture medium (Figure 4 d), while its background release without compression is almost negligible (Figure 4 e). In addition, the released fibroblasts retain high viability after tens of cycles of 20% strain, even after the CCS is completely smashed (Figure 4 e). It is worth mentioning that the mechanically modulated, ultrahigh release precision of 1400 cells mL −1 per cycle by 1 cm 3 CCS outperforms that of any on‐demand delivery systems so far. 45, 46 In contrast to conventional strategy, 47, 48 we did not use additional linkers or adhesive proteins to tether the hydrophobic BPB, hydrophilic albumin, or cells. Retention of the cargos in CCS probably relies on weak bonds like hydrogen bonds and intermolecular forces since both HA and starch molecules have an abundant amount of surface hydroxyl groups. The above results indicate that the weakly bonded cargos can be readily released when the gates of moisture and strain are open and vice versa. For loading highly water‐soluble entities or special cargos in CCS, additional linkage is certainly useful and can be designed depending on specific applications. In summary, on‐demand ultrahigh precision logic delivery is achieved in mechanically modulated CCS that has a three‐level hierarchy of porosity and resilience inspired by leuconoid sponge. The capacity of logic delivery of molecules and cells with ultrahigh precision and repeatability makes CCS a truly on‐demand delivery system that enables temporal, spatial, and quantitative controls of cargos via mechanical stimuli. CCS is also a versatile platform that can mechanically adapt to the compliance of host tissue or environment by adjusting its moisture content. Such bioinspired mechano‐active ceramic sponges are thus promising for delicate programmable and gate controlled delivery, opening a wide range of applications from environmental to biomedical applications. 49, 50, 51, 52, 53, 54 For instance, CCS is expected to be an ideal tissue engineering scaffold for bones, cartilage, and muscle repair, where patient's movement acts as external stimuli to actively regulate the highly precise release kinetics of drugs (e. g. , bone morphological proteins, antibiotics, anti‐osteoporosis drugs, etc. ) or stem cells. CCS may also rely on intrinsic mechanical effects, such as hemokinesis, heartbeat, and peristalsis to achieve precise temporal control of drugs or cell releases. Additional applications of CCS also include logic sensors and programmable delivery devices for electrical, energy, and environmental engineering. Experimental Section Materials : Food‐grade cornstarch was purchased from Weimeisi Co. , Ltd (Shanghai, China). HA was synthesized in house according to a method described elsewhere. 55 Triton X‐100 was purchased from Sinopharm Chemical Reagent Co. , Ltd. (Shanghai, China). BPB was procured from Sigma (St. Louis, MO). Fabrication of CCS : Cornstarch and DI water was homogeneously mixed in a beaker to obtain 10 wt% starch suspension. Different amounts of HA powders were then added into the starch suspension and mixed uniformly to form ceramic slurries. For different purposes, the solid contents of HA (the ratio of HA mass to total solid mass) varied from 71. 4% to 85. 9%. The slurry was later heated to 90 °C in a water bath and a surfactant of Triton X‐100 was added. The heated suspension became a highly viscous gel and then was vigorously stirred by an overhead stirrer till air bubbles fully infiltrated into the suspension to obtain foam. The foam was removed from the water bath, cooled down to room temperature, and then set for 24 h. The foam was further dried to porous composite scaffolds with desired moisture contents, depending on different applications. Porous HA ceramics were sintered from dried CCS with 80% ceramic content. CCSs with the dimension of 15 × 15 × 15 mm 3 were sintered at 1250 °C using a muffle furnace (SJKQ‐1700, Dingan Tec, Suzhou) for 2 h. Preparation of BPB‐ and BSA‐Loaded CCS : 8 g starch, different amounts of hydroxyapatite and appropriate volume of surfactant were mixed with 72 mL 0. 1 wt% BPB solution in a 200 mL beaker to obtain the uniform slurry, and then the slurry was heated to 90 °C in a water bath to allow starch to gelatinize. After that the slurry was foamed and set for 24 h to get the stable porous network. For BSA loading, the BPB solution was replaced by DI water. After foaming at 90 °C, the beaker was placed at room temperature until the temperature of the slurry dropped to 50 °C, 3. 5 g of BSA (Sigma, St. Louis, MO) was added in the beaker and the slurry was foamed. The foam was set for 24 h to get the stable porous network. BPB‐ and BSA‐loaded CCS could be cut to different shapes and dried by dehydration (dried by gradient of ethanol solution) or freeze drying (−54 °C) depending on different applications. Characterization of CCS : Microstructure and morphology of CCS were characterized by scanning electron microscopy (SEM, FEI Quanta 250, acceleration voltage of 1. 5 kV under a vacuum of 1. 56 × 10 −4 Pa). The sample was fractured and the facture surface for observation was sputtering‐coated with Au‐Pd. Dried CCSs with the dimension of 10 × 10 × 10 mm were scanned at the speed of 0. 7° s −1 by Micro‐Computed Tomography (Micro‐CT, Skyscan1176) with the precision of 8. 8 µm, and then a 3D microstructure of CCSs was obtained by reconstructions of micro‐CT data. Apparent density (ρ app ) of CCS and porous HA ceramic was calculated by its weight and volume that can be directly measured. Theoretical density (ρ th ) of the porous composites was calculated by the densities of the HA and starch according to their proportions in the composites. The porosity ( p ) of CCS was thus calculated by the equation (1) p = 1 − ρ app ρ th × 100 % Mechanical Characterization : For uniaxial compression tests, CCS were cut into cubes with dimension of 10 × 10 × 10 mm 3 and tested on a mechanical tester (HY‐1080, testing range 0–500 N with precision of 0. 01 N, Hengyi company, Shanghai) operating at a crosshead speed of 1 mm min −1. From the stress–strain curve, the maximum stress before failure was determined as compressive strength and the linear range in the stress curve before failure was used to calculate compressive modulus. For the testing of the compressive loops of CCS, CCSs with 45% moisture content were cut into cubes with dimension of 10 × 10 × 10 mm 3 and the cubes were pressed to a desired strain of CCS (1%, 3%, 5%, and 10%) and then unloaded to their initial position. Resilience was characterized using a CCS sample with 45% water contents using a microscope. A CCS cube was pressed to designated strains (3% and 10%), and released, and this process was video‐recorded by a microscope. The resilience was then calculated by comparing the position of the rebounding surface to its initial position. BPB Release Behavior of CCS : Release behavior of CCS loaded with BPB was studied by compressing CCS cubes (10 × 10 × 10 mm 3 ) at different strains (0. 1%, 0. 2%, 0. 5%, 0. 8%, 1%, 2%, 3%, 5%, 6%, and 8%) in a series of ethanol/water solution (water contents 0, 20%, 40%, 60%, 80%, and 100%). After pressing for twenty times, the solution containing released BPB was retrieved and spectrophotometrically measured on a microplate reader (at 450 nm on BioTek MQX200R). Optical densities (OD) of the solution were compared to a standard OD curve of solution containing known concentrations of BPB (see Figure S4 and Supporting Information for more information). Meanwhile, a CCS cube that was placed in the same ethanol/water solution but not pressed was used as a control to calculate the BPB released from CCS without mechanical stimulation. For cyclic release tests, CCS was compressed at designated strains for 50 consecutive cycles (defined as one set of compression) and the solution containing released BPB was removed for spectrophotometical measurement. Then test resumed for another set and repeated up to six sets in total. A parallel test of the CCS without compression was used to determine the background release. The amount of BPB released in each set (background release was subtracted) was averaged by 50 cycles and then the total volume of CCS to obtain the net release amount per cycle per cm 3 of CCS. All the releases tests were repeated at least three times. Calculation of the Amount of BPB Released from CCS : 100 µL of BPB solution with different concentration was placed in a 96‐well plate, and then the solution was spectrophotometrically measured on a microplate reader at a wavelength of 450 nm (BioTek MQX200R). The as‐measured OD value was corrected by subtracting the background of DI water. After that, the standard curve of BPB concentrations versus OD values was obtained. The experiments were repeated at least three times. In order to calculate the amounts of the BPB released from the porous composite scaffolds with different compressive strains, the as‐measured OD value was corrected by subtracting the background of DI water. The concentration of the BPB solution (c) could be calculated by comparing it with the standard curve. Assuming the total volume of the BPB solution released and the volume of the CCS cubes were v 1 and v 2, respectively. Then the amount of BPB released from per cm 3 CCS could be calculated as (2) m = c v 1 v 2 Measurement of the Amount of BSA Released from CCS : For cyclic release of BSA test, BSA‐loaded CCS was compressed at designated strains (3%) for ten consecutive cycles (defined as one set of compression) and the solution containing released BAS was then removed for spectrophotometrical measurements (100 µL to 96‐well plate, 570 nm) before treated with a Micro BCATM Protein Assay Kit (Prod#23235, Thermo Scientific). After, the test was resumed for another set and repeated up to 15 sets in total. A parallel test of the CCS without compression was used to determine the background release. The amount of BSA was measured following the instructions of the protein kit, and then the amount of BSA released in each set subtracted the background release was averaged by ten cycles and then total volume of CCS to obtain the net release amount per cycle per cm 3 of CCS. All the releases tests were repeated at least three times. Cell Release Test : Cell‐loaded CCS was immersed in a cell culture medium (Dulbecco's modified eagle medium (DMEM) with 10% fetal bovine serum (FBS) and 1% penicilin‐streptomycin (P/S)) and then subjected to cyclic compressions at a strain of 20%. CCS was compressed for 20 consecutive cycles (defined as one set of compression) and the cell culture medium containing released cells was removed for a 24 h culture and then cell count by the Live/Dead Viability/Cytotoxicity Kit (Thermo Fisher Scientific, L‐3224) according to its instructions. Then test resumed for another set and repeated up to four sets in total. A parallel test of the CCS without compression was used to determine the background release of cells. Live/Dead cells after staining were imaged by fluorescence microscopy (ZEISS, AxioCamHRc) and the remaining number of live cells was counted. The number of live cells released in each set subtracted by the number of background release, averaged over 20 cycles and then total volume of CCS to obtain the net released cell number per cycle per cm 3 of CCS. All the releases tests were repeated three times. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201600480
| 2,017
|
Advanced Science
|
Draw‐Spinning of Kilometer‐Long and Highly Stretchable Polymer Submicrometer Fibers
|
A new method is developed to directly spinning perfectly uniaxial fibers in an ultrafast manner. Besides, this method can tune the fibers' diameter through adjusting processing parameters such as the feeding rate of precursors. Uniaxial nylon 66 fibers prepared via this method show superior mechanical properties due to the alignment in each level of the structure.
|
In this study, we report draw‐spinning method that can produce aligned polymer ultrathin fibers with diameter down to 200 nm. Apart from creating perfectly aligned fibers, this method can draw fibers at a high speed and can thus reorient molecular chains. Tensile tests show that nylon 66 fiber bundles fabricated by draw‐spinning exhibit >400% elongation at break and ≈250 MPa strength, which are superior to the corresponding values for conventionally fabricated fibers and bulks (usually <30% and <100 MPa, respectively). The superior performance can be attributed to the relative motions of building blocks in each of the four levels of the structure: the bundle, the fibers within the bundles, microfibrils within the fibers, and aligned molecular chains within the microfibrils. We confirm that each level within the hierarchy is highly aligned and propose a “tetra‐slip” system. Fibers have been existing inside plants and playing structural roles to give mechanical support in harsh living conditions and produced by animals for specific applications. 1, 2, 3 For example, to withstand huge bending forces from the wind, the inner fibrous structure within bamboos evolves to be highly uniaxial and various spider silks can perform a variety of roles such as webs and draglines. 4, 5 Fibers of plants, such as flax, hemp, and jute, have been applied to daily usage. 6 With the development of chemical synthesis, polymers such as nylon and kevlar were invented and fabricated into structural fibers. Except for direct applications, carbon fibers and glass fibers can be made into composites for reinforcement. Regardless of their component, fibers with superior mechanical performance have been a persistent topic. Apart from facilitating the evolution of component materials, adjusting the structure can be effective as well. 7 In this regard, fiber diameter and alignment are two of the key factors. However, combining the structural advantages of these factors necessitates the improvement in fiber spinning methods. 8 Industrial approaches, such as wet spinning, dry spinning, gel spinning, and melt spinning, are relatively fast, but the fiber products usually have diameters of dozens or hundreds of micrometers. Although direct drawing fibers from the precursor is quite simple, it has a low yield rate, and thus it has a low practical value. 9, 10, 11, 12 Electrospinning, which originated from the 1940s, is probably the most well‐established processing method for nanofibers. 13 Researchers have attempted to obtain aligned fibers by modifying the conventional setup of this process. 14, 15, 16, 17, 18 However, due to the electrostatic repulsion among the charged fiber segments, the general products of electrospinning are nonwoven mats. 19 Thus, the previous work rarely attained perfect alignment and precise positioning of the fibers. Except mechanical performance, 20 profound practical values of aligned fibers lie in various fields. For example, in tissue engineering, highly aligned fibers can be applied in cell growth and drug delivery. 21, 22 Periodic structures exhibit structural colors and generate fascinating applications in photonic crystals, and surfaces can be tuned to be hydrophobic by coating nanofibers. 23 Generally speaking, aligned fibers can probably play a part in most anisotropic applications. Plenty work has been done to improve the mechanical performance of nylon 66 by adding other tougher materials. MWCNT (Multi‐wall Carbon Nanotube)‐strengthened uniaxial electrospun nylon 66 fibers exhibit elongation at break of 130% and ultimate strength of 100 MPa, 24 which are six and nine times higher than those of the crosslink neat nylon 66 electrospun fibers. 25 Other significant efforts include glass fiber‐reinforced nylon 66 via injecting molding and electrospinning nylon 66/organoclay nanocomposite. 26, 27 In this communication, we present the general procedures of draw‐spinning and describe how it controls the diameter of aligned fibers; then, taking this structural advantage, we prepared nylon 66 and polyethylene oxide (PEO) fibers with different diameters for tensile tests; in the Experimental Section, we analyzed the alignment within each level of the tested bundles and proposed a “tetra‐slip” model to explain the mechanical results. The basic setup for draw‐spinning is composed of two main parts. One part is a syringe that serves as the reservoir for the precursors and is loaded on a syringe pump, and the other is a collector, which is typically a rotating substrate ( Figure 1 A; Figure S1, Supporting Information). The rotating substrate provides a force to draw fibers directly from the reservoir. To supply on demand, raw material feeding is controlled by the syringe pump to balance the take‐up speed. As for the collector, a 2D plate, a cylindrical roller, or even an object with any shape can be used to provide support for draw‐spun fibers. As shown in Figure 1 B, a roller with a diameter of ≈16 cm was used in our experiment to collect the draw‐spun fibers. To initiate draw‐spinning, the precursor was prepared first. It was then loaded to a syringe, and the syringe with the precursor is placed in the syringe pump. While the pump is switched on, the liquid pool on the nozzle was manually drawn into a fiber. The rotating substrate was connected afterward. The draw‐spinning process stabilizes when the combination of the processing parameters are optimized. Continuously spinning PEO fibers at 1 m s –1 for 90 min resulted theoretically in one ≈5 km‐long submicrometer fiber (Figure 1 C; Movies S1 and S2, Supporting Information). The fibers are highly aligned and the diameters are uniformly distributed. Two intrinsic properties of the precursor must be considered to reach an acceptable spinability. One is the molecular weight of the polymer, and the other is the surface tension of the solvent. High molecular weight ensures the entanglement of long molecular chains in less concentrated polymeric solutions. Fibers are formed only when the entangled molecular chains are consecutively drawn in this method, as poor entanglement caused either by low molecular weight or by low concentration reduces spinability. Another reason for the breakage of the fibers is surface tension. Solvents with low surface tensions or surfactants, such as sodium dodecyl sulfate and Triton X‐100, can be used to prevent breakage. The state of the precursor also influences the surface tension. Most organic solvents are volatile, but rapid solvent evaporation could unfavorably affect spinability. When the solvent evaporates, the solid residue could disrupt the feeding of the liquid precursor in the nozzle. In addition, the dried cone is extremely tough to be drawn into the fibers. Adjusting the feeding rate can balance the take‐up speed such that no redundant material accumulates on the nozzle. Moreover, the setup is highly adaptable. Using four‐nozzle needles can increase the production speed fourfold (≈5 m s –1 in Figure S2, Supporting Information). Figure 1 Basic setup, products, and procedures of draw‐spinning. A) Schematic illustration of the basic draw‐spin setup. B) Digital photo of the actual draw‐spin setup in our lab. C) Fiber bundle removed from the roller (The lower row: SEM images of the as‐prepared fibers). Spinning parameters influence the morphology of the fiber in most cases. The influential parameters include the voltage in electrospinning to control the drawing force, the molecular weight, the concentration of the precursors, the travelling distance, and the flux rate. 28, 29, 30, 31, 32 In the present study, we evaluated two draw‐spinning parameters that affect the size of the fibers, the drawing speed and the precursor concentration. The draw‐spinning process is a balanced result between feeding and consuming rates of the precursors. Under the same feeding rate, the consuming rate can fluctuate within a certain range while maintaining spinability. Deformation of liquid pools and change in fiber diameter can be observed when the feeding and consuming rates are mismatched. Investigations on the diameter's dependence on drawing speed were performed on PEO fibers. The substrate was a silicon wafer attached to a roller with a diameter of ≈3 cm. While maintaining the feeding rate at 0. 13 mL h –1 and changing the rotating speed, we obtained the magnified digital photos of the deformed liquid cones. As shown in Figure 2 A, from top to bottom, the rotating speeds are 150, 300, and 450 RPM (the take‐up speeds were 23. 6, 47. 1, 70. 7 cm s –1, respectively). As a result, the shape of the liquid pool on the 0. 06‐mm nozzle transformed from “jumping sphere” to “constant jet. ” When the amount of the precursor is fixed, an increased rotating speed directly leads to an increase in production rate, resulting in thinner fibers, as shown in the scanning electron microscopy (SEM) images in the right‐hand column, and the scatter plot regarding the statistics are depicted in Figure 2 B and Figure S3 (Supporting Information). Movie S3 (Supporting Information) provides a dynamic demonstration of the balance and mismatch during draw‐spinning. During traveling from the nozzle to the substrate or after depositing on the collector, the solvent in the as‐prepared fibers tends to evaporate, and the solute thus becomes the final product. The diameter of the draw‐spun fibers was hypothesized to be closely related to precursor concentration. To test this hypothesis, highly concentrated (3. 2 wt%) PEO/acetonitrile precursor was prepared. By consecutively performing the “spin‐dilute” operation several times, the lowest spinable concentration was reached at ≈0. 5 wt%, and several data points were acquired, as shown in Figure 2 C. A positive correlation between the diameters of the draw‐spun fibers and the polymeric ratios of the precursor was observed, and the diameter can be adjusted from several micrometers to ≈200 nm. Profound meanings in patterning arrays and grids are present. Massive efforts have been made to overcome the randomness of electrospinning. The lowest pitch between the near‐field electrospun fibers is ≈5 µm, but at 2 mm s –1. 33 A high speed version of this method can reach 0. 5 m s –1, but the lowest spacing is 100 µm. 34 Draw‐spinning can achieve each of this result without compromising the other. To arrange fibers into an array, we used another syringe pump to introduce translational motion into the system. By mounting the motor‐powdered substrate onto the syringe pump, we integrate rotating motion and linear motion to wind one fiber into an array (Figure S4 and Movie S4, Supporting Information). SEM images of the perfectly aligned PEO fibers with different spacings (6, 12, and 18 µm) are shown in Figure 2 D. Another proof for the recurring structures is the structural colors caused by the diffraction grating effect. 35 This result also shows the potential of the method in coating (Figures S5 and S6, Supporting Information). After an array is prepared, rotating the substrate and spinning again deposit another layer of fiber array on the previous array, resulting in a grid (Figures S7 and S8, Supporting Information). Besides polymer fibers, we also achieved the patterning of brittle ceramic fibers and successfully assembled metal fibers into grids (Figure S9, Supporting Information). One of the envisioned field of fibers, especially the fiber mesh is flexible electronics. 36, 37, 38 The potential advantage is to reach bulk performance in conductivity, especially when the junctions are welded together and while the smallest amount of raw materials are used, improving the transparency of the electrode. In addition, because mechanical strength is strongly related with structures, the square holes are expected to deform in order to survive external stress. Thus, the mesh must be a good starting point toward a flexible, stretchable, and transparent electrode. 39, 40, 41 Figure 2 Diameter control and fiber arrays. A) Diameter and feeding rate. Magnified digital photos of the deformed cones caused by the “feeding‐consuming” mismatch (left column) and SEM images of the corresponding products (right column). B) Scattered points showing the diameter of the fiber products fabricated under corresponding processing parameters. C) Upward trend of the diameter change when polymeric concentration increases. D) Fiber arrays of varying spacings: 18, 12, and 6 µm. As mentioned previously, spinability can be maintained even when a mismatch between feeding and consuming is present, although it influences the diameter of the fibers. Thus, we fixed the flux rate while changing the rotating speed to produce a bundle. With the flux rate and fabricating time fixed, we produced two bundles consisting of fibers with different diameters. Specifically, we used a roller to collect nylon 66 fibers (mixed with PEO at the weight ratio of 10:1). The product is a fiber “ring, ” as removed from the roller. It was then folded twice and twisted. Draw‐spun nylon 66 fiber bundle can stretch from its original length of 5–25 cm ( Figure 3 A; Movie S5, Supporting Information). To compensate the fourfold increase along the axis, the average diameter shrank from 3. 0 to 1. 6 µm (Figure 3 B). The notable mechanical properties are the Young's modulus, the tensile strength and the elongation at break. By contrast, the elongations at break of the electrospun nylon 66 fibers are mostly below 100%. Since the chemical components of the samples in the published work are different from our samples, we prepared nylon 66 thick fibers (diameter: ≈75 µm) following the wet‐spinning procedures for tensile tests. As shown in Figure 3 C, the elongation and the strength of our samples can reach 400% and 230 MPa, respectively, superior to most published results, plotted as scattered points. The diameter mainly has influences on the strength and the modulus. The faster spun samples exhibit the strength three times higher than its slower spun counterparts and the modulus can be improved from 0. 81 to 1. 35 GPa. In Figure 3 D, comparable results can be found in draw‐spun PEO fibers. We also investigated the mechanical performances of fibers with different proportion of nylon 66 and PEO, and found that the strength increases with the addition of nylon 66 (Figure S10, Supporting Information). Tolerance against crack propagation is one of the mechanical advantages of the fiber bundles. When the bulk counterparts are stretched, the stress concentration starts when the components are interlocked, and then cracks propagate to cause failure. However, failure of a single fiber in a bundle does not lead to the breakdown of the entire system. The support for this theory can be found from the “staircases” in the strain–stress curves shown in the left plot of Figure 3 D. Figure 3 Tensile tests of draw‐spun nylon 66 fibers. A) Photos of a bundle before and after the test. B) Diameter distribution of fibers: tunable average diameter and the shrinkage of diameter after tensile tests. C) Strain–stress curves of nylon 66 fibers. Right: Young's modules region. D) Strain–stress curves of PEO fibers. Right: Young's modules region. The reason for the huge discrepancy between the actual and theoretical mechanical performance is the folding of the polymeric chains. As soon as they are extruded, industrial spun fibers immediately go through a drawing process to increase their strength but at a sacrifice of stretchability. For a better analysis of the mechanical advantage, we revealed the four‐level alignment and propose a “tetra‐slip” system. On macroscopic levels, slipping level‐1 occurs between the four strands, and inside each strand, slipping level‐2 occurs between the draw‐spun fibers ( Figure 4 A, B). At microscopic levels, the nylon 66 microfibrils are oriented, most likely because of the shear force when they are drawn. By selectively dissolving PEO in the mixture using acetonitrile, we showed the shape of the nylon 66 microfibrils. Figure 4 C shows that in draw‐spun products, the microfibrils are elongated along the shear drawing force. For comparison, we cast the same precursor on glass films by dipping, and after the selective dissolution, the microfibrils are randomly distributed. Fourier transformed infrared spectroscopy was conducted to verify the effectiveness of the selective dissolution (Figure S1, Supporting Information). During stretching, slipping occurs between the oriented microfibrils. In conventional materials, microfibrils interlock each other. Finally, innermost slipping is present between the aligned linear molecular chains. Raman spectroscopy characterized the orientation of the molecular chains. 42 The Roman intensity was acquired twice for each sample, which was rotated 90 °C horizontally for the second measurement. In this way, the fiber axis was positioned parallel or perpendicular to the exciting laser beam. A large difference between the two sets of data indicates high heterogeneity or intensive orientation of the molecular chains. Figure 4 D shows the alignment of the molecular chains in the following order: cast films, as‐prepared draw‐spun fibers, and draw‐spun fibers after tensile tests. During stretching, the components in the tetra‐slip system relocate and reorient, thus effectively avoiding stress concentration. Figure 4 Tetra‐slip system as an explanation for the mechanical advantages of draw‐spun fiber bundles. A) Twisted four strands. B) Perfectly aligned fibers. C) Selective dissolution of PEO for nylon 66 microfibrils. Left: oriented nylon 66 microfibrils in draw‐spun fibers; Right: randomly distributed microfibrils in cast films. D) Raman proof of molecular orientation. To conclude, draw‐spinning possesses several advantages over current processing methods for micro‐/nanofibers. First, single continuous fibers with limitless length and tunable diameter can be fabricated through this method. Second, the fibers can be arranged into arrays and meshes. Third, this method is adaptable and versatile, and it can be integrated with current processing methods or followed with well‐established post‐treatments seamlessly to create novel structures and expand the draw‐spinable systems. Small fiber diameter and better fiber alignment improve mechanical properties. We conducted tensile tests on draw‐spun nylon 66 fiber bundles to investigate the influence of fiber diameter on mechanics. Because of the structural advantages, our product is tenfold higher than most nylon 66 samples when elongated, demonstrates superior performance in strength. The mechanical improvements can be explained by our tetra‐slip system. Experimental Section Sample Preparation : Polymer solution containing a certain weight ratio (ranging from 4. 0 to 0. 8 wt%) of PEO ( M v = 8 000 000, Sigma‐Aldrich) was prepared using acetonitrile (Analytical Reagent, Sinopharm Chemical Reagent Co. , Ltd, China) as the solvent. The nylon 66 precursor was prepared using nylon 66 and PEO (three weight ratios were chosen in our experiment, 10:1, 6:1, and 3:1) as the solutes. Nylon 66 was added into the ≈1 wt% PEO/formic acid (chemically pure, Sinopharm Chemical Reagent Co. , Ltd, China) and stirred until dissolution. Draw‐Spinning Procedures : Precursors were loaded into a syringe, which was then mounted onto a syringe pump. A rotating roller was used as the collector of draw‐spun fibers and by manually drawing the liquid cone on the nozzle into a fiber onto the rotating roller, the precursor can be constantly spun into fibers. To construct arrays, the translational motion was added to the rotating roller. Sample Characterization : Morphology and microstructures of fibers were observed using a field‐emission electron scanning microscopy (LEO‐1530, Zeiss, Germany). Energy Dispersive Spectroscopy signals of the fiber arrays were collected using detectors from Oxford Instrument (X‐Max N Silicon Drift Detector). Optical transmittance was measured using a UV–vis spectroscopy (SHIMADZU UV‐2600). Fourier‐transformed infrared spectrum was obtained using an IR spectrometer from Bruker Corp (VERTEX 70v). Powder samples were mixed with KBr at weight ratio of 1:100; bulk samples were examined with the help of ZnSe as the attenuated total reflectance crystal. Raman intensity was performed using a Raman spectrometer with a 600 gr mm –1 grating and the laser emitting at 532 nm (LabRAM HR Evolution, HORIBA Jobin Yvon, French). Tensile Tests : Tensile tests were performed using a Zwick universal testing machine, Zwick Roell Z005, under room temperature and at moisture of ≈30%. The effective area of the samples was calculated via multiplying the average cross sectional area of an individual fiber by the overall amount of fibers. For each sample, more than three measurements were conducted to ensure the repeatability and the plotted curves are representative. The test speeds were 130% strain per minute. Results at other speeds are available in Supporting Information. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700029
| 2,017
|
Advanced Science
|
Recent Progress on Piezoelectric and Triboelectric Energy Harvesters in Biomedical Systems
|
Implantable medical devices (IMDs) have become indispensable medical tools for improving the quality of life and prolonging the patient's lifespan. The minimization and extension of lifetime are main challenges for the development of IMDs. Current innovative research on this topic is focused on internal charging using the energy generated by the physiological environment or natural body activity. To harvest biomechanical energy efficiently, piezoelectric and triboelectric energy harvesters with sophisticated structural and material design have been developed. Energy from body movement, muscle contraction/relaxation, cardiac/lung motions, and blood circulation is captured and used for powering medical devices. Other recent progress in this field includes using PENGs and TENGs for our cognition of the biological processes by biological pressure/strain sensing, or direct intervention of them for some special self‐powered treatments. Future opportunities lie in the fabrication of intelligent, flexible, stretchable, and/or fully biodegradable self‐powered medical systems for monitoring biological signals and treatment of various diseases in vitro and in vivo.
|
1 Introduction Over last decades, implantable medical devices (IMDs) have experienced tremendous growth, becoming indispensable medical tools for improving the quality of life and prolonging the patient's lifespan. Currently, the IMDs have been implanted in various parts of human body as artificial treatments and diagnostic tools, including sensors, pacemakers, implantable cardioverter defibrillators, cochlear implant and stimulators for deep brain, nerve and bone. 1 These implantable electronic devices can provide diagnosis (e. g. heart rate, blood pressure and temperature monitoring) for a number of diseases related to the heart, brain and some other important organs and support real‐time treatment (e. g. stimulation of muscle and nerve system). For instance, a cardiac pacemaker can help to correct abnormal heart rhythms by using electrical stimulation to contract the patients' cardiac muscle to relieve heart blockage or sick sinus syndrome. 2 Additionally, to achieve better quality of life and enhance the survival rateof patients worldwide, IMDs have also contributed significantly to our cognition of the biological processes existing in the human body, including the complicated mechanisms of neural communication, memory and control, which significantly deepen our understanding of how these processes are affected by differernt diseases and treatments. 3 In spite of substantial advancement in the manufacture and application of IMDs since the first in vivo cardiac pacemaker borned in 1958, the present IMDs are still faced with numerous challenges. 4 There is a strong demand to design IMDs with diminishing size and weight for the sake of minimize their impact on daily human activities and increase comfort for the users. The batteries usually occupy the weight and size of the IMDs. However, they are limited by current technology and difficult to achieve miniaturization and weight reduction. The lifetime of battery is another challenge need to be overcome. When used to generate electrical pulse in deep brain stimulators (DBS) and cardiac pacemakers, the lifetime of those batteries is predetermined (e. g. 3 to 5 years for DBS), at the end of which these IMDs have to be replaced by surgery, leading to high cost of money to the patients and the social healthcare system. 5 In vivo energy generation and internal charging by the natural body activity and physiological environment has been reported recently. Further miniaturization can be realized by means of self‐powered implantable devices that harvesting energy from natural sources or artificial power around the patient for sustaining the device directly. Variouse methods to reclaim energy from electrical, thermal, chemical and mechanical processes in vivo has been demonstrated, for example electric potentials from inner ear, glucose oxidation, vibration of organs and muscle contraction. 6 In all these biological power sources, mechanical energy was considered as one of the most popular and sufficient power in a living creature. By harvesting in vitro and in vivo biomechanical energies, self‐powered medical electronics for nearly lifetime can be attained. Attractive approaches of self‐powered biomedical systems have been recently investigated by integration of energy harvesting devices, which can convert biomechanical energy from movements of human body (including body motion, blood circulation and the contraction/relaxation of cardiac, lung and muscle) into electricity. 7 In this mechanical energy conversion process, many techniques such as piezoelectric effect, triboelectric effect, magnetostrictive effect and electromagnetic induction, can be used. 8 However, another type of bulky energy‐harvesters utilized for implantable energy sources were limited, due to discrepant contact with the curved, corrugated and irregularly shaped surfaces of organs such as the lung, brain, eye and heart. Moreover, energy harvesters on rigid and thick substrates are unsuitable for converting the subtle movements of internal musclesand organs to generate electric power. Extremely flexible and lightweight energy harvesters are settled conformally on muscle and organ surfaces. These devices are fabricated on plastic thin films such as polyethylene terephthalate (PET), polyimide (PI), and polydimethylsiloxane (PDMS) widely used in flexible electronics because of their appropriate flexibility and strength. Recent development in biomechanical energy harvesters based on piezoelectric effect is an significant progress for solving the aforementioned issues. Several researchers have reported thier pliable piezoelectric energy harvesters with high performance called piezoelectric nanogenerators (PENGs), 9 which use organic or inorganic materials such as polyvinylidene fluoride (PVDF), 10 poly(vinylidenefluorideco‐trifluoroethylene) (P(VDF‐TrFE)), 11 ZnO, 12 BaTiO 3 (BTO), 13 Pb(Zr x Ti 1–x )O 3 (PZT), 14 and (1–x) Pb(Mg 1/3 Nb 2/3 )O 3 ‐xPbTiO 3 (PMN‐PT). 15 These functional devices can generate electric power under tiny irregular deformation and mechanical vibration revealing a tremendous potential to be applied in differernt medical devices in vivo. The fast growing triboelectric energy conversion devices, defined as triboelectric nanogenerator (TENG), have also shown many advantages, such as high efficiency, light‐weight, low costand easy fabrication, and provided new options for biomechanical energy harvesting. 16 The mechanism of the TENG relies on a conjunction of triboelectrification and electrostatic induction between two contacted materials. Based on the triboelectric series, 17 any two different materials with disparate tendency to lose or gain electrons in a frictional contact process have the potential to be used for TENG. Since there is a broad range of materials selections, we can take into account the output ability, flexibility, biocompatibility and cost of the device, which make TENG an ideal candidate for biomedical applications. Several research groups have employed high‐performance flexible and/or implantable TENGs for harvesting mechanical energy derived from the human motions or in vivo physiological movements. Together with the PENGs, TENGs are becoming most promising biomechanical energy harvester and providing great chance for building self‐powered biomedical systems. Besides energy harvesters, PENG and TENG can be as mechanical nanosensors in biomedical field. By specific structural design, these devices can be very sensitive to detect small scale mechanical motions (e. g. the vibration of carotid artery caused by phonation and bloodstream). Recently, some studies reported the delicate biosensing applications of PENG and TENG. For example, piezoelectric device was reported as a electromechanical biosenser to monitor the volume change of PC12 cells and as bio‐inspired artificial hair cells to detect displacement of vibrational in 15 nm range. 18 In addition, a triboelectric based bionic membrane sensor (BMS) can monitor the throat sound in high‐frequency and continuously detect the human arterial pulse wave in low‐frequency by one single device, exhibiting great potential in wearable medical/health monitoring. 19 These progresses could be developed to real‐time sensing of cardiac function or blood pressure and in vivo monitoring the recovery of damaged sensorium in future. This review introduces a overview of recent progress of biomechanical energy harvesters, nanosensors and stimulator. Particularly, this manuscript presents applications of piezoelectric and triboelectric devices as energy harvester and self‐powered sensers according to their remarkable flexibility, biocompatibility and cost‐effective compared to other types of mechanical energy conversion means. Moreover, the applications of piezoelectric and triboelectric based devices in self‐powered cardiac pacemaker, nerves/muscles stimulator, and nanoscale sensor for monitoring biomedical pressure/strain changes, are discussed ( Figure 1 ). Figure 1 Recent applications of PENGs and TENGs in biomedical field. 2 Energy Conversion Mechanism 2. 1 Piezoelectricity Appling PENGs to harvest mechanical energy has received a great amount of attention due to its diversity of sophisticated design for direct conversion of mechanical energy into electric power for variouse integrated applications. The nature of the piezoelectric effect is closely related to the generation of electric dipole moments in solids. The dipole moments may either be induced by ions on crystal lattice sites with asymmetric charge surroundings (as in ZnO, BaTiO 3 and PZT) or may be carried by molecular groups directly (as in cane sugar). Take the ZnO crystal of wurtzite‐structured as example, the tetrahedrally coordinated Zn 2+ and O 2− are accumulated layer‐by‐layer along the c‐axis ( Figure 2 A). At its original state, the charge center of the anions and cations coincide with each another. Once applying an external force, the structure is deformed (stretching or compressing). Therefore, the negative and positive charge centers are separated and form an electric dipole leading to a piezoelectric potential (Figure 2 B). If an external load is connected to the deformed material, the free electrons are driven to partially screen the piezoelectric potential and flow through the external circuit to realize a new equilibrium state. 20 Therefore, a current pulse flowing through the external circuit is continuously generated when the piezoelectric potential is altered sequentially by applying a dynamic external force (Figure 2 C, D). 21 This primary mechanism of piezoelectric potential generation applies to various piezoelectric materials. PENGs mainly consist of piezoelectric materials and flexible substrates. Two core factors for developing PENG are material choosing and device structural design. Figure 2 Mechanism of piezoelectricity. A) Atomic model of the wurtzite‐structured ZnO. B) Different piezopotential in tension and compression modes of the PENG. C) Numerical calculation of the piezoelectric potential distribution in a ZnO nanowire under axial strain. Reproduced with permission. [[qv: 20a]] Copyright 2009, AIP Publishing LLC. D) Band diagram for the charge outputting and flowing processes in the PENG. Reproduced with permission. [[qv: 20b]] 2. 2 Triboelectricity & Electrostatic Induction The triboelectric effect which has been known for thousands of years, is the electrically charging process when two different materials contact each other through friction. Despite this is one of the most common phenomena that seen every day, the mechanism of triboelectrication is still not very clear. The process of chemical bonds formed between some parts of the surface in two different contacted materials and charges movinge from one material to another due to different stability to gain electron is generally believed. The charges transferred between two materials can be molecules, ions and electrons. When two materials are separated, some of the bonded atoms tend to maintain the additional transferred electrons, and some tend to give them away, which possibly give rise to the opposite charges on differernt friction materials. The opposite charges on both friction surfaces can generate a triboelectric potential, which can drive electrons in the back electrode to flow in order to balance the created electric potential drop. On the basis of this principle, four kinds of TENGs with different modes have been invented ( Figure 3 ), as elaborated in the following: Figure 3 The four fundamental modes of TENGs: A) vertical contact‐separation mode; B) Lateral‐sliding mode; C) freestanding triboelectric‐layer mode, and D) single‐electrode mode. 2. 2. 1 Vertical Contact‐Separation Mode The first design of TENG was shown in Figure 3 A as an example. Two differernt dielectric films place as top and bottom friction layer, and electrodes are deposited on outer surfaces of the friction film. The opposite charge is created between the surfaces of two contacted dielectric films. When the external force is released, the two dielectric films are separated with each other, accompanying with the electric potential drop between the friction surfaces. After connecting the two electrods with a load circuit, free electrons would flow from an electrode to another, driven by the electrostatic field. Once the gap becomes closed, the electric potential created by triboelectric effect disappears and the electrons flow backward. 22 2. 2. 2 Lateral Sliding Mode The original structure of lateral sliding mode and vertical contact‐separation mode have the same starting position. A relative parallel sliding creates triboelectric charges on the both surfaces, when two friction surfaces contacted (Figure 3 B). 23 Therefore the electrons on two electrodes are driven to flow by the triboelectric charges, which is generated by a lateral polarization introduced along the sliding direction. With the periodic process of sliding apart and closing, this lateral sliding mode TENG generates an AC output. The sliding can be a planar motion, a disc rotation 24 or a cylindrical rotation. Related studies of mechanism have been used to improve the basic mode and grating structured TENG. 2. 2. 3 Single‐Electrode Mode These two kinds of modes mentioned above contain two friction layers and two electrodes interconnected by an external circuit. Such TENGs can work independently and be freely moved. While in some other cases, a mobile object can be one electrod of the TENG, which do not need to be electrically connected to the load, such as a human jogging on the ground. For scavenging mechanical energy in these cases, single electrode TENG is created. The bottom electrode is grounded in single‐elecrtod mode TENG (Figure 3 D). The local electrical field distribution will be changed with the top object approaching or departing from the bottom film, resulting in electron flow between the ground and the bottom electrode to match the potential change. In both contact‐sliding mode and contact‐separation mode, this energy harvesting strategy has been applied. 25 2. 2. 4 Freestanding Triboelectric‐Layer Mode In natural environment, a moving object, such as our cloths, gloves and shoes, is usually charged by contacting with air or ground. The electric charges can maintain on the surface for hours. Due to the density of electric charges achieves a maximum, there is no necessary to do friction or contact during this time. A pair of symmetric electrodes is designed under a dielectric film and the width of the electrodes and the gaps are of the same order with the moving object. With the object approaching to and departure of the electrodes, an asymmetric electric charge distribution on the surface of materials can be createed, which causes the electrons flowing from one electrodes to another to screen the local potential distribution (Figure 3 C). Between the paired electrodes, an AC output is generated by the oscillation of the electrons. Because there is no direct friction or touching between the moving object and the top dielectric layer of the electrodes, free rotation is possible without direct mechanical contact, and abrasion of the dielectric layer can be completely reduced. This approach will grealy enhance the durability and prolong the service life of TENGs in rotation mode. Harvesting energy from human walking, automobile and other moving objects is feasible and convenient. 26 3 Materials and Device Design 3. 1 Nanowire Devices The name of piezoelectric nanogenerators (PENGs) was defined by Wang's group in 2006 when they found piezoelectricity in ZnO nanowire. 9 After that, numerous researches have been implemented about the working mechanism, structural design and modeling, and output optimization of the PENGs. Until recently, various kinds of flexible PENGs have been developed, which could be used for harvesting varies mechanical energies from either the environment or human bodies. The output electrical energy has been increased from several millivolts to several hundred volts, which is enough for driving a small devices with low power consumption, such as liquid crystal display (LCD), light‐emission diode (LED), and some wireless signal transmitting devices. In this section, the author briefly reviewed the working mechanism, modeling/simulations, and the experimental progress of piezoelectric nanogenerators according to the structure of the nanogenerators including the vertical aligned nanowire arrays, the lateral‐aligned nanowire networks, and some other similar structures, such as nanobelts and nanoribbons. 3. 1. 1 Lateral Nanowires PENGs built with single ZnO nanowire or lateral nanowires arrays have been well established, providing various delicate nanoscale devices for multiple applications ( Figure 4 ). In a typical case, both ends of a single ZnO nanowire was fixed by silver paste as electrodes. The ZnO nanowire and two electrodes were packaged on a flexible substrate laterally. A uniaxial tensile strain was induced by the bending of the substrate and cause a piezoelectric potential along the ZnO nanowire, which drove the electrons to flow between the external circuit. A pulse current signal was generated by ZnO nanowire‐based PENG in a repeated bending‐releasing process (Figure 4 A). However, the output performance of PENG based on single ZnO nanowire was relative low. The open‐circuit voltage ( V oc ) was about 20–50 mV and short‐circuit current ( I sc ) was about 400–750 pA, which limited their applications of energy harvesting (Figure 4 B). 27 For enhancing the output of ZnO based PENG, a direct approach was to integrate numerous lateral ZnO nanowires in a single device. As shown in Figure 4 C, nearly 700 rows of ZnO nanowires were integrated on a single flexible substrate in a parallel configuration. 28 Each row in this configuration contained about 20, 000 ZnO nanowires. The whole device exhibited great flexibility. When this device was deformed by a linear motor at a straining rate of 2. 13%, it reached about 1. 2 V V oc and 26 nA I sc as an average, respectively. The slightly different of the magnitudes of the PENG output peaks is noted as the different straining rates in the releasing and stretching processes. Figure 4 PENGs based on lateral ZnO nanowires. A, B) single nanowire based PENG fixed on flexible substrate. Reproduced with permission. 27 Copyright 2009, Nature Publishing Group and American Chemical Society. C) Structure and optical images of a flexible lateral‐nanowire‐array integrated PENG. Reproduced with permission. 28 Copyright 2010, Nature Publishing Group. D) Fabrication process of a flexible PENG based upon a lateral ZnO NW array. Reproduced with permission. 12 Copyright 2010, American Chemical Society. While for powering electrical devices, the mentioned output is still need to be improved and PENG with higher output is desired, which means the total number as well as the density of integrated ZnO nanowires need to be further increased. The lateral ZnO nanowire array was utilized in a sweeping‐printing method to fabricate high‐output PENG by Zhu et al. (Figure 4 D). 12 In a standard process, uniform ZnO nanowires were synthesized firstly on silicon substrates in physical vapor deposition (PVD) method. By sweeping on a flexible Kapton film, the vertically aligned nanowires were transferred to substrate surface by shear force. The avergae density of as‐transferred nanowires was 1. 1 × 10 6 cm –2. Next, 600 rows of Au electrodes in stripe‐shaped with space of 10 µm in width were deposited on top of the lateral nanowires arrays by conventional photolithographic procedures. At last, PDMS was used to package the entire device, which could further protect the device from invasive environment of humidity and corrosive chemicals, and enhance the mechanical robustness. The identical growth and alignment of nanowires insureed the direction of the piezoelectric potentials and a successful superposition effect of the electric energy generation of all integrated nanowires. It can reach up to 2. 03 V of V oc and an 11 mW cm −3 of output power density. Furthermore, the commercial LEDs were powered by the electrical energy generated by the PENG, demonstrating the great potential of PENG as a power source of small electronic devices. However, the photolithographic fabrication method limited the scale‐up and industrial production of this kind of PENG, for the process was technical, complex and expensive. To address this problem, a facile fabricating process of PENG was developed by coating conical ZnO nanowires into a flat and stretchable polymer film. This device generated a macroscopic electric potential by the superposition of piezoelectric potential in each conical nanowire between top and bottom electrodes. The generated energy was sufficient to drive a LCD and was reported as 2 V in voltage and 50 nA in current. 29 Besides ZnO, many other nanowires made up of different materials also have shown great performance when fabricated into PENGs. Wu et al. described a series of work about robust and low cost PENG made by ZnSnO 3 nano‐belts. 30 One of the PENG was fabricated by the integration of thousands of randomly distributed lateral ZnSnO 3 nanowires. Due to the spontaneous polarization of ZnSnO 3 belts along the z‐axis, a substantial piezoelectric potential across the PENG's thickness was generated under a compressive strain of about 0. 1% in the unipolar assembly of the triangular‐beltsdue to. A maximum output voltage and current reacheed up to 5. 3 V and 0. 13 µA, respectively. 31 Dagdeviren and her co‐workers fabricated a conformal PENG based on lateral integrated PZT nano‐ribbons for harvesting mechanical energy from motions of the organs, such as diagrams, heartand lung. An output voltage of 4 V was achieved in this report, providing evidence that PENG can significantly yield electric energy from in vivo organs. Moreover, BaTiO 3 nanowires synthesized via a simple and low temperature hydrothermal method were used to develop a lead‐free, flexible PENG by Park et al. During regular and periodical bending and releasing, the BaTiO 3 nanowires–PDMS composite device successfully generated about 7. 0 V of output voltageand 0. 36 µA of output current. 32 3. 1. 2 Vertical Nanowires Similar to the above mentioned lateral nanowires devices, the vertically grown ZnO nanowire arrays that have consistent polar directions can also be used to convert mechanical energy into electric power ( Figure 5 ). One of the most acceptable methods to fabricated vertical ZnO nanowires is solution‐based wet chemistry method (Figure 5 C). The nanowire arrays can be synthesised at about 80 °C on various substrates. The uniformity and periodicity of these nanowires can be well controlled by pre‐sputtering a layer of seeds to the substrate. 33 This technique provides a low cost, precticaland large scale approach to fabricate PENGs. In this context, a large number of flexible PENGs were developed, benefiting the design of self‐powered portable electronics. A fully flexible PENG was fabricated by Choi et al. based on vertically aligned ZnO nanowires in 2010. 34 The working principle of this kind of PENGs was just similar with that of the former mentioned PENGs based on the lateral ZnO nanowires. The piezoelectric potential generated electron flow, which was controled by Schottky barrier between the interface of metal electrode and ZnO nanowires (Figure 5 B). However, when built into a flexible device, the electrodes in this kind of energy harvester may appear some problems of mechanical durability, which make against the lifetime and stability of the device. Therefore, in the manuscript, they provided a potential solution by using single walled carbon nanotube (CNT) networksas conductive electrodes. Due to its improved contact between the nanowires and electrodes, the output of this PENG was significantly enhanced, and meanwhile its durability and stability was exceptional good compared with former devices. The same group further developed a flexible and transparent PENG consisting vertically grown ZnO nanowires using the conventional solution growth method and 2D graphene electrode. The 2D graphene electrode has exhibited extraordinary electrical and mechanical properties in comparison with other electrode materials, which made it an attractive candidate in the application of flexible electronics. Due to its outstanding properties, the as fabricated graphene‐based PENG showed a stable and reliable output current density of 2 µA cm –2. When the device was bended or rolled for several times, the output was without any fluctuations. The mechanical strength as well as its improved output performance provided this device a great potential for reliable biomedical applications. In 2011, Hu and her co‐workers reported a series work to demonstrate a new PENG fabricated with vertical ZnO nanowires arrays. An insulating layer of a thin layer of poly(methyl methacrylate) (PMMA) coated on top of the ZnO nanowires arrays was the most striking advance. The metal electrode was deposited on the PMMA layer. By coupling of the piezoelectric effect and the electrostatic induction, electron flows went back‐and‐forth in the external load induced by an iterative mechanical deformation applied on the device. The output voltage and the corresponding current were up to 20 V and 8 µA, respectively. However, this high output PENG typically was 1–3 mm in thicknesses, which will normally result in a large percentage of consumption of the mechanical energy inputed. Therefore, this PENG was not conducive for small‐scale mechanical energy conversion when applied in biomedical systems. In consideration of this drawback, Lee et al. developed a super‐flexible PENG with same principle by using an Al foil (18 µm) as both the substrate and the electrode (Figure 5 D). The device was responsive to tiny motions on human face. 35 In a typical demonstration, the device was driven by blinking motion, generating an output voltage and current of 200 mV and 2 nA, respectively. Thus, it had great potential to be as active biomechanical sensor. Furthermore, a self‐powered device of multilayered structures without number and size restrictions can be easily integrated based on this kind of PENG. The integration will significantly enhance the output performance of the PENG and efficiency of converting the biomechanical energy. For example, when used in the condition of human walking, a 3. 2 V and a 0. 195 µA of output voltage and current can be reached, respectively. Figure 5 PENGs based on lateral ZnO nanowires. A) Experimental setup, internal structure, and mechanism of PENG fabricated by ZnO NWs vertically grown radially around textile fibers. Reproduced with permission. 36 Copyright 2008, Nature Publishing Group. B) PENG prepared by growing ZnO NW arrays on graphene substrate. Reproduced with permission. 34 C) PENG based on vertical ZnO nanowire arrays that grown by low‐temperature hydrothermal decomposition, which covered in PMMA by spin coating. Reproduced with permission. 28 Copyright 2010, Nature Publishing Group. D) Super‐flexible ZnO based PENG with Al foil as both the substrate and the electrode. Reproduced with permission. 35 Many research works have shown us that the vertical ZnO nanowires can be fabricated on various substrates, such as polymers fibers and metals. In 2008, a fiber like PENG for harvesting low frequency vibration/friction energy was fabricated by using vertical ZnO nanowireson fibers. By entangling two fibers and brushing the nanowires rooted on them with respect to each other, mechanical energy was converted into electricity becaus of coupling the piezoelectric–semiconductor process by brushing the nanowires on two entangled fibers (Figure 5 A). 36 This work established a methodology to scavenge body‐movement energy using fabrics, showing the distinct advantages of designing wearable, adaptable and flexible power source for smart applications in personal electronic. Lee et al. also demonstrated a fiber PENG to convert low frequency mechanical energy into electricity, based on ZnO nanowires and a conducting fiber coated by PVDF. The folding‐releasing motion of human elbow can drive this device to produce an output voltage of 0. 1V and power density of 16 µW cm –3. Similarly, Lee et al. and Li et al. also fabricated a flexible, fiber nanogenerator based on ZnO. 37 This device demonstrated an output voltage and average current density of 3. 2 V and 0. 15 µA cm –2, respectively. A ultrasensitive sensors for human heart can be composed by this PENG. 3. 2 Thin Film Devices 3. 2. 1 PZT Based PENGs PZT is a traditional piezoelectric material that is preferred by researchers for its higher electromechanical coupling coefficients than many other conventional piezoelectric materials such as BaTiO 3. 14, 31, 38 For a long time, due to its fragile nature, PZT thin film was considered inflexible and not stretchable. A structural failure will be caused by a stretch in slight scale and 0. 2% is the maximum safe strain range of PZT. Researchers have proposed various innovative designs of PENGs based on the different morphology and characteristics of PZT to overcome this undesirable property. A transfer printing method was used by Qi et al. to transferred PZT nanoribbons onto flexible rubber or plastic substrates in large scale and high production. 39 An 1 cm 2 PENG can generated voltage and current output up to 0. 25 V and 40 nA, respectively, which was prepared by printed nanoribbons of PZT. The same group further developed a stretchable PENG by integrated the wavy/buckled PZT ribbons with PDMS thin film to increase ability of strain, which can be stretched up to 8% strain and can facilitate its integration with flexible structure. A PENG consisting of 10 PZT ribbons and a current of 60 pA can be ganerated by a device with ten PZT ribbons. Park et al. reported a large‐area and highly efficient, PENG with PZT thin‐film on flexible substrates. An inorganic‐based laser lift‐off (ILLO) process was used in this method. 40 The size limitation of PZT thin film was eliminated by the ILLO process, resulting in an output increase of the flexible PENG ( Figure 6 A). Additionally, the excimer laser and the dry‐type transfer technique advanced the possibility of commercial fabrication of PENG, comparing with other etching method. The large area roll‐to‐roll process and laser irradiation were applied in display industry such as of low‐temperature poly‐silicon (LTPS). During a periodical process of bending and releasing, this device can reach up to ≈200 V in voltage and 150 µA cm −2 in currant, respectively, representing an excellent improvement comparing with other flexible PENGs. A 3. 5 cm × 3. 5 cm large‐area PENG was reported to generate a 250 V of voltage and an 8. 7 µA of current, just being bended by a tiny motion of fingers. 14 This PENG was fabricated by the ILLO process and could lighten 105 LEDs. By applying graphene as the electrodes, a flexible and semi‐transparent PENGs was demonstrated by Kwon et al. The output performance of voltage, current density and power density were ≈2 V, ≈2. 2 mA cm −2 and ≈88 mW cm −3, respectively. Figure 6 PENGs with thin film structure. A) A schematic illustration of the fabrication process of the flexible and large‐area PZT thin‐film PENG using the ILLO process. Reproduced with permission. 14 B) Thin‐film PENG based on laterally‐aligned PZT nanofibers. Reproduced with permission. 41 Copyright 2010, American Chemical Society. C) Flexible BaTiO 3 thin film PENG. Reproduced with permission. [[qv: 44b]] Copyright 2010, American Chemical Society. D) Thin film (P(VDF‐TrFE)) based PENG on flexible substrates. Reproduced with permission. 51 Copyright 2011, Elsevier. E) Porous PENG based on PVDF film. Reproduced with permission. [[qv: 49b]] PZT nanofibers are applied to develop high output PENGs as well. A high‐performance PENG was demonstrated by Cheng et al. fabricated by PZT nanofibers of laterally‐aligned structure. 41 The PZT electrospining nanofibers was ≈60 nm in diameter and ≈500 µm in length, which can generate a 1. 63 V and 0. 03 µW of output voltage and power, respectively (Figure 6 B). Qin et al. prepared a high output PENG fabricated by electrospun ultra‐long PZT nanofibers. 42 They demonstrated the capability of this integrated PENG by bending, stretching, or twisting into a large degree and there was not damage to the structure of PENG. The output of voltage and current produced by this PENG were 209 V and 53 µA, respectively. The reported PENGs have presented a particularly efficient for electrical‐to‐mechanical energy conversion that demonstrate a promising improvement toward the realization of self‐powered systems for wearable and implantable electronics. [[qv: 42b]] The PZT based implantable energy harvesting devices can also be powered by ultrasonic, which is an effective way of converting acoustic energy into electricity. 44 Shi et al. reported a piezoelectric ultrasonic energy harvester (PUEH) based on microfabricated PZT diaphragm array with a size of 5 mm × 5 mm, which has advantages of extra wide operation bandwidth and controlled outputs. [[qv: 43c]] The output performances can be maintained under an appropriate level by adjusting new ultrasound frequency. For instance, when changing the frequency of an ultrasound with 1 mW cm –2 intensity from 250 kHz to 240 kHz, the output power density of PUEH can be increased from 0. 59 µW cm –2 to 3. 75 µW cm –2 at 1 cm distance away from the ultrasound source. This type of device has shown a great potential as power source for battery‐less implantable medical devices and systems. 3. 2. 2 BaTiO 3 Based PENG BaTiO 3 has earned attention for its outstanding ferroelectric properties and biocompatible characteristics which are crucial for biomedical applications. 45 A flexible perovskite BaTiO 3 thin film PENG was reported by Park et al. Piezoelectric BaTiO 3 thin film was deposited on a Pt/Ti/SiO 2 /Si substrate by radio frequency magnetron sputtering and polarized between a high electric field (100 kV cm –1 ) (Figure 6 C). This BaTiO 3 ribbons based flexible PENG can produce a 1 V of output voltage and 7 mW cm –3 of power density, when bended periodically. To achieve higher flexibility and large‐area fabrication, BaTiO 3 nanoparticles were utilized to fabricate PENGs. Park et al. mixed the BaTiO 3 nanoparticles and carbon nanomaterials in PDMS to gain a piezoelectric nanocomposite (p‐NC), which was spin‐coated onto substrates of flexible plastic and cured under a selected temperature. 13 After electrods deposition and high‐voltage polarization, electric signals were generated from the device by mechianical bending or human body movements. Kim and his co‐workers developed a “layer‐by‐layer” method (LBL) to synthesize BaTiO 3 nanoparticles (BTONP) based piezoelectric thin film. A multilayer nanocomposite was prepared by oleic acid (OA) ligands stabilized BaTiO 3 nanoparticles (>20 nm in diameter) and polymers which was functionalized by carboxylic acid (COOH), such as poly(acrylic acid) (PAA). The high affinity between the BTONPs and the ‐COOH groups helped the organizztion of the OA‐BTONP/PAA nanocomposite. By altering the bilayer number, OA‐BTONP size or inserted polymer type, the performance of this piezoelectric and ferroelectric thin films can be precisely controlled. Assembled by LBL assembling method in nonpolar solvent media, the quantity of adsorbed OA‐BTONP could be effectively increased and significantly increased the output power of PENG. Moreover, a bio‐inspired flexible PENG based on BaTiO 3 was presented in 2013. 46 In this work, anisotropic BaTiO 3 nanocrystals were deposited on a viral template via the metal ion precursors self‐assembling. The filamentous virus provided template for the deposition of an anisotropic, entangled, highly crystalline BaTiO 3 nanostructures. This flexible nanogenerator based on virus‐enabled piezoelectric structure can generate electricity of ≈6 V and ≈300 nA without additional structural stabilizers, indicating the high output related to the importance structure in nanoscale. This method of using bio‐template is facile and with particularly enlightening meanings for flexible PENG designing and applications. 3. 2. 3 PVDF‐Based PENGs PVDF and its copolymer P(VDF‐TrFE) are polymeric piezoelectric material, with high piezoelectric coefficient (d33 = 32. 5 pC/N). Its flexibility, transparency, adequate mechanical strength, and high chemical resistance present many advantages in piezoelectric relative applications. Chemical stability and biocompatibility of polymers is beneficial to the biomedical application for in vivo biological sensors and energy harvester. However, a disadvantage should be noticed that to achieve a good performance the aligning of mechanical stretching with the dipoles of β‐phase PVDF is required. There have been proposed various strategies to fabricate flexible PENGs based on PVDF and its co‐polymers. Chang et al. place piezoelectric PVDF nanofibers on easily acquired substrates to direct‐write thin film device by utilizing a near‐field electrospinning (NFES) and then processed with electrical poling. 47 Under mechanical stretching, the PENG has shown consistent and repeatable electrical outputs voltage of 30 mV and current of 3 nA, respectively. The as synthesized nanofibers were arranged in either serial or parallel connections to form thin film like structure in order to augment the total output. In another case, Hansen et al. used electrospinning technique to synthesize and pattern an aligned nanofibers arrays with two electrodes, following with an in‐plane poling process. 48 The PVDF nanofibers were oriented by packaged in PDMS to form a flexible thin film device. As the deforming of the device under alternating compressive or tensile force, it can produce output up to 20 mV and 0. 3 nA respectively. Sun et al. reported a method to convert the energy from low‐speed airflow to electricity by resonant oscillation of piezoelectric PVDF microbelts. 49 The electrical energy generated by this PVDF based PENG from low speed airflow was sufficient to operation low power consumption electronic devices, demonstrating its capabilities for the energy harvesting between inhalation and exhalation process. In order to enhance the output performance, PVDF films with well controlled nanostructures were desired. Nanoporous PVDF based PENG was developed to address this issue. 50 Cha et al. presented a porous PENG based on PVDF which was fabricated by a ZnO nanoparticle‐assisted preparation method. This nanogenerator with porous PVDF thin film produced an enhanced piezoelectric energy with power density of 0. 17 mW cm –3, which was 5. 2 times and 6 times increase of output voltage and current than the PVDF nanogenerators of bulk materials. In 2014, Mao et al. reported a large‐area sponge like PENG based on the porous PVDF thin‐film. [[qv: 49b]] This wafer‐scale porous PVDF thin films were fabricated by a simple casting‐etching process (Figure 6 E) and can be directly integrated into an electronic device, such as a cell phone, to convert environmental energy to electricity. Thier method provide a promising technique for integrating self‐powered electronics. Recently, Cho et al. reported a high performance P(VDF‐TrFE) based PENG through a surface morphology engineering using solvent annealing method for simple and cost‐effective fabrication at room temperature (Figure 6 D). This surface morphology engineered PENG presented 8 times enhanced output voltage and current because of well‐aligned electrical dipoles. 51 In addition to these aforementioned materials, many piezoelectric materials, such as GaN, 52 PMT‐NT, 53 2D‐MoS 2 54 and various of their composites, 55 are possessing special properties to be applied in various flexible devices. This technique of PENG is scalable and integratable to fabricate practicable and advanced biomedical electronics for healthcare. 3. 3 Triboelectric Devices 3. 3. 1 Materials The material properties including friction, work function, electron affinity and so on, play important roles in TENGs' output performance. Basically, nearly all materials can exhibit triboelectricity. So, pairing the right materials can achieve output in maximum. The triboelectric series is a guidance, 17 in which the capability of a material to gain or lose an electrons is shown as a qualitative indication. As we know, almost all kinds of materials, such as metal, polymer, silk and wood, have triboelectrification effect. Thus, they are all potential slections for producing TENGs. In consideration of some other aspects, such as mechanical properties, stability and biocompatibility, a variety of commercially purchased organic materials (e. g. PTFE, PET, PI, PDMS, PMMA) and inorganic materials (e. g. ITO, Al, Cu, Au, Ti, TiO 2, Si) have been utilized for fabricating TENG and showed outstanding performance. 16, 19, 56 Recently, various of advanced materials also have been studied, including graphene, 57 carbon nanotube, 58 paper, 59 nano‐Ag ink, 60 degradable polymers etc 61 ( Figure 7 ). Figure 7 Device structure design and material selection of flexible TENGs. A) The typical arch shaped TENG. Reproduced with permission. [[qv: 22b]] Copyright 2012, American Chemical Society. B) Stacked arch‐shaped TENGs. Reproduced with permission. 67 Copyright 2013, Elsevier. C) Zigzag TENGs. Reproduced with permission. 68 Copyright 2013, American Chemical Society. D, E) Stretchable TENGs. Reproduced with permissions. 72, 73 Copyright 2016, the American Association for the Advancement of Science and the American Chemical Society. F) Fiber shaped TENG. Reproduced with permission. [[qv: 73b]] Copyright 2016, The Nature Publishing Group. G) Transparent polymer based TENG. Reproduced with permission. 63 Copyright 2012, the American Chemical Society. H) Graphene based TENG. Reproduced with permission. [[qv: 56a]] I) CNT based TENG. Reproduced with permission. 58 J) Bio‐degradable TENG. TENG based on carbon‐based materials have been reported recently (graphene, carbon nanotube, etc. ). The conductivity, stretchability, triboelectric properties, friction and their roughness in nano‐scale might facilitate the application in TENG. 62 The graphene‐based TENG (G‐TENG) was reported by Kim et al. [[qv: 56a]] By transfered layer by layer, different number of layers of graphene (from 1–4 layers) were stacked for fabricating the TENG (Figure 7 H). It was revealed that the output voltage and current was decreasing with the increasing of graphene layers. For example, the current density were 500, 250, 160, and 100 nA cm −2 corresponding to the layers of graphene from 1, 2, 3, and 4 in the G‐TENGs. The different electronic interactions between stacked graphene layers were interpreted as the main reason. A roll‐to‐roll method was applied to produce the G‐TENG by Liu et al. to improve the fabrication proces. [[qv: 56b]] A CNT‐based TENG was developed by Bao et al. (Figure 7 I), which could generate tens of volts and tenths to several µA cm −2 of output voltage and current density. 58 As for some environment and human friendly considerations, biodegradable materials were also employed for TENG construction. Kim et al. applied silk fibroin film to build a biocompatibility and eco‐friendly bio‐TENG. The output performance of bio‐TENG was 16 V for voltage and 2. 5 µA for current. [[qv: 60b]] Zheng et al. tested the triboelectrical properties of various degradable polymers and based on this result they fabricated fully biodegradable TENG (BD‐TENG) (Figure 7 J). The BD‐TENG was implanted in sub dermal region for biomechanical energy harvesting. After a period of implantation, the BD‐TENG can be dissolved in body leaving without any residue. Paper was also involved in the fabrication of TENG for its popular, cheap, lightweight, disposable, and environmentally friendly. [[qv: 60a]] Structure change will lead to an improvement of the output. Fan et al. studied the influence of varies micro‐fabricated surface structures of PDMS on TENG's output, and found that the TENG with pyramid shaped surface structure can produced highest output compared with that with cubic or liner shaped surface structure. 63 Recently, Zhao et al. developed a very simple and rapid material surface processing method which can efficiently create liner structure on many friction films and result in an increased electrical output. 64 Besides, surface charging and chemical modification for the friction material have also been proposed by researchers to not only improve the output performance of TENGs, but also increase their stability in various environmental conditions. 65 3. 3. 2 Device Structure Design Biomechanical energy is small scale, randomness, and multi‐directional. Therefore, TENGs used for biomechanical energy harvesting usually need to be flexible for shape‐adaptive energy harvesting. In 2012, an all‐polymer based flexible TENG with characteristics of simple, cost‐effective for mechanical energy harvesting was firstly demonstrated, through a coupling of triboelectrification and electrostatic induction. 63 The TENG consisted of two different polymer films with metal electrodes deposited on their back sides. An output voltage and a power density of 3. 3 V and 10. 4 µW cm −3, respectively, was generated by this typical TENG. This most basic working mode of TENG was later defined as vertical contact‐separation mode (Figure 3 A). Since the first demonstration of TENG, the arch shaped structure has become one of the most popular structure due to its simple fabrication, high performance, and universal feasibility. [[qv: 22b, 65]] In this structure, TENG with an arch‐shaped gap between a polymer film and a metal foil was fabricated (Figure 7 A). In a typical demonstration, the TENG with arch‐shape structure had the V oc and I sc about 230 V and 94 µA, respectively. The energy conversion efficiency reached up to 39%. Inspired by this structure and its working mechanism, various derived structures are proposed. For example, arch‐shaped and anti‐arch‐shaped TENGs were consisted to alternatively‐stacked structure and electrically connected in a parallel configuration, which has been proved that can significantly enhance the output performance (Figure 7 B). 67 The zigzag structure was another variation of the stacked arch‐shaped structure (Figure 7 C). 68 A device with a 6‐layered construction can generate a current of 656 µA. TENGs with spacer layer is another deformation structure of arch shape. Instead of utilizing the bending of the material itself to form an arch shape gap, a spacer was introduced between two friction layers, which brought a stable gap to TENG. The gap width was determined by the thickness of spacer layer, so its controllability and durability were improved a lot. The existence of spacer facilitated the versatile design of TENG based systems, because shape and the thickness of TENG can be easily controlled. [[qv: 22a, 68]] The lateral sliding mode in plane was designed as another mode of the flexible TENGs (Figure 3 B). Two films with complementary micro‐sized linear grating arrays were fabricated to a thin‐film‐based micrograting triboelectric nanogenerator (MG‐TENG), which can convert the mechanical energy into electricity from the motion of relative sliding between two micrograting films. 70 The output current and the output power of the device were up to 10 mA and 3 W, respectively, with the conversion efficiency of ≈50%. In order to harvest mechanical energy in differernt directions, a TENG with a sandwiched structure and checker‐like electrodes was designed by Xi et al. 71 This TENG generated a voltage of 210 V and showed great performance in either sliding directions. Furthermore, they also demonstrated to light up LEDs by the mouse operation energy when integrated the sliding mode TENG into a mouse pad or sliding panel. Considering the rapid development of deformable and stretchable electronics, a power source structural and functional suitable for this class of electronics was needed. Yi et al. reported a method for TENGs and highly deformable and stretchable self‐powered sensors. 72 A shape‐adaptive triboelectric nanogenerator (saTENG) unit with conductive liquid as electrodes can be used to harvest various mechanical energy effectively in different motions (Figure 7 D). The saTENG can sustain a 300% strain in maximum and be adapted to almost all of the curvilinear and three‐dimensional surface, beacause of its admirable flexibility. It increased the possibility of employing the saTENG to be wearable energy sources or self‐powered device for sensoring human body motion. A stretchable supercapacitors was combined with a stretchable TENG to develop a stretchable, soft and fully packaged self‐charging power system by the same group. 73 Because of the fully soft structure, this system can subject to large‐degree deformation and convert energy from various motion, including human body movement (Figure 7 E). This self‐charging power system shows wide applications to harvest all kinds of biomechanical energy and fabricate personal wearable self‐powered electronics. Fiber‐based triboelectric generator (FB‐TENG) have the advantage of harvesting the biomechanical energy from human motion in three‐dimensional. 74 A low cost and facile FB‐TENG was introduced by Zhou et al. [[qv: 73a]] This wearable TENG is capable of scavenging energy from human motions and body vibration to electric power by using carbon nanotubes, polytetrafluoroethylene (PTFE) aqueous suspension and cotton threads. The power density can reach up to ≈0. 1 µW cm −2. Recently, Chen et al. have reported a micro‐cable power textile for simultaneously harvesting energy from ambient sunshine and mechanical movement (Figure 7 F). [[qv: 73b]] Solar cells fabricated from lightweight polymer fibres into micro cables were then woven via a shuttle‐flying process with fibre‐based TENG to create a smart fabric. A single layer of such fabric was 320 µm thick and can be integrated into various cloths, curtains, tents and so on. This hybrid power textile, fabricated with a size of 4 cm by 5 cm, was demonstrated to charge a 2 mF commercial capacitor up to 2 V in 1 min under ambient sunlight in the presence of mechanical excitation, such as human motion and wind blowing. The textile could continuously power an electronic watch, directly charge a cell phone and drive water splitting reactions. Wang et al. demonstrated a Fiber based TENG for continuously power wearable electronics only by human motion, realized through with optimized materials and structural design. Fabricated by elastomeric materials and a helix inner electrode sticking on a tube with the dielectric layer and outer electrode, the TENG has desirable features including flexibility, stretchability, isotropy, weavability, water‐resistance and a high surface charge density of 250 mC m –2. With only the energy extracted from walking or jogging by the TENG that is built in outsoles, wearable electronics such as an electronic watch and fitness tracker can be immediately and continuously powered. [[qv: 73c]] 4 Recent Progress in Biomedical Applications 4. 1 Power Source for Biomedical System The recent development of biomedical systems, especially the microelectronic devices for healthcare and other medical applications illustrates the potential of such a futuristic concept. 75 Some possible applications are exciting and extensive, but they still need reliable and safe energy sources to operate. Considering the application in biology and medicine, batteries can be dangerous, bulky, and difficult to change. For example, replacing a pacemaker battery is alway problematic for causing additional pain and cost. However, great progress has been made in energy harvesters over the past decade. In human body, where there is little light and there are only small thermal gradients, mechanical energy harvesting is an ideal method to provide power for implantable medical devices. Biomedical systems are consuming less and less power, and new harvesting technologies like PENGs and TENGs have the potential to supply the power needed for the safe operation of medical devices ( Figure 8 ). Figure 8 In vivo energy harvesting by PENGs and TENGs. A ) The first demonstration of in vivo biomechanical‐energy harvesting using a single nanowire based PENG. Reproduced with permission. 76 B) Conformal energy harvesting from heart/lung by PZT based PENG. Reproduced with permission. 31 Copyright 2014, National Academy of Sciences. C) Flexible PVDF based PENG for harvesting energy from ascending aorta. Reproduced with permission. 78 Copyright 2015, Elsevier. D) The first demonstration of implantable TENG for harvesting biomechanical energy. [[qv: 68b]] E) Wireless cardiac monitoring system powered by iTENG. 82 Harvesting mechanical energy using PENG has been demonstrated some time ago. The continuous optimization in flexibility and output performance of the piezo‐devices made them possible for biomedical applications. In 2009, Yang et al. converted biomechanical energy from muscle movement into electricity by a ZnO based PENG. [[qv: 27b]] This device was a single nanowire generator that consisted of a lateral ZnO nanowire with two ends affixed on a flexible substrate. Muscular movement drove the back and forth deformation of the whole device. The electrons flowing in the external circuit was launched by the piezoelectric potential generated generated inside the ZnO nanowires. By integrating multiple nanowires, the output V oc can be increased and four ZnO nanowires in a series connection can produced an output voltage of 0. 15 V. Scavenging low‐frequency energy from a running hamster or a tapping finger was successfully demonstrated revealing the potential application for harvesting regular and irregular biomotion by nanogenerators. Based on the same mechanism, Li et al. developed an implantable PENG based on single ZnO nanowire to harvest energy in breath and heartbeat in a living rat. 76 This is the first demonstration of harvesting in vivo biomechanical‐energy by an PENG (Figure 8 A). Typically, the average magnitude of the in vivo voltage and current signals were about 1 mV and 1 pA, respectively. Besides energy harvesting, some crucial results for in vivo application were also addressed in this work, such as the biocompatibility, toxicity, encapsulation, anchoring and the potential for active sensing of in vivo biological motions. This study not only shows the potential of scavenging in vivo mechanical energy by PENG, but also starts a new trend for pursuing sustainable power source of implantable medical devices. Although such a model highlights the potential for self‐powered biomedical devices, there is still an important practical challenge in the efficient output power. Dagdeviren et al. utilized a thin‐film PZT nano‐ribbon based PENG for conformal energy harvesting and storage from motions of the diaphragm, lung, and heart. Figure 8 B presents schematic diagrams of the flexible PZT thin‐film PENG. 31 A bridge rectifier and a micro‐battery can be integrated with the flexible PZT mechanical harvester on a flexible substrate. The device was further packaged by biocompatible encapsulation layer to harvest energy directly from the movement of bovine heart and lung in vivo. During the cardiac contraction and relaxation, the conformal contact of the whole device with the heart without any destruction was significant importance for in vivo applications of energy harvester. The output V oc was up to 5 V and greater by three orders of magnitude than previous in vivo results, when the in vivo PENG was attached on the right ventricle (RV) of a bovine heart. In this work, simultaneous energy harvesting and storage were achieved by fully integration of PENG, rectifiers and micro‐batteries. Thier research indicates strong feasibility of achieving self‐powered implantable biomedical devices. This work also provides an interesting demonstration of five independent PZT harvesters connected serially in a multilayer‐stacked flexible PENG. This multilayered flexible PENG could generate an 8. 1 V of voltage and 1. 2 µW cm −2 of power density, which were much higher than the output of the single‐layer device. In 2015, a similar work was presented by Lu et al. 77 They developed an ultra‐flexible PZT based energy harvester, which had the similar structure of above‐mentioned device. It can harvest the mechanical energy derived from cardiac motions by integrated with the heart. For simulating the scenario of applying mechanical energy harvester in body, in vivo tests were performed in differernt settings of the experimental animals, for example opened/close chest, awake and under anesthesia. When the PENG was fixed from left ventricular apex to right ventricle, this ultra‐flexible piezoelectric device could generate a 3V voltage, which demonstrated the potential applications of harvesting the in vivo mechanical energy from cardiovascular system and body motion for powering the implantable medical devices and sensors sustainably. Apart from the biomechanical energy that can be easily through of, such as human motion, heartbeat and breath, some special tissues or organs also contains a wealth of mechanical energy. However, harvesting this kind of mechanical energy needs more sophisticated design of devices. Zhang et al. reported a PVDF based flexible and implantable PENG, which wrapped around the ascending aorta and scavenged the energy from the pulsation. 78 This research reveals the in vitro and in vivo studies of harvesting the energy in artery pressure with a flexible and implantable PENG (Figure 8 C). In the in vitro study, the max output power (P max ), voltage (V max ) and current (I max ) and of the PENG were 681 nW, 10. 3 V and 400 nA respectively. The quantity of electric charging by one pulse was about 7–9 nC. When implanted the PENG in vivo by wrapping around the ascending aorta of a porcine, the V max and V max of the implanted PENG were 1. 5 V and 300 nA under the heart rate of 120 bpm and the blood pressure of 160/105 mm Hg. The instantaneous output power could reach up to 30 nW with a long‐lasting duration of 700 ms and 77. 8% duty ratio. The implanted PENG could charge for a 1 µF capacitor to 1. 0 V within 40 s. Currently, in order to improve the efficiency of energy conversion, TENGs were introduced for biomechanical energy harvesting. Since 2012, a series of work reported that energy from human motion has been successfully collect under in vitro conditions by TENG, such as walking, running, finger tapping or elbow bending, which demonstrated its potential for biomedical applications. 79 In 2014, Zheng et al. first demonstrated the in vivo application of TENG. [[qv: 68b]] In this work, an implantable triboelectric nanogenerator (iTENG) was implanted in a living rat to harvest energy from its periodic breathing (Figure 8 D). The iTENG was fully encapsulated to protect its inner structure from physiological environment. The implantation site was near the chest, which allowed the iTENG to convert the mechanical energy from the normal inhalation and exhalation of rats into electricity and used to power a prototype pacemaker. Through theoretically calculation, the energy harvested during 5 breaths can generate a single pulse to stimulate the rat's heart. This was a significant progress of charging an implanted medical devices by a TENG. Shi et al. reported a packaged self‐powered unit based on piezoelectric and triboelectric hybridized nanogenerators (PTNG). 80 The PTNG was fabricated with a BaTiO 3 @PDMS film as a contact layer based onvertical contact‐separation working mode. The polarized BaTiO 3 @PDMS film was bended to generate a piezoelectric potential and contacted with aluminum foil to generate a triboelectric effect simultaneously. Both of the piezoelectric potential and triboelectric effect was coupled and increased the output performance of the PTNG. The power density of the PTNG was increased by 80% and reach up to 97. 41 mW m –2, compared with the TENG in control. The PTNG was further integrated with several power management components to form a packaged self‐powered system. The self‐powered system was utilized to harvest and store the biomechanical energy, and then to successfully power a temperature senser that was implanted in vivo. It has to be noted that, the entire packaged self‐powered system was a viable “Plug and Play” mobile energy source by designing the universal connections of plug and socketfor the system. Tang et al. demonstrated that flexible TENGs attached to human articulations could generate enough electricity power a medical laser. [[qv: 78c]] The aim was to use this laser to accelerate bone remodeling treatments. Such a flexible TENG does not have the same ability to power a device over time compared to a battery, because the specifications and behaviors of TENGs are different: batteries deliver continuous power until they are exhausted, while flexible TENGs deliver a pulse of power every time they are activated. Producing a continuous light emission during 60s when powered by a battery was equal to a series of 100 light pulses when powered by a flexible TENG. Tang et al. showed that even if the TENG powered laser was a little less efficient than the battery‐powered laser, both treatments can successfully accelerate bone remodeling. This finding is important because it proves that TENGs with a small energy buffer such as a capacitance can be sufficient to power some small biomedical devices. Recently, Zheng et al. developed a reliable encapsulation method of TENG, which provide the TENG outstanding stability in harsh environment. 81 Base on this technology, the same group designed a novel iTENG with significantly improved in vivo output and reliability. 82 A ‘ Keel structure ’ was introduced to facilitate the contact and separation process of TENG in complicated environment of living body (Figure 8 E). Driven by the heartbeat of an adult Yorkshire porcine, the V oc and I sc can reach up to 14 V and 5 µA, which were improved by factors of 3. 5 and 25, respectively, compared with the reported in vivo output performance of PENG and TENG. The in vivo performance of the iTENG was evaluated for over 72 h of implantation, revealing the outstanding stability of iTENG for generating electricity continuously in a living animal. The iTENG was then applied to power a wireless transmission system (SWTS) and the in vivo related signals of heartbeat was successfully transmitted, showing the feasibility of iTENG for powering the electric medical sensors in mobile and real‐time wireless monitoring. 4. 2 Active Pressure/Strain Sensor The human‐body activity induced pressure is distributed in a large range from a low‐pressure regime to a high‐pressure regime (10–100 kPa). 83 Physiological activity in different positions of human body will generate differeent pressure signals. For instance, intraocular pressure and intracranial pressure can belong to the low pressure regime. The medium‐pressure regime is mainly generated by the respiration motion, heart beating, blood pressure wave, jugular venous pulse, radial artery waves and phonation vibration. The body weight brings the high pressure regime under the feet. The monitoring of these pressures by suitable pressure sensors is important for the application in diagnostics of heart failure, cardiovascular disease, respiratory disorders and damaged vocal cords. It was also crucial for monitoring the sleep apneahyponea syndrome, sports injuries in athletes and high‐risk diabetic foot ulceration. For developing high sensitive, durable, stretchable and flexible pressure sensors, piezoelectric and triboelectric materials and devices have been paid much attention to, recently. According to the structures and materials of the active sensors, each of them has its own advantages in sensing mechanisms. 4. 2. 1 Active Pressure Sensors PENG based pressure sensors is fabricated and the generated electric charge is proportional to the applied pressure when a piezoelectric material is stressed. For the development of a flexible piezoelectric pressure sensor, PVDF and its co‐polymer P(VDF‐TrFE) are ideal candidate materials for their flexibility and the ease of the fabrication. 84 Although inorganic materials have inferior flexibility comparing with organic materials, various inorganic materials also exhibit distinct mechanic flexibility when processed into ultrathin films or nanowires. Therefore, ultrathin films or nanowires based on piezoelectric inorganic materials have been attracted much attention and researched comprehensively for flexible and stretchable pressure sensers. A PENG based pressure sensor was reported by Zirkl et al. and Graz et al. , which was integrated with an amplifying element of transistor for reading the signal from the sensor. [[qv: 83b, c]] The sensing element was a PENG based on P(VDF‐TrFE) or P(PVDF‐TrFE)/PbTiO 3 nanocomposite. When applied a pressure on the PENG integrated sensor, it responsed a voltage signal. This voltage signal from PENG controlled the magnitude of current in the integeated transistor and swiched on/off the gate of the transistor. However, some limitations of this pressure sensors with an amplifying element of transistor have to be noted here. The fabrication process is more complicated and the energy consumption is high, which are caused by the additional inner connections of PENG and transistor element. Therefore, another approach of employing a piezoelectric polymer direct integrated into a transistor as a gate dielectric layer is demonstrated in presure sensors. Many groups, such as Tien and co‐workers, [[qv: 83a, 84]] Kim et al. , [[qv: 83d]] and Trung et al. , 86 directly coupled a P(VDF‐TrFE) and P(VDF‐TrFE)/BaTiO 3 nanocomposite as gate dielectric layers with an organic field‐effect transistor(OFET) ( Figure 9 C). The equivalent voltage or remnant polarization inside the piezoelectric polymer was changed by the direct piezoelectric effect under external force. Meanwhile, the applied external force also led to a change of source‐drain read‐out current by modulating the intensity of the charge that accumulated between the semiconductor/dielectric interface. Figure 9 Active pressure/strain sensor based on PENG and TENG. A) Conformal modulus sensor based on PZT. Reproduced with permission. 90 Copyright 2015, Nature Publishing Group. B) Active‐matrix strain sensor based on a PENG‐powered graphene transistor. Reproduced with permission. 96 C) Structure of an OFET composed of microstructured P(VDF‐TrFE). Reproduced with permission. [[qv: 83a]] D) Implantable active sensor based on TENG. Reproduced with permission. 92 Copyright 2016, American Chemistry Society. In addition, Persano et al. introduced a high‐performance pressure sensor, which was fabricated by P(VDF‐TrFE) nanofibers. 87 By the electrospinning method, a large‐area free‐standing P(VDF‐TrFE) fibers was achieved in a flexible PENG. This as fabricated PENG based sensor showed exceptional piezoelectric characteristics with a high sensitivity to pressure, which could monitor small pressure about 0. 1 Pa in minimum. These progress displayed the wide potential application of this active pressure sensor from self‐powered micro‐mechanical elements to sensitive impact bio‐detectors. Cheng et al developed a thin film PENG based on PVDF for implantable and self‐powered monitoring of blood pressure. 10 An excellent linear relationship was achieved between the output voltage of device and blood pressure, with a high‐sensitivity of 173 mV/mm Hg, which was significantly improved than reported results. 88 The device also showed excellent stability for over 50, 000 operating cycles. An in vivo experiment in adult Yorkshire porcine showed a sensitivity of 14. 32 mV mm –1 Hg. Based on these characteristics of the device, they established an implantable, self‐powered and visualized blood pressure monitoring system. A PZT conformal pressure sensor was introduced by Dagdeviren et al. 89 A sheet structure of PZT array was connected with the gate electrodes of a metal‐oxide‐semiconductor field‐effect Transistor (MOSFET) to form a stretchable, lightweight and ultrathin devices. This device showed a fast response time of less than 0. 1 ms, a high detection resolution with detection limit of 0. 005 Pa, high stability and high sensitivity. After stretched at 30%, there was no impact to the performance of the sensor. This sensor was also softly laminated on the wrist, neck and throat to monitor human blood pressure. A conformal modulus sensor (CMS) with ultrathin, stretchable networks of sensors and mechanical actuators based on PZT was introduced by the same group as well (Figure 9 A). 90 The sophisticated sensor demonstrated to measure the viscoelasticity of the epidermis in the surface regions. These conformal avctive sensor systems can be compactly attached with soft tissues, organs, and various biological substrats of live creatures, which facilitate their biomedical applications. Since the first proof‐of‐concept demonstration, the TENGs have been proved to be one of the most convenient way to demonstrate the pressure sensing capabilities. Such self‐powered active sensor can detect gentle external mechanical force like a 20 mg falling feather with 0. 4 Pa in contact pressure and a 8 mg water droplet with 3. 6 Pa in contact pressure through relative deformation between two sheets of polymer by pressure. For biomedical pressure sensing, the devices should be deformed to adapt the complex shapes of human body and even soft biological tissues and organ systems. Yang et al. assembled serpentine‐patterned electrodes and a wavy‐structured polymer film together to develope a TENG with new stracture. 91 Owing to the unique design, this flexible TENG (FTENG) could be operated at both stretching and compressive movement, which was ideal for strain monitoring. A high‐output power density of 5 W m −2 was delivered under the traditional compressive mode. The FTENG could sustain 22% tensile strain under stretching and generate almost 70 times of power output comparing with the TENG of planar structure. For the superior feature of the reliable output performance on curved surfaces, the FTENG could be employed as pressure sensor to detect tiny motions of muscles, joints and prominentia laryngea with close fitting to human skin. This work presents a progress of self‐powered pressure sensing and orients to the new generation of bio‐integrated systems. Very recently, Ma et al. proposed a self‐powered and multifunctional implantable triboelectric active sensor (iTEAS) based on implantable TENG, which can provide accurate, continuous, and real time monitoring of multiple physiological and pathological signals. 92 As demonstrated in large‐scale animals, the device can monitor heart rates, reaching an accuracy of ≈99%. So, cardiac arrhythmias such as atrial fibrillation and ventricular premature contraction were successfully detected in real‐time (Figure 9 D). Furthermore, this device also provided a novel method of monitoring respiratory rates and phases by analyzing variations of the its output peaks. Blood pressure can also be independently estimated and the velocity of blood flow calculated with the aid of a separate arterial pressure catheter. The in vivo biocompatibility of the device was examined after 2 weeks of implantation, proving suitability for practical use. As a multifunctional biomedical monitor that is exempt from needing an external power supply, the proposed iTEAS holds great potential in the future of the healthcare industry. 4. 2. 2 Active Strain Sensors A strain sensor or strain gauge isused to measur the deformation of objects. Usually, a patterned metal foil on a flexible substrate constitutes a typical strain sensor for monitoring deformation on an object. 93 Recently, according to the huge requirements of electric skin and variouse motion monitoring for rehabilitation assistance and developing diagnosis, the interest of developing flexible and stretchable strain sensors integrated on cloth or human body was increased. Body motion monitoring of human is of great importance and can be sorted out to differernt categories: large‐scale and small‐scale motions. The large‐scale motions is the movements of limbs and the small‐scale motions is the subtle movements, such as speaking, swallowing and breathing. 94 These movements are crucial and indicating the conditions, damages and diseases of human body. For example, applying strain sensors to diagnose respiratory disorders, damaged vocal cords and monitoring Parkinson's disease. Piezoelectric strain sensors rely on PENGs that made of piezoelectric materials, which can convert mechanical energy into electricity. And they have the advantage of low power consumption, high sensitivity and ultrafast response. 95 Recently, P(VDF‐TrFE) polymer material and Zinc oxide naowires have been used in developing piezoelectric strain sensors. 95, 96, 97 An strain sensor with active‐matrix was reported by Sun et al. , which was based on a PENG‐powered coplanar‐gate graphene transistor. 96 Due to the external strain, the P(VDF‐TrFE) generated a piezopotential, which was coupled to the channels of graphene transistor by dielectric ion gel effectively (Figure 9 B). The piezopotential modulated the channel conductance of the graphene transistor. This strain sensors showed ultrasensitivity and high flexibility to detect ≈0. 008% of minimum detectable strain, which were promising for human‐activity monitoring. ZnO nanowire is semiconductive and piezoelectric material, which is also used to fabricate strain sensors. Under applied tensile strain, a piezoelectric potential and charges were generated between two ends of the ZnO nanowire. The applied external strain can control the height of the potential barrier and adjust the transport behavior of the charge carriers. 98 A single ZnO piezoelectric fine wire (PFW) was fabricated to build a flexible strain sensor by Zhou et al. [[qv: 96e]] This piezotronic strain sensor showed great stability, high sensitivity, and very short response time. The flexible ZnO PFW strain sensor is of great potential to be applied in measuring cell mechanical stress and strain. In addition, a single ZnSnO 3 nanowire was developed to a strain sensor with high‐sensitivity by Wu et al. [[qv: 30b]] The sensitivity of a ZnSnO 3 NW‐based strain sensor was 19 times highter than Si devices and 3 times higher than those of ZnO and CNT nanowires. However, the piezotronic nanowires of ZnO or ZnSnO 3 are usually attached on the surface or inset the substrat of strain sensors, which brings some disadvantages of detecting strains at discrete points. Strain sensors were developed based on a flexible substrate with in situ synthesised ZnO nanowires. Vertically aligned ZnO nanowire arrays were fabricated to a strain sensor reported by Zhang et al. , which had higher sensitivity compared with the single ZnO nanowire based strain sensor device. [[qv: 96b]] Moreover, another flexible strain sensor was developed on carbon fiber textile with a ZnO nanowire film grown. 99 Gullapalli et al. demonstrated a piezoelectric strain sensor based on paper matrix with ZnO nanostructures embedded in, for improving the flexibility of entire device. 95 This strain sensor presented quite low power consumption and great strain sensitivity. Generally, piezoelectric strain sensors based on nanowire materials showed high sensitivity, fast response, and low power consumption. However, they still have some limitations, especially in flexibility, stretchability, and detection range. These disadvantages limite the applications of piezoelectric strain sensors in measuring curved objects or the irregular surface on human body for monitoring human‐activity, healthcare devices and wearable electronics. A conformal piezoelectric strain sensor with better flexiblity and stretchability is crucial for biomedical applications. 4. 3 Direct Stimulation of Living Cell, Tissue and Organs Some illnesses such as chronic pain, abnormal heart rate, and Parkinson's syndromes can be eased or cured by stimulating the spinal cord, heart, and the brain at an in vivo state using electrical pulse. 100 Most biomedical stimulation devices consumed their embedded battery power providing such functional electrical stimulations for particular muscles and nerves generally. 101 To solve the energy limitation of electrical medical devices that the flexible PENGs and TENGs generate electric power and supply a functional electrical puls to stimulate muscles and nerves is of prime importance. [[qv: 42a]] Lately, Hwang et al. proposed a method to use a flexible PMN‐PT thin film nanogenerator to output momentary electric energy and stimulate the heart of a living rat directly ( Figure 10 A). 15 The PMN‐PT thin film PENG stimulated the rat heart with electric pulses and the electrocardiogram (ECG) was simultaneously recorded by attached sensing electrods. Figure 10 A showed the artificial heartbeat generated with the assistance of PENG in an animal experiment with chest laparotomy. Before stimulating, the the ECG signals showed typical P, T waves, and regular QRS complex with a heart rate of 6 beats per second. A few micro joule of electric power was required for triggering the action potential for artificially stimulating the living heart. 102 When the flexible PENG was bent and released periodically, corresponding electrical signals were recorded in the ECG with the cardiac bioelectricity of rat. The electric energy generated by the flexible PENG was 2. 7 µJ and capable of exciting the artificial cardiac action potential of the rat. Figure 10 Direct stimulation on cell, tissue and organ by PENG and TENG. A) A schematic of the experimental setup for artificial cardiac pacemaking using the electric output from the flexible PMN‐PT thin‐film PENG. Reproduced with permission. 15 B) Self‐powered deep brain stimulator using the electric output from thin‐film PENG. Reproduced with permission. 103 Copyright 2015, Royal Society of Chemistry. C) Electrical stimulation for neuron orientation based on BD‐TENG. D) A self‐powered neural differentiation system with a step‐driven TENG as the electrical simulation power source. Reproduced with permission. 106 Copyright 2016, American Chemistry Society. The same group also presented a high‐performance flexible PENG enabled by a single crystalline thin film of Pb(In 1/2 Nb 1/2 )O 3 –Pb(Mg 1/3 Nb 2/3 )O 3 –PbTiO 3 (PIMNT) on a plastic substrate and used as a self‐powered deep brain stimulator (DBS) in vivo. 103 A modified Bridgman method with a stress‐controlled nickel exfoliation process was used to grow and transfer thin film of PIMNT onto a flexible substrate, without any mechanical damages. This PENG stimulator can generate a V oc and an I sc of 11 V and 285 mA, respectively, when bent periodically on a linear stage. These results matched with load impedance for the electrode of practical DBS and were adequate to stimulate nerve nuclei. A maximum output current of 0. 57 mA was measured when fixed and bended by a human finger. Meanwhile, the PIMNT thin film PENG could light on 120 green LEDs and easily charge capacitors for the output power was about 0. 7 mW. A demonstration of control body motions in a live rat was verified by stimulating the primary motor (M1) cortex. When the PIMNT powered the real‐time DBS to activate M1 cortex with functional electric potential, the muscle contraction and the motion of the forelimb was induced (Figure 10 B). Various electric stimulation technologies were also applied in cell manipulation, which provided an exciting route for tissue engineering and many of them have been proved effective in research and clinical settings. 104, 105 Guo et al. combined a highly electrically conductive rGO−PEDOT hybrid scaffold with a step‐driven TENG as the electrical simulation power source to build a self‐powered neural differentiation system. 106 The electrical outputs of 250 V and 30 µA were achieved by driving the TENG with human step motion, which was sufficient to stimulate the MSC cells. The rGO−PEDOT hybrid microfiber can not only enhance the proliferation of MSCs but also function as a medium for step‐driven TENG pulse electrical simulation signals, which can induce MSCs to differentiate into neural cells. The study realized an enhancement of MSC neural differentiation on the rGO−PEDOT hybrid microfiber under TENG‐driven electrical pulse simulation. This work shows a significant potential application of a self‐powered TENG electrical stimulation system for the assistance of nerve regeneration (Figure 10 C). Zheng et al. reported a biodegradable triboelectric nanogenerator (BD‐TENG) for short‐term stimulation of neural cells in vivo. [[qv: 60a]] The V oc and I sc of BD‐TENG were ≈40 V and ≈1 mA, respectively. When two complementary micrograting electrodes were powered by BD‐TENG, an1 Hz and 10 V mm –1 DC‐pulsed electric field (EF) was generated. This DC‐pulsed EF oriented the growth direction of nerve cells cultured on the micrograting electrode successfully, which was crucial for neural repair (Figure 10 D). Several unique advantages were exhibited by the BD‐TENG here. When it was implanted in vivo, various kinds of biomechanical energy in respiratory motion, heartbeat, and pressure of blood vessels can be converted into electrical power. For example, a BD‐TENG is implanted under the skin of the left thorax of a rat, the inhalation and exhalation of the rat can result in an alternative expansion and contraction of the thorax, which deforme the BD‐TENG and make two fraction layers to contact and separate periodically. During this course, the electric potential caused by the contact electrification and electrostatic induction drives the electrons flowing back and forth in the external circuit along with the respiratory motion. So a continuous AC output is generated with the respiratory movement continues. The satisfactory biocompatibility and light‐weight, cost‐effective, and designable size will further promote the electric potential's in vivo application. The low frequency and the relatively small amplitudes are suitable for in vivo electrical stimulation. Furthermore, fully implantable stimulation or diagnostic devices could be fabricated if integrated with a specially designed electrode or wireless transmission component. The implanted devices can be left behind in the body even when the therapeutic or diagnostic process is completed. The whole device can be degraded and absorbed gradually without any residue. The BD‐TENG revealed many outstanding advantages and tremendous potential as aremarkable power source candidate for transient in vivo medical devices. 5 Conclusions and Outlook Biomedical systems and nanotechnologies are revolutionizing healthcare and medicine; their synergy could be extremely powerful, and they could play key roles in near‐term medical technologies. Taking into account the decreasing power consumption of microchips and the increasing efficiency of nanomaterials‐based mechanical energy harvesters such as PENGs and TENGs, which should be possible to power autonomous biomedical systems. Reaching energetic independence from current bulky batteries is a first important step for their development. However, to reach this purpose, the optimization of the output performance and power management of nanogenerators need to be addressed in order to increase the energy conversion rate and efficiency in use. Furthermore, to realize their full potential in the field of healthcare, nanogenerators need to continue to evolve more high flexibility, sensitivity, elasticity, stretchability, durability and biocompatibility, to be fully operational in the human body. Autonomous biomedical systems with self‐powered and active sensing properties will be the future development direction of the medical field. For easily using ex vivo and in vivo, they need to be wearable and implantable. In order to achieve this, they will need to be fully flexible to fit the shape of organs, including skin, closely. Thus, the devices will be more discrete and comfortable for the patients, and will be better adapted to their targeted tissue/organ, which will increase their sensing ability and the amount of energy harvested. Flexible electronics are currently used, flexible microelectronics are under development, 107 and flexible PENGs and TENGs are progressing. Therefore, flexible autonomous biomedical devices are possible, and the challenge will be the integration of all the parts into one small, flexible device. Then the device will need to be embedded into a flexible, biocompatible package to be patched on or implanted into a patient. Improving and facilitating this step is another important goal for biomedical systems. Fully biodegradable, high‐performance electronics and sensors, defined as transient electronics are another class of novel devices that under fast growing. They are fabricated with biodegradable organic/inorganic materials that are used widely in medical devices. The integration of these devices could expand their functional capabilities in medicine, especially in the cases that a medical device is just need to be temporarily implanted in human body, for example, some bio‐sensors, stimulators and drug delivery systems. Suitable power source that functionally and structurally compatible with transient electronics is of great importance. The newly appearanced BD‐TENG is a potential solution. Future works in this field will focus on the performance improvement, structural integration and optimization and intelligent control of their dynamic properties in vivo, such as the operating lifetimes and the absorption efficiency.
|
10. 1002/advs. 201700191
| 2,017
|
Advanced Science
|
Precise Protein Photolithography (P
|
Precise patterning of biomaterials has widespread applications, including drug release, degradable implants, tissue engineering, and regenerative medicine. Patterning of protein‐based microstructures using UV‐photolithography has been demonstrated using protein as the resist material. The Achilles heel of existing protein‐based biophotoresists is the inevitable wide molecular weight distribution during the protein extraction/regeneration process, hindering their practical uses in the semiconductor industry where reliability and repeatability are paramount. A wafer‐scale high resolution patterning of bio‐microstructures using well‐defined silk fibroin light chain as the resist material is presented showing unprecedent performances. The lithographic and etching performance of silk fibroin light chain resists are evaluated systematically and the underlying mechanisms are thoroughly discussed. The micropatterned silk structures are tested as cellular substrates for the successful spatial guidance of fetal neural stems cells seeded on the patterned substrates. The enhanced patterning resolution, the improved etch resistance, and the inherent biocompatibility of such protein‐based photoresist provide new opportunities in fabricating large scale biocompatible functional microstructures.
|
1 Introduction Precise patterning of micro and nanostructures using polymer‐based biomaterials has widespread applications including drug release, degradable implants, tissue engineering, and regenerative medicine. 1, 2, 3, 4 In this context, natural silk proteins obtained from cocoons of the silkworm Bombyx mori provide “green” alternatives to synthetic materials with advantages such as superior mechanical properties (strength and toughness), outstanding biocompatibility and biodegradability, and controllable water‐solubility and degradation rate. 1, 5, 6, 7 Natural silk fibers from B. mori cocoons exist in a self‐assembled fibrous configuration, in which a mechanically robust protein–fibroin (≈75%, w/w) comprises the core, surrounded by a glue protein–sericin (≈25%, w/w). 8 To date, several techniques, such as Electron beam (E‐beam) writing, 9, 10 imprinting, 11, 12 molding, 13, 14 electrospinning, 15 embossing, 16, 17 inkjet printing, 18 and photolithography (P 3 ) 19, 20 have enabled the development of a variety of material formats, including hydrogels, fibers, particles, and films using fibroin or sericin as well as their blends with other materials. 3, 21, 22 Photolithography, in particular, remains one of the most appealing techniques for scalable biomanufacturing as it is Complementary Metal Oxice Semiconductor (CMOS)‐compatible and can rapidly fabricate high fidelity micro/nanopatterns in parallel—in contrast, scanning‐probe lithography and electron beam lithography for biomanufacturing use serial manufacturing techniques. 23, 24, 25, 26, 27 Patterning of silk microstructures using UV‐P 3 has been successfully demonstrated where either silk fibroin or sericin was chemically modified to be photoreactive and then served as the photoresist. 28, 29 Cell culture studies have been conducted to verify the biocompatibility of silk protein resists after the chemical modification and lithographic process. 28, 29 Though very promising, compared to their commercial counterparts based on synthetic polymers, current silk protein resists still suffer from issues such as relatively low resolution and pattern contrast in terms of lithographic patterns, mainly due to the inevitable wide molecular weight distribution (ranging from a few tens to a few hundreds of kDa for both silk fibroin and sericin proteins) during the degumming process for protein extraction. 30 Such limits hinder their practical use in precision biopatterning and the semiconductor industry where reliability and repeatability are paramount. Proteins with more uniform molecular structures (such as well‐defined chain lengths and molecular weights) and preferably more active group sites for further functionalization have yet to be explored for high‐performance protein‐based photolithography. In this study, we report on a precise protein photolithography for high‐performance biopatterning using the well‐defined silk fibroin light chain as the basic resist material. Silk fibroin is mainly composed of two components, namely heavy chain (H‐fibroin, ≈85%, w/w) and light chain ( l ‐fibroin, ≈15%, w/w), which are linked by a single disulfide bond between Cys‐c20 of H‐fibroin and Cys‐172 of l ‐fibroin. 31, 32, 33 Compared to silk fibroin and sericin proteins, l ‐fibroin has a well‐defined molecular weight of ≈26 kDa. 34 It also has a higher proportion of undifferentiated and hydrophilic amino acid composition than H‐fibroin, which facilitates facile chemical modification for the synthesis of a variety of biologically and chemically functional photoresists. 35, 36, 37 2 Results and Discussion Figure 1 illustrates the material synthesis, functionalization, and photolithographic results of UV‐reactive silk l ‐fibroin (UV–LC) resists. The B. mori silkworm cocoons were first cut into small pieces and degummed for 60 min to remove sericin using a previously reported process 38 (Figure 1 a, b). Formic acid was used to break the covalent disulfide bonds between H‐fibroin and l ‐fibroin, and to separate silk fragments based on their different solubilities in formic acid without causing severe protein degradation. 39, 40 The soluble fractions (i. e. , l ‐fibroin) were harvested and air‐dried (Figure 1 c). The l ‐fibroin was modified to be photoreactive by conjugating a photoreactive reagent of 2‐isocyanatoethyl methacrylate (IEM) to l ‐fibroin's side groups, yielding a photocrosslinkable UV–LC precursor (Figure 1 d). The UV–LC precursor was then dissolved in 1, 1, 1, 3, 3, 3‐hexafluoro‐2‐propanol (HFIP, Sigma Aldrich, St. Louis, MO). An organic photoinitiator of Irgacure 2959 (Sigma Aldrich, St. Louis, MO) was added 0. 5% (w/v) into the UV–LC precursor solution to generate (and transfer) reactive species (free radicals in this case) when exposed to UV radiation (Figure 1 e). The UV–LC resist solution (2%, w/v) was spin coated on a silicon or glass substrate to form a resist layer with a controllable thickness ranging from 50 nm to several micrometers which was then exposed through a photomask (Figure 1 f). In this case, the UV–LC resist acted as a negative photoresist which can be crosslinked due to IEM in the presence of UV light (followed by the development step) to generate wafer‐scale micropatterns on silicon and glass substrates via standard UV photolithography (Figure 1 g). UV–LC microstructures were tested as cellular substrates and for the spatial guidance of fetal neural stems cells which were seeded on micropatterned surfaces and incubated for 3 d. Cells tended to preferentially attach to the UV–LC protein patterns in comparison to the surrounding surface (i. e. , silicon in this case) (Figure 1 h, more details in Figure 5, also see Supporting Information). Note that the sensitivity of UV–LC resists can be readily tuned by regulating the IEM molecules conjugated into l ‐fibroin. Additionally, the presence of unmodified amino acids can enable further functions (e. g. , association with favorable cellular interactions and the production of multifunctional biomaterial architectures) via concurrent or subsequent modification strategies. 41 In this study, the IEM molecules were intentionally designed to exceed the population of available amino acids conversion to fully occupy nearly all active group sites on the protein chains to better investigate the underlying mechanism of photo‐only‐induced formation of crosslinked silk micro/nanostructures. Figure 1 Synthesis of the UV‐reactive silk l ‐fibroin (UV–LC) and the result of photolithography using UV–LC as a negative resist. a) B. mori cocoons are degummed for 60 min to obtain b) silk fibroin, and c) the l ‐fibroin is then separated from the silk fibroin using formic acid; d) photoactive l ‐fibroin (UV–LC precursor) is obtained by conjugating IEM to the l ‐fibroin; e) by adding the photoinitiator (Irgacure 2959), the UV–LC resist can be synthesized; f) photolithography using UV–LC resist; g) optical images of the fabricated patterns (linewidth: 5 µm, zoom‐in image) shows that UV–LC has better lithographic performance than UV–Silk30. Scale bar: 50 µm. h) Dark‐field stereomicroscopic photograph of double immunofluorescence staining with Nestin (green fluorescence) and nuclear staining (blue DAPI staining) of fetal neural stems cells that were guided to be cultured on a micropatterned UV–LC resist (dash line) on a silicon substrate. Scale bar: 100 µm. A variety of photoreactive fibroin (UV–Silk) resists with varied degumming conditions (thus varied protein chain lengths and molecular weight distributions) have been prepared for comparison using a previously reported method. 29 Note that UV–Silk resists consist of both H‐fibroin and l ‐fibroin fragments, while UV–LC only has l ‐fibroin. Figure 2 a schematically shows the (simplified) molecular structures of some example UV–Silk and UV–LC precursors, including UV–Silk made of silk fibroin fibers degummed for 30 min (UV–Silk30), degummed at high temperature (121 °C) and pressure (25 psi) for 4 h (UV–Silk high temperature and pressure [HTP]), and l ‐fibroin protein, respectively. In general, longer degumming time results in shorter but more uniform silk fibroin fragments. Therefore, compared to UV–Silk30, the H‐fibroin fragments in UV–SilkHTP are generally shorter but more uniform due to the high temperature and pressure treatment conditions during its extended degumming process. 42 In comparison, UV–LC provides a promising route serving as the basic molecular blocks for precise protein photolithography thanks to its well‐defined and evenly distributed protein chains. Figure 2 Characterization and analysis of patterns fabricated by protein photolithography using different types of silk‐based materials (e. g. , UV–Silk30, UV–SilkHTP, and UV–LC). a) Schematic comparison between structures of UV–LC and UV–Silk (including both UV–Silk30 and UV–SilkHTP) precursor, where UV–Silk30 has longer protein chains than UV–HTP. UV–LC contains only l ‐fibroin; b) morphological characterization (using an optical microscope and AFM, scale bar: 200 µm) of micropatterns fabricated by protein photolithography using UV–Silk30, UV–SilkHTP, and UV–LC. It shows that the UV–LC can achieve better resolution and surface smoothness than UV–Silk30 and UV–SilkHTP; c, d) quantitative analysis of resolution and surface roughness of micropatterns fabricated using various UV–Silk and UV–LC. The result is consistent with the observations from optical and AFM images. The surface morphology and fidelity of as‐fabricated micropatterns on isopropyl alcohol (IPA) cleaned silicon substrates using UV–LC and UV–Silk protein resists were characterized and compared using atomic force microscopy (AFM) and an optical microscope, showing that lithographic performances including the spatial resolution, pattern sharpness, and surface morphology/roughness strongly depend on the molecular structures of as‐used protein resists (Figure 2 b). AFM results show that UV–LC has the best surface roughness with a root mean square roughness of ≈2. 3 nm while UV–SilkHTP (≈8. 5 nm) is better than UV–Silk30 (≈23. 8 nm), over an area of 5 × 5 µm. We postulate that, during drying, silk fibroin proteins spontaneously form micro and nanoscale wrinkled patterns guided by a diffusion‐limited aggregation process 43 which has been observed in the assembly of a range of materials including colloids, polymer thin films, 44 peptides, 45, 46 and proteins. 47 This is partially due to the polarity mismatch between photoreactive silk resists consisting of strongly polar side groups, such as hydroxyl, carboxyl, and amino groups (thus strongly polar) and the IPA‐treated silicon (weakly polar) substrate, which can be improved by appropriate surface treatment of silicon substrates (details in the Supporting Information). The mismatch increases with the protein chain length and uneven distribution of the molecule weight where the internal molecular polarity difference within the protein chain becomes more predominant due to increased ratio between the hydrophobic H‐Fibroin fragments and the hydrophilic l ‐Fibroin fragments (Figure 2 c). Figure 2 d shows that, under the same lithographic conditions (i. e. , exposure duration and development time), the resolution (which was determined by the minimum distinguishable feature size in our case, see the Supporting Information) of UV–Silk resists improves with the increasing degumming time, due to the decreased (and more uniformly distributed) molecular weight. The UV–SilkHTP and UV–LC provide better lithographic performances in terms of resolution, yielding minimum feature sizes of 1. 54 and 1. 51 µm, respectively, close to the minimum designed feature size of the mask (1. 5 µm), which was chosen based on the capabilities of our current photolithography setup. The underlying crosslinking mechanism via the conjugation of the multifunctional acrylate moiety (i. e. , IEM in this case) to silk fibroin proteins (including both H‐fibroin and l ‐fibroin) and l ‐fibroin only—at macro and nanoscale—was investigated via both Fourier transform infrared spectroscopy (FTIR) and the scattering‐type scanning near‐field optical microscopy (s‐SNOM), respectively. Three characteristic peaks (red curve, Figure 3 b) were found in Amide I (1600–1700 cm −1 ), Amide II (1500–1600 cm −1 ), and Amide III (1200–1300 cm −1 ) bands for bulk silk fibroin proteins measured by FTIR in the attenuated total reflection (ATR) mode with an aperture size of several tens of micrometers (Figure 3 a), which gradually decreased after the introduction of the photoactive component IEM (Figure 3 b). Three prominent peaks (blue and pink curves for UV–Silk and UV–LC, respectively) surged at 1720 cm −1 (C=O stretch), 1635 cm −1 (terminal C=C stretch), and 1160 cm −1 (C—O stretch), which overlapped with the characteristic peaks of pure IEM. Figure 3 Structural characterization of the UV–Silk and UV–LC using FTIR and s‐SNOM. a) Schematic of ATR–FTIR setup, where the sample is illuminated from the back of the ATR crystal; b) FTIR spectrum of IEM, silk fibroin protein, UV–Silk, and UV–LC. The peaks vanish in the UV–Silk and UV–LC, indicating the binding of IEM on silk fibroin and l ‐fibroin; c) schematic of the s‐SNOM system. An infrared laser is focused onto the AFM tip, and the scattered signal is collected by the detector; d, e) IR nanoimaging and absorbance (acquired by s‐SNOM measurement performed at 1635 cm −1 ) of UV–Silk30, UV–Silk90, UV–SilkHTP, and UV–LC with various exposure time. The disappearance of the absorbance with increasing exposure time indicates the increasing crosslinking degree of IEM until about 90 s, after which time all the available IEM active conjugated acrylate group sites are crosslinked. s‐SNOM was employed to provide direct imaging and chemical identification of the thin protein layers at the nanoscale, to understand the variation of local chemical composites during crosslinking under UV exposure, and to overcome the special resolution and thickness limits of FTIR spectroscopic study. 48 In this work, s‐SNOM (NeaSNOM, Neaspec GmbH, Germany) has been utilized for high‐resolution optical images and spectroscopic information to map out the chemical and mechanical properties of protein patterns at the nanoscale with a spatial resolution of ≈20 nm. In our setup, s‐SNOM is coupled to a tunable IR quantum cascade laser (QCL, Daylight Solutions Inc. , USA) covering the broad IR spectra of the Amide I and II bands over the range from 1450 to 1750 cm −1 (Figure 3 c, also see the Supporting Information). The near‐field phase spectrum resembles the molecular absorbance band, while the near‐field amplitude spectrum acquires a dispersive line shape similar to a far‐field reflectivity spectrum. 49 The crosslinking degree was measured by using absorbance phase images, where the absorbance intensity at characteristic peak of crosslinking is inversely proportional to the degree of crosslink. Absorbance phase images of UV–Silk30, UV–Silk90, UV–SilkHTP, and UV–LC patterns on a silicon substrate were captured at 1635 cm −1, which corresponds to IEM‐induced photocrosslinking from terminal C=C group sites. As shown in Figure 3 d, at 1635 cm −1, the phase image exhibited a strong contrast between silk and silicon (silicon is used as the reference for IR imaging) in UV–Silk30 micropatterns that were exposed for 20 s. The contrast gradually weakened with the longer exposure time indicating an increased crosslinking degree. IEM‐induced crosslinking degrees within various protein micropatterns were obtained and evaluated quantitatively (Figure 3 e). The absorbance intensity of UV–Silk30 samples decreased monotonically with increasing exposure time until 90 s, indicating an increase in crosslinking degree due to IEM. The crosslinking was found to be saturated after 90 s exposure which expended all available active conjugated acrylate group sites and remained nearly constant thereafter. It is found that UV–Silk resists with lower molecular weights are easier to crosslink partially due to their higher degrees of molecular mobility and more uniform protein chain lengths. Compared to UV–Silk resists, UV–LC resist shows considerably higher sensitivity thanks to its shorter protein chain length and more available IEM side groups. One main use of photolithography is to pattern a resist layer which can serve as a temporary mask when etching an underlying layer. Therefore, a systematic study on the use of UV–Silk and UV–LC resists as the etching mask for pattern transfer was conducted. We have found that there are at least three factors that play an important and synergistic role in the etching performance (i. e. , etching resistance) of the silk‐based microstructures, namely, (1) the average molecular weight (i. e. , average protein chain length, which is determined by the degumming process (for UV–Silk resists) and protein separation process (for UV–LC resist)); (2) the photoinduced crosslinking due to IEM; and (3) the crosslinking due to the formation of beta sheets. We first investigated the dependence of etching performance on the photoinduced crosslinking within the protein matrix and the protein chain lengths. As shown in Figure 4 a, the etching rates of both UV–Silk30 and UV–LC resists decreased monotonically with the increased exposure time (and thus the increased crosslinking degree before saturation) and reached plateaus at ≈25. 4 and ≈72. 4 nm min −1 after UV exposure of 90 and 30 s, respectively. The etching resistance of UV–LC was initially better than UV–Silk30 since the crosslinking degree in UV–LC was considerably higher than UV–Silk30 under the same exposure conditions. With the increased exposure time, the molecular weight of silk protein chains became to play a more important role and UV–Silk30 showed a better etching resistance when both resists were fully crosslinked. Figure 4 Etching rate measurements and schematic structures of various UV–Silk and UV–LC. a) Etching rate measurement of the UV–Silk30 and UV–LC with increasing exposure time. The etching rate of UV–LC decreases faster than the UV–Silk30 with increasing exposure time but reaches a constant rate that is higher than UV–Silk30; b) etching rate comparison between various UV–Silk (including both UV–Silk and methanol treated UV–Silk) and UV–LC at two exposure times (20 and 120 s). All the samples with 20 s exposure times have larger etching rate than the samples with 120 s exposure times. UV–Silk has an increasing etching rate with increasing degumming time because the mechanical strength is better with longer chain length. UV–LC has slightly less etching rate because its highly defined molecular structure help form better crystalline structure; c) Young's modulus of UV–Silk and UV–LC and ratio of Young's modulus before and after methanol treatment. It shows the similar trend with the data of etching speed, where the largest Young's modulus value corresponds to slower etching rate. It also shows no obvious change on Young's modulus by treating with methanol, suggesting no formation of beta sheet structures in UV–Silk and UV–LC resists induced by the methanol treatment; d) schematic structure and the corresponding etching rate of the photocrosslinked UV–Silk and UV–LC with 20 s exposure time and UV–Silk30 exposed for 120 s. For UV–Silk30, the etching rate decreases with longer exposure time because of the increased crosslinking degree. With the same exposure time, the etching rate increases with increasing degumming time because of the shorter chain length, and thus less mechanical strength. With 20 s exposure (partially crosslinking) UV–LC has less etching speed because its highly defined molecular structure helps it form better IEM‐induced crystalline structure. We designed two sets of experiments to systematically investigate the etching performance among a variety of silk resists (including both UV–Silk and UV–LC ones) that were (1) partially crosslinked (for 20 s exposure so that all resists were “underexposed”) and (2) fully crosslinked (for 120 s exposure so that all resists were “overexposed”) (Figure 4 b). It has been found that, under same exposure conditions (for both partially and fully crosslinking cases), the etching rate of UV–Silk resists increased with the degumming time. This is mainly due to the reduced protein chain length during the prolonged degumming process which weakens the mechanical strength of the as‐prepared protein resist and causes the increase in the etching rate (Figure 4 c). A schematic illustration of the underlying mechanism is given in Figure 4 d. We then compared the etching performances of UV–LC resist to UV–Silk ones. In the partially crosslinking case, the UV–LC resist showed the best etching performance due to its significantly higher crosslinking degree than all UV–Silk resists (also see Figure 3 e). However, for the fully crosslinking case, a competing mechanism becomes more noticeable between the crosslinking degree and the molecular weight on the etching performance. Generally, UV–LC resist has much lower molecular weight but higher crosslinking degree than UV–Silk resists under same exposure conditions. Therefore, when fully crosslinked, UV–Silk resists with relatively short degummed time showed better etching resistance than UV–LC, due to their much higher molecular weights and better mechanical strengths. UV–Silk90 shows a comparable etching resistance to UV–LC as it has higher molecular weight but less crosslinking degree. UV–SilkHTP shows the highest etching rate as it has much lower average molecular weight than other UV–Silk counterparts due to the excessive degumming time under high temperature and pressure. Furthermore, we have found that the influence of the secondary structure of beta sheets within silk resists on their etching performances is considerably minor compared to the other two factors, namely, the average molecular weight and IEM‐induced crosslink. It is well known that methanol treatment can promote the formation of beta sheets within the silk protein matrix. 6 No noticeable variation was found in terms of the etching resistance or Young's modulus before and after the methanol treatment for all silk resist samples (Figure 4 b, c). We attribute this to the fact that the degree of IEM substitution was designed to exceed the population of amino acids conversion so to occupy almost all the active group sites on the protein chains, which hindered the formation of betasheet structures in the protein resist matrix. One of the most compelling attributes of silk materials is their abilities to allow for the incorporation of functional elements such as labile biological components with retention of bioactivity to generate functional material formats. 50 The effectiveness of UV–LC photolithography to large scale reproduce microscale geometries and topologies allows for functional components to be generated from silk. We therefore explore the doping and stabilization of UV–LC patterns with an enzyme of horseradish peroxidase (HRP) and the effects of the UV–LC photolithography process on its bioactivities as proof‐of‐principle demonstrations. As shown in Figure 5 a, the enzymatic activity of the HRP‐doped UV–LC resist was quantitatively assessed by a colorimetric enzyme‐linked immunosorbent assay (ELISA) for HRP/3, 3′, 5, 5′‐tetramethylbenzidine (TMB) after UV exposure. The bioactivity test shows that as‐prepared patterns possess bioactivity of embedded biological molecules to some extent (i. e. , HRP‐dope UV patterns turn blue after exposure to TMB) during UV–LC photolithography. Finally, UV–LC microstructures were fabricated and examined using a standard immunofluorescence assay as biocompatible cellular substrates (Figure 5 b–d). Fetal neural stem cells were seeded on nonpatterned (i. e. , a uniform coating of UV–LC resist w/o UV exposure) and patterned surfaces using UV–LC photolithography and incubated for 3 d. As shown in Figure 5 e, f, cells were well anchored to the UV–LC substrates in both cases and tended to preferentially attach to UV–LC patterned compared to the surrounding surface (i. e. , silicon in this case), showing that UV–LC micropatterns have good biocompatibility and can be used for precisely spatial cell guidance. Figure 5 a) Bioactivity evaluation of HRP‐doped silk resist and HRP enzyme after UV exposure. The ELISA test shows that enzyme activities are negatively affected during UV exposure (as shown in the photos on the right) and silk resists can help to stabilize the bioactivities to some extent during UV exposure; b) portion of the designed photomask. c–f) Double immunofluorescence staining with Nestin (green fluorescence) and nuclear staining (blue DAPI staining) of fetal neural stem cells cultured on UC–LC substrates showing the spatial guidance of cell seeding. Scale bar: 100 µm. 3 Conclusion In conclusion, we report on a precise protein P 3 for wafer‐scale, high‐performance biopatterning using chemically modified well‐defined silk l ‐fibroins as the photoresist material. The lithographic and etching performance of UV–LC and UV–Silk resists have been evaluated systematically and the underlying mechanisms have been thoroughly discussed. A general guidance on the synthesis and the use of silk l ‐fibroin resist has been provided. The inherent biocompatibility and the enhanced patterning resolution along with the improved surface roughness and etching performance of such protein‐based resists offer new opportunities in fabricating large‐scale high‐precision biocompatible functional micro/nanostructures. 4 Experimental Section Synthesis and Purification of Light Chain Proteins : 60 min degummed silk fiber was weighed and dispersed in 98–100% formic acid at a range of concentrations (0. 01–8%, w/v) for 30 min. The mixture was then centrifuged at 4000 rpm for half an hour to sediment the undissolved material. The supernatant was filtered using glass fiber filters to remove any remaining suspended particles/fibers. Then, the soluble fractions were left under a flow of air at room temperature to evaporate to constant weight. Note that the degumming conditions significantly affect the performance of UV–Silk resist but have much less effect on UV–LC resist. Synthesis of Photosensitive l ‐Fibroin (UV–LC) : The l ‐fibroin photoresist was synthesized via chemical conjugation between l ‐fibroin and photocrosslinkers (2‐Isocyanatoethyl methacrylate, IEM) in an anhydrous environment. l ‐fibroin was suspended at 1% (w/v) in a solution of 1 m LiCl/Dimethyl Sulfoxide (DMSO) and stirred at 65 °C in a dry N 2 atmosphere for 40 min. Immediately after, the IEM was added at a stoichiometric equivalent to reactive hydroxyl‐containing amino acids and reacted for 5 h at 65 °C. The product was precipitated out, centrifuged, washed and freeze‐dried, sequentially. For comparison, silk fibroins under 10, 30, 60, 90 min degumming time, as well as SilkHTP were prepared and synthesized to be photoreactive following the similar procedure in the Supporting information. Photolithography : Microscale patterns of fibroin were fabricated using photolithography. A solution of 2% (w/v) photoresist was prepared using HFIP (Sigma Aldrich, St. Louis, MO) as a solvent and 0. 5% (w/v) of Irgacure 2959 (Sigma Aldrich, St. Louis, MO) as a photoinitiator. The principle of solvent selection is discussed in the Supporting Information. The photoresist solution was then cast at 0. 5 mg per substrate. Contact photolithography was conducted using a photomask under a UV exposure (Lumen Dynamics OmniCure 1000, 320–500 nm filter). The unexposed and uncrosslinked protein photoresist was developed using deionized water (18. 2 m Ω cm) for 2 h followed by copious rinsing with deionized water and ethanol. Substrates with the developed protein patterns were then dried in a gentle stream of dry N 2. FTIR : To confirm the methacrylate conjugation, FTIR was conducted on unmodified silk fibroin proteins and l ‐fibroin film using a Nicolet iS10 FTIR spectrometer. Cast films (5. 0 mg) were analyzed in ATR mode using a Ge ATR crystal, and data was collected between 4000 and 1000 cm −1, for 32 scans at a resolution of 1 cm −1. s‐SNOM : A commercially available scattering‐type s‐SNOM (Neaspec GmbH, Germany) was utilized with a QCL IR laser (MIRCat, Daylight solutions Inc. , USA) tunable between 1495 and 1790 cm −1. During instrument operation, the laser was attenuated to ∼10 mW such that the detector yields a nominal signal of 1. 5 V. The AFM was operated in tapping mode with 65 nm tapping amplitude. Gold‐coated AFM tips with about 250 kHz resonance (Tap300G‐B‐G, budgetsensors. com) were used to achieve decent IR near‐field signal. The IR near‐field signal was detected simultaneously with AFM signals, using pseudoheterodyne technique and a lock‐in amplifier. The lockin frequency was set at the second or third harmonics of the tip tapping frequency which yield background‐free near‐field amplitude and phase information with a spatial resolution down to 10 nm. The image was scanned at 3. 3 ms per pixel for a 500 × 500 pixel sized image. 51 Bioactivity Test of Silk Protein Photoresist : HRP (Sigma Aldrich) was mixed with aqueous UV–LC solution to a final concentration of 0. 2 unit mL −1. The HRP‐containing UV–LC solution was spin coated onto a quartz substrate to a thickness of 100 nm, and floor UV exposure treatment using the previously described method. The remaining resist was exposed to 3, 3′, 5, 5′‐TMB solution (Sigma Aldrich) to test the activity of the HRP stabilized within the UV–LC resist. Etching Process : All etching process was carried out in a March anisotropic reactive ion etch plasma system. The reactive ion etcher was capable of etching oxides to remove organic contamination from sample surfaces, as well as, tetrafluoromethane (CF 4 ) to etch silicon and patterned silk proteins on the silicon substrate. The unit was outfitted with 4 mass flow controllers for reproducible gas flow. O 2 gas was dissociated in a microwave discharge (100 W) 20 cm upstream of the etching chamber for 10 min to remove the contamination from glass surface. The flow rate of the gas (89. 0 L min −1 ) was controlled using mass flow controller. After the cover glass was cleaned, a series of patterned silk protein samples were placed on the holder in the device. CF 4 gas was dissociated in the microwave discharge of 50 W with flow rate of 48. 0 L min −1 for 60 s to etch the samples. The etch rate of samples was obtained by measuring thickness change rate via AFM. Cell Culture : Human fetal neural stem cells were subcultured when they were over 90% confluent. Cells were plated in a medium consisting of Dulbecco's Modified Eagle Medium (DMEM)/F12 (GIBCO), 10% fetal bovine serum (GIBCO), penicillin/streptomycin (GIBCO), and 3. 5 × 10 −3 m glucose (Sigma), supplemented with B27 (GIBCO), 10 ng mL −1 EGF (Invitrogen), and 10 ng mL −1 FGF2 (Invitrogen). Cells were maintained at 37 °C in humidified air with 5% CO 2. Immunofluorescence Assay : For cell culture staining, the cultures were fixed in 4% Polyformaldehyde (PFA) in Phosphate Buffer Saline (PBS) for 15 min at room temperature. Cells were first washed three times by PBS and then pretreated in 0. 1% Triton X‐100 in PBS for 15 min, followed by incubation in 4% normal donkey serum, and 0. 1%Triton X‐100 in PBS for 30 min. Primary antibodies (first ab source rabbit, Nestin abcom) were incubated with cultures overnight at 4 °C in 2% normal donkey serum, and 0. 1% Triton X‐100 in PBS. After additional washing in PBS, the samples were incubated with appropriate secondary antibodies (second ab donkey anti rabbit 488 flour) conjugated to Alexa Fluor 488 Alexa Fluor. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201700278
| 2,017
|
Advanced Science
|
Netrin‐1 Promotes Inflammation Resolution to Achieve Endothelialization of Small‐Diameter Tissue Engineering Blood Vessels by Improving Endothelial Progenitor Cells Function In Situ
|
Abstract The transplant of small‐diameter tissue engineering blood vessels (small‐diameter TEBVs) (<6 mm) in vascular replacement therapy often fails because of early onset thrombosis and long‐standing chronic inflammation. The specific inflammation state involved in small‐diameter TEBVs transplants remains unclear, and whether promoting inflammation resolution would be useful for small‐diameter TEBVs therapy need study. The neural protuberant orientation factor 1 (Netrin‐1) is found present in endothelial cells of natural blood vessels and has anti‐inflammatory effects. This work generates netrin‐1‐modified small‐diameter TEBVs by using layer‐by‐layer self‐assembly to resolve the inflammation. The results show that netrin‐1 reprograms macrophages (MΦ) to assume an anti‐inflammatory phenotype and promotes the infiltration and subsequent efflux of MΦ from inflamed sites over time, which improves the local microenvironment and the function of early homing endothelial progenitor cells (EPCs). Small‐diameter TEBVs modified by netrin‐1 achieve endothelialization after 30 d and retain patency at 14 months. These findings suggest that promoting the resolution of inflammation in time is necessary to induce endothelialization of small‐diameter TEBVs and prevent early thrombosis and problems associated with chronic inflammation. Furthermore, this work finds that the MΦ‐derived exosomes can target and regulate EPCs, which may serve as a useful treatment for other inflammatory diseases.
|
1 Introduction Cardiovascular disease is a leading cause of death that creates a significant burden on patients and requires replacement therapy to treat the damaged vessels. Small‐diameter tissue engineering blood vessels (TEBVs) have been used with increasing frequency for vascular replacement therapy, and the construction of small‐diameter TEBVs that undergo endothelialization in vivo (without endothelial cells, ECs, seeded in vitro) has been at the forefront of the field. 1 However, the grafts have been leading to high incidences of thrombosis, intimal hyperplasia, and calcification by chronic graft rejection due to innate immune responses and chronic inflammation, 2, 3 which has become the primary challenge for the clinical applications of small‐diameter (1. 5–2 mm or less) TEBVs. 4, 5 Recent studies have found that acute inflammation initiates the regenerative response, and resolution by macrophages (MΦ) is necessary for survival of regenerative cells. 6, 7 For example, CD86‐MΦ induces high expression of proinflammatory cytokines and reactive oxygen species, whereas CD163‐MΦ induces high expression of scavenging molecules and anti‐inflammatory cytokines and promotes tissue remodeling. 8 Exosomes released by MΦ or other inflammatory cells represent a new component of the inflammatory microenvironment that play an important role in the cell's ability to communicate with its environment. 9, 10, 11, 12 Endothelial progenitor cells (EPCs) are precursor stem cells of ECs that can be mobilized from the bone marrow and targeted to specific sites to promote angiogenesis and early endothelialization of small‐diameter TEBVs. 13, 14 Studies have shown that reduced inflammatory response can promote mobilization and targeting of early EPCs. 15, 16 However, excessive or sustained inflammation can reduce the presence of EPCs in the blood and cause problems such as thrombosis and intimal hyperplasia. 17 Therefore, it is critical to promote inflammation resolution of small‐diameter TEBVs and provide a favorable microenvironment for the growth and differentiation of EPCs into ECs. Previous studies have found that netrin‐1 is present in the ECs of natural blood vessels and can promote angiogenesis through uncoordinated‐5 homolog B (UNC5B) and deleted in colorectal cancer (DCC) receptors. 18, 19, 20 Additionally, netrin‐1 prevents ischemia and exerts anti‐inflammatory effects by binding to the adenosine receptor A2b. 21 In the present study, we hypothesized that netrin‐1 could reprogram MΦ to exert anti‐inflammatory effects and promote inflammation resolution to establish a favorable microenvironment for EPCs, thereby achieving rapid endothelialization and improving long‐term patency rate of small‐diameter TEBVs. After testing and verifying our hypothesis and the underlying mechanism in vitro, we constructed the small‐diameter TEBVs from decellularized rat arteries and modified them for the controlled release of netrin‐1 by chitosan. The modified small‐diameter TEBVs were then transplanted into the carotid arteries of rats for subsequent evaluations of patency and endothelialization, and the relationships between these parameters were examined. 2 Results 2. 1 Netrin‐1 Reprograms MΦ into CD163 via the A2b Receptor In Vitro After being stimulated by lipopolysaccharide (LPS) in vitro, flow cytometry analysis showed that the rat peritoneal MΦ can be transformed into proinflammatory MΦ, as demonstrated by the strong expression of CD86, a costimulatory receptor necessary for T cell activation. 22 MΦ were reprogrammed to be anti‐inflammatory with 250 ng mL −1 netrin‐1 and high expression of CD163, a scavenger receptor involved in endocytosis of hemoglobin/haptoglobin, and the transformation rate was up to 69. 8%. However, the effect of netrin‐1 was suppressed after the addition of 100 × 10 −6 m MRS, an A2b receptor antagonist, at an inhibition rate of 60. 9% ( Figure 1 A, C). Immunofluorescence analysis showed that the two types of MΦ had no significant difference in morphology; however, the inflammatory stimulation was regulated by netrin‐1, the transformation rate of CD86‐MΦ changed from 70. 5% to 10. 5%, and that of CD163‐MΦ changed from 7. 5% to 66. 0%. MRS inhibited these effects (Figure 1 B, C). In addition, we found 250 ng mL −1 was the appropriate concentration for netrin‐1 in MΦ reprogramming experiments, and higher concentration had no significant influence on reprogramming efficiency (Figure S1, Supporting Information). Figure 1 Netrin‐1 reprograms MΦ to express CD163 by enhancing mitochondrial energy metabolism. A) Flow cytometry analysis showed the cell surface expression of the CD86 and CD163. B) Immunofluorescence staining of the peritoneal MΦ. CD86 and CD163 are stained with red and green fluorophore, respectively; and nuclei are stained with 4′, 6‐diamidino‐2‐phenylindole (DAPI). C) Statistical results of CD86‐ or CD163‐positive MΦ. The percentage of two phenotypes of MΦ was tested and analyzed by flow cytometry and immunofluorescence. D) Release of cytokines was analyzed in cell culture media according to each group using enzyme linked immunosorbent assay (ELISA). E) Western blot analysis of PPARγ and STAT6 expression in MΦ treated with LPS or netrin‐1. F) Model for regulation of MΦ reprogramming by LPS to support proinflammatory CD86‐MΦ or netrin‐1 to activate STAT6 and PPARγ to promote expression of anti‐inflammatory CD163‐MΦ. All data are representative of ten independent experiments. Where applicable, results are expressed as mean ± SEM, * p < 0. 05, ** p < 0. 01 versus LPS, ## p < 0. 01 versus LPS+Netrin‐1 ( n = 10). To determine whether cytokine production of MΦ is associated with polarization change, the expression of interleukin‐1β (IL‐1β), tumor necrosis factor‐α (TNF‐α), transforming growth factor‐β (TGF‐β), and IL‐10 was quantified in supernatants of MΦ from each group. Consistent with the proinflammatory and anti‐inflammatory phenotypes of CD86‐MΦ and CD163‐MΦ, our results showed that netrin‐1 suppressed the LPS‐induced increase in IL‐1β and TNF‐α levels in vitro. In contrast, secretion of immunomodulatory TGF‐β and the anti‐inflammatory cytokine IL‐10 was markedly increased in CD163‐MΦ compared with that in CD86‐MΦ (Figure 1 D). Similar differences in supernatant levels in CD86‐ and CD163‐MΦ have previously been reported. 23 Our results also showed that the anti‐inflammatory effect of netrin‐1 could be blocked by MRS. The signaling pathway through which netrin‐1 suppresses inflammation is not known. It was previously reported that peroxisome proliferator‐activated receptor (PPAR) pathways might be primarily involved owing to their known role in MΦ polarization. 24 We tested PPARγ and STAT6 expression by western blot analysis and found the levels of both proteins were increased following netrin‐1 treatment and decreased following MRS treatment (Figure 1 E). The possible regulatory pathways of LPS and netrin‐1 is shown in the schematic diagram (Figure 1 F). 2. 2 MΦ‐Derived Exosomes Are Actively Incorporated by EPCs In Vitro and In Vivo We observed a transfer of membrane vesicles from MΦ to EPCs based on the coculture of PKH26‐labeled primary MΦ with EPCs for 24 h ( Figure 2 A). Several multivesicular bodies (MVBs) were identified in the cytoplasm of MΦ by transmission electron microscopy (TEM), carrying bilipid membrane‐bound exosome‐like vesicles. The MVB membranes then invaginated and initiated the biogenesis of exosomes as previously shown, 25 MVBs fused to the cell membrane and released the exosome‐like vesicles to the extracellular space (Figure 2 B). The morphology and phenotypes of isolated vesicles were identified as described previously. 26 First, we found that the vesicle diameter was 108 nm on average and ranged from 50 to 150 nm (Figure 2 C). Second, the morphology of the MΦ‐derived vesicles was observed by TEM, and the vesicles were found to be round in shape with a diameter of ≈100 nm (Figure 2 D). Finally, the exosome marker proteins CD63, HSP90, and TSG101 were detected in the MΦ‐derived exosomes (Figure 2 E). These properties suggest that the transferred membrane vesicles from MΦ to EPCs in our experiments were indeed exosomes. Figure 2 MΦ secrete exosomes that enter endothelial progenitor cells in culture; A) A 24 h coculture assay of MΦ and EPCs at 37 °C. PKH‐26 stained MΦ were cultured in upper compartment, EPCs were adherent cultured in the lower compartment and exosomes were transferred through the 0. 4 µm. Representative image of n = 10 experiments. B) Transmission electron micrograph (TEM) of MΦ secreting exosomes i) cytoplasm with MVBs enclosing numerous bilipidic layer‐bound exosomes (red arrows), ii) inward invagination (red arrows) in the MVB membrane demonstrates the beginning of exosomes biogenesis, iii) MVB fusing with cytomembrane, and iv) exosomes are secreted out from the macrophage. C) Size distribution measurement of isolated exosome population demonstrates a single peak at nearly 100 nm and free contamination. D) TEM analyzed MΦ‐derived exosomes. The image shows small vesicles of ≈100 nm in diameter. E) Western blotting reveals expression of the MΦ exosome markers CD63, HSP90, and TSG101. F) In the coculture assay at 37 °C, MΦ‐derived exosomes were taken up by EPCs in each group. Images were taken up at various time intervals 0. 5, 1, 2, 6, and 12 h. In the coculture assay at 4 °C, images were taken up at 12 h. From the in vitro transwell experiment, we found that exosomes were taken up by MΦ 1–2 h after application and accumulated in EPCs over time. There was no significant difference in exosome number and uptake time between groups. Exosome uptake was impaired after incubation at 4 °C, which confirmed that exosome uptake was mediated by a biologically active process (Figure 2 F). The in vivo experiment revealed the accumulation of exosomes in EPCs of the spleen and peripheral blood (PBL) within 24 h after injection; 5. 06% PBL EPCs and 4. 89% spleen EPCs had phagocytosed labeled exosomes (Figure S2, Supporting Information). 2. 3 Reprogrammed MΦ Transfer Functional lncRNAs to EPCs through Exosomes, Altering the Migration, Proliferation, and Tube Formation of EPCs Exosomes are useful for regulating cell–cell communication and we have found MΦ‐derived exosomes that enter EPCs in vitro and in vivo. We then pretreated the EPCs with MΦ‐derived exosomes of different phenotypes to determine whether the EPCs specific effects on inflammation were caused by these exosomes. EPCs migration is a critical initiating event for rapid endothelialization. 15 Using a Transwell migration assay, we found that the number of migrated cells increased 0. 92–1. 14‐fold following netrin‐1 treatment compared to the control, whereas the number of migrated cells treated with LPS significantly increased 1. 82–2. 58‐fold compared to the control, and 0. 35–0. 62‐fold compared to the cells treated with netrin‐1 ( Figure 3 A, B). EdU incorporation, reflecting EPCs proliferation, significantly increased by 1. 46–2. 2‐fold following netrin‐1 treatment compared to the control, whereas cell proliferation decreased by 0. 4–0. 7‐fold following LPS treatment (Figure 3 C, D). Netrin‐1 promoted exosome‐induced EPCs tube formation compared to the control, whereas LPS inhibited tube formation at all time intervals and most notably at 24 h (Figure 3 E). We evaluated vessel parameters including the following: the number of closed loops, branching points, the tubes, and the total length of the tubes. The results showed they all decreased significantly in the LPS group especially at 24 h and increased in the netrin‐1 group. The tubes number and length of tubes was also strong decreased at 12 h in the LPS group, and netrin‐1‐exo obviously inhibited the decrease (Figure 3 F). Figure 3 Exosomes secreted by netrin‐1‐modified MΦ affect EPCs migration, proliferation, and tube formation in vitro. A) The exosomes influenced the migration of EPCs. The medium in the lower chamber contained 100 µg mL −1 exosomes derived from LPS or netrin‐1‐stimulated MΦ or blank control in the transwell assay. B) The absolute number of EPCs migrating to the lower chamber in each group. C) DAPI (blue) was used to stain nuclei of all cells and EdU (red) was incorporated into EPCs that were proliferating in each group. D) The percentage of actively proliferating EPCs in each group. E) EPCs were cultured in 96‐well plates previously coated with Matrigel and incubated in each group. Images were taken at 6, 12, and 24 h. F) Several parameters of the tube formation assay including loops, branching points, tubes and tubes length were quantificated using ImageJ. G) Relative expression level of Tie‐1ASlncRNA, lnc‐Ang362, lncRNA‐MALAT1, and lncRNA‐Cox2 in exosomes of MΦ in each group. Total RNA was extracted from exosomes and quantitative real time polymerase chain reaction (qRT‐PCR) was performed. Representative image of n = 10 experiments. ** p < 0. 01 versus control, ## p < 0. 01 Netrin‐1‐exo versus LPS‐exo. To identify lncRNA that might control EPCs function, we measured the expression of several angiogenesis related lncRNA (Tie‐1ASlncRNA, lnc‐Ang362, lnc‐Cox2, MALAT1, Meg3, and Long Noncoding RNA (MANTIS)) in exosomes derived from MΦ. The results revealed that several of them were statistically different, Meg3 expression decreased in CD163‐MΦ‐derived exosomes, and increased in CD86‐MΦ‐derived exosomes. MALAT1 expression increased in CD163‐MΦ‐derived exosomes and decreased in CD86‐MΦ. LncRNA‐Cox2 expression increased in CD86‐MΦ‐derived exosomes and remained unchanged in CD163‐MΦ‐derived exosomes (Figure 3 G). 2. 4 Construction and Characterization of the Nanostructured Small‐Diameter TEBVs by Netrin‐1 To study the effect of netrin‐1 on patency of small‐diameter TEBVs, we employed a rat model and transplanted the small‐diameter TEBVs into rats. The schematic diagram shows the layer‐by‐layer self‐assembly modified netrin‐1 both existed on the surface and inside the chitosan nanoparticles in small‐diameter TEBVs ( Figure 4 A). The interaction between the carboxylic acid of collagen and the amine of chitosan creates an ionic complex and improves stability and strength (Figure S3A, Supporting Information). 27 Scanning electron microscopy (SEM) was used to visualize the luminal surface of the vessel scaffold during the modification process from decellularization to deposition of collagen and incubation of chitosan nanoparticles on small‐diameter TEBVs (Figure 4 B). The immunofluorescence result showed N ‐succinimidyl 3‐(2‐pyridyldithio) propionate successfully couple netrin‐1 to the collagen surface both inside and outside of the vessel scaffold based on the receptor‐ligand interaction, and small‐diameter TEBVs in control group that only crosslinked two layers of collagen and modified without netrin‐1 had no fluorescence (Figure 4 C and Figure S3B, Supporting Information). In vitro assessments indicated that the chitosan nanoparticles released netrin‐1 at a consistent rate (24. 82 pg mL −1 per day) over 30 d (Figure 4 D); however, netrin‐1 could no longer be detected at 60 d. The surface materials in the modified small‐diameter TEBVs, which were crosslinked only to a layer of collagen on the decellularized arteries, had been washed away by blood flow at 1 month after implantation. The control small‐diameter TEBVs resisted blood flow but the cells failed to attach to them, whereas the netrin‐1 modified small‐diameter TEBVs resisted blood flow and promoted endothelialization 1 month after implantation (Figure 4 E). Figure 4 Layer‐by‐layer self‐assembly of small‐diameter TEBVs resists blood flow washing. A) Layer‐by‐layer self‐assembly was used to modify the decellularized vascular scaffold. B) SEM image of the luminal surface of the vessel scaffold after the decellularization process i), the surface modification of the first layer of collagen ii), and the surface modification of chitosan nanoparticles in the second layer of collagen iii). C) The luminal surface and the abluminal surface of the netrin‐1 and control small‐diameter TEBVs were immunostained for netrin‐1 (red). D) Controlled release of netrin‐1 from chitosan nanoparticles localized at the scaffolds for 4 weeks. E) SEM image of the luminal surface of small‐diameter TEBVs cross‐linked with only a layer of collagen i), control small‐diameter TEBVs without netrin‐1 modification ii), and netrin‐1‐modified small‐diameter TEBVs iii) at 1 month after transplantation. 2. 5 Netrin‐1 Reprograms MΦ Infiltrated in Small‐Diameter TEBVs and Promote Inflammation Resolution In Vivo Representative confocal images revealed the presence of CD86‐MΦ and CD163‐MΦ infiltrated on the intimal layer of small‐diameter TEBVs in control, netrin‐1, and netrin‐1 + MRS groups. We found that MΦ were infiltrated from 1 to 3 d after implantation and accumulated on intima over time, peaking after 7 d in three different groups. MΦ concentration decreased from 7 to 14 d and nearly disappeared at 30 d in the group treated with netrin‐1, whereas MΦ (especially CD86‐MΦ) were still present in the control and netrin‐1 + MRS groups ( Figure 5 A). Further quantitative analysis showed that the netrin‐1 group had the lowest amount of CD86‐MΦ and the greatest amount of CD163‐MΦ after 3, 7, and 14 d compared with the other two groups (Figure 5 B, C), indicating that reprogramming and phenotype transformation of MΦ occurred in the netrin‐1 group. Conversely, in the control and netrin‐1 + MRS groups, CD86‐MΦ were widespread and significantly more numerous than in the netrin‐1 group, especially at 3 and 7 d after implantation. Figure 5 Netrin‐1 reprograms MΦ infiltrated in small‐diameter TEBVs and promotes inflammation resolution in vivo. A) Immunofluorescence staining of CD86 (red) and CD163 (green) colocalized with DAPI (blue) on the intima in the small‐diameter TEBVs in each group at various time intervals 1, 3, 7, 14, and 30 d after implantation. B) Quantitative comparison of the absolute numbers of CD86‐MΦ at high magnification in the control, Netrin‐1, and Netrin‐1+MRS groups. C) Quantitative comparison of the absolute numbers of CD163‐MΦ at high magnification in each group. D) ELISA showed that netrin‐1 significantly decreased the levels of all of the proinflammatory cytokines IL‐1β, TNF‐α, and IL‐6 and increased the anti‐inflammatory cytokine TGF‐β, IL‐4, and IL‐10. MRS inhibited the effects of netrin‐1. All data are representative of ten independent experiments. Results are expressed as mean ± SE, * p < 0. 05, ** p < 0. 01 versus control, ## p < 0. 01 versus Netrin‐1. The expression patterns of three proinflammatory cytokines IL‐1β, TNF‐α, and IL‐6 were all reduced in cells treated with netrin‐1 at 3, 7, 14, and 30 d, and the expression patterns of three anti‐inflammatory cytokines TGF‐β, IL‐4, and IL‐10 were increased in cells treated with netrin‐1 at 3 and 7 d after implantation. There was no significant difference after 1 d between the three groups, indicating that small‐diameter TEBVs caused inflammation at 1–3 d after implantation. The results also revealed continuous inflammation in control and MRS groups because both proinflammatory and anti‐inflammatory cytokines were high at 14 and 30 d in the two groups and low in netrin‐1‐modified groups (Figure 5 D). Taken together, these results indicated that netrin‐1 induced higher levels of anti‐inflammatory cytokines and lower levels of proinflammatory cytokines. 2. 6 The Inflammation Resolution Microenvironment Provided by Netrin‐1 Promotes EPCs Differentiation, and Small‐Diameter TEBVs Achieve Quicker and Better Endothelialization Thirty days after transplantation, more evidence of endothelialization was observed on the intimal layer of small‐diameter TEBVs in netrin‐1 groups compared to the control, as observed by SEM ( Figure 6 A). The numbers of ECs had no significant difference during the initial 7 d after implantation, but more ECs were prevalent on the intimal layer of small‐diameter TEBVs at 14 d in netrin‐1 compared to the control and MRS groups. Additionally, we found that endothelialization was achieved by day 30 in the netrin‐1 group, while there were few viable cells on the intima of other groups. H&E staining showed similar results and demonstrated that small‐diameter TEBVs kept patent at 30 d in the netrin‐1 group, but there was obvious intimal hyperplasia at 30 d in small‐diameter TEBVs in the control and MRS group (Figure 6 B). Figure 6 Endothelialization process of small‐diameter TEBVs. A) SEM images of the intimal layer of the different TEBV groups at various time intervals 1, 3, 7, 14, and 30 d. The quantitative comparison of ECs in each group. Results are expressed as mean ± SE, ** p < 0. 01 versus control, ## p < 0. 01 versus Netrin‐1 ( n = 10). B) Representative H&E images corresponding to SEM images in (A). All data are representative of ten independent experiments. In addition, we also found that netrin‐1 could reduce platelet adhesion to the A2b receptor in the acute stage of small‐diameter TEBVs implantation prior to endothelialization. We performed a set of flow‐chamber experiments to determine if netrin‐1 altered the adhesion of platelets to the luminal surface of small‐diameter TEBVs. Significantly fewer platelets were observed on the surface of netrin‐1‐modified small‐diameter TEBVs than on control or A2b‐blocked small‐diameter TEBVs (Figure S4A, B, Supporting Information). Treatment with netrin‐1, but not with netrin‐1 and the A2b‐blocker, significantly increased cyclic Adenosine monophosphate (cAMP) levels in cultured platelets. Therefore, netrin‐1 release caused a decrease in the risk of thrombosis (Figure S4C, Supporting Information). 2. 7 Netrin‐1‐Modified Small‐Diameter TEBVs Improve Long‐Term Patency We used the laser Doppler ultrasound imaging and microcomputed tomography angiography (CTA) to assess the blood flow and long‐term patency rate at 2, 6, and 14 months. The results showed that 90% of the A2b‐blocked small‐diameter TEBVs had been obstructed at 2 months after implantation, and the average blood flow rate was 1. 01 mL min −1, decreasing to 0. 32 mL min −1 at 14 months. However, the blood flow rate of small‐diameter TEBVs in the netrin‐1 group remained steady at 5. 35–5. 55 mL min −1 up to 14 months. H&E staining also showed that there was no smooth muscle cell (SMC) excessive proliferation and no intimal hyperplasia in netrin‐1‐modified small‐diameter TEBVs at 14 months. CTA results showed netrin‐1‐modified small‐diameter TEBVs still kept patent at 14 months ( Figure 7 A). Netrin‐1‐modified small‐diameter TEBVs maintained normal pulsation, whereas A2b‐blocked small‐diameter TEBVs became occluded (Videos S1 and S2, Supporting Information). Doppler ultrasound image revealed the high patency of the small‐diameter TEBVs and the blood flow velocity under vascular contraction and relaxation (Figure 7 B, C and Video S3, Supporting Information). Figure 7 Netrin‐1‐modified small‐diameter TEBVs accomplish endothelialization in time and keep patent for 14 months. A) Dissection, H&E staining, and CTA revealed the characteristics of the small‐diameter TEBVs among different groups at different time points; the red arrows indicate the small‐diameter TEBVs in vivo. B) Laser Doppler ultrasound imaging assessment of netrin‐1‐modified small‐diameter TEBVs transplanted for 14 months and synchronization of pulsation with adjacent host aorta. C) Blood flow volume of small‐diameter TEBVs in different groups were determined by Doppler after being transplanted into rats. * p < 0. 05 versus netrin‐1‐modified small‐diameter TEBVs. Values represent the mean ± SD ( n = 10). D) SEM image of the luminal surface of the small‐diameter TEBVs among different treatment groups at 2, 6, 14, and 24 months. All data are representative of ten independent experiments. SEM images revealed that there were few ECs grown on A2b‐blocked small‐diameter TEBVs, and they were disorganized at 2 months after implantation, whereas the netrin‐1‐modified small‐diameter TEBVs had a greater number of ECs and the majority achieved endothelialization. Unlike the critical intimal hyperplasia present in the A2b‐blocked group, the netrin‐1‐modified small‐diameter TEBVs had favorable endothelization at 6 months and the ECs showed polar growth in the direction of the blood flow. At 14 months after implantation, A2b‐blocked small‐diameter TEBVs were totally implosive, while netrin‐1‐modified small‐diameter TEBVs achieved complete endothelialization and the ECs began to fuse and secrete extracellular matrix (ECM). At 24 months after implantation, we could see that the ECs had successfully grown on the small‐diameter TEBVs, and antenna and intercellular junctions were evident at higher magnifications (Figure 7 D). 3 Discussion The effect of inflammation on vascular remodeling and early monocyte recruitment on small‐diameter TEBVs has been studied previously; 28, 29 however, there have been no studies regarding the regulation of inflammation to promote quick endothelialization of small‐diameter TEBVs. First, there is a need for a method to deliver netrin‐1 and control its release through the resistance caused by blood flow that also allows small‐diameter TEBVs to capture EPCs from circulation to promote endothelialization. The results showed that the multilayered collagen‐chitosan nanoparticles containing netrin‐1 successfully bonded to small‐diameter TEBVs, which also enhanced their strength and stabilization due to the ionic bond between the carboxylic acid of collagen and amino group of chitosan. 30, 31 In vivo, the results showed netrin‐1 was released evenly for 30 d. Exogenous netrin‐1 is able to suppress inflammation by engaging A2b, which may play a role in attenuating neutrophil transmigration. 21 In this study, the results showed that treating MΦ with netrin‐1 induced CD163 expression and suppressed LPS‐induced CD86 expression. The western blot results further proved that netrin‐1 activates STAT6, PPARγ, and glucocorticoid pathways, known anti‐inflammatory MΦ‐polarizing signaling pathways. MRS inhibited the reprograming effect of netrin‐1, indicating that the A2b receptor is important for the effect of netrin‐1 on MΦ. Increasing evidence indicates that the microenvironment plays a significant role in vascular disease treatments and EPC survival and function; 32, 33, 34, 35 In vitro, we observed exosome synthesis and secretion and found that exosomes derived from distinct phenotypes of MΦ rapidly entered and delivered their content to EPCs through an active process. The results revealed that EPCs acquired changes in migratory, proliferative, and tube formation properties in response to their communication with MΦ exosomes. Interestingly, the results also showed that CD86‐exosomes had stronger ability of inducing EPCs migration than CD163‐exosomes, confirming that moderate inflammation can promote EPCs homing mediated by exosomes. In addition, CD163‐exosomes remarkably upregulated the proliferation and tube formation of EPCs, whereas CD86‐exosomes significantly downregulated these properties, which confirmed that intense inflammation discourages the survival and function of EPCs. The expression of Meg3 and lncRNA‐MALAT1 was significantly changed in exosomes of CD163‐MΦ and CD86‐MΦ, which had been shown to play an important role in regulating EC function and vessel growth. 36, 37, 38 In vivo, the results showed there was a large number of inflammatory MΦ infiltrating the TEBVs in control group and high expression of proinflammatory cytokines in blood plasma at 3–7 d after implantation. The inflammation then became chronic at 14 d and the small‐diameter TEBVs gradually underwent thrombosis. The MΦ that infiltrated netrin‐1‐modified small‐diameter TEBVs were reprogrammed to express CD163 with a high content of anti‐inflammatory cytokines in the plasma, and then the MΦ evacuated from the sites, inflammation gradually subsided, and small‐diameter TEBVs achieved better endothelialization at 14 and 30 d after implantation. These compelling findings suggest that netrin‐1 is capable of reprogramming MΦ, thereby promoting inflammation resolution of small‐diameter TEBVs. The SEM and H&E staining results indicated that the netrin‐1‐modified small‐diameter TEBVs maintained optimal conditions for EPCs function and accomplished endothelialization in time at the intimal surface. They had significantly less intimal hyperplasia and more ECs on the intimal surface compared to the control and MRS groups at each time interval. Endothelialization on the vascular cavity surface was significantly increased and there was no sign of thrombosis after 30 d in the netrin‐1 group. The CTA and Doppler ultrasound results showed good long‐term patency of netrin‐1‐modified small‐diameter TEBVs. In summary, netrin‐1‐modified small‐diameter TEBVs can get early inflammation resolution, which contributes to homing EPCs survival and quick endothelialization. Additionally, patients who need small‐diameter TEBVs often have the disease like diabetes or atherosclerosis, 39, 40 so we need further study the inflammatory condition and resolution ways of small‐diameter TEBVs in these disease state. 4 Conclusion These findings suggest that promoting the resolution of inflammation in time is necessary to induce endothelialization of small‐diameter TEBVs in the stages immediately following implantation and to prevent early thrombosis and problems associated with chronic inflammation. Furthermore, we find that netrin‐1 promotes MΦ reprogramming to induce inflammation resolution and the releasing of exosomes from MΦ to target and regulate EPCs, which improves the long‐term patency rate of small‐diameter TEBVs. In the clinical vascular transplantation, inflammation always induces thrombosis and block of vascular graft. So our study will provide new perspective for vascular graft long‐term survives in host, which may serve as a useful treatment for other inflammatory diseases. 5 Experimental Section Detailed methods are provided in the Supporting Information. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700310
| 2,017
|
Advanced Science
|
Sequentially Programmable and Cellularly Selective Assembly of Fluorescent Polymerized Vesicles for Monitoring Cell Apoptosis
|
Abstract The introduction of controlled self‐assembly into living organisms opens up desired biomedical applications in wide areas including bioimaging/assays, drug delivery, and tissue engineering. Besides the enzyme‐activated examples reported before, controlled self‐assembly under integrated stimuli, especially in the form of sequential input, is unprecedented and ultimately challenging. This study reports a programmable self‐assembling strategy in living cells under sequentially integrated control of both endogenous and exogenous stimuli. Fluorescent polymerized vesicles are constructed by using cholinesterase conversion followed by photopolymerization and thermochromism. Furthermore, as a proof‐of‐principle application, the cell apoptosis involved in the overexpression of cholinesterase in virtue of the generated fluorescence is monitored, showing potential in screening apoptosis‐inducing drugs. The approach exhibits multiple advantages for bioimaging in living cells, including specificity to cholinesterase, red emission, wash free, high signal‐to‐noise ratio.
|
1 Introduction Construction of self‐assembled nanomaterials aiming to biomedical applications is attractive ever‐growing interests, including various biosensing/imaging probes, drug delivery systems, tissue engineering materials for the purpose of diagnosis and therapy. 1, 2, 3, 4, 5 On demand of precise medicine with sensitive/selective diagnosis and targeting therapy, development of activatable theranostic nanoagents that can undergo an intrinsic evolution upon cell uptake is highly imperative. 6, 7, 8, 9 Compared to the general way that self‐assembled nanoagents were beforehand produced in inanimate environments, the introduction of controlled self‐assembly into living things paves an alternative avenue to generate smart biomedical materials. 10, 11, 12, 13 In these examples, molecular building blocks undergo self‐assembly following initial cellular uptake and subcellular activation. 14, 15, 16 From the viewpoint of diagnostics, Rao's group imparted a biocompatible/bioorthogonal cyclization‐mediated in situ self‐assembly of small‐molecule probes for imaging protease activity both in vitro and in vivo. 17, 18 Wang's group fabricated a novel photoacoustic contrast agent from an enzyme‐activated building block for specific and sensitive bacterial infection detection. 19 From the viewpoint of therapeutics, enzyme‐instructed intracellular molecular self‐assembly exhibits superiorities of not only selectively killing cancer cells but also overcoming multidrug resistance, 20, 21, 22 which was recently demonstrated by Xu's and Liang's groups concurrently. Enzymes, as an endogenous stimulus with desired specificity, have been commonly used to convert nonassembling precursors into self‐assembling blocks in living things, 14, 15, 16, 23 which could alternatively achievable by some biocompatible exogenous stimuli too. However, self‐assembly under integrated control of both endogenous and exogenous stimuli, especially in the form of sequential input, is unprecedented and ultimately challenging. In this work, we wish to report a sequentially programmable self‐assembling strategy for fluorescent polymerized vesicles ( Scheme 1 ). To the best of our knowledge, this work represents the first example of programmable self‐assembly in living cells under sequentially integrated control of both endogenous and exogenous stimuli, i. e. , enzymatic reaction and photopolymerization, respectively. Such programmable self‐assembly, relying on operations involving two or more stimuli arranged in a specific sequence, 24, 25 can foreseeably improve selectivity and modularity of in situ and/or in vivo prepared smart biomaterials that can and will only respond to multiple stimuli in a predefined cascade. 26, 27, 28 Moreover, fluorescence generation accompanied with the programmable self‐assembly endowed us capability of visualizing on site and in time information on biological structures and processes. Consequently, a proof‐of‐principle example of biomedical applications was demonstrated for monitoring the cell apoptosis process with cholinesterase as biomarker, which further allows apoptosis‐related drug screening. The fluorescent probe generated in situ by programmable self‐assembly exhibits multiple advantages, including specificity to cholinesterase, red emission over 600 nm, wash free, high signal‐to‐noise ratio for bioimaging in living cells. Scheme 1 Schematic illustration of fabricating fluorescent polymerized vesicles (red PDA) from DC by a sequentially programmable control in the order of enzyme, light, and heating inputs, which cannot be otherwise achieved, even with the right combination of stimuli in different orders. BChE represents the abbreviation of butyrylcholinesterase. Methoxy poly(ethylene glycol)‐conjugated diacetylene (PEG‐DA) was doped on account of solubility, degree of hydration, stability, as well as thermochromism. 2 Results and Discussion To meet the requirement of enzyme‐responsivity in self‐assembly, the aggregation behaviors of enzymatic substrates and products have to be distinct from each other, e. g. , enzymes convert nonassembling substrates into self‐assembling products or vice versa. The building precursor diacetylene‐appended choline (DC) forms the micellar assembly in the 20 × 10 −3 m 4‐(2‐hydroxyethyl)piperazine‐1‐erhanesulfonic acid (HEPES) buffer (pH = 7. 4, 150 × 10 −3 m KCl) with a critical micelle concentration value of 14 µm (Figure S4, Supporting Information). The hydrolyzed product diacetylene acid (DA) has been well reported to form vesicular aggregation in aqueous solution. 29, 30, 31 According to the calculation of critical packing parameter (CPP), 32 DA with a CPP value of 0. 94 prefers to form spherical vesicle while DC with a CPP value of 0. 33 prefers to form spherical (or cylindrical) micelle. 33 However, DA cannot be well hydrated in HEPES buffer even with harsh sonication at 80 °C for 30 min owing to its poor water solubility. Poly(ethylene glycol) (PEG) doping was then implemented to address this issue. Methoxy poly(ethylene glycol)‐conjugated diacetylene (PEG‐DA) was prepared according to a literature procedure. 34, 35 Doping PEG‐DA to DA generates a DA vesicle coated with PEG on its surface. PEG incorporation can therefore improve the solubility of DA and make hydration of DA vesicle occur easily in mild conditions. 36 A comprehensive screen of the molar ratio of PEG‐DA and DA shows that 5% PEG‐doping is optimal for operating the programmable self‐assembly in both inanimate environments and living cells on accounts of solubility, hydration, photopolymerization, and colorimetric response (see the Supporting Information). Dynamic light scattering (DLS) and cryoelectron microscopy (cryo‐EM) and small‐angle X‐ray scattering (SAXS) measurements were employed to identify the different self‐assembling morphology and size between DC and DA. The DLS examinations revealed that 95% DC (5% PEG‐doping) forms small micellar aggregation, giving an averaged diameter of 8 nm, whereas 95% DA forms large vesicular aggregation with an averaged diameter of 242 nm ( Figure 1 ). UV irradiation of 95% DA for 5 min generates 95% blue polydiacetylene acid (PDA) with the diameter changing slightly to 215 nm. Heating transfers 95% blue PDA to the red one with a diameter of 242 nm. DLS and static light scattering measurements were combined to provide a profile of the assembling morphology as this is reflected by the ratio R g / R H. 37 The ratio R g / R H is calculated as 0. 91 (Figure S10, Supporting Information), indicating the vesicular morphology. Transmission electron microscope (TEM) images show spherical morphology of both 95% DC and 95% polydiacetylene choline (PDC) and the sizes of 95% DC and 95% PDC are consistent with DLS results. (Figure S11, Supporting Information) Cryo‐EM images show the hollow spherical morphology of both 95% DA and 95% PDA, reinforcing the vesicular structure (Figure S12, Supporting Information). SAXS measurements give the same wall thickness of 4. 7 nm for 95% DA and 95% PDA, which is in good accordance with two DA lengths (Figure S13, Supporting Information). All these results revealed that the enzyme product DA forms the large unilamellar vesicular aggregation whereas the enzyme substrate DC forms the small micellar aggregation. Neither the photopolymerization nor the thermochromism procedure affects the assembling size and morphology of DA vesicle appreciably. Figure 1 DLS data of a) 95% DC micelle, b) 95% DA vesicle, c) 95% blue PDA vesicle, and d) 95% red PDA vesicle in HEPES buffer (0. 2 × 10 −3 m in diacetylene unit). The distinguishable CPP values lead to not only the different assembly morphologies but also, more importantly, the distinct photopolymerization induced chromogenesis between DC and DA. After photopolymerization by UV irradiation, the color of DA vesicles turned deep blue ( Figure 2 a), which is characteristic of a highly conjugated system, i. e. , the formation of PDA. The blue polydiacetylene is susceptible to a wide range of external stimuli (temperature, pressure, light, etc. ), 38, 39, 40 leading to chromic changes. 29, 30, 41, 42 The color of PDA changes from blue to carmine red when heated at 70 °C. Moreover, the colorimetric response of PDA from blue to red is accompanied by fluorescence generation (Figure 2 b). Fluorescence, as one noninvasive optical technique, is preferable to observe the self‐assembly entities in living cells. In contrast to DA, photopolymerization cannot be obtained for DC with the same irradiation condition. Photopolymerization of DC was merely obtained when extending the irradiation time to several hours, forming PDC in pale yellow. The discriminative photopolymerization between DA and DC is ascribed to that the bilayer arrangement of DA vesicle is much more favorable for photopolymerization than the curvature arrangement of DC micelle. 43, 44, 45, 46 Figure 2 a) UV–vis spectra of DC, PDC, DA, blue PDA, and red PDA (0. 2 × 10 −3 m in diacetylene unit) in 10 × 10 −3 m phosphate buffer saline (PBS, ) (H = 8. 0, considering the solubility of DA). PDC and blue PDA were produced by UV irradiation at 254 nm of DC for 5 h and DA for 5 min, respectively. Blue PDA was converted to red PDA when heated at 70 °C. The insets show the sequential color changes of the DA solution after photopolymerization and thermochromism. b) Fluorescence spectra of blue and red PDA in PBS buffer (0. 2 × 10 −3 m in diacetylene unit), λ ex = 550 nm. c) UV–vis spectra of 95% DC in HEPES buffer (0. 2 × 10 −3 m in diacetylene unit) under the programmable control of the predefined cascade, where the stimuli are in the sequence of enzyme incubation (2 d), UV irradiation (5 min) and thermochromism. d) Schematic illustration showing the color changes of the 95% DC solution by multiple stimuli. For the process of programmable self‐assembly, the optimized 95% DC was subjected to the cascade stimuli, in the order of enzymatic hydrolysis, photopolymerization, thermochromism, monitored by UV–vis spectroscopy (Figure 2 c). 5 U mL −1 butyrylcholinesterase (BChE) was employed according to the average activity of cholinesterase present in human tissues. 47, 48 Incubation with BChE for 2 d converted DC to DA, 49, 50, 51, 52 and then the DC micelles changed to the DA vesicles. Irradiation at 254 nm for 5 min led to the formation of the blue‐form PDA, and followed by a warming process at 37 °C, the desired fluorescent polymerized vesicles, red‐form PDA, were expectedly constructed, exhibiting red emission over 600 nm (Figure S14, Supporting Information). The thermochromism kinetics at 37 °C was monitored in real time (Figure S15, Supporting Information), giving that the blue‐form PDA isomerized to the red form in a few minutes. The colorimetric response occurs quickly around body temperature (37 °C) to make the construction of fluorescent polymerized vesicles practically operational in living cells, which benefits from the PEG doping that could tune the colorimetric response temperature of PDA vesicles (Figure S8, Supporting Information, the PDA vesicle is more thermosensitive with PEG doping). 36 The formation processes were further monitored by DLS (Figure S16, Supporting Information). As mentioned above, 95% DC shows very weak scattering intensity with the average diameter of 8 nm. After incubation with BChE, the scattering intensity increases dramatically, giving an average diameter of 96 nm, which definitely indicates the enzymatic conversion from the DC micelle to the DA vesicle. Further UV irradiation does not lead to appreciable size change (99 nm). 53 Mass spectrometry (MS) and Zeta potential measurements also gave the indicative information about the enzymatic conversion. MS clearly shows the peak of choline, the other product of enzymatic hydrolysis, after incubation of DC with BChE (Figure S17a, Supporting Information). DC with the quaternary ammonium head group shows a positive Zeta potential of +36 mV, while DA with the carboxylic group shows a negative Zeta potential of −50 mV. Upon incubation with BChE, the Zeta potential of DC changes to −17 mV (Figure S18, Supporting Information). To test the specificity of BChE‐triggered assembly, we investigated the programmable control of 95% DC by replacing BChE with other enzymes such as exonuclease I, trypsin, and α‐chymotrypsin (Figure S19a, Supporting Information). No photopolymerization induced chromogenesis was observed upon incubation with all control enzymes. To verify that the protein BChE itself is not a factor contributing to the chromogenesis, a control experiment was further carried out in which the same amount of denatured BChE (treated in boiling water for 1 h) was added to 95% DC, and no chromogenesis was observed either. Moreover, with the addition of tacrine, a BChE inhibitor, the BChE‐triggered chromogenesis was quenched to extent minimal level. Thus, the results clearly establish that it is the specific enzymatic activity of BChE, responsible for converting DC to DA, which serves as an indispensable prerequisite to build the fluorescent polymerized vesicles. To test the necessity of programmable control, stimuli with an alternative sequence were imposed to 95% DC (Figure S19b, Supporting Information). When DC was imposed to UV irradiation first and then BChE incubation, followed by irradiation again, no appreciable chromogenesis was observed. The result demonstrates undoubtedly that implementing the multiple stimuli in the predefined cascade, that is, enzymatic hydrolysis followed by photopolymerization and thermochromism, is indispensable for fabricating the fluorescent polymerized vesicles while stimuli in any other sequence are invalid. A cholinesterase‐rich cell line (HepG2) was then employed to evaluate the feasibility of the programmable self‐assembly strategy in living cells. 47 It is prerequisite to examine the biocompatibility of the 95% DC micelles and the results show the low cell toxicity at experimental conditions (Figure S22, Supporting Information). The programmable fabrication of fluorescent polymerized vesicles in living cells was monitored by confocal laser scanning microscopy (CLSM). Fluorescent imaging shows the accumulation of bright red fluorescence inside cells ( Figure 3 a). The fluorescence outside cells remains dim during the whole experiment, which provides a clear background and enables the visualization of the fabrication of fluorescent polymerized vesicles inside living cells. Figure 3 a) CLSM images of HepG2 cells incubated with 95% DC and 100% PDC, as well as PC‐3 cells incubated with 95% DC for 12 h at 37 °C (the concentration is 0. 05 × 10 −3 m in diacetylene unit), followed by UV irradiation. The scale bar is 40 µm. b) CLSM images of HepG2 cells incubated with 95% DC for 12 h at 37 °C without (−) or with (+) paclitaxel to trigger the cell apoptosis, followed by UV irradiation. The cell nuclei were stained with DAPI. The scale bar is 30 µm. It is clearly observed that intense red fluorescence was distributed in cytoplasm, whereas nearly no detectable signal was internalized by the cell nucleus, which implies that the fluorescent polymerized vesicles were generated in cytoplasm where cholinesterase is expressed. Figure 3 b further validates the intracellular distribution of fluorescent polymerized vesicles, where the cell nuclei were costained with 4′, 6‐diamidino‐2‐phenylindole (DAPI). The red fluorescence was mainly observed in cytoplasm around but not in nuclei. Note that no autofluorescence from the cell itself can be detected under the same experimental conditions (Figure S23, Supporting Information). Notably, cholinesterase has been demonstrated as another kind of enzymatic biomarkers of cell apoptosis besides caspases. 54, 55 Accompanying with the apoptosis process, cholinesterase is overexpressed and transfers from cytoplasm to cell nucleus. 56 So far, assays for caspases have received considerable attention for apoptosis imaging; 18, 57 however, developing cholinesterase assay principles for apoptosis imaging is still on demand. Only a few examples were reported 58, 59 but suffered from puzzles of background fluorescence and washing difficulty. To our gratifying surprise, the present programmable self‐assembly is envisaged to be a smart strategy for imaging of live cell apoptosis with high fluorescence contrast. The precursor DC shows definitely no photopolymerization induced chromogenesis, which promises no interference of fluorescence background, and also, the fluorescent polymerized vesicles generated by cascade stimuli are light‐up probes with superiority of wash free. We therefore treated the cells with paclitaxel to trigger the cell apoptosis. Definitely, much brighter fluorescence was observed in the apoptotic cells than normal cells (Figure 3 b). More importantly, the red fluorescence was observed in both cytoplasm and nucleus regions, indicating the transference of fluorescent polymerized vesicles from cytoplasm to cell nucleus. Both phenomena are indicative of the overexpression and transference of cholinesterase induced by apoptosis. As a result, the present system provides a new protocol for monitoring the cell apoptosis process involved in the overexpression of cholinesterase, which further allows in situ screening and quantification of apoptosis‐inducing agents. Comparing with the commercially available methods, it is a complementary approach to monitor cell apoptosis with cholinesterase as a biomarker, which may lead to a new understanding of mechanisms of apoptosis or drug action. Moreover, although sophisticated, the present strategy is still endorsed with multiple advantages, including specificity to cholinesterase, red emission, wash free, high signal‐to‐noise ratio for bioimaging in living cells. As a blank control, the HepG2 cells were incubated with PDC, followed by the same operating procedure. No appreciable fluorescence was monitored by CLSM (Figure 3 a), which is in accordance with the aforementioned result in the inanimate environments. That is, construction of fluorescent polymerized vesicles could be realized in both the inanimate environments and living cells, by only implementing the multiple stimuli in the predefined cascade, but not for any other undefined sequence of control. Cholinesterase is also distributed in serum 47 and 10% fetal bovine serum (FBS) is used in cell culturing process. It is therefore necessary to examine the influence of cholinesterase in cell medium on the programmable assembly process. 95% DC was incubated in cell medium at 37 °C for 12 h, and followed by the same operating procedure. No red fluorescence was observed (Figure S24, Supporting Information). This result reinforced that in cell experiments the fluorescent polymerized vesicles were fabricated in living cells but not in cell medium. Enzymatic conversion is the rate controlling process in constructing fluorescent polymerized vesicles, and such a programmable self‐assembly in living cells is cholinesterase dependent. We further examined the programmable self‐assembly strategy in PC‐3 cell line, which is relatively cholinesterase deficient. 60 Comparing with HepG2 cells, very weak red fluorescence was observed in PC‐3 cells (Figure 3 a). That is, the programmable construction of fluorescent polymerized vesicle is with cellular selectivity, depending on the activity of cholinesterase. 3 Conclusion In summary, we reported a sequentially programmable self‐assembling strategy that fluorescent polymerized vesicles were constructed in not only inanimate milieu but also, more importantly, in living cells by implementing the programmable control in a strictly choreographed sequence of operations, that is, enzymatic conversion followed by photopolymerization and thermochromism. The synergistic contribution of endogenous enzymatic reaction and exogenous photopolymerization imparts the self‐assembling strategy not only with cellular selectivity but also promised high degree of spatial‐temporal control, which may find use in various biomedical applications, such as bioimaging/sensing, regulation and perturbation of cellular processes, tissue engineering and beyond. Moreover, the programmable self‐assembly was accompanied with fluorescence generation, which therefore demonstrated a novel assay principle for imaging of apoptosis and in situ evaluation of apoptosis‐inducing agents in living cells. The same programmable self‐assembling strategy may be amendable to other enzymatic targets, and opens up new avenues for generating smart biomaterials in vitro and in vivo. 4 Experimental Section Materials Preparation : All chemicals used are reagent grade unless noted. 10, 12‐Pentacosadiunoic acid, DA was purchased from Alfa Aesar. BChE was purchased from equine serum (246 U mg −1 ), trypsin (from bovine pancreas), and tacrine (9‐amino‐1, 2, 3, 4‐tetrahydroacridine hydrochloride hydrate) were purchased from Sigma‐Aldrich. Exonuclease I was purchased from Takara. α‐Chymotrypsin was purchased from Aladdin. All of these were used without further purification. Methoxy PEG‐DA was synthesized and purified according to procedures reported previously. 34, 35 The synthesis and characterization of DC were shown in the Supporting Information. FBS and Roswell Park Memorial Institute (RPMI) 1640 cell culture medium were purchased from Gibco. DAPI was purchased from Sigma. A General Procedure for the Fabrication of Assemblies : Preparation of the DA vesicles: PEG‐DA was dissolved in buffer at a concentration of 1. 0 × 10 −3 m. A solution of 1. 0 × 10 −3 m DA in chloroform was dried for several hours under vacuum. Then a certain amount of PEG‐DA solution and buffer were added until final concentrations of 0. 1 × 10 −3 or 0. 2 × 10 −3 m in diacetylene unit. The samples were sonicated at 60−80 °C for 30 min, subsequently cooled to room temperature. Preparation of the DC Micelles : A certain amount of PEG‐DA solution was added to the DC solution, and buffer was added until got final concentrations of 0. 2 × 10 −3 m in diacetylene unit. The sample was sonicated at 60 °C for 30 min and subsequently cooled to room temperature. A General Procedure for the Programmable Self‐Assembling Process in Inanimate Milieu : 0. 2 × 10 −3 m 95% DC solution was prepared as described above. The solution was incubated with 5 U mL −1 BChE at 37 °C for 2 d to ensure enzymatic conversion DC to DA, and then was cooled at 4 °C overnight. The sample was sequentially irradiated with UV light (254 nm, 120 W) for 5 min at room temperature, leading to the formation of the blue‐form PDA, and followed by warming up to 37 °C for 15 min to generate the red‐form PDA. A General Procedure for the Programmable Self‐Assembling Process in Living Cells : The HepG2 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) each supplemented with 10% FBS in 5% CO 2 at 37 °C. The PC‐3 cells were cultured with RPMI 1640, which was supplemented with 10% FBS, at 37 °C in a humidified atmosphere with 5% CO 2. Both HepG2 and PC‐3 cells were incubated with 95% DC or 100% PDC (0. 05 × 10 −3 m in diacetylene unit) at 37 °C for 12 h and then were cooled at 4 °C for 2 h. The cells were sequentially irradiated with UV light (254 nm, holding‐lamp, 14. 6 mJ cm −2 ) for 15 min at room temperature followed by warming up to 37 °C for a few minutes. A General Procedure for the DAPI‐Labeled Process : The HepG2 cells were cultured in DMEM each supplemented with 10% FBS in 5% CO 2 at 37 °C. HepG2 cells were incubated with 95% DC (0. 05 × 10 −3 m in diacetylene unit) at 37 °C for 12 h and then were cooled at 4 °C for 2 h. Alternatively, 0. 05 × 10 −3 m paclitaxel was added at 9 h to trigger the cell apoptosis within the 12 h DC incubation at 37 °C. The cells were sequentially irradiated with UV light (254 nm, holding‐lamp, 14. 6 mJ cm −2 ) for 15 min at room temperature followed by warming up to 37 °C for a few minutes. HepG2 cells were washed three times with sterile 1 × PBS buffer and fixed using 1 mL 4% paraformaldehyde for 15 min at room temperature. After that, the HepG2 cells were washed with 1 × PBS buffer and treated with 1 mL of 0. 1% Triton X‐100 for 10 min at room temperature. Then, the cells were also washed three times carefully and stained with DAPI for 10 min at room temperature. All cells were imaged by CLSM. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201700401
| 2,017
|
Advanced Science
|
3D Printing of Lotus Root‐Like Biomimetic Materials for Cell Delivery and Tissue Regeneration
|
Abstract Biomimetic materials have drawn more and more attention in recent years. Regeneration of large bone defects is still a major clinical challenge. In addition, vascularization plays an important role in the process of large bone regeneration and microchannel structure can induce endothelial cells to form rudimentary vasculature. In recent years, 3D printing scaffolds are major materials for large bone defect repair. However, these traditional 3D scaffolds have low porosity and nonchannel structure, which impede angiogenesis and osteogenesis. In this study, inspired by the microstructure of natural plant lotus root, biomimetic materials with lotus root‐like structures are successfully prepared via a modified 3D printing strategy. Compared with traditional 3D materials, these biomimetic materials can significantly improve in vitro cell attachment and proliferation as well as promote in vivo osteogenesis, indicating potential application for cell delivery and bone regeneration.
|
1 Introduction With natural evolution for millions of years, organisms have achieved multifunctional and sophisticated structures, textures or patterns spontaneously for survival, of which the unparalleled advantages have inspired the biomimetic synthesis of materials and structures in recent years. By imitating the specific hierarchical structures and synthetic process of the corresponding organisms from nanoscale to microscale, scientists have synthesized various materials with special structures and novel properties, 1, 2, 3, 4, 5 such as ultrastrong and stiff layered composites inspired by nacre, 6, 7, 8, 9, 10, 11, 12 high adhesion materials, 13, 14 and other special biomimetic materials. 15, 16, 17, 18, 19, 20, 21 Taking advantage of the wisdom of natural organisms, smart materials with improved properties have been prepared and used in many fields. Lotus root, a common vegetable, has unique structure of many parallel channels penetrating itself. Moreover, the parallel multichannel structure maintains in lotus petioles, which connects the lotus root and leaves to form an expedite space ( Figure 1 a, inset). This structure significantly enlarges the contact area and keeps low flow resistance which could effectively promote air (CO 2, O 2 ) and moisture (H 2 O) exchange with external environment while reduces the weight of lotus root itself (Figure 1 a). It is an ideal structure model with low density, high porosity, and low flow resistance, which is promising to be applied in bone tissue engineering (Figure 1 b). The regeneration of large bone defects is still a major clinical challenge. Implantation of synthetic porous scaffolds into large bone defects is an expected approach for guiding and stimulating the formation of new bone tissue. 22, 23 In addition, vascularization plays an important role in the process of new bone formation. It is known that blood vessels are initially formed as endothelial cells organizing into microtubes. Previous studies demonstrated that microfluidic system, a set of microchannels, can be used to induce endothelial cells to form rudimentary vasculature in vitro. 24, 25 It has been indicated in previous studies that incorporating parts of channels through 3D scaffolds could promote oxygen/nutrient perfusion and induce tissue ingrowth along these channels. 26, 27 Therefore, it is reasonable to speculate that materials with lotus root‐like structure (multichannel) possess better angiogenic and osteogenic bioactivity for the regeneration of large bone defects (Figure 1 b). The 3D printing technique, which shows distinct advantages in preparing porous scaffolds with designed macropores for bone regeneration, has drawn much attention in recent years. 28, 29, 30, 31, 32, 33, 34 However, until now, most of 3D printing scaffolds are stacked by solid struts without channel structure (Figure 1 c–f). This simple structure limits the delivery of oxygen and nutrition and further the formation of new bone tissue in the center of defects. 35, 36, 37 Meanwhile, although many 3D printing bioscaffolds with high porosity were prepared for tissue regeneration, these micropores in scaffolds cannot form channel structure, hindering the formation of rudimentary vasculature and interior new bone tissues. 38, 39, 40, 41 Considering the benefits of hollow‐channel materials to vascularization and inspired by the advantages of the unique structure in lotus root, we fabricated the lotus root‐like biomimetic materials with parallel multichannels structure via a modified 3D printing strategy (Figure 1 d–g; Figure S1, Supporting Information) in this study. The physicochemical properties of the lotus root‐like biomimetic materials could be effectively controlled and the lotus root‐like structure could be applied as a passageway to support cell delivery and the formation of new blood vessels and bone tissue in the inner of materials (Figure 1 b). We emphatically explored the preparation and properties of the lotus root‐like biomimetic materials in tissue engineering in comparison with traditional 3D scaffolds. It is believed that the prepared lotus root‐like biomimetic materials will be ideal materials for cell delivery and tissue regeneration. Figure 1 The feasible applications and fabrication of lotus root‐like biomimetic materials inspired by lotus root microstructure. a) The schemata of the functions of lotus root microstructure and the same microstructure in lotus petiole (the inset). b) The schemata of the application in tissue regeneration of lotus root‐like biomimetic materials. c) Traditional 3D printer nozzle with simple shell structure, d) the embedded structure of the modified 3D printer nozzle inspired by lotus root microstructure and e) 3D printing process. f) The traditional 3D printing scaffolds packed by solid struts. g) The lotus root‐like biomimetic materials packed by struts with different numbers of channels. 2 Results and Discussion 2. 1 Fabrication and Characterization of the Lotus Root‐Like Biomimetic Materials To explore the function of lotus root‐like structure in the biomimetic materials, we prepared biomimetic scaffolds with different numbers of channels: 1 channel‐struts‐packed scaffolds, 2 channel‐struts‐packed scaffolds, 3 channel‐struts‐packed scaffolds, and 4 channel‐struts‐packed scaffolds (named as 1CSP, 2CSP, 3CSP, and 4CSP scaffolds, respectively). Traditional solid struts‐packed (TSSP) scaffolds were prepared as control materials. Using this printing strategy, we directly prepared the lotus root‐like biomimetic materials at one go‐off with different raw materials including inorganic ceramic, metal, and polymer materials ( Figure 2 a). The biomimetic materials of different chemical compositions have various morphology and properties (Figure S2, Supporting Information). We can prepare the lotus root‐like biomimetic materials with different shapes (i. e. , cube, disk, and rod) to satisfy various research requirements (Figure 2 b; Figure S3, Supporting Information). The number and size of channels and the dimension of struts can also be well controlled (Figure 2 c; Figure S4, Supporting Information). In this study, silicate‐based bioceramic, akermanite (AKT, Ca 2 MgSi 2 O 7 ) was chosen as representative biomaterial to prepare lotus root‐like biomimetic scaffolds owning to its good biocompatibility with bone marrow stromal cells according to our previous work. 28, 38, 42 Figure 2 Morphology regulation and control of the lotus root‐like biomimetic materials. a) Biomimetic materials of different chemical compositions (e. g. , ceramics, metal, and polymer) with lotus root‐like structure can be prepared by the modified 3D printing strategy. b) 3D micro‐CT images of biomimetic materials with different shapes (i. e. , cube, disk, and rod) and different numbers of channels (e. g. , 4CSP representing four channels in one strut). The shapes of biomimetic materials can be well controlled by the predesigned programs for 3D printing. c) The number and channel size can be well controlled by designing corresponding nozzle with embedded structure. As shown in the optical microscopy and scanning electron microscopy (SEM) images in Figure 3 a, b and Figure S5 (Supporting Information), all the scaffolds are packed by struts (Ø1. 5 mm) with different numbers of channels (Ø400–600 µm). The SEM image in Figure 3 c, optical microscopy images in Figure S6 (Supporting Information), and micro‐CT images of vertical section in Figure 2 b all indicated that the channels in lotus root‐like structure were completely open as compared to TSSP scaffolds without round channels. The surface microstructure of the channels in the sintered biomimetic scaffolds is dense (Figure 3 d). X‐ray diffraction (XRD) analysis indicates that sodium alginate and Pluronic F‐127 which were used as the binder of bioceramic ink, showed no obvious effects on the final crystal phase composition of lotus root‐like bioceramic scaffolds, which maintains pure crystal phase of Ca 2 MgSi 2 O 7 (JCPD 79–2425) (Figure 3 e). Generally, 3D scaffolds prepared by ink‐jet 3D printing technique are packed layer by layer. We adjusted the packing patterns by writing corresponding program to 3D printing system and prepared three kinds of biomimetic scaffolds with different packing patterns (i. e. , cross packing pattern, quartet close packing pattern, and hexagonal close packing pattern) (Figure 3 f; Figure S7, Supporting Information). With the increase of channels, the total porosity, and specific surface area significantly increased. The channel numbers can observably influence the specific surface area of the sintered scaffolds. In addition, both the channel numbers and packing patterns showed obvious impacts on porosity and compressive strength of the sintered scaffolds. The 4CSP scaffolds with cross packing pattern obtained the highest porosity of up to 80% while the porosity of TSSP scaffolds with the same packing pattern was only 58%. The biomimetic scaffolds with hexagonal close packing pattern had the best compressive strength (range of 30–46 MPa) but lower porosity than scaffolds with other two packing patterns (Figure 3 g, h). The specific surface area of 4CSP scaffolds (3. 86 ± 0. 39 × 10 −3 m 2 g −1 ) was twice higher than that of TSSP scaffolds (1. 40 ± 0. 05 × 10 −3 m 2 g −1, Table S1, Supporting Information). Thus, much more surface area was available in the biomimetic scaffolds for cell and tissue attachment. The porosity, specific surface area, and mechanical strength of lotus root‐like biomimetic scaffolds can be well controlled by modulating the channel number or packing patterns, which is of great importance to satisfy the different mechanical requirements for human body. Therefore, we realized the controllable preparation of the biomimetic materials in both chemical compositions and physical structures. Figure 3 Characterizations of biomimetic silicate‐based bioceramic (Akermanite, (AKT), Ca 2 MgSi 2 O 7 ) scaffolds with lotus root‐like microstructure. a) Optical microscope and b) SEM images show the lotus root‐like biomimetic structure. The materials are packed by struts (Ø1. 5 mm) with different numbers of channels (Ø400–600 µm). c) SEM image for cross‐section showing that the hollow channels are completely open. d) SEM image for the inner surface microstructure of the channels. e) XRD analysis demonstrating the pure crystal phase of silicate‐based bioceramic scaffolds. f) Packing patterns (i. e. , cross packing pattern, quartet close packing pattern, and hexagonal close packing pattern) have significant influence on the g) porosity and h) compressive strength of the lotus root‐like biomimetic materials. Porosity and mechanical property can be well controlled by predesign of packing pattern and number of channels. ( n = 5, ** P < 0. 01, *** P < 0. 001. ) 2. 2 In Vitro Bioactivity Analysis of the Lotus Root‐Like Biomimetic Materials The porous architecture and the porosity of the scaffolds play a critical role in promoting nutrient diffusion, blood vessel ingrowth, and tissue regeneration. 36, 43 A potential application of the lotus root‐like biomimetic materials is bone regeneration. In this study, rabbit bone marrow stem cells (BMSCs) were seeded on the lotus root‐like biomimetic materials (1CSP, 2CSP, 3CSP, and 4CSP) with TSSP group as control. The attachment and morphology of BMSCs on the struts' surface of TSSP group and biomimetic groups were observed by SEM and confocal laser scanning microscopy ( Figure 4 a–e; Figure S8, Supporting Information). As shown in Figure 4 a, b and Figure S8a (Supporting Information), all scaffolds support BMSCs attachment and the cells closely adhere to the scaffolds by numerous filopodia after 3 d of culture. It is found that BMSCs adhere not only on the outer surface but also on the inner surface of lotus root‐like channels. As shown in Figure 4 c–e and Figure S8b (Supporting Information), the cytoskeleton of BMSCs adhering on the scaffolds was stained in green with fluorescein isothiocyanate (FITC) after culturing for 3 d. The confocal laser scanning microscope (CLSM) images demonstrated that BMSCs not only attached uniformly on the surface of the scaffolds but also penetrated into the channels and attached on the walls of lotus root‐like structures (see Movies S1–S3, Supporting Information). More BMSCs were delivered in the biomimetic groups than that of TSSP group. The amount of the delivered BMSCs showed positive correlation with the number of channels in the biomimetic groups. In addition, with increasing number of hollow channels, biomimetic materials showed significant improvement on cell initial attachment at hour 8, 16, and 24 and proliferation activity at day 3 and day 7 (Figure 4 f, g). The lotus root‐like structure in the biomimetic materials may be beneficial for enhancing oxygen and nutrient distribution in the inner of scaffolds. The lotus root‐like channels of the biomimetic scaffolds can be used for delivering cell and nutrition in tissue regeneration. Figure 4 BMSCs cultured in TSSP, 1CSP, 2CSP, 3CSP, and 4CSP‐AKT bioceramic scaffolds for different time periods. a, b) SEM images of BMSCs attached in the channels of biomimetic scaffolds after culturing for 3 d. b) BMSCs adhered on the scaffolds via numerous filopodia as shown by the yellow arrows. c–e) The CLSM images for the morphology and cytoskeleton of BMSCs on the surface of struts and channels in TSSP, 1CSP, 2CSP, 3CSP, and 4CSP scaffolds after culturing for 3 d. d) Surface magnified image and e) 3D image shows that BMSCs penetrated into channels and attached on the inner walls of channels. f) The amount of adhered BMSCs after 4, 8, 16, and 24 h culturing and g) the proliferation activity of BMSCs in different scaffolds after 1, 3, and 7 d of incubation respectively, detected by the CCK‐8 assay. The initial adhered cells and their proliferation activity enhanced with the increase of the channel numbers in the biomimetic scaffolds. ( n = 6, ** P < 0. 01, *** P < 0. 001. ) 2. 3 In Vivo Bioactivity Analysis of the Lotus Root‐Like Biomimetic Materials To investigate the effect of lotus root‐like biomimetic scaffolds on the vascularization and bone regeneration, the rat muscle model and rabbit calvarial defects model were applied to evaluate both the angiogenesis and osteogenesis processes. Four weeks after implantation in the rat muscle, samples were perfused by microfil to label the blood vessels and taken for undecalcified sections, partial sections were stained with DAPI (a specific coloring agent to stain cell nucleus into blue) to detect the bioactivity of biomimetic scaffolds, other unstained sections were used to detect the newly formed blood vessels. The DAPI‐stained sections showed that the channels throughout the struts in biomimetic materials were filled with cells ( Figure 5 a; Figure S9a, Supporting Information). The perfused microfil in the channels could be clearly detected in the histological images by different colors (blue filate tissues represent blood vessels, see Figure 5 b, c). In the histological images, many blood vessels could be found in the lotus root‐like channels of the biomimetic groups while there were no blood vessels in TSSP group due to its solid struts (Figure S9b, Supporting Information). These results demonstrate that the lotus root‐like structure in the scaffolds enhances the angiogenic process at the early stage of tissue regeneration. In addition to the rat muscle implantation model, the rabbit calvarial defect model was applied to testify significantly improved osteogenesis capacity of the biomimetic materials as compared to the TSSP group. More newly formed bone tissue was observed in 3CSP group compared with 1CSP and TSSP groups after 12 weeks of implantation in the 3D micro‐CT images. The images of calvarial defect's surface showed that the bone defects in 3CSP group healed well and the peripheral bone grew tightly with the scaffolds, as compared to that of 1CSP and TSSP groups (Figure 5 d). The micro‐CT images of cross‐sections showed that newly formed bone tissues had grown into the channels (Figure 5 e). Besides, Van Gieson's staining results displayed that the newly formed bone tissue in the TSSP group was mainly detected in the periphery of the defects, while more newly formed bone was mainly detected in both the periphery and center of the bone defects in 1CSP and 3CSP groups, especially in the 3CSP group after implantation for 12 weeks (Figure 5 f; Figure S10, Supporting Information). Moreover, according to the quantitative analysis of micro‐CT, a significantly higher BV/TV value for new bone volume was revealed for the 3CSP (24. 6% ± 3. 05%) group as compared with 1CSP (16. 2% ± 0. 503%) and TSSP (12. 0% ± 1. 17%) groups at week 12. Under a histomorphometric assay, the higher percentage of new bone area was observed in the 3CSP group (19. 55% ± 2. 88%) compared with 1CSP (7. 41% ± 1. 43%) and TSSP groups (4. 27% ± 0. 939%) at week 12 (Figure 5 g). Therefore, this lotus root‐like structure can successfully induce blood vessels and new bone tissues to grow into the inner of the biomimetic materials and effectively promote the bone defect healing. The lotus root‐like biomimetic materials present better angiogenic and osteogenic stimulatory capability than traditional 3D printing scaffolds according to in vivo bioactivity analysis. Figure 5 Characterizations of the lotus root‐like biomimetic scaffolds to enhance in vivo angiogenesis in rat muscle implantation and osteogenesis in rabbit calvarial defects. a) Fluorescence image of histological sections of biomimetic scaffolds stained with DAPI. b, c) The sections from microfil‐perfused samples were used to detect the new blood vessels, b) optical microscope image of 3CSP biomimetic scaffolds with blood vessels perfused by microfil, c) the magnified image of blood vessels (in blue) in the lotus root‐like structure. d) Typical 3D reconstruction micro‐CT images of the edges between materials and rabbit calvarial defects, and e) micro‐CT cross‐section images of rabbit calvarial defect regions (red for new bone tissues, green for materials). f) The undecalcified histological sections stained with Van Gieson's picrofuchsin, newly formed bone tissues (in red) can be well observed (blue arrows point to the new bone). g) Micro‐CT reconstruction analysis of the volume ratio of the newly formed bone to the defect regions (BV/TV) and histological morphometric analysis of the area of the newly formed bones in the whole defect regions at week 12. The 3CSP biomimetic materials showed significantly improvement in bone regeneration as compared to 1CSP and TSSP materials. ( n = 6, * P < 0. 05, ** P < 0. 01, and *** P < 0. 001. ) 3 Conclusions Inspired by the root of the natural lotus plant, we successfully prepared the biomimetic materials with lotus root‐like structure via a modified 3D printing strategy, which breaks the limitation of traditional 3D printing method. We are able to prepare the lotus root‐like biomimetic materials with different raw materials including ceramics, metal and polymer. Furthermore, their shape, packing pattern, porosity, specific surface area, mechanical property, and the lotus root‐like structure (the size and number of hollow channels and the size of struts) can be well controlled. Our results suggest that the porosity and specific surface area could be distinctly improved in the biomimetic materials. Compared to traditional 3D printing materials, the lotus root‐like biomimetic materials significantly improved in vitro BMSCs attachment and proliferation as well as in vivo osteogenesis and angiogenesis, indicating that the lotus root‐like biomimetic materials are more suitable for cell delivery and regeneration of large bone defects. 4 Experimental Section Materials : The AKT (Ca 2 MgSi 2 O 7 ), Ca, Mg, and Si‐containing bioceramic powders were synthesized by a sol–gel process using tetraethyl orthosilicate ((C 2 H 5 O) 4 Si, TEOS), magnesium dinitrate hexahydrate (Mg(NO 3 ) 2 ·6H 2 O), and calcium nitrate tetrahydrate (Ca(NO 3 ) 2 ·4H 2 O). 44 The other ceramic and metal (ZrO 2, Al 2 O 3, and Fe) powders were purchased from Kunshan Chinese Technology New Materials Co. , Ltd. In order to avoid blockage of the printing nozzle during the 3D printer working, the synthetic powders were ground to a particle size less than 74 µm by sieving through 200 meshes. To prepare the printable ink, 3. 50 g of corresponding (AKT, ZrO 2, Al 2 O 3, and Fe) powders was mixed with 0. 10 g of sodium alginate powder (Alfa Aesar) and 1. 54 g of Pluronic F‐127 (20 wt%) (Sigma‐Aldrich) aqueous solution and then stirred until homogeneous paste was achieved. Design and Preparation of 3D Printer Nozzles and Printable Ink : To design new printing nozzle for lotus root‐like biomimetic materials, different numbers of parallel needles were embedded into traditional nozzle and the number of needles embedded in the nozzle matched to the number of channels (embedded 1, 2, 3, and 4 needles for 1CSP, 2CSP, 3CSP, and 4CSP scaffolds, respectively). Simultaneously, a hole was bored through the center of every embedded nozzle to make sure the inks could be ejected out smoothly (Figure S1, Supporting Information). To prevent severe deformation or collapse of channels during the printing process, in this study, certain amounts of sodium alginate (2. 0 wt%) and Pluronic F‐127 polymer solution (e. g. , 20 wt%) with the under‐printed powders (ceramic/metal/polymer powders) were applied to prepare the inks with suitable rheological characteristics and mechanical stability. Fabrication and Characterizations of the Lotus Root‐Like Biomimetic Materials : The 3D printing system was developed by Fraunhofer IWS (Dresden, Germany) based on the Nano‐Plotter device from GeSiM (Grosserkmannsdorf, Germany). After the scaffolds were printed, the ceramic and metal scaffolds were dried overnight at room temperature and then sintered at suitable conditions (AKT 1350 °C for 3 h, ZrO 2, and Al 2 O 3 1520 °C for 3 h, Fe 1300 °C for 3 h in Ar atmosphere) to remove the sodium alginate and F‐127 phases, and the ceramic and metal particles were densified to form the final lotus root‐like biomimetic materials with certain longitudinal shrinkage and volume shrinkage. The raw polymer biomimetic materials of sodium alginate went through freeze drying process via a freeze drying machine (SRK, GMBH, Germany) to form the final scaffolds. The lotus root‐like materials still maintained the designed biomimetic structures, including macropores (outside of the struts) and open channels (inside of the struts), despite 15–25% longitudinal shrinkage and around 22–58% volume shrinkage occurring for the whole materials after high temperature sintering or freeze drying (Table S2, Supporting Information). The morphology of the sintered materials with different physicochemical properties was observed by optical microscopy (S6D, Leica, Germany). The 3D images of biomimetic materials were reconstructed by micro‐CT (SKYSCAN1172, SKYSCAN, Belgium). The macropore and microstructure of the sintered scaffolds were characterized by SEM (JSM‐6700F, JEOL, Japan). XRD (D8ADVANCE, Bruker, Germany) analysis indicated that alginate and Pluronic F‐127, as the solution of bioceramic ink, have no obvious effect on the sintered crystal phase composition of AKT bioceramics. The compressive strength of the obtained materials (10 × 10 × 15 mm) with different packing patterns and porosities were tested using a computer‐controlled universal testing machine (AG‐I, Shimadzu, Japan) at a cross‐head speed of 0. 5 mm min −1. Five samples were tested for each kind of scaffold. The porosity was measured by Archimedes' principle, a liquid displacement method. In brief, the ceramic scaffolds were first dried at 100 °C overnight, weighed, and marked as M 1. Then, the ceramic scaffolds were immersed in water and placed under vacuum until no bubbles appeared. The weight of the scaffolds with water‐filled pores was marked as M 2. Finally, the ceramic scaffolds were immersed in water, and the buoyant weight was marked as M 3. The porosity ( P ) was calculated using Equation (1) (1) P = M 2 − M 1 / M 2 − M 3 × 100 % In Vitro Bioactivity Analysis of the Lotus Root‐Like Biomimetic Materials : BMSCs were isolated from the femurs of rabbits (one month old) and cultured in Dulbecco's modified Eagle's medium (DMEM, HyClone, China) supplemented with 10% fetal calf serum (Invitrogen) and penicillin–streptomycin (Invitrogen). For the evaluation of cell attachment, BMSCs were seeded in the lotus root‐like biomimetic scaffolds and TSSP scaffolds at an initial density of 1 × 10 4 cells per scaffold and placed in 24‐well culture plates for culture. After incubation for 3 d, the scaffolds were rinsed with phosphate‐buffered saline (PBS) and fixed with 2. 5% glutaraldehyde. The additional glutaraldehyde was removed by washing with PBS, followed by sequential dehydration in graded ethanol (30, 50, 70, 90, 95, 100, and 100 v/v%). The specimens were dried in hexamethyldisilazane for 30 min before SEM analysis. To observe the cells attached in the interior of the scaffolds, the cells were fixed with 4% paraformaldehyde solution and permeabilized in 0. 1% Triton‐X‐100, and stained with FITC‐labeld Phalloidin solution (stock solution in methanol diluted in 1:100, Cytoskeleton Inc. , USA) at room temperature for 45 min. Finally, after a brief PBS wash, DAPI solution (5 mg mL −1 ) was added for counterstaining cell nucleus. Confocal images were obtained on a confocal laser scanning microscope (Leica TCS SP8). CCK‐8 assay was performed to assess the initial attachment and proliferation activity of BMSCs. Briefly, BMSCs were seeded in the lotus root‐like biomimetic scaffolds and TSSP scaffolds in 24‐well culture plates and cultured for 4, 8, and 16 h and 1, 3, and 7 d. The growth medium was replaced by culture medium with CCK‐8 (1 mL, V CCK‑8 : V DMEM = 1:9) at each time point for 30 min. The absorbance was measured at λ = 450 nm in a microplate reader (Epoch microplate spectrophotometer, BioTek Instruments, USA). All the results are presented as optical density values minus the absorbance of blank wells. In Vivo Bioactivity Analysis of the Lotus Root‐Like Biomimetic Materials : All experiments were performed in compliance with the relevant laws and institutional guidelines, and the Animal Care and Use Committee of Shanghai Jiaotong University approved the experiments. All animal experiments and the related experimental protocols, used in the present study, were approved by the Animal Care and Experiment Committee of the Ninth People's Hospital. To evaluate both of angiogenesis and osteogenesis processes, all total of 15 white rats and 9 New Zealand white rabbits (male, 12‐month‐old average) were taken for the model and rabbit calvarial defects model. At week 4 after the rat muscle implantation, the rats were sacrificed and perfused with microfil (Flow Tech, Carver, MA, USA). All samples were extracted and immersed in 10% buffered formaldehyde solution for 48 h and then dehydrated and embedded in polymethyl methacrylate (PMMA) for preparing the undecalcified sections. The tissue sections were made using a microtome (Leica, Nusseloch, Germany). Without staining, the sections were observed under a microscope to record the blue microfil for new blood vessels. And then, the sections were stained with DAPI (a specific coloring agent to stain cell nucleus into blue) to detect the amount of cells growing into the lotus root‐like structures. Nine New Zealand white rabbits (male, 12‐month‐old average) were taken for the rabbit calvarial defects implantation. After the general anaesthesia, an incision was made on the skull, and then the surface periosteum and Ø10 mm bone defects were completely removed from the both side of the skull, respectively, using electric drill. Three kinds of 3D printed scaffolds (disk, Ø10 × 3 mm), including 1CSP scaffold, 3CSP scaffold and TSSP scaffold, were implanted into the calvarial defects ( n = 6). After the implantation for 12 weeks, these rabbits were sacrificed and the samples were extracted for bone regeneration evaluation. First, the samples were scanned by the Micro‐CT device and reconstructed into the three dimensional images to display the gross morphology. Afterward, the samples were dehydrated and embedded in PMMA. The undecalcified sections were made using a microtome (Leica, Nusseloch, Germany) and stained with the Van Gieson's picrofuchsin dye. On the stained sections, the percentage of the newly formed bone area in the whole defect regions was calculated. Statistical Analysis : All data were displayed as the mean ± standard deviation. The statistical analysis was performed using an SAS 8. 2 software. Results were analyzed for significance (* P < 0. 05, ** P < 0. 01, and *** P < 0. 001) by Student's t‐ test (two groups) or one‐way ANOVA followed by Tukey's post hoc test for multiple comparisons. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700402
| 2,018
|
Advanced Science
|
A Modular Strategy to Engineer Complex Tissues and Organs
|
Abstract Currently, there are no synthetic or biologic materials suitable for long‐term treatment of large tracheal defects. A successful tracheal replacement must (1) have radial rigidity to prevent airway collapse during respiration, (2) contain an immunoprotective respiratory epithelium, and (3) integrate with the host vasculature to support epithelium viability. Herein, biopolymer microspheres are used to deliver chondrogenic growth factors to human mesenchymal stem cells (hMSCs) seeded in a custom mold that self‐assemble into cartilage rings, which can be fused into tubes. These rings and tubes can be fabricated with tunable wall thicknesses and lumen diameters with promising mechanical properties for airway collapse prevention. Epithelialized cartilage is developed by establishing a spatially defined composite tissue composed of human epithelial cells on the surface of an hMSC‐derived cartilage sheet. Prevascular rings comprised of human umbilical vein endothelial cells and hMSCs are fused with cartilage rings to form prevascular–cartilage composite tubes, which are then coated with human epithelial cells, forming a tri‐tissue construct. When prevascular– cartilage tubes are implanted subcutaneously in mice, the prevascular structures anastomose with host vasculature, demonstrated by their ability to be perfused. This microparticle–cell self‐assembly strategy is promising for engineering complex tissues such as a multi‐tissue composite trachea.
|
1 Introduction Patients suffering from tracheal stenosis have a significantly reduced quality of life. The diseased region of the trachea cannot be resected when more than half of the airway is affected in adults. 1 Thus, tracheal tissue engineering has the exciting potential to fill this gap and will have a tremendous impact for the patients in need. A variety of tracheal replacement strategies have been developed in vitro and implanted in vivo, both in animals and human patients, including cell‐free artificial prostheses, 2 autografts, 3 intact allografts or decellularized allografts often seeded with the recipient's own cells, 4 high‐cell density constructs derived from mature cell sources, 5 and primarily scaffold‐based tissue engineered constructs. 6 Despite the broad range of approaches, each has its own shortcomings ranging from the technical, such as restenosis, [[qv: 1c, 3b]] to the practical, such as lack of tissue availability. The adult trachea has 18–22 tracheal cartilage rings, which keep the airway patent during respiration. [[qv: 1a]] The tracheal lumen is epithelialized with a ciliated mucosa which acts as a protective barrier, humidifies inspired air and clears secretions from the lungs. [[qv: 1a]] A vascularized elastic tissue encases the cartilaginous rings and houses blood vessels that provide nutrients and oxygen to tracheal tissues. [[qv: 1a]] Keeping these attributes in mind, a functional tracheal replacement must therefore meet critical design criteria that include: (1) having radial rigidity, (2) possessing a luminal epithelium, and (3) supporting neovascularization to restore an open and functional airway while avoiding restenosis, bacterial infections, and ischemic necrosis. 7 A modular tracheal engineering approach that employs scaffold‐free 3D tissue building blocks is an attractive option to address these three functional requirements. The ultimate goal is to fuse three types of engineered tissues into a continuous tracheal replacement by employing a custom‐built culture system, coupled with localized bioactive factor delivery, to provide spatial and temporal control over tissue formation. To address the requirement for mechanical integrity of the engineered trachea, our group recently demonstrated an approach for the production of stiff, human mesenchymal stem cell (hMSC)‐derived cartilage tubes by fusing cartilaginous rings in a custom culture system. 8 A key feature of this system is the incorporation of growth factor releasing gelatin microspheres that allow for controlled chondrogenic signal presentation to drive cartilage formation. In this work, engineered cartilaginous rings and tubes were shown to have mechanical properties similar or better than comparably sized native rat trachea segments. 8 This study demonstrates that the engineered tissues can be formed in clinically relevant specific geometries and sizes that match human tissue dimensions by modulating the thickness and diameter of the engineered cartilage rings and tubes. To address the second element of a mature engineered trachea, an epithelium must be able to be cultured on the luminal surface of the construct. [[qv: 1c]] To that end, we use a human bronchial epithelial cell line (hBECs) to engineer an epithelial—cartilaginous bilayer as proof‐of‐principle to demonstrate that the engineered cartilage tissue can support an epithelial lining. Epithelialized cartilage bilayers were cultured either submerged in medium or at an air–liquid interface (ALI), which has been shown in previous studies to be beneficial for respiratory epithelial maturation. 9 Next, vascularization is essential for the success of any tissue engineered construct and is currently a major limitation in the field of tissue engineering. The diffusional limitation of oxygen requires all metabolically active cells within the body to reside within ≈150–200 µm of a capillary, 10 thus necessitating the microvascularization of thick tissue engineered constructs such as a trachea. Guiding the tissue engineered construct through the initial phases of vasculogenesis might accelerate neovascularization and anastomosis with host vasculature upon implantation. 11 Human umbilical vein endothelial cells (HUVECs) have been cocultured with various stromal supporting cells for vasculogenic purposes. 12 Here, HUVECs were cocultured with hMSCs in vasculogenic media to generate prevascular tissue rings, where the cells self‐assembled into early microvascular structures. These were then fused with cartilage rings to form prevascular—cartilage composite tissue tubes. Finally, these composite tissues were (1) seeded with human tracheal epithelial cells (HTEs) to engineer a tri‐tissue construct with spatial control of tissue composition and phenotype and (2) implanted subcutaneously in mice to demonstrate the ability of the tissues to anastomose with host vasculature. The system presented herein has the potential to allow for the fabrication of many different complex, vascularized tubular tissues and organs. By adjusting the composition of the individual tissue rings, structures such as blood vessels, gastrointestinal tract, and ureters may be developed. This paper describes the application of this base technology to respiratory airway engineering for functional tracheal replacement. A schematic depicting fabrication of each of these tissues is presented in Figure 1. Figure 1 Schematic overview of the cartilage (Part Ia and Ib), epithelial–cartilage (Part II) and prevascular–cartilage and tri‐tissue (Part III) engineering components of this work aimed toward the generation of a multi‐tissue tracheal construct for airway repair. 2 Results 2. 1 Part Ia—Cartilage Rings with Defined Wall Thickness 2. 1. 1 Macroscopic and Histological Assessment With the goal of controlling the thickness of cartilage rings, the number of hMSCs used to form the rings was varied from 0. 1 to 0. 4 million cells per ring, and TGF‐β1 was presented either in the media or from incorporated microspheres (MS). Rings assembled from a greater number of hMSCs had a higher frequency of ring formation, were grossly ( Figure 2 A) and quantitatively thicker (Figure 2 B) and were heavier (Figure 2 C). Ring thickness could be modulated by 25–35% by increasing cell number. Samples that did not form rings resulted in C‐shaped constructs or 1–3 individual aggregates. Microsphere‐containing rings were significantly thicker than hMSC‐only constructs and all engineered rings were thicker than cartilage ring segments from rat tracheas. Glycosaminoglycan (GAG; a prevalent polysaccharide in hyaline cartilage) staining was strong in all engineered tissues and appeared similar to staining in rat tracheal sections (Figure 2 A). Type II collagen, the main collagenous component of hyaline cartilage ECM, was also apparent in both types of engineered cartilage (Figure 2 A). Compared to hMSC‐only constructs, hMSC + MS rings appeared to have less of a fibrous capsule on the outer edge of the tissue. Figure 2 Analysis of tissue engineered cartilage rings formed with different cell numbers. A) Macroscopic images of rings were acquired, GAGs were stained with Safranin O (pink/red), immunohistochemistry was performed for human type II collagen (red) and the frequency of tissue engineered cartilage ring formation with varied cell number were reported compared to rat trachea. MS, microspheres; Fast Green counterstain (blue/green). Black scale bars are 200 µm; white scale bars are 2 mm. Images in a single row are the same magnification. B) Ring thickness was measured in tissue engineered hMSC (light gray) and hMSC + MS (dark gray) cartilage rings ( N = 3) and compared to rat tracheal segments (black, N = 4). C) Construct wet weight, D) GAG content, E) DNA content, F) GAG normalized to DNA, G) GAG normalized to tissue wet weight and H) DNA normalized to tissue wet weight were acquired from cartilage rings ( N = 3–4) and, for (F), (G) and (H), rat tracheal segments ( N = 4). I) Failure load during pull‐to‐failure uniaxial testing (pictured in I, inset) and J) ultimate tensile strength (UTS; load normalized by area) were measured in engineered rings ( N = 3) and compared to rat tracheal segments (black, N = 4). MS = microspheres. a, b, c, d: hMSC groups without common letter differ ( p < 0. 05); 1, 2, 3, 4: hMSC + MS groups without common number differ ( p < 0. 05); #: significantly different than hMSC group ( p < 0. 05); *: rat trachea is significantly different compared to all other groups ( p < 0. 05); R: significantly different than rat trachea ( p < 0. 05). Data shown as mean ± SD. 2. 1. 2 Biochemical Analysis By measuring the amount of GAG and DNA in the variable thickness cartilage rings, quantitative differences in tissue composition as a function of the number of cells used during fabrication were elucidated (Figure 2 C–H). Rings composed of more cells contained more GAG (Figure 2 D) and DNA (Figure 2 E). Chondrogenesis was not affected by cell number as the GAG/DNA (Figure 2 F) and GAG/wet weight (Figure 2 G) values were the same in rings made with varying cell numbers. As expected, hMSC + MS rings weighed more (Figure 2 C) and contained significantly more GAG (Figure 2 D, except 0. 1 million cells per ring) and GAG/DNA (Figure 2 F) than hMSC‐only rings. Compared to rat tracheal segments, engineered tissues had significantly more GAG/DNA (Figure 2 F), more GAG/wet weight (Figure 2 G), and less DNA/wet weight (Figure 2 H). This was anticipated because the tracheal segments are composed of other tissues in addition to cartilage. 2. 1. 3 Mechanical Analysis All engineered rings required a significantly greater load to rupture under uniaxial tension compared to rat tracheal rings (1. 21 ± 0. 16 mm vertically) (Figure 2 I), demonstrating biomechanical functionality and potential ability to maintain tracheal luminal patency. In addition, the ultimate tensile stress (UTS, failure load normalized by loaded area) was at least as high as that of the rat trachea in all groups (Figure 2 J) signifying that the engineered cartilage tissue is at least as strong as the native rat tracheal tissue. Similar failure loads and UTS values across the different thickness rings (except the UTS of 0. 2 million cells per ring without microspheres) indicate that geometry does not affect the mechanical properties of the engineered cartilage for the range of wall thickness examined. 2. 2 Part Ib—Cartilage Tubes with Defined Lumen Diameters 2. 2. 1 Macroscopic and Histological Assessment To demonstrate the scalability of this system, cartilaginous rings comprised of hMSCs and TGF‐β1‐microspheres were fabricated with three different inner diameters and then fused into cartilage tubes with defined lumen dimensions. Rings assembled in larger diameter wells had a lower frequency of ring formation even though the cell and microsphere numbers were linearly scaled with the diameter of the wells ( Figure 3 A). Cartilage rings of each diameter successfully fused into firm cartilage tubes. The lumen of the tubes was smooth while the outer wall was ribbed. Rabbit tracheal sections resembled the dimensions and gross morphology of 6 mm engineered cartilage tubes. The wall thickness of the 2, 6, and 12 mm diameter engineered cartilage tubes was not affected by the lumen size (Figure 3 B). Additionally, 6 mm cartilage tube walls were only 16% thicker than rabbit tracheal walls, which had similar lumen diameters. GAG staining was strong in all engineered tubes but appeared weaker than staining in rabbit tracheal sections (Figure 3 A). Type II collagen stained uniformly with strong intensity in the engineered cartilage tubes (Figure 3 A). Figure 3 Macroscopic and histological images, and biochemical and mechanical analysis of cartilage tubes of 2, 6, and 12 mm inner diameters and the frequency of formation of tissue engineered rings used to make cartilage tubes. A) Macroscopic images of tubes were acquired, GAGs were stained with Safranin O (pink/red), immunohistochemistry was performed for human type II collagen (red) and the frequency of ring formation of 2, 6, and 12 mm inner diameter cartilage tubes was recorded. Tissue tubes were cut to show transverse and longitudinal sections. MS, microspheres; Fast Green counterstain (blue/green). Black scale bars are 200 µm; white scale bars are 4 mm. Images without scale bars in a single row are the same magnification. B) Tube wall thickness was measured in five‐ring engineered cartilage tubes (gray, N = 3) and compared to rabbit tracheal segments (black, N = 6). C) Construct wet weight, D) GAG content, E) DNA content, F) GAG normalized to DNA, G) GAG normalized to tissue wet weight, and H) DNA normalized to tissue wet weight were acquired from tissue engineered two‐ring cartilage tubes ( N = 4–5). I) Load required to collapse 80% of the five‐ring engineered tube lumen diameter and J) tube recoil after luminal compression (I inset) were measured ( N = 3) and compared to rabbit tracheal segments (black, N = 6). MS = microspheres. a, b, c: hMSC + MS groups without common letter differ ( p < 0. 05); *: rabbit trachea is significantly different compared to all other groups ( p < 0. 05). MS, microspheres. Data shown as mean ± SD. 2. 2. 2 Biochemical Analysis Biochemical evaluation of cartilage tubes (Figure 3 C–H) revealed that larger diameter tubes were heavier (Figure 3 C) and contained more GAG (Figure 3 D) and DNA (Figure 3 E). Interestingly, tubes with larger lumen diameter (6 and 12 mm) had decreased GAG/DNA (Figure 3 F) and GAG/wet weight (Figure 3 G) compared to 2 mm lumen diameter tubes while maintaining similar amounts of DNA normalized to wet weight (Figure 3 H) indicating somewhat decreased chondrogenesis in the larger tubes. 2. 2. 3 Mechanical Analysis To evaluate the biomechanical properties of engineered cartilage tubes and compare their behavior to native rabbit tracheas, the lumen of each tube or trachea was compressed by applying a perpendicular load (Movies S1–S3, Supporting Information, correspond to the testing of 2, 6, and 12 mm inner diameter tubes, respectively). All engineered tubes required a significantly smaller load to collapse 80% of the lumen diameter compared to rabbit tracheal segments of similar length (Figure 3 I). The load at 80% collapse was significantly larger for 2 mm diameter tubes compared to 6 and 12 mm tubes. Subsequently, the compressive load was removed and the outer diameter of each tube before and after collapse was compared to calculate the recoil of engineered tubes and rabbit tracheas. All luminally collapsed tubes recoiled to nearly 100% of the original outer diameter with no significant differences between the groups (Figure 3 J). 2. 3 Part II—Epithelialized Cartilage Bilayers 2. 3. 1 Histological Assessment and Immunohistochemistry As a proof‐of‐principle of establishing an epithelial lining on the tissue engineered tracheas, epithelized cartilage (EC) bilayers were engineered by seeding hBECs on the surface of 2‐week‐old microsphere‐containing hMSC sheets. Bilayer tissues cultured at ALI, which is typically employed for respiratory epithelial culture, were compared to bilayers cultured submerged in media for 4 and 7 days (d). Hematoxylin and eosin (H&E) staining showed uniformly distributed chondrocytes within a homogeneous ECM that stained positive for GAG in the microsphere‐containing hMSC layer of all single and bilayer tissues ( Figure 4 A). However, GAG staining appeared less intense in the cartilage layer of the EC bilayer sheets compared to cartilage only sheets (C sheets) alone in 50:50 basal pellet medium (BPM):bronchial epithelial growth medium (BEGM) and C control sheets in 100% BPM. Additionally, the cartilage layer of the epithelial–cartilage bilayers cultured at ALI showed a time‐dependent decrease in GAG staining from 4 to 7 d (Figure 4 A). Figure 4 Histologic staining and thickness measurements of epithelialized cartilage sheets and controls. Epithelial–cartilage bilayer sheets and controls were stained with A) hematoxylin & eosin, Safranin O for GAG (pink/red) and B) antitype II collagen (red) and anticytokeratin (red) for epithelial cells. ALI, sheets cultured at air liquid interface for 4 or 7 d; submerged, sheets cultured submerged in media for 7 d. BPM, chondrogenic basal pellet media; BEGM, bronchial epithelial growth media. Fast green counterstain (blue/green). Black scale bars are 40 µm; white scale bars are 200 µm. Images without scale bars in a single row are the same magnification. C) Cartilage layer thickness and D) epithelial layer thickness were measured from histological images of cartilage (C) sheets (dark gray and patterned (control)), epithelial (E) sheets (white (control)) and epithelial–cartilage (EC) bilayers (light gray) ( N = 3). ALI, sheets cultured at air–liquid interface for 4 or 7 d; Subm. ; sheets cultured submerged in media for 7 d. BPM, chondrogenic basal pellet media; BEGM, bronchial epithelial growth media. *: significantly different compared to control cartilage sheets in (C) and control epithelial sheets in (D) ( p < 0. 05); a, b, c: Bilayer sheets without common letter differ ( p < 0. 05). Data shown as mean ± SD. Immunostaining for type II collagen was strong in all cartilage layers further confirming cartilaginous ECM synthesis (Figure 4 B). Unlike GAG deposition, however, there were differences in the distribution of type II collagen between the experimental groups and the control C sheets in BPM. Control C sheets in BPM had more uniform type II collagen deposition throughout their thickness compared to C sheets in 50:50 BPM:BEGM and EC bilayer sheets, which had decreased type II collagen on the upper surface of the hMSC layers. Cartilage layers were negative for epithelium‐specific cytokeratin across groups. In contrast, localized cytokeratin staining was observed in all epithelial layers of EC bilayer sheets comparable to control E sheets. 2. 3. 2 Thickness Quantification of Cartilage and Epithelial Components To evaluate the quality of the epithelial and cartilage layers by another metric, the thickness of each tissue portion was quantified as a measure of spatial control over cell phenotype and tissue development. The cartilage layers of all C sheets and EC bilayer sheets cultured in the 50:50 mixture of BPM and BEGM were thinner compared to controls submerged in normal BPM, but differences were significant only for the C sheets ALI 4 d, EC bilayer ALI 7 d, and submerged 7 d groups (Figure 4 C). Neither the ALI culture conditions nor duration significantly affected the thicknesses of hMSC layers in C sheets cultured in mixed media (Figure 4 C). The cartilage layer of the EC bilayers showed a slight but not significant time‐dependent decrease in thickness from 4 to 7 d at ALI (Figure 4 C). Moreover, all epithelial layers of EC bilayer sheets were significantly thinner compared to control E sheets cultured on cell culture inserts at ALI for 7 d (Figure 4 D). Epithelial layers at ALI showed a significant time‐dependent increase in thickness from 4 d to 7 d (Figure 4 D). Compared to ALI 7 d, E sheets submerged for 7 d were significantly thinner (Figure 4 D). 2. 4 Part III—Prevascular–Cartilage Composite Tubes and Tri‐Tissue Tracheas 2. 4. 1 Macroscopic and Histological Assessment While localized tissue differentiation and maturation are critical for the complex spatial organization of a trachea, coherent tissue fusion between the incorporated tissue types is also important for organ functionality. Timing may play a critical role in the success of tissue fusion 13 as well as development of tissue‐specific phenotypes. As a result, cartilage and prevascular rings were cultured individually for varying time periods prior to stacking them for fusion into a composite tube to determine the impact of timing of fusion on resultant tissue structure and differentiation. With the goal to establish custom‐patterned, localized cartilage and prevascular soft tissue segments within the engineered tracheas, two prevascular rings (V) comprised of hMSCs and HUVECs were fused with three cartilaginous microsphere‐containing hMSC rings (C) in an alternating fashion to create prevascular soft tissue–cartilage composite tubes (CVCVC). Control cartilage‐only tubes were composed of three cartilage rings (CCC). Individual rings cultured for 2 or 4 d in their respective media were successfully fused into continuous tissue tubes. Tubes were harvested 2 weeks after cell seeding ( Figure 5 ; Figure S1, Supporting Information). Rings that were cultured individually for 6 d or longer did not fuse into tissue tubes. Figure 5 Macroscopic images and photomicrographs of Safranin O staining for GAG (pink/red), and human type II collagen (red), human type I collagen (red) and CD31 (brown) immunohistochemistry of prevascular–cartilage tissue engineered tubes and controls. Images and longitudinal sections of cartilage tubes in chondrogenic media (CCC), cartilage tubes in mixed chondrogenic and endothelial media (CCC–MM) and prevascular–cartilage tubes in mixed media (CVCVC) stacked on day 2 and day 4 after ring formation, and axial sections of prevascular rings (V) in endothelial media are depicted. Fast green counterstain (blue/green) for Safranin O, type I and II collagen; hematoxylin counterstain (blue) for CD31. Black scale bars are 200 µm; white scale bars are 2 mm. Images without scale bars in a single row are the same magnification. All cartilage containing tubes had a pearly white surface and maintained an open lumen upon harvest. In the control cartilage‐only tubes that were cultured in mixed media (CCC‐MM; 50:50 BPM:endothelial growth medium‐2 (EGM‐2)) and the CVCVC groups but not in the cartilage‐only tubes cultured in 100% BPM (CCC), rings fused at day 4 resulted in qualitatively thicker‐walled constructs compared to rings fused at day 2. The walls of all constructs were grossly thicker than walls of rat tracheas. Cartilaginous components of all the tubes stained positively for GAG and type II collagen content with minimal type I collagen staining. Rat tracheal cartilage exhibited the strongest Safranin O staining. Remaining gelatin microspheres stained with type I collagen antibody, as expected. GAG and type II collagen staining in cartilage‐only tubes grown in BPM was the best distributed from the lumens to the outer tube edges whereas CCC‐MM and CVCVC tubes had noncartilaginous fibrous capsules on the outer surfaces. CD31 staining (Figure 5 ; Figure S1, Supporting Information) showed that endothelial cells remained localized to the prevascular ring portions and some regions exhibited endothelial cell organization into prevascular structures. However, these structures were not as extensive as those in prevascular rings (V) grown for the duration of the experiment on glass tubes in EGM‐2, in which more of the endothelial cells were incorporated into prevascular structures. The majority of endothelial cells in prevascular rings grown in EGM‐2 were incorporated into complex prevascular plexuses. All cartilage only tubes, whether grown in BPM or mixed media, were negative for CD31 expression. Interestingly, more advanced plexus formation in the prevascular ring portion was noted in prevascular–cartilage composite tubes in another experiment (Figure S2, Supporting Information). Additionally, prevascular rings cultured in agarose wells for the duration of the experiment showed even more robust prevascular plexus formation (Figure S2, Supporting Information) compared to prevascular rings cultured on glass tubes (Figure 5 ). 2. 4. 2 Biochemical Analysis The quality of the cartilage tissue in the composite tubes was also analyzed biochemically. CVCVC constructs, which were comprised of more cells, had significantly more DNA than cartilage‐only constructs ( Figure 6 A). Surprisingly, CCC‐MM groups also had significantly more DNA than CCC tubes despite having the same number of incorporated cells. The GAG and GAG/DNA contents (Figure 6 B, C) of constructs grown in BPM:EGM‐2 mixed media (CCC‐MM and CVCVC) was significantly less than those of CCC tubes grown in BPM, which corroborates histological findings (Figure 5 ). Additionally, cartilage‐only and CVCVC day 4 constructs in mixed media had significantly more GAG and GAG/DNA compared to day 2 constructs cultured in the same media. All engineered tubes except for CVCVC day 2 had significantly more GAG/DNA than native rat tracheas (Figure 6 C). Figure 6 Biochemical and mechanical analysis of prevascular–cartilage tubes and controls. A) DNA, B) GAG and C) GAG normalized to DNA were acquired from tissue engineered cartilage only tubes in chondrogenic media (CCC, light gray, N = 3–4), cartilage tubes in mixed endothelial and chondrogenic media (CCC—MM, medium gray, N = 4) and prevascular–cartilage tubes in mixed media (CVCVC, dark gray, N = 4). C) Rat tracheal GAG/DNA was used for comparison (black, N = 4). D) Tube outer diameter, E) load required to collapse 80% of lumen diameter and F) tube recoil after luminal compression were measured in engineered tubes ( N = 3). D2, Rings fused after 2 d of individual culture, D4, Rings fused after 4 d of individual culture. A–C) a, b, c: Day 2 groups without common letter differ ( p < 0. 05); 1, 2, 3: Day 4 groups without common number differ ( p < 0. 05); #: significantly different than Day 2 group ( p < 0. 05); R: significantly different than rat trachea. D–F) x, y, z: groups without common letter differ ( p < 0. 05). Data shown as mean ± SD. 2. 4. 3 Mechanical Analysis CCC and CVCVC tubes were mechanically characterized for luminal rigidity (Figure 6 D–F). CVCVC day 2 tubes had the smallest outer diameter out of all the tubes analyzed (Figure 6 D). CCC day 2, CCC day 4, and CVCVC day 4 tubes had outer diameters of ≈4 mm. Evaluation of tube mechanical properties showed that CVCVC tubes bore significantly less load at 80% luminal collapse compared to CCC tubes (Figure 6 E). Additionally, CVCVC day 4 tubes bore significantly more load at 80% luminal collapse than CVCVC day 2 tubes. All engineered tubes recoiled to nearly 100% of the original diameter (Figure 6 F). 2. 4. 4 Tri‐Tissue Trachea Formation After fusing two cartilage rings and one vascular ring as before, CVC tubes were cultured in a suspension of human tracheal epithelial (HTE) cells for 24 h. Staining for GAG and CD31 demonstrated that cartilage matrix and endothelial prevascular structures, respectively, were still present after the additional culture in epithelial media ( Figure 7 ). Importantly, cytokeratin staining demonstrated the presence of adhered epithelial cells (Figure 7 ) to the surface of the CVC tubes. Figure 7 Histologic staining of tri‐tissue tracheas. CVC tubes were suspended in epithelial cell media in the absence (− HTE) or presence (+ HTE) of primary human tracheal epithelial (HTE) cells. Samples were harvested at 24 h, sectioned and stained with Safranin O (pink/red) with a fast green (blue/green) counterstain as an indication of cartilage extracellular matrix. To determine epithelial attachment, sections were stained for cytokeratin (red). Finally, sections were stained for CD31 (brown) and counterstained with Mayer's hematoxylin (blue) to identify endothelial cells. The left column scale bar for −HTE images is 2 mm and for the right column, it is 200 µm. Images in +HTE group are at same magnification as corresponding −HTE images. Dotted black box regions are shown in high magnification. 2. 4. 5 Subcutaneous Implantation in Athymic Mice After 19 d of in vitro culture, prevascular–cartilage composite tubes were implanted subcutaneously in athymic mice to assess vascularization of the engineered tissues. In addition to the CCC and CVCVC tubes described above, one additional group was included in this study. This new, third group consisted of three cartilage rings, initially cultured in BPM, separated by two vascular rings, composed only of MSCs, initially cultured in vasculogenic media. The vascular rings were seeded into the agarose molds and cultured like the vascular rings previously described but without the HUVECs. Upon fusing, these tubes were termed CVCVC‐noH and cultured in mixed media. Tissue engineered constructs that were fixed at day 0 before implantation and those harvested from the mice 15 and 42 d after implantation were processed for histology and immunohistochemistry ( Figure 8 ). The tubular structure was well maintained, and connective tissue had grown into the lumens of all samples. Although mechanical testing was not possible due to low sample numbers, explants were very firm to the touch and qualitatively stiffer than all in vitro‐cultured CVCVC or cartilage‐only tubes described previously. Composite neotracheas with intervening prevascular rings (with HUVECs) and control prevascular rings (without HUVECs) showed slightly more deformation from their original shape than tubes comprised solely of cartilage rings. The cartilage rings in composite tubes were slightly displaced with respect to each other along the longitudinal axis resulting in an offset tube. All samples showed the presence of GAG by Safranin O staining at day 0 and day 15 harvest, with CCC tubes exhibiting the strongest staining (Figure 8 A). However, the staining for GAG was reduced in all samples at 42 d ( N = 1). Alizarin red S staining demonstrated progressive calcium deposition within the samples over the course of the in vivo culture (Figure 8 A). No calcium staining was seen at time 0. Frozen sections of day 42 explants were stained with DAPI and fluorescently imaged for perfused lectin staining of human endothelial cells to show human cell‐derived vasculature (Figure 8 B) that anastomosed with the host. No lectin staining was seen in CCC and CVCVC‐noH samples. However, staining was seen in the CVCVC tube, and some HUVECs appeared to form structures with lumens. Figure 8 Histologic staining of tracheal tubes implanted subcutaneously in mice. Three types of tubes (CCC, CVCVC‐noH, and CVCVC) were implanted subcutaneously in mice for 0, 15, or 42 d. A) Longitudinal sections were stained with H&E, Safranin O (pink/red) with a Fast Green (blue/green) counterstain and alizarin red S (mineral is red). B) Tissue sections from day 42 samples were counterstained with DAPI and visualized for fluorescent FITC–UEA‐1 (green) perfusion staining. Black and red scale bars are 100 µm; white scale bars are 2 mm. Inset in (B) CVCVC group is 3× magnification of image. 3 Discussion 3. 1 Part I—Cartilage Rings and Tubes with Defined Dimensions Tracheal cartilage provides mechanical support to the airway, which is imperative for proper airway function. Our group has demonstrated that scaffold‐free, hMSCs with localized chondrogenic growth factor (TGF‐β1)‐delivering microspheres can be employed to engineer cartilaginous 2 mm inner diameter rings and tubes which are similarly sized to native rat tracheas. 8 The microspheres promote chondrogenic differentiation of the stem cells through spatiotemporally controlled release of the growth factor. 14 In this work, we hypothesized that the number of hMSCs and hMSCs with TGF‐β1‐laden microspheres used to fabricate 2 mm inner diameter rings can be varied to engineer rings with different wall thicknesses. At the same time, constructs with diameters >2 mm are required for testing in a larger animal model, like the rabbit (≈6 mm inner diameter 15 ), and eventually translating for human use (≥12 mm inner diameter[[qv: 1a]]). The size of the custom culture wells and tube holders were modified to alter the diameter of resultant rings and tubes. By changing the cell and microsphere numbers, cartilage ring (hMSC or hMSC + MS) thickness could be altered by 25–35%. However, even the thinnest rings were still not as thin as the rat trachea (Figure 2 B). It is unclear if a thicker‐walled tracheal replacement will impede tracheal function in a rat airway defect model. Furthermore, the incorporation of additional tissue types and altered culture conditions, which will be necessary to functionalize the cartilage tube, will likely reduce the thickness of the cartilage. Finally, the customizable annular culture wells can easily be altered to obtain narrower troughs or modified trough geometry (e. g. , U‐bottom vs. V‐bottom) to produce thinner rings or even other geometrical shapes if desired. In addition, the ultimate tensile stress (UTS, failure load normalized by loaded area) of all rings, except the 0. 2 million cells per ring without microspheres condition, were similar among the groups as well as compared to the rat trachea, which means that engineered rings may fulfill the mechanical support requirement when used in rat airway repair. Next, cartilaginous rings and tubes of larger diameters were engineered by modifying the cell culture wells and tube holders to produce 6 and 12 mm inner diameter constructs in addition to 2 mm constructs. Surprisingly, the number of hMSCs with microspheres required to reliably form complete tissue rings did not scale linearly with the circumference of the post. In fact, seeding 2. 4 million hMSCs in 12 mm wells (6 × 0. 4 million cells per ring) resulted in only 25% ring formation frequency. As a result, the cell number was increased by 25% in all sizes of rings (0. 5 million per 2 mm ring, 1. 5 million per 6 mm ring and 3 million per 12 mm ring) while keeping the microsphere‐to‐cell ratio constant. Still, the frequency of ring formation decreased with increasing diameter (Figure 3 A). It is possible that the smaller curvature (1 per radius; 1 mm −1 for 2 mm, 0. 33 mm −1 for 6 mm, 0. 17 mm −1 for 12 mm) affected the ability of the hMSCs and microspheres to self‐assemble into continuous rings and ultimately elaborate chondrogenic matrix resulting in lower GAG/DNA and GAG/WW (Figure 3 F, G). In fact, previous reports have shown that geometrical shape cues, like curvature, can direct hMSC differentiation. 16 Nevertheless, chondrogenesis was not severely affected as GAG/DNA content only decreased by 23% and 25% for 6 mm and 12 mm tubes, respectively, compared to 2 mm tubes, while type II collagen staining remained strong. Tube wall thickness data indicate that 6 mm tubes (1. 16 mm average thickness) may be of suitable size for the rabbit airway (0. 97 mm average thickness) repair and 12 mm tubes (1. 26 mm average thickness) may be adequately sized for use in humans whose tracheal cartilage is ≈1 mm thick. 17 However, the 6 mm engineered cartilage tubes required only 8% of the rabbit tracheal load to collapse 80% of their respective lumen diameters and 12 mm cartilage tubes are likely substantially weaker than human tracheas (Figure 3 I). As expected, the tube structural stiffness is proportional to the ratio of the radius to the wall thickness with greater ratios leading to more compliant tubes. Mechanical properties of engineered tracheal tubes containing multiple cell types were further weakened (Figure 6 E). However, it is important to note that both the cartilage‐only tubes (Part I) and the prevascular–cartilage tubes (Part III) are immature constructs that were only grown for 3 and 2 weeks, respectively. So, it is not surprising that the neotracheas are weaker than their native counterparts, which had developed for much longer. In addition, it is unclear whether the weaker luminal collapse properties of the tracheal replacement would hinder airway function because healthy infant tracheas can safely collapse up to 50%, mostly by invagination of the posterior smooth muscle wall, during the first milliseconds of deep inspiration, while normal breathing results in minimal deformation. 18 Importantly, all engineered tubes recoiled to nearly 100% of original diameter after luminal collapse just like the native rabbit trachea (Figure 3 J). If necessary, mechanical properties of the tubes may be improved by increasing the wall thickness by using more hMSCs and microspheres or lengthening the culture time to elaborate more mature cartilaginous matrix. Taken together, these findings demonstrate the flexibility of the customizable culture system and show that cartilage tissue rings and tube with specific dimensions can be successfully engineered. 3. 2 Part II—Epithelialized Cartilage Bilayers Tracheal epithelium performs a crucial role in the innate host defense by (1) providing a protective physical barrier and (2) producing mucus to allow the body to clear infectious agents and environmental toxins. 19 These functions can be temporarily or permanently compromised in patients with airway stenosis and malignant tumors or after tracheal resection. 20 Hence, one of the main challenges for the translational application of tissue‐engineered tracheas is the ability to efficiently restore a functional epithelial cell layer. Isolated epithelial cells are typically cultured on protein‐coated cell culture inserts. Upon reaching confluence, the medium from inside the cell culture insert is removed, thereby exposing the epithelial sheet to air. This ALI culture model mimics the in vivo distribution of the airway lining; the apical side of epithelial cells faces the airway lumen and the basal side of the cells is attached to the basement membrane. [[qv: 9c, 21]] We hypothesized that our established scaffold‐free, high‐cell density cartilaginous sheets 22 may support epithelialization. In proof‐of‐principle studies, we sought to engineer epithelial–cartilage bilayer sheet constructs. In this work, we were able to successfully engineer cohesive epithelial–cartilaginous bilayers. Qualitative histological analysis showed that while the 50:50 mixture of BPM and BEGM required for the maintenance of the two distinct cell types slightly reduced GAG content and altered type II collagen distribution in the underlying cartilaginous sheets relative to positive controls, the thickness of the epithelial layer seeded on top increased over time when exposed to air. Differences in GAG deposition in the C sheets compared to EC groups cultured in mixed media may be a result of hBEC–hMSC cell–cell communication, nutrient availability or diffusion limitations. Similarly, the E sheet positive control was thicker than the epithelial layers on the EC constructs, suggesting that nutrient availability, nutrient and oxygen diffusion or the presence of hMSCs altered epithelial layer development. Furthermore, expression of cytokeratins, an epithelial cell marker indicating structural cell integrity, 23 was robust in the epithelial portion of the EC bilayer sheets and similar to that of the E control sheets cultured on standard cell culture inserts. The successful epithelial cell growth and differentiation at ALI and submerged in media achieved here is promising for engineering tubular tracheal replacements with a luminal epithelial coating. 3. 3 Part III—Incorporation of Prevascular Rings into Tubes and Tri‐Tissue Tracheas Similar to nearly all tissues in the body, native trachea has a vascular supply to facilitate oxygen and nutrient delivery and metabolic waste removal, 10 critical for the survival of the investing soft tissue and epithelium lining the lumen. [[qv: 4g]] To develop a tissue engineered construct of clinically relevant size, the construct must be able to quickly develop a microvascular supply to survive after implantation. Neovascularization in replacement tracheas has been explored in only a few studies whose constructs contained vascular progenitor or endothelial cells obtained from adipose, 24 bone marrow, 25 or skin 26 tissue. In vitro approaches to support vascularization of replacement tracheas have mostly relied on scaffold materials with potential angiogenic properties (e. g. , decellularized allogeneic trachea 27 ) or delivery of growth factors to drive angiogenesis. 28 However, relying solely on host blood vessel infiltration may be insufficient in an orthotopic tracheal replacement because transplanted cells in the interior of the construct may die before becoming adequately microvascularized. Therefore, in this work, a prevascularization approach was developed to provide the framework for a microvasculature. By seeding endothelial cells with stromal supporting hMSCs in a ring mold and culturing for 2 weeks in EGM‐2, the endothelial cells were able to self‐assemble into prevascular structures (Figure 5 ) reminiscent of the prevascular cords and plexuses formed from blood islands in the mesenchymal condensations during embryonic vasculogenesis. 29 Furthermore, with such a high cell number, the interior of these constructs may have been at low oxygen tension, a known provasculogenic stimulus, 30 that may have enhanced prevascular structure formation. The prevascular soft tissue rings were successfully stacked with cartilage rings resulting in ring fusion into tubes with similar architecture to native trachea (Figure 5 ). The presence of a prevascular network in the composite tubes will allow for the support of a luminal epithelium, a clear advance beyond cartilage only tubes. Unlike previous endothelialization approaches for tracheal tissue engineering, this strategy is advantageous because it avoids extensive material processing required for the use of decellularized materials, [[qv: 27a]] the need to reseed endothelial cells into the decellularized tissues, 25 the need for ectopic implantation prior to orthotopic use 31 and the use of angiogenesis‐inducing drugs. 32 Additionally, it allows for precise spatial control over prevascular tissue generation. 33 Stacked prevascular and cartilaginous rings fused together to form continuous multi‐tissue constructs. While the essential features of each tissue type were maintained throughout the coculture period, the resultant phenotype of each tissue type was diminished somewhat compared to that resulting from individual culture in specific medium for each respective tissue type. Prevascular rings cultured for the duration of the experiment in EGM‐2 showed a high degree of endothelial organization into prevascular plexus‐like structures (Figure 5 ). However, in composite tubes, there was a lower degree of endothelial cell incorporation into such structures, likely due to the decreased concentration of angiogenic growth factors in the mixed media. The GAG and GAG/DNA content in the prevascularized tubes cultured in mixed media was less than in tracheal tubes cultured in 100% BPM as demonstrated quantitatively by biochemical analysis (Figure 6 A–C) and qualitatively by Safranin O staining (Figure 5 ). This decrease in GAG content was demonstrated to be, at least partially, due to the media and not the presence of prevascular rings because a similar decrease in GAG content was seen in cartilage‐only tubes cultured in the same 50:50 mixed media used with the prevascularized tubes. The weakened load bearing properties of prevascular tubes (Figure 6 E) may be a result of the decreased GAG content from the culture in mixed media, which would decrease the mechanical properties of the engineered cartilage component. 34 Nevertheless, tube recoil after luminal collapse testing was minimally affected (Figure 6 F). To demonstrate the capability of this system to simultaneously support a further, third, physiologically relevant tracheal tissue type, CVC tubes were cultured in a suspension of epithelial cells to form tri‐tissue tracheal constructs. Importantly, the cartilage and prevascular phenotypes were maintained (Figure 7 ) through this additional period of culture while allowing for the attachment of epithelial cells on the surface of the CVC constructs (Figure 7 ). The epithelial cell coverage was not complete, but rather discontinuous. However, in areas of coverage, many cells were seen in juxtaposition with each other as opposed to individually attached cells, which may allow for the establishment of a continuous epithelium to provide an immunologic barrier function. These results demonstrate the feasibility of tri‐tissue tissue engineering in the current system. The composite prevascular–cartilage tracheal constructs were then implanted subcutaneously in mice (Figure 7 ). Critical to their function of supporting an open airway, the engineered tracheas maintained their tubular architecture and their mechanical integrity was enhanced throughout in vivo culture. Substantial GAG staining was observed at the time of implantation and after 15 d in all groups. However, at the last time points, less GAG staining was evident. Progressive calcification of the constructs was noted (Figure 7 A) with the staining appearing principally in the peripheral regions of the cartilage segments of the tubes. It may be that the chondrogenically driven hMSCs are contributing directly to tissue calcification themselves 35 or by recruitment of host osteoprogenitor cells. 36 It has been shown that control of hMSC fate is challenging as chondrogenically driven hMSCs can eventually lead to calcified tissue. 37 It should be noted, however, that some calcification is seen in about half of native tracheas in the elderly without deleterious effects. 38 Another key challenge upon implantation is host integration. If the constructs can hold a suture, then connecting them end‐to‐end via sewing to the healthy native tissue, potentially with a biocompatible sealant like fibrin glue, may be a simple and viable solution. The intravenous injection of FITC–UEA‐1 allowed for the identification of human endothelial cells that were incorporated into perfused vasculature within the engineered neotracheas (Figure 7 B). Predictably, no such staining was found in samples that lacked HUVECs. However, in the tracheal constructs with prevascular segments containing HUVECs, lectin‐labeled cells were seen in the explanted tissue. These positively staining cells were only seen in the prevascular segments of the tracheas indicating that there was no vascular invasion into the cartilage segments from the prevascular rings. The labeled endothelial cells were variably seen in circular or short cord‐like structures. The circular assemblies of stained HUVECs likely indicate patent lumens that were perfused with lectin containing host blood. While vascular lumen formation was not observed in vitro, the development of such mature, stable structures may have required the perfusion present in vivo. 39 The spread HUVECs that do not enclose a lumen may be sections through the walls of vessels not oriented perpendicular to the plane of the section. Alternatively, as these stained, spread cells were often in the vicinity of lumen forming HUVECs, the vasculature formed by the transplanted cells may be leaky 40 and as there are no lymphatics to drain the vascular filtrate, the lectin may simply be staining unincorporated HUVECs in the vicinity of the new HUVEC‐derived vasculature. To enhance vascularization in the future, it may be valuable to vary chondrogenic to vasculogenic media ratios during in vitro culture. In addition, microparticles presenting vasculogenic growth factors may be incorporated into prevascular rings, which may allow for a more uniformly distributed delivery of these factors than is possible by exogenous delivery alone to thicker constructs. 4 Conclusion In this work, a modular, scaffold‐free approach for engineering complex tubular hollow organs was presented via the formation of a tri‐tissue engineered trachea. The three distinct tissues (i. e. , cartilage, epithelial, and vascular) with defined spatial placement provide for luminal rigidity, a respiratory epithelium and prevascular structures to facilitate perfusion after implantation and anastomosis with host vasculature. Moving forward, prevascular–cartilaginous tubes may be epithelialized on the lumen with the aid of a tubular organ bioreactor. [[qv: 4d, f, 41]] This modular, tubular tissue and organ engineering approach may also find great utility for regenerating other tissues such as large blood vessels and segments of the gastrointestinal (i. e. , esophagus, intestines) and urinary (i. e. , ureters) tract. 5 Experimental Section Experimental Design : Four research objectives were examined in this body of work (Figure 1 ). The goal of Part Ia was to tune the thickness of engineered cartilage rings. Part Ib aimed to develop cartilage rings and tubes with custom‐defined lumen diameters. In Part II, a respiratory epithelium was engineered on the cartilaginous surface of cartilage tissues. Lastly, Part III focused on developing multi‐tissue type tubular constructs comprised of prevascular rings fused with cartilaginous rings, which were ultimately seeded with epithelial cells. Cell Culture : Bone‐marrow‐derived hMSCs from a single donor were isolated using a Percoll gradient (Sigma‐Aldrich, St. Louis, MO) and the differential adhesion method, and then expanded in Dulbecco's modified Eagle's medium—low glucose (DMEM‐LG; Sigma‐Aldrich) containing 10% prescreened fetal bovine serum (Gibco Qualified FBS; Life Technologies, Carlsbad, CA or Sigma Premium FBS; Sigma‐Aldrich) and 10 ng mL −1 fibroblast growth factor‐2 (FGF‐2, R&D Systems, Minneapolis, MN) as previously described. 8 Passage 2‐3 hMSCs were used in this study. BEAS‐2B human bronchial epithelial cells (hBECs) (ATCC; Manassas, VA) were cultured in bronchial epithelial cell growth medium (BEGM; Lonza, Walkersville, MD). Prior to use, culture flasks were coated overnight with 0. 01 mg mL −1 fibronectin (Sigma‐Aldrich), 0. 03 mg mL −1 type I collagen (Advanced BioMatrix, San Diego, CA), and 0. 01 mg mL −1 bovine serum albumin (Thermo Fisher Scientific, Waltham, MA) at 37 °C. The next day, flasks were washed with PBS and hBECs were plated at 3. 3 × 10 3 cells cm −2 for expansion. Primary HUVECs (ATCC) were cultured in EGM‐2 (Lonza; Basel, Switzerland) and used at passage 3‐4. Human tracheal segments obtained at necropsy were stored at 4 °C in 50% Dulbecco's modified Eagle's medium‐high glucose (DMEM‐HG, Hyclone; South Logan, UT):50% Ham's F‐12 (Hyclone) supplemented with 2. 5 × 10 −3 m l ‐glutamine (Sigma‐Aldrich), 5 µg mL −1 insulin (Sigma‐Aldrich), 5 µg mL −1 transferrin (Sigma‐Aldrich), 5 × 10 −6 m hydrocortisone (Sigma‐Aldrich), and 2. 5 µg mL −1 amphotericin (Sigma‐Aldrich). The tracheal segments were trimmed of excess connective and fatty tissue and treated with 0. 1% protease XIV (Sigma‐Aldrich) at 4 °C for 16 h. The epithelial cells were isolated by gentle scraping of the luminal surface with a plastic coverslip. Isolated cell clumps were treated 5–7 min with Accutase (Sigma‐Aldrich) to dissociate. Cells were washed, resuspended in epithelial proliferation media (75% Ham's F‐12:25 % DMEM, supplemented with 5% FBS (Sigma‐Aldrich), 24 µg mL −1 adenine (Sigma‐Aldrich), 8. 4 ng mL −1 cholera toxin (Sigma‐Aldrich), 10 ng mL −1 epidermal growth factor (Sigma‐Aldrich), 0. 4 µg mL −1 hydrocortisone (Sigma‐Aldrich), 10 × 10 −6 m Y‐27632 (Selleck Chemicals; Houston, TX), and 5 µg mL −1 insulin), and seeded onto a lawn of irradiated 3T3 fibroblasts. Media was changed daily and cultures were passaged 2–3 times (1:5) to produce 20–50 × 10 6 epithelial cells. Epithelial cells were harvested for use by differential trypsinization to first remove irradiated 3T3 fibroblasts and then to collect the primary human tracheal epithelial cells. Microsphere Synthesis and Characterization : Gelatin microspheres (11. 1 w/v% Type A; Sigma‐Aldrich) were engineered as previously described. 22 In this work, microspheres were crosslinked with 1 w/v% genipin (Wako Chemicals USA Inc. , Richmond, VA) for 2. 25–2. 5 h. The percentage of crosslinked amine groups in the polymer was assessed by incubating microspheres for 1. 75–2. 75 min in ninhydrin solution. 8 Experiments in this study used two batches of microspheres, which contained similar amounts of crosslinked amine groups in the polymer: 23. 6 ± 4. 7% and 25. 7 ± 2. 2%. The diameters of gelatin microspheres used in Parts I and III were 54. 4 ± 40. 4 µm ( N = 354) and 43. 3 ± 30. 0 µm ( N = 353) in Part II. For growth factor delivery, microspheres were loaded with 400 ng TGF‐β1 (PeproTech, Rocky Hill, NJ) per mg microspheres. Preparation of Custom Geometry Culture Wells : Annular wells were machined as previously described[[qv: 13a]] or 3D printed (Objet 260 Connex, Stratasys) in three sizes: 2, 6, or 12 mm diameter posts surrounded by a 2 mm wide trough. Polydimethylsiloxane (PDMS; Sylgard 184, Dow Corning, Midland, MI) was cured in the molds and served as negative molds for casting 2% w/v agarose (Denville Scientific Inc. , Metuchen, NJ) culture wells based on previously described methods. [[qv: 13a]] Prior to cell seeding, culture wells that were used to engineer cartilage rings were incubated overnight in chondrogenic BPM comprised of Dulbecco's modified Eagle's medium‐high glucose (DMEM‐HG; Sigma‐Aldrich), 1% ITS+ Premix (Corning Inc. , Corning, NY), 10 −7 m dexamethasone (MP Biomedicals, Solon, OH), 1 × 10 −3 m sodium pyruvate (HyClone Laboratories), 100 × 10 −6 m nonessential amino acids (Lonza Group, Basel, Switzerland), 37. 5 µg mL −1 ascorbic acid‐2‐phosphate (Wako Chemicals USA Inc. ) and 100 U mL −1 penicillin–streptomycin (Corning Inc. ). 8 Culture wells that were used to engineer prevascular rings were incubated in endothelial basal medium (EBM; Lonza Group). Assembly of Cartilage Rings and Tubes (Part I) : In Part Ia, hMSCs (0. 1 × 10 6 –0. 4 × 10 6 cells) with or without 0. 75 mg TGF‐β1 laden microspheres/1 × 10 6 cells in 50 µL media were seeded and cultured in 2 mm culture wells based on previously described methods. 8 Rings with microspheres (“hMSC + MS”) were cultured in BPM and hMSC‐only groups (“hMSC”), which did not contain microspheres, were cultured in BPM supplemented with 10 ng mL −1 TGF‐β1. Tissue rings were grown in a humidified cell culture incubator at 37 °C and 5% CO 2 for 21 d with media changes every 2 d. In Part Ib, 0. 5 × 10 6 hMSCs, 1. 5 × 10 6 hMSCs, and 3 × 10 6 hMSCs with 0. 75 mg TGF‐β1‐loaded microspheres/1 × 10 6 cells were seeded and cultured in 2, 6, and 12 mm culture wells in BPM, respectively. On day 2–3, cartilage rings were removed from the annular wells and stacked onto 2, 6, or 12 mm outer diameter borosilicate glass tubes (Adams & Chittenden Scientific Glass, Berkeley, CA) resulting in two‐ and five‐ring tissue tubes in each diameter. Tissue tubes were cultured horizontally on custom engineered polycarbonate holders in deep reservoirs (Axygen Scientific, Union City, CA) containing 100 mL BPM. Tissue tubes were grown in a humidified cell culture incubator at 37 °C and 5% CO 2 for 21 d with media changes every 3 d. Assembly of Epithelial–Cartilage Bilayers (Part II) : Cartilaginous sheets were formed by seeding 0. 6 × 10 6 cells with 0. 75 mg TGF‐β1 laden microspheres/1 × 10 6 cells onto cell culture inserts (6. 5 mm diameter, 3 µm pore size; Corning). These were grown for 2 weeks with media changes (1 mL, BPM) every 2 d. Epithelial–cartilage bilayers were formed by seeding 8. 3 × 10 4 hBECs per sheet in 100 µL of BEGM on top of the hMSC+MS sheets at 2 weeks (“EC bilayer sheets”). Cartilage only sheets without an epithelial layer (“C sheets”) and epithelial only sheets without a cartilage layer (“Control E sheets”) served as controls. Control hBEC sheets were seeded on cell culture inserts (6. 5 mm diameter, 0. 4 µm pore size; Corning) previously coated overnight with 0. 01 mg mL −1 fibronectin, 0. 03 mg mL −1 type I collagen, and 0. 01 mg mL −1 bovine serum albumin at 37 °C and rinsed with PBS. In all groups, seeded hBECs were allowed to settle and adhere for 3 d, after which the insert medium was removed in some of the groups to expose the constructs to an air–liquid interface (ALI) for 4 or 7 d (“ALI 4 d” and “ALI 7 d”). Some of the sheets remained submerged for 7 d (“Submerged 7 d”). Starting on Day 3 after hBEC seeding, all tissues were cultured in a 50:50 mixture of BPM and BEGM (“50% BPM 50% BEGM”) except cartilage control sheets which continued to be cultured in 100% BPM (“Control C sheets”). In summary, the experimental groups included: (1) Control C sheets (submerged; BPM; 7 d), (2) Control E sheets (ALI; 50% BPM 50% BEGM; 7 d), (3) C sheets (submerged; 50% BPM 50% BEGM; 7 d), (4) C sheets (ALI; 50% BPM 50% BEGM; 4 and 7 d), (5) EC bilayer sheets (submerged; 50% BPM 50% BEGM; 7 d), and (6) EC bilayer sheets (ALI; 50% BPM 50% BEGM; 4 and 7 d). Sheets were grown in a humidified cell culture incubator at 37 °C and 5% CO 2 with media changes every 2 d. Assembly of Prevascular–Cartilage Composite Tubes and Tri‐Tissue Trachea Formation (Part III) : 2 mm cartilage rings were assembled as in Part Ia using 0. 4 × 10 6 hMSCs and 0. 3 mg TGF‐β1 laden microspheres. Prevascular rings were made from HUVECs (0. 4 × 10 6 cells) and hMSCs (0. 4 × 10 6 cells) suspended in 50 µL EGM‐2 and seeded in 2 mm ring‐shaped culture wells. On day 2, some cartilage and vascular rings were transferred onto 2 mm outer diameter glass tubes suspended in media. All cartilage rings and most prevascular rings that were transferred to glass tubes were used to make tubes on day 2 (“D2”) or day 4 (“D4”) after ring formation. Prevascular–cartilage tubes were formed by stacking three cartilage and two prevascular rings in an alternating sequence on day 2 or day 4 to make five‐ring tissue engineered tracheal tubes (“CVCVC D2” and “CVCVC D4”) and cultured in a 50:50 mixture of BPM and EGM‐2. As controls, at each stacking time point, three cartilage rings were stacked without intervening prevascular rings and cultured in BPM (“CCC D2” and “CCC D4”) or 50:50 BPM‐EGM‐2 mixed media (“CCC‐MM D2” and “CCC‐MM D4”). Some prevascular rings were cultured individually on glass tubes in EGM‐2 until the end of the experiment (“V”). Alternatively, some prevascular rings composed of 0. 2 × 10 6 HUVECs and 0. 2 × 10 6 hMSCs were maintained in agarose molds for 14 d without being transferred to glass tubes on day 2 (“V–agarose”, Figure S2, Supporting Information). In an additional experiment, prevascular–cartilage tubes included two cartilage rings and one prevascular ring stacked at day 2 (“CVC D2”, Figure S2, Supporting Information). These were cultured in a 50:50 mix of BPM and EGM‐2 media. All tissue rings and tubes were grown in a humidified cell culture incubator at 37 °C and 5% CO 2 for 15 d following ring formation with media changes every 2 d. For tri‐tissue trachea formation, CVC tubes were formed as described earlier for CVCVC tubes, only with two cartilage rings and one vascular ring. These constructs were then placed in closed, perforated 1. 5 mL microcentrifuge tubes containing a 0. 5 mL suspension of 0. 5 × 10 6 HTE cells in epithelial proliferation medium and cultured on a rotisserie shaker (Barnstead Thermolyne, Dubuque, IA) for 24 h. Medium was replaced every 6 h. Tissue Harvest : Cartilage rings and 2, 6, and 12 mm tubes were harvested 3 weeks after ring formation (Part I). Epithelial–cartilage bilayers and their cartilage‐only controls were harvested 7 and 10 d after bilayer creation, and epithelial‐only controls were harvested after 7 d of culture (Part II). Epithelial control sheets (no cartilage) were maintained on the cell culture insert membrane during tissue harvest and sectioning to limit damage to the thin sheets. Prevascular–cartilage tubes and their controls were harvested 15 d post cell seeding in ring molds, and tissue tubes to be seeded with epithelial cells were switched to an epithelial cell suspension for the final 24 h, 14 d post cell seeding in ring molds (Part III). All macroscopic images of tissues were taken with a Galaxy S4 phone camera (Samsung, Seoul, Korea). Tracheas from healthy rabbits (male New Zealand white rabbits, 9 months old ( N = 6); Covance) and rats (male NIH Nude rats, 14–15 weeks old ( N = 4); Taconic, Hudson, NY) were used for comparison. Use of harvested tissues from animals sacrificed for unrelated studies was approved by the Institutional Animal Care and Usage Committee (IACUC) at Case Western Reserve University (CWRU). Biochemical Analysis : Tissue rings ( N = 3–4 from Part Ia) and tubes ( N = 4–5 two‐ring tissue tubes from Part Ib, N = 3–4 tubes from Part III and N = 4 rat tracheal segments) were digested in papain solution (Sigma‐Aldrich) at 65 °C. 42 GAG and DNA contents were measured using dimethylmethylene blue (DMMB; Sigma‐Aldrich) 43 and PicoGreen (Invitrogen, Carlsbad, CA) assays, respectively. 44 Histology and Immunohistochemistry : Cartilaginous components of tissues from each Part ( N ≥ 2) were evaluated for GAG via Safranin O (Acros Organics, Thermo Fisher Scientific) staining with a Fast Green (Fisher Chemical) counterstain and type II collagen deposition (ab34712 at 1:200 dilution; Abcam, Cambridge, UK) with a Fast Green counterstain as previously described. 8 Sections of epithelial–cartilage bilayer sheets and their controls were stained with hematoxylin & eosin (H&E) ( N = 3). Cytokeratin was detected using an anti‐pan‐cytokeratin antibody (sc‐81714 at 1:100 dilution; Santa Cruz Biotechnology, Santa Cruz, CA) with a Fast Green counterstain ( N = 3). Type I collagen was visualized in prevascular–cartilage tissues using an anti‐type I collagen antibody (ab21287 at 1:250 dilution; Abcam) with a Fast Green counterstain ( N = 2). In vitro‐cultured engineered prevascular–cartilage tubes and controls were stained for CD31, an endothelial cell marker, to evaluate prevascular structure formation ( N = 2). Cryosectioned samples were stained using a CINtec Histology Staining Kit (Roche, Mannheim, Germany) and antihuman CD31 primary antibody (M0823 at 1:100 dilution; Dako, Carpinteria, CA) and counterstained with Mayer's hematoxylin (Thermo Fisher Scientific). The tri‐tissue tube ( N = 1) was stained with Safranin O for cartilage, pan‐cytokeratin antibody for epithelial cells and CD31 antibody for endothelial cells. Stained tissue sections were imaged using an Olympus BX61VS microscope (Olympus, Center Valley, PA) with a Pike F‐505 camera (Allied Vision Technologies, Stadtroda, Germany). Cartilage and Epithelial Layer Thickness Analysis in Epithelial–Cartilage Bilayers : Cartilage sheets, epithelial sheets, and epithelial–cartilage bilayers ( N = 3 per group) stained with H&E were used to assess thickness. For quantification of the thickness of the cartilage portion in each cartilage sheet and epithelial–cartilage bilayer, a 10× magnified image of the center of each construct was acquired. The thickness of the cartilage portion was measured in three regions of interest (left, central and right) within this image. For quantification of the thickness of the epithelial portion within each epithelial control sheet and CE bilayer, 3 40× magnified images were acquired. The thickness of the epithelial portion was measured in three regions of interest (left, central, and right) within each image. The measurements were performed using ImageJ software (NIH, Washington, DC, USA). Tissue Dimension Measurements and Biomechanical Analysis : Tissue engineered rings composed of different cell numbers ( N = 3) and rat tracheas ( N = 4) were analyzed using uniaxial tension to failure testing and the ring wall thicknesses were measured as previously described. 8 Tissue engineered tubes ( N = 3) and rabbit tracheas ( N = 6) (Part Ib and III) were evaluated via luminal collapse and recoil as before with slight modifications. 8 Freshly harvested five‐ring cartilaginous tubes of 2, 6, and 12 mm inner diameters and prevascular–cartilage 2 mm inner diameter tubes and their cartilage‐only controls were compressed by their respective luminal size at a rate of 0. 5 mm min −1. Rabbit tracheas were evaluated in a similar manner by collapsing the lumen by each respective lumen diameter (5. 17 ± 0. 35 mm). Maximum load at 80% luminal collapse and the outer diameter after recoil were used for comparison. Engineered (2, 6, and 12 mm diameter) and native tracheal tube wall thicknesses were measured with calipers. Subcutaneous Implantation of Tracheal Constructs : The surgical procedures used in this study were conducted according to a protocol approved by the IACUC of CWRU which adhered to the NIH Guide for the Care and Use of Laboratory Animals. Nine week old athymic mice (NCR nu/nu) from the CWRU Athymic Animal Facility were anesthetized using ketamine (160 mg kg −1 )/xylazine (16 mg kg −1 ), and tracheal constructs were implanted subcutaneously on the dorsa of mice (3 or 4 constructs per mouse). The incisions were closed and the mice were administered 0. 1 mg kg −1 buprenorphine at 0 and 12 h postsurgery. Three different constructs were implanted: (1) tubes consisting of three cartilage rings prepared as above and cultured in BPM (CCC), (2) tubes consisting of three cartilage rings alternating with two prevascular rings prepared as above but with only hMSCs in the prevascular rings (without HUVECs) and cultured in 50:50 BPM:EGM‐2: (CVCVC‐noH), and (3) tubes consisting of three cartilage rings separated by two prevascular rings prepared as above and cultured in 50:50 BPM:EGM‐2 (CVCVC). All tubes were made by stacking rings at day 4 and culturing for a further 15 d prior to implantation. Tubes were collected on the day of implantation ( N = 1 per group) and harvested on days 15 ( N = 2 for CCC and N = 3 for CVCVC‐noH and CVCVC per group) and 42 ( N = 1 per group) postimplantation. Prior to the 42 d harvest FITC‐UEA‐1 (Vector Labs, 100 µg in 100 µL of PBS) was perfused via tail vein injection to label human endothelial cells that were incorporated into perfused vasculature. Half of each tube was processed via paraffin sections which were stained with H&E, Safranin O, and alizarin red S (Sigma‐Aldrich). The other half of each tube was cryosectioned and sections were stained with DAPI (Thermo Fisher Scientific) to evaluate for fluorescent lectin staining. Bright‐field images were acquired as described earlier. Fluorescent images were taken on an Eclipse TE300 (Nikon, Tokyo, Japan) equipped with a Retiga‐SRV digital camera (Qimaging, Burnaby, BC, Canada). Statistical Analysis : Statistical analysis of tissue engineered constructs and rat and rabbit tracheas was conducted using 1‐way ANOVA with Tukey's post hoc tests performed when p < 0. 05 (InStat 3. 06 software; GraphPad Software Inc. , La Jolla, CA). Means of all values are reported with errors signifying standard deviation. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700499
| 2,018
|
Advanced Science
|
3D Fabrication with Integration Molding of a Graphene Oxide/Polycaprolactone Nanoscaffold for Neurite Regeneration and Angiogenesis
|
Abstract Treating peripheral nerve injury faces major challenges and may benefit from bioactive scaffolds due to the limited autograft resources. Graphene oxide (GO) has emerged as a promising nanomaterial with excellent physical and chemical properties. GO has functional groups that confer biocompatibility that is better than that of graphene. Here, GO/polycaprolactone (PCL) nanoscaffolds are fabricated using an integration molding method. The nanoscaffolds exhibit many merits, including even GO nanoparticle distribution, macroporous structure, and strong mechanical support. Additionally, the process enables excellent quality control. In vitro studies confirm the advantages of the GO/PCL nanoscaffolds in terms of Schwann cell proliferation, viability, and attachment, as well as neural characteristics maintenance. This is the first study to evaluate the in vivo performance of GO‐based nanoscaffolds in this context. GO release and PCL biodegradation is analyzed after long‐term in vivo study. It is also found that the GO/PCL nerve guidance conduit could successfully repair a 15 mm sciatic nerve defect. The pro‐angiogenic characteristic of GO is evaluated in vivo using immunohistochemistry. In addition, the AKT‐endothelial nitric oxide synthase (eNOS)‐vascular endothelial growth factor (VEGF) signaling pathway might play a major role in the angiogenic process. These findings demonstrate that the GO/PCL nanoscaffold efficiently promotes functional and morphological recovery in peripheral nerve regeneration, indicating its promise for tissue engineering applications.
|
1 Introduction Although peripheral nerves exhibit some self‐healing potential after mild and moderate trauma as they spontaneously start new axons sprouting after injury, 1 successful reinnervation cannot be achieved, especially for long‐range nerve defects and it calls for implantation of a nerve graft to bridge the gap. 2 Conventional surgical treatment has plateaued because the gold‐standard protocol‐autologous nerve transplantation causes unavoidable secondary damage to the donor site. 3 Tissue engineering and bioactive materials are ideal alternatives for use in the circulatory, digestive, respiratory and nervous systems. 4, 5, 6, 7 Many studies have reported that nerve guidance conduits (NGCs) exhibit extensive capabilities for repairing large nerve defects in the peripheral nervous system. 8, 9, 10 NGCs are expected to direct cell migration in a targeted manner with their physical and chemical characteristics and to facilitate cell proliferation and differentiation. Synthetic materials such as polycaprolactone (PCL) possess several advantages such as biodegradability, nontoxicity, and structural stability. PCL has been tested in various applications, and has been shown to positively influence cardiovascular, nervous and soft tissues. 11, 12, 13 Another important aspect of nerve tissue engineering is emphasized on inherent electrical excitability of nerve cells. Electrical stimulation can promote neurite extension and axonal regrowth. Conductive scaffolds have better electrical conductivity, biocompatibility and lipophilicity for cell adhesion. 14, 15, 16 Among them, graphene is an extremely important nanomaterial due to its exceptional physical and chemical properties. 17, 18 Moreover, graphene can interact with biomolecules such as proteins, polypeptides and nucleic acids in regenerative medicine. 19 Graphene oxide (GO) is an extremely oxidized graphene derivative with a 2D structure consisting of a single layer with multiple functional groups. 20 It exhibits amphiphilic characteristics because of the presence of epoxide, hydroxyl and carboxylic acid functional groups in the same plane. 21 It has been previously reported that the application of potassium permanganate and sulfuric acid in graphite has previously been reported to facilitate GO nanosheets separations. 22 Unmodified graphene areas produce a large aromatic interface of free π electrons and the unique physical and chemical nature of this interface allows GO to interact with proteins, peptides, and DNA. 23 Aligned poly‐ l ‐lactide scaffolds coated with GO have demonstrated an excellent ability to mediate neurite growth and differentiation. 24 In addition, GO/polyacrylamide composite hydrogels with 0. 4% GO content can improve the attachment and proliferation of Schwann cells. 25 GO and PCL can also be successfully combined to fabricate biocompatible nanofibrous scaffolds. Furthermore, the addition of moderate amounts of GO to PCL significantly strengthened the attachment and proliferation of mesenchymal stem cells. 26 In addition, researchers exert external electrical stimulation to graphene and other conductive materials and improve nerve regeneration. Heo et al. fabricated graphene/polyethylene terephthalate film to induce nerve differentiation under electrical stimulation. They effectively strengthened cell‐cell communication, which might result from changes in the intercellular coupling with endogenous cytoskeletal proteins. 27 Song et al. fabricated polypyrrole/PLCL (Poly (l‐lactic acid caprolactone)) conduit and used it to influence PC12 cells and dorsal root ganglion in vitro and peripheral nerve repair in vivo under electrical stimulation. 28 Therefore, we wonder whether GO/PCL scaffold alone can also improve peripheral nerve regeneration and why it is capable of doing so. Furthermore, we assume that GO/PCL scaffold might conduct the bioelectricity within peripheral nerves to accelerate nerve repair process without external electrical stimulation. Apart from the excellent electrical conductivity, biocompatibility and physical adsorption of GO, recent studies have made breakthroughs in elucidating pioneering roles of GO in angiogenesis. Angiogenesis is a basic process in tissue regeneration. Many strategies of nerve tissue engineering emphasize on mimicking native nerve characteristics including architecture, and protein composition. It serves as a bridge to direct reinnervation and provide necessary nutrition for nerve regeneration. GO can induce angiogenesis and contribute to nutrient formation and transportation in bone regeneration. 29 Low concentrations of GO can be pro‐angiogenic by regulating the intracellular expression of reactive oxygen species (ROS) and reactive nitrogen species (RNS). 30, 31 We also evaluate this important property of GO after long‐term nerve restoration in vivo and reveal the underlying mechanism in this study. In this study, we used a novel integration molding method to fabricate GO/PCL nanoscaffolds. Biocompatibility, cellular proliferation, and neural differentiation were evaluated using rat Schwann 96 cells (RSCs), a very commonly studied cell type in peripheral nerve regeneration experiments in vitro. 32 This was the first study to extensively evaluate the long‐term performance of GO‐based nanoscaffolds in peripheral nerve restoration. In vivo, the GO/PCL NGCs could efficiently heal a 15 mm sciatic nerve defect in a Sprague Dawley rat model by 18 weeks after injury. We also validated the influence of these innovative GO/PCL conduits on angiogenesis and explored the underlying mechanism. 2 Results and Discussion In this study, an integration molding method was used to fabricate GO/PCL nerve nanoscaffolds ( Figure 1 ). A tubular mold consisting of four tubes of concentric circles had been previously prepared. Two complex tubes were located between the inner tube and outer‐most tubes to form a concentric circle structure. A mixed solution of GO and PCL was injected into the space between the outer‐most and second outer‐most layers. After solidifying, the second outer‐most layer was removed, and the GO/PCL solution was injected into the space between the second and third outer‐most layers. The procedure was repeated a third time between the third outer‐most layer and the inner‐most layer. Finally, a 3D printer was used to create multiple aligned pores in the surface of the GO/PCL conduit. The multi‐layered structure strengthened the mechanical properties of the nerve conduit. In addition, the several porous layers facilitated biodegradation and optimized the long‐term in vivo performance of the NGC because the macropores between the different layers increased the possibility of internal contact with body fluid. Another major challenge in nerve tissue engineering is how to treat long‐range nerve defect over a long‐term regeneration. A strategy to do this is a controlled release system and a slow but steady biodegradation substrate material. In this study, the 3D printing enabled us to fabricate the conduit with a bottom‐up style, which indicates integrated multi‐layered fabrication. In addition, the printer also permitted digital control of mixed solution injection, and this resulted in an even distribution of different biomaterials in the conduit. Furthermore, the 3D printing enabled us to create a conduit with certain volume of different elements and to fabricate it from any angle, position, or plane. With high resolution, 3D printing could better improve RSC proliferation, attachment and neural expression. Figure 1 Schematic illustration of GO/PCL nanoscaffold fabrication by the integration molding method A) and NGC implantation in the rat model B). We prepared a tubular mold that included four tubes of concentric circles. Two complex tubes are located inside the inner‐most tube and the outer‐most tube, forming a concentric circle structure. A GO and PCL mixture was injected into the space between the outer‐most layer and the second outer‐most layer. After solidifying, the second outer‐most layer was removed, and the GO/PCL mixture was injected into the space between the second outer‐most layer and the third outer‐most layer. The same procedure was repeated again between the third outer‐most layer and the inner‐most layer. Finally, a 3D printer was used to create multiple aligned pores in the surface of the GO/PCL conduit. We characterized the GO/PCL nanoscaffold morphology by scanning electron microscopy (SEM) and optical imaging (Sirion 200/Instrumental Analysis Center (IAC), Figure 2 ). The nanoparticles and the stiff surface are shown at various magnifications. The macroporous structure could be identified at a lower magnification, revealing a clear and oriented alignment. The presence and distribution of GO nanoparticles in the tubular PCL structure were evaluated by transmission electron microscopy (TEM). The GO was confirmed to have an oriented distribution in the cross‐section of the GO/PCL nanoscaffolds (Figure 2 ). The diameter of the pores was just several micrometers due to the adjustable porosity of the 3D printer. The multiple aligned pores in the conduit enable exchanges of water, oxygen and other nutrients. At the same time, it prevents alien cells such as fibroblasts from entering the conduit and interfering with normal and functional nerve regrowth. Figure 2 Characterization of the GO/PCL NGC. Optical images of the GO/PCL NGC A). SEM images showing the nanoporous structure of the GO/PCL NGC B, C) and the multilayered structure of an ultra‐thin section D). TEM images showing the uniform distribution of GO nanoparticles in the PCL scaffolds E, F). Mechanical and electrical properties, i. e. , scaffold thickness, elongation at break, elastic modulus, and electrical conductivity of the GO/PCL and PCL scaffolds G). We further evaluated two mechanical properties of the 3D conduit, the scaffold thickness and the elastic modulus. The scaffold thickness was similar for both materials. Therefore, the addition of GO did not influence this property. However, the average elastic modulus of the GO/PCL conduit was 48. 32 MPa, in contrast to 31. 77 MPa for the PCL conduit. The elongation at break results also indicated that the addition of GO improved the mechanical strength of the nanoscaffolds. The reason for the improved elastic modulus and increased elongation at break of the GO/PCL nanoscaffolds is that the continuity of the PCL matrix interferes with the even distribution of the GO nanoparticles. In addition, GO can interact with the molecular structures of PCL, leading to increased intermolecular strength, 33 which was confirmed by the transmission FTIR spectroscopy in this study (Figure S1, Supporting Information). The mechanical test confirmed that the porous 3D GO/PCL conduit could maintain its tubular structure and allow nerve growth. The electrical conductivity of different scaffolds was also evaluated. The GO/PCL conduit displayed a relatively high conductivity of 4. 55 × 10 −4 S/cm, while the electrical conductivity of the PCL conduit was 0 S/cm. According to the above results, the GO/PCL NGC exhibited excellent mechanical and topological properties, including a multi‐layered design, ideal rigidity and flexibility, microporosity for the free exchange of nutrients, relatively high electrical conductivity, and even GO nanoparticle distribution for efficient drug delivery. The mechanical performance was improved by 3D printing multi‐layered fabrication in comparison with non‐layered counterparts (Table S1). To determine the optimal GO percentage in the PCL scaffolds, we cultured Schwann cells on 0. 5% GO/PCL, 1% GO/PCL, 2% GO/PCL, 4% GO/PCL, and PCL scaffolds in 24‐well plates for 1, 3, 5, and 7 days. The medium was replaced every two days. Twenty microliters of CCK‐8 solution was added to 200 µL of medium in each well, and cells were further cultured in a 5% CO 2 incubator for 4 hours. Then, 100 µL of medium from each well was transferred to a new 96‐well plate and the absorbance at a wavelength of 450 nm was determined using a multifunctional microplate reader (Thermo 3001, Thermo Fischer Scientific, USA). A tissue culture plate (TCP) was used as a control. This procedure was repeated for three independent samples for each scaffold. The results showed greater cell proliferation on the 1% GO/PCL scaffold than on the other scaffolds ( p < 0. 05) and that this proliferation was similar to that on PCL and TCP ( p > 0. 05) ( Figure 3 ). The fate of cell proliferation is determined by the nanoscaffold. The optimal amount of GO in the PCL scaffold would promote the greatest extent of RSC proliferation, which is important for further evaluating cell attachment, viability, and neural expression. Therefore, the 1% GO/PCL scaffold was selected for further evaluation. We found that GO was relatively biocompatible at a low level. A recent study on the influence of GO on cell proliferation indicated that the application of more than 2% GO in cell culture could lead to significant cytotoxicity. 34 This finding was consistent with our study, and we found that GO/PCL could be a biocompatible material at relatively low concentrations of GO. Figure 3 Cell viability as assayed by LIVE/DEAD cell staining on GO/PCL scaffolds A–C), PCL scaffolds D–F) and TCP G–I). Live cells (green fluorescence, A, D, and G). Dead cells (red fluorescence, B, E, and H). Merged images (C, F, and I). The scale bar is 50 µm. CCK‐8 assay for RSCs cultured on GO/PCL scaffolds with different concentrations of GO, PCL scaffolds and TCP at 24, 72, 120, and 168 h J). * P < 0. 05 compared with 0. 5% GO/PCL. # P < 0. 05compared with 2% GO/PCL. Δ P < 0. 05 compared with 4% GO/PCL. Relative cell viability was evaluated by the LIVE/DEAD cell staining for 1% GO/PCL scaffolds, PCL scaffolds and TCP K). To verify the viability of RSCs on the different nanoscaffolds, cell viability was assayed using a LIVE/DEAD cell kit for mammalian cells (Invitrogen) according to standard protocols. The images displayed high percentages of live cells and negligible differences were found in cell viability among the GO/PCL, PCL, and TCP groups (Figure 3 ). Cell attachment is important for ideal cell viability. SEM, immunofluorescence and real‐time quantitative PCR (qPCR) were adopted to test cell attachment and morphology on the different nanoscaffolds. After being cultured on the different scaffolds for 4 d, the RSCs were ready for SEM observation. The SEM images revealed that most of the scaffolds were covered with cells exhibiting normal structures and extended protuberances ( Figure 4 ). Figure 4 SEM images showing RSC morphology on GO/PCL and PCL scaffolds. RSCs were cultured on GO/PCL, and PCL nanoscaffolds for 4 d before observation. GO/PCL scaffold A–C). PCL scaffold D–F). The scale bars are 100 µm A, D), 50 µm B, E), and 20 µm C, F), respectively. Cell morphology and attachment were evaluated by staining actin cytoskeleton with phalloidin. Immunofluorescence showed the attachment and morphology of the cells on the GO/PCL and PCL scaffolds. The cells exhibited normal spindle‐like shapes ( Figure 5 ). This result further demonstrated the functional bioactive environment provided by the GO/PCL nanoscaffolds. Figure 5 Immunofluorescence staining for Tuj1 A, B), Ki67 C, D), and phalloidin E, F). All samples were washed three times, fixed with 4% paraformaldehyde for 20 min at 25 °C and blocked with BSA overnight. DAPI staining appears blue. GO/PCL scaffolds A1–A3, C1–C3, and E1–E3). PCL scaffolds B1–B3, D1–D3, and F1–F3). The scale bar is 50 µm. We also comprehensively evaluated the proliferation and attachment of cells on the GO/PCL scaffolds. Ki67 is a nuclear protein that is highly involved in cell proliferation. Ki67 immunofluorescence indicated greater cell proliferation on the GO/PCL scaffolds (Figure 5 ) than on the PCL nanoscaffolds. qPCR and flow cytometry (FCM) both showed that the Ki67 expression of cells cultured for 4 d on the GO/PCL nanoscaffolds was increased by 1. 8‐fold compared with those cultured on the PCL nanoscaffolds ( Figure 6 and Figure S2, Supporting Information). Western blot also showed higher expression levels of various proteins in cells cultured on the GO/PCL scaffolds than in those cultured on the PCL scaffolds (Figure 6 ). Some adhesion‐related genes, such as N‐cadherin, vinculin, and integrin, were also analyzed by qPCR. N‐cadherin is a transmembrane protein that forms adherence junctions between cells for interlinking and, thus, plays a notable role in cell adhesion. The expression of N‐cadherin by cells cultured on GO/PCL nanoscaffolds cultured for 4 d was 1. 2‐fold greater than that of cells cultured on PCL nanoscaffolds (Figure 6 ). Vinculin serves as a significant link between the actin cytoskeleton and adhesion molecules. The expression of vinculin by cells cultured on the GO/PCL nanoscaffolds cultured for 4 d was increased by 1. 2‐fold compared with that of cells cultured on the PCL nanoscaffolds (Figure 6 ). Integrin also has important implications for cell‐extracellular matrix (ECM) connections. The expression of integrin by cells cultured on the GO/PCL nanoscaffolds cultured for 4 d was increased by 1. 9‐fold compared with that of cells cultured on the PCL nanoscaffolds (Figure 6 ). These results confirmed the good ability of RSCs to adhere to the GO/PCL nanoscaffolds. Figure 6 Western blot and qPCR results for S100 A), GFAP B), nestin C), Tuj1 D), Ki67 E), GAP‐43 F), N‐cadherin G), vinculin H), and integrin I) expression of RSCs on the GO/PCL and PCL scaffolds. Three independent replicates were included for each group. Relative mRNA expression is shown compared with GAPDH. * P < 0. 05 compared with PCL. To confirm the impact of the GO/PCL nanoscaffold on RSC neural expression, we assessed the neural specific proteins, glial fibrillary acidic protein (GFAP), β III tubulin (Tuj1), and nestin, as well as the Schwann cell marker S100. GFAP and nestin are both intermediate filament proteins highly expressed by nerve cells. Tuj1 can distinguish neurons from glial cells as a neuron‐specific protein. The immunofluorescence results are displayed in Figures 5 and 7. In addition, we performed qPCR to analyze and quantify the expression of GFAP, Tuj1, nestin, S100, and growth‐associated protein (GAP‐43). GAP‐43 is a growth protein that is highly expressed in nervous tissues and is vital to axonal and synaptic development. The expression levels of GFAP, Tuj1, nestin, GAP‐43, and S100 were increased by 1. 1‐fold, 2. 4‐fold, 1. 8‐fold, 3. 5‐fold, and 1. 8‐fold respectively, on the GO/PCL nanoscaffolds compared with the PCL nanoscaffolds. For further validation, the relevant Western blot results are also shown (Figure 6 ). In summary, the GO/PCL scaffolds increased the expression of neuronal markers to a small extent. Figure 7 Immunofluorescence staining for S100 A, B), GFAP C, D), and nestin E, F). All samples were washed three times, fixed with 4% paraformaldehyde for 20 min at 25 °C and were blocked with BSA overnight. DAPI staining appears blue. GO/PCL scaffolds (A1–A3, C1–C3, and E1–E3). PCL scaffolds (B1–B3, D1–D3, and F1–F3). The scale bar is 50 µm. We further included a long‐term in vivo Sprague Dawley rat model in this study. Rats were randomly divided into three groups a GO/PCL conduit group, a PCL conduit group and an autograft group. All observations and examinations were completed at weeks 6, 12, and 18 postoperatively. No animals suffered an infection through postoperative day 7. No rats showed signs of edema, ulcers or failed surgical wound healing. None of the nerve conduits had degraded by 18 weeks after surgery. The GO/PCL conduit at implantation and the regenerated nerves at 18 weeks postoperatively in rats that received the GO/PCL conduit are displayed in Figure 8. Figure 8 Morphological evaluation of sciatic nerve and muscle regeneration at 18 weeks postoperatively. Optical images of the GO/PCL NGC at implantation A) and at 18 weeks after surgery B), as well as a dissected regenerated nerve section from the in vivo study C). Toluidine blue staining of a GO/PCL conduit D), a PCL conduit E), and an autograft F) at 18 weeks postoperatively. All samples were dissected from 15 mm sections of regenerated nerves. Ultra‐thin 5 µm thick sections were created using a cryostat. The gastrocnemius muscle from the injured side was also collected at 18 weeks postoperatively. The results for the GO/PCL conduits G), PCL conduits H), and autografts I) are displayed. The scale bar is 100 µm. For the functional and electrophysiological evaluations, we used walking track analysis and electrophysiological assessment methods. At 6 weeks after surgery, the sciatic function index (SFI) of the GO/PCL group (−33. 2) was notably higher than that of the PCL group (−39. 1, p < 0. 05) and lower than that of the autograft group (−29. 0, p < 0. 05). Similar results were observed at 12 weeks postoperatively. The SFI of the GO/PCL, PCL and autograft groups was −9. 5, 14. 4, and 6. 5, respectively. At 18 weeks, the SFI of the GO/PCL group was not significantly different from that of the autograft group ( p > 0. 05) (Figure S3, Supporting Information). Significantly less gastrocnemius muscle regeneration was observed in the PCL group than in the GO/PCL and autograft groups ( p < 0. 05). No notable difference was found between the GO/PCL and autograft groups ( p > 0. 05). The results of the electrophysiological analysis at 6 weeks postoperatively revealed that the nerve conduction velocity (NCV) of the GO/PCL group (13. 2 m s −1 ) was notably higher than that of the PCL group (11. 2 m s −1, p < 0. 05) and significantly lower than that of the autograft group (17. 3 m s −1, p < 0. 05). At 12 weeks after surgery, the NCV was 20. 8, 17. 2, and 25. 3 m s −1 in the GO/PCL, PCL, and autograft groups, respectively. At 18 weeks, the NCV of the GO/PCL group (33. 4 m s −1 ) was similar to that of the autograft group (37. 3 m s −1, p > 0. 05) and superior to that of the PCL group (29. 3 m s −1, p < 0. 05). At 6 weeks after surgery, the distal compound motor action potential (DCMAP) of the PCL group (7. 1 mV) was significantly lower than that of the GO/PCL group (9. 3 mV, p < 0. 05) and the autograft group (12. 3 mV, p < 0. 05). This trend was also observed at 12 weeks postoperatively. At 18 weeks, the DCMAP of the GO/PCL group (25. 1 mV) was not significantly different from that of the autograft group (27. 4 mV, p > 0. 05) (Figure S3, Supporting Information). TEM and toluidine blue staining were performed to validate the morphological improvement in the different groups. The representative sections that were included in the following experiments are displayed in Figure 8 C. In general, the total number, area, diameter, and thickness of regenerated nerves and myelinated axons of the GO/PCL group were markedly higher than those of the PCL group ( p < 0. 05) at 6, 12, and 18 weeks after surgery and were similar to those of the autograft group at 18 weeks despite significant differences at 6 and 12 weeks after surgery ( Figures 8 and 9 and Figures S4 and S5, Supporting Information). Toluidine blue staining also revealed significantly more nerve fibers in the GO/PCL group than in the PCL group at 18 weeks after implantation ( p < 0. 05, Figure 8 ). TEM further indicated that the thickness of myelinated axons in the GO/PCL group was significantly greater than that of the PCL group at 18 weeks ( p < 0. 05). However, there was no notable difference in the myelinated axon thickness between the GO/PCL and autograft groups at 18 weeks ( p > 0. 05, Figure 9 ). Figure 9 TEM images for transverse sections of regenerated nerves from a GO/PCL conduit A–C), a PCL conduit D–F), and an autograft G–I) at 18 weeks postoperatively. We evaluated cross‐sections from different samples and used uranyl acetate and lead citrate for staining. All specimens were observed by TEM. The scale bars are 10 µm A, D, and G), 2 µm B, E, and H) and 1 µm C, F, and I), respectively. The sciatic nerve controls the gastrocnemius muscle. Therefore, the functional regeneration of the sciatic nerve can be reflected by atrophy of the gastrocnemius muscle. 35 Random fields of view in images of different tissue sections were selected to identify muscle fibers using Image‐Pro Plus software. The percentage of muscle fiber area (Pm) was assessed using the equation Pm = Am/At × 100%, where Am is the mean area of muscle fibers, and At is the total area of the field. 36 The mean area of muscle fibers was much larger in the autograft and GO/PCL conduit groups than in the PCL group ( p < 0. 05). The mean area of muscle fibers was not significantly different between the GO/PCL and autograft groups ( p > 0. 05, Figure 8 and Figure S4, Supporting Information), indicating that GO/PCL can reverse muscle atrophy to some extent in long‐range defects of the sciatic nerve. Microvessels are vital to Schwann cell migration and peripheral nerve regeneration. To verify the influence of the different scaffolds on angiogenesis in the process of nerve regeneration, various markers were used including CD34 and CD31. CD34 is a hematopoietic transmembrane protein that is closely associated with vascular‐associated tissue. CD31 is also known as platelet endothelial cell adhesion molecule (PECAM‐1) which actively participates in endothelial cell junction formation and angiogenesis. The microvessel density (MVD) was assessed by immunostaining for CD34, as shown in Figure 10 and Figure S6 in the Supporting Information. The MVD was significantly lower in the PCL group than in the GO/PCL and autograft groups ( p < 0. 05), and while it was slightly higher in the autograft group than in the GO/PCL group, this difference was not significant ( p > 0. 05, Figure 11 ). Figure 10 Assessment of angiogenesis in sciatic nerve regeneration at 18 weeks postoperatively. Immunohistochemistry staining for CD31 in regenerated nerve samples from a GO/PCL conduit A), a PCL conduit B), and an autograft C) at 18 weeks after surgery. CD31 is important for endothelial cell intercellular junctions and is extensively involved in angiogenesis. CD31 + cells are indicated by arrows in each picture. CD34 is a transmembrane protein that is associated with vascular tissues. Immunofluorescence staining for CD34 in regenerated nerve samples is also shown for a GO/PCL conduit D), a PCL conduit E), and an autograft F) at 18 weeks after surgery. The scale bars are 100 µm A–C) and 50 µm D–F), respectively. Figure 11 Quantification of CD31 + ‐ and CD34 + cells based on various measurements of sciatic nerve samples from the GO/PCL conduit, PCL conduit and autograft groups at 6, 12, and 18 weeks postoperatively. CD31 area (mm 2 )/region of interest (ROI) (mm 2 ) A). VLS area (mm 2 )/ROI (mm 2 ) B). VLS number/ROI (mm 2 ) C). Average VLS area (mm 2 ) D). Quantification of CD34 + region via MVD (per mm 2 ) E). Quantification of muscle fiber area (%) (F). * P < 0. 05 compared with autograft; # P < 0. 05 compared with PCL conduit. CD31 staining was performed via immunohistochemistry assays as previously described. 37 The CD31 + area, vessel‐like structure (VLS) area and density ((VLS area+ CD31 + area)/total scaffold area) were better in the autograft and GO/PCL groups than in the PCL group. There were no significant differences in these metrics between the autograft and GO/PCL groups ( p > 0. 05, Figures 10 and 11 and Figure S6, Supporting Information). After demonstrating the angiogenic capability of the GO/PCL NGC in sciatic nerve regeneration, we further considered the potential mechanism behind this important phenomenon. We purified proteins from the regenerated sciatic nerves of the Sprague‐Dawley rats for western blot at 18 weeks postoperatively in the three experimental groups. We evaluated the expression of AKT, p‐AKT, endothelial nitric oxide synthase (eNOS), p‐eNOS, vascular endothelial growth factor receptor (VEGFR) 2, and p‐VEGFR2 with β‐actin as a loading control. The WB pictures were displayed in Figure 12. From the pictures, there were no obvious differences in AKT expression among the GO/PCL, PCL and autologous groups. p‐AKT expression was significantly lower in the PCL group than in the GO/PCL and autologous groups ( p < 0. 05). Similarly, there were no significant differences in eNOS or VEGFR2 among the three groups ( p > 0. 05). The expression levels of p‐eNOS and p‐VEGFR2 were notably elevated in the GO/PCL and autologous groups compared with the PCL group, reflecting the potential role of GO instead of PCL in regulating the eNOS and VEGF signaling pathway activation. Comparing all the evaluated proteins revealed that GO contributed to a definitive improvement in AKT signaling pathway activation, thus, initiating downstream NO activation as confirmed by the upregulation of p‐eNOS, finally leading to the stimulation of VEGF expression and contributing to angiogenesis in the injured peripheral nerve. Figure 12 Western blot results for AKT, p‐AKT, eNOS, p‐eNOS, VEGFR, and p‐VEGFR expression in regenerated nerves from the GO/PCL conduit, PCL conduit and autograft groups at 18 weeks after surgery. The relative expression shown was normalized to that of β‐actin. From left to right: GO/PCL conduit, autograft and PCL conduit. * P < 0. 05 compared with autograft; # P < 0. 05 compared with PCL conduit. The regeneration process of long‐range nerve defects primarily relies on a physical bridge between two nerve stumps and the chemical guidance of bioactive molecules and proteins. 38 NGCs have been widely accepted and used as a successful biomaterial in studies of peripheral nerve injury. 39 Masand fabricated a peptide‐modified nerve conduit with polysialic acid and a human natural killer cell epitope and found that it was highly effective for promoting myelination, axonal growth and motor neuron recovery in a mouse femoral nerve defect model. 40 Huang used an active silk conduit in rat sciatic nerve repair and discovered that long gaps, including 11 and 13 mm gaps, could be successfully repaired by 12 weeks after injury. 41 Wu fabricated a bioactive polyurethane nanoscaffold and discovered that it upregulated neurotrophin expression by activating voltage‐gated calcium channels to improve peripheral nerve regrowth. 42 Wang fabricated nerve scaffolds with different stiffnesses and evaluated their roles in peripheral nerve regeneration. In their research, higher PCL concentrations in the poly(propylene fumarate)‐co‐PCL scaffold better improved peripheral nerve function and structural repair. 43 However, to achieve fully functional and structural recovery, an advanced bioactive scaffold is needed to provide an ideal environment for nerve tissue regrowth. Conductive nanoscaffolds have the advantage of promoting electrical signal transduction in cell‐cell junctions. 44 GO is a representative new conductive material with the ability to promote cell attachment, 45 proliferation, 46 and more importantly, angiogenesis. 47 The integration molding method enables GO and PCL to be mixed to fabricate an integrated scalable nanoscaffold, which imparts great convenience and efficiency. 48 In addition, integrated fabrication can successfully solve many problems of traditional electrospinning method, including the uneven distribution of GO nanoparticles, weak rigidity and elasticity, and imbalanced quality control. The rough and wrinkled surface allows GO to mimic the natural state of the ECM. Thus, RSCs can adhere, proliferate and migrate on GO/PCL nanoscaffolds. In our study, vinculin, N‐cadherin and integrin were used to evaluate cell adhesion on the GO/PCL nanoscaffolds. The results showed that the GO/PCL nanoscaffolds contributed to favorable RSC attachment due to excellent simulation of the cell‐matrix interface. Apart from the cell morphology on the scaffolds, we also observed cross‐sectional images of GO/PCL sheets. Very few cells grew within the scaffolds and exhibited disorganized shapes (Figure S7, Supporting Information). This revalidated that RSCs exhibited firm attachment to the GO/PCL nanoscaffold. The multiple pores of the GO/PCL nanoscaffolds permit the free exchange of nutrients, such as necessary proteins, oxygen and water, thus facilitating the delivery of energy for nerve regeneration. It was previously reported that porous nanoscaffolds could mediate cellular functions in vitro, such as proliferation and migration. 49 Ki67, which is an important indicator of the proliferative state of cells, was used in our study to show that the GO/PCL scaffolds promoted cell proliferation. The enhanced proliferation was further demonstrated by western blot, qPCR and immunofluorescence assays. GO has multiple functional groups that can reduce its cytotoxicity to different cells and tissues. However, the functional groups also sacrifice some of the mechanical and electrical properties of GO, thus diminishing the ability of the material to maintain neural cells. In this study, Tuj1, GFAP, GAP‐43 and nestin were used to evaluate the influence of the GO/PCL scaffolds on RSCs. The results revealed slightly elevated levels in these proteins in cells cultured on the GO/PCL nanoscaffolds compared with those cultured on PCL nanoscaffolds, which might result from relatively low concentration of GO in the scaffold. At the meantime, we found better RSC performance of proliferation, angiogenesis and neural expression in GO/PCL and PCL scaffolds compared with other traditional substrate materials in nerve regeneration, such as PLGA (poly(lactic‐co‐glycolic acid)) and collagen (Figures S8 and S9, Supporting Information). This revalidated that PCL was a common and advantageous system and the addition of GO further improve the excellent performance in nerve tissue engineering. This is the first study to evaluate the ability of GO to promote long‐term peripheral nerve regeneration in vivo. A rat model of peripheral nerve injury was established by creating a 15 mm defect in the sciatic nerve to determine whether GO could fully repair the injury within 18 weeks. The functional and electrophysiological evaluation confirmed that the GO/PCL scaffolds better promoted axonal regeneration than did the PCL scaffolds. In the histological examination, the total number, area, diameter and thickness of regenerated nerves and myelinated axons were higher in the GO/PCL group than in the PCL group. Moreover, the gastrocnemius muscle, which is manipulated by the sciatic nerve, also exhibited less atrophy in the GO/PCL conduit group than in the PCL conduit group, indicating the role played by GO in neural repair. We also analyzed high GO concentration in PCL scaffold for long‐term peripheral nerve regeneration. The morphological results showed significant toxicity of 2% and 4% GO in PCL led to poor nerve regrowth (Figure S10, Supporting Information). Angiogenesis is a necessary process during which new blood vessels appear and grow in the original vessel channel. Angiogenesis plays a vital role in many important physiological mechanisms, such as cardiovascular diseases, wound healing and cancer. 50, 51, 52 In peripheral nerve injury, angiogenesis also serves as an important agent for nerve repair as it can offer a physical link between nerve stumps, provide injured nerves with fresh nutrients and help remove wastes produced in the physiological process, during which ROS are activated as a result of regional inflammation. 53 Some studies have reported that low ROS concentrations may be pro‐angiogenic. However, high ROS concentrations may be anti‐angiogenic. 54 Furthermore, it is difficult to manipulate ROS. In addition to ROS, NO, regulates angiogenesis in various cardiovascular diseases, 55 and NOS facilitates NO release from L‐arginine. Of the three isoforms of NOS, eNOS produces NO in blood vessels and participates in mediating vascular function. In addition, the activity of PI3K and AKT is correlated and these molecules further activate eNOS in the mediation of angiogenesis, which has become widely studied by many researchers. 56, 57, 58 In this study, we focused on the degree of angiogenesis and its underlying mechanism in a rat sciatic nerve injury model. After 18 weeks, regenerated sciatic nerves from the three experimental groups were dissected for evaluating angiogenesis. CD31 and CD34 were selected as immunochemistry markers. CD31, also termed PECAM‐1, was discovered on the superficial layer of platelets, neutrophils, and some T cells and has been confirmed to be present in endothelial cell intercellular junctions. 59 CD34, which belongs to a family of a single‐pass transmembrane proteins, was found in association with early hematopoietic and vascular‐associated tissue. 60 Different MVDs were found in the three groups by CD34 staining. Furthermore, CD31 contributes to the strong angiogenic power of GO in the process of sciatic nerve regeneration. By measuring the VLS area and density, the degree of angiogenesis of the regenerated nerves in the GO/PCL conduits was further quantified. We then sought to identify the potential mechanism behind the pro‐angiogenic characteristic of the GO/PCL nanoscaffolds. The AKT protein expression level did not significantly differ among the GO/PCL, PCL and autograft groups. However, the p‐AKT expression level was markedly higher in the GO/PCL group than in the PCL group, indicating initiation of the AKT signaling pathway by GO. Similarly, p‐eNOS was highly upregulated in the GO/PCL group compared with the PCL group, while eNOS was not. This finding is consistent with previous publications showing that eNOS is a downstream protein regulated by the AKT signaling cascade. 61, 62 VEGF is widely considered a major element of angiogenesis initiation. The binding of VEGF with VEGF receptors (e. g. , VEGFR2) induces endothelial cell proliferation, adhesion and migration and further mediates activation of the angiogenic process downstream of NO to improve vessel permeability. We also confirmed p‐VEGFR2 activation in vivo by western blot. Thus, AKT‐eNOS‐VEGF signaling was confirmed to be involved in the angiogenic ability of GO in the process of peripheral nerve regeneration. Several other molecules stimulate angiogenesis, including bFGF, Ang1, Ang2, and ephrin. The mechanism discussed above is only one possible signaling cascade. In future work, we will focus on determining other leading factors that may govern the healing and regrowth of vessels, which has major implications for peripheral nerve regeneration. Material fate is very vital to a successful nerve scaffold and functional nerve regeneration. We also evaluated the gradual GO release and PCL degradation in the long‐term in vivo study. The GO was released along with PCL degradation due to its nanoscale. Nanosized GO hardly diffused to the environment in the long‐term in vivo experiment from PCL scaffold. In this way, we displayed the PCL biodegradation and GO release at different time points in vivo. The details were shown in Table S2 and Figure S11 in the Supporting Information. To correlate GO release profile with tissue reconstruction efficiency, we performed the graph concerning the correlation of GO release and peripheral nerve and vascular regeneration at 6, 12, and 18 weeks after surgery. The increase of GO release is accompanied by increased nerve morphological recovery and angiogenic restoration (Figure S12, Supporting Information). This validates our previous results and statement that GO/PCL scaffold can improve long‐range peripheral nerve defect effectively and this phenomenon is closely related with GO release to the surrounding environment. In conclusion, we successfully created GO/PCL NGCs using the integration molding method. Our NGC combines an excellent conductive material GO and a biodegradable stiff material PCL fabricated by 3D integration molding method. Our scaffold design takes biochemical cues, electrical cues and topographical cues into account, which are major factors to make a successful nerve guidance conduit in the peripheral nerve tissue engineering. 63 Our in vitro study indicated that GO promoted cell attachment, cell proliferation and neural property maintenance. The in vivo study confirmed that the GO/PCL NGC and autologous nerves promoted healing to a similar extent in a 15 mm sciatic nerve defect model. Therefore, GO/PCL NGCs address current drawbacks associated with nerve guidance and neurite regeneration and have huge potential for use in peripheral nerve regeneration. 3 Conclusion In this study, the effects of GO/PCL nanoscaffolds in nerve repair were evaluated both in vitro and in vivo, and the results showed excellent functional and morphological recovery equivalent to those of autografts. We focused on the pro‐angiogenic characteristic of GO and the potential mechanism behind this key phenomenon. The nanoscaffolds directly contributed to successful functional restoration in a long nerve defect model. We plan to continue studying the complex interplay between GO nanomaterials and nerve regeneration. 4 Experimental Section Integration Molding Method for Producing GO‐Coated PCL Nanofiber Scaffolds and Conduit : GO and PCL were purchased from Sigma Aldrich. The GO nanoparticles were mixed with PCL in a uniform solution and sonicated for 5 min. The injectable suspension was then contained in a previously designed mold and further prepared by a jet spraying process. Using a nozzle and compressed air, the solution was sprayed from a collection container. The dichloromethane evaporated with the formation of the GO/PCL membrane. The integration molding method was used to create a multi‐layered GO/PCL conduit. Finally, a 3D printer was used to create various evenly distributed pores in the surface of the GO/PCL conduit. Characterization of the GO/PCL Conduit : The GO/PCL nanoscaffold was examined by SEM to evaluate the surface structure and by TEM to examine the GO nanoparticle distribution. Samples were coated with gold prior to observation. A Sirion 200/IAC SEM system was used for observation at an accelerating voltage of 5 kV. Images were captured at 100×, 2000×, and 5000× magnification. A JEOL JEM‐2010 (HT) electron microscope was used for TEM. The images were selected at different magnifications and random fields of view for the final assessment. The molecular orientation of two scaffolds was evaluated using transmission FTIR spectrometer (Nexus670, ThermoNicolet). In addition, the surface elastic modulus and elongation at break were measured by nanoindentation (Nano Indenter G200, Agilent, USA). At least six indentations were recorded for the final statistical evaluation. In addition, the conductive capability was measured via a four‐point probe method using a Hall Effect Test System (DX3000, Dexing Magnet Tech, China). RSC Culture and Proliferation Assay : RSCs were purchased from the cell bank of the Chinese Academy of Sciences (Shanghai, China). RSCs were cultured in high‐glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Gibco, USA) and 1% penicillin/streptomycin solution (Gibco, USA). Cells were incubated in a humidified atmosphere at 37 °C and 5% CO 2. The GO/PCL nanoscaffolds were sterilized by a 4 h immersion in ethyl alcohol and overnight exposure to ultraviolet light. Cells were seeded on the nanoscaffolds at a density of 2 × 10 4 cm −2. CCK8 was used to assess cell proliferation. To determine the optimal GO percentage in the PCL scaffolds, the cells were cultured on 0. 5% GO/PCL, 1% GO/PCL, 2% GO/PCL, 4% GO/PCL, and PCL scaffolds in 24‐well plates for 1, 3, 5, and 7 d. The medium was replaced every 2 d. Twenty microliters of CCK‐8 solution was added to 200 µL of medium in each well, and cells were further cultured in a 5% CO 2 incubator for 4 h. Then, 100 µL of medium from each well was transferred to a new 96‐well plate and the absorbance at a wavelength of 450 nm was determined using a multifunctional microplate reader (Thermo 3001, Thermo Fischer Scientific, USA). TCP was used as a control. Three independent samples were evaluated with RSCs for each nanoscaffold. Cell Viability Assay : RSCs were cultured on 1% GO/PCL scaffolds, PCL scaffolds and TCP. After 24 h of culture, cells on the different nanoscaffolds were washed gently with phosphate‐buffered saline (PBS). Then, a LIVE/DEAD cell staining kit (Invitrogen) was used to quantify cell viability according to standard protocols. Finally, all samples were observed by an inverted phase contrast microscope. Cell Morphology : RSCs were cultured on GO/PCL and PCL nanoscaffolds for 4 d. The medium was replaced every 2 d. The morphology of the RSCs on the different scaffolds was observed by SEM (Hitachi). First, the medium was removed from the cell culture and replaced with fresh Dulbecco's PBS (Gibco, USA). Cells on the scaffolds were then fixed with 2. 5% glutaraldehyde for 12 h at 4 °C. After the fixation solution was removed, 1% osmium acid was added, and the cells were cultured for 2 h at 4 °C. The samples were dehydrated by a graded ethanol series (30%, 50%, 70%, 80%, 90%, 95%, 100%) twice for 20 min each. Lyophilization and gold coating were employed for the final analysis. The samples were observed by SEM to evaluate cell morphology and attachment on the scaffolds. Immunofluorescence : After 4 d of cell culture, the cells on the scaffolds were gently washed with PBS three times, fixed in 4% paraformaldehyde for 30 min, and immersed in 0. 1% Triton X‐100 (Sigma) for 5 min. The samples were then blocked with 5% bovine serum albumin (BSA) and incubated with primary antibodies overnight at 4 °C followed by 2 h of incubation with secondary antibodies at room temperature. Finally, 4, 6‐diamidino‐2 phenylindole (DAPI) (1:500, Gibco, USA) was used to stain the nuclei. The primary antibodies were anti‐nestin (1:100, Abcam, USA), anti‐Tuj1 (1:500, Abcam, USA), anti‐GFAP (1:1000, Abcam, USA), anti‐Ki67 (1:250, Abcam, USA), and anti‐S100 β (1:100, Abcam, USA). The secondary antibody was Alexa Fluor 488‐conjugated mouse anti‐rabbit IgG (1:200, Gibco USA). For the cell attachment analysis, F‐actin was stained after the cells were cultured for 4 d on different scaffolds (GO/PCL and PCL). The procedures were similar to those described above. Phalloidin conjugated to Alexa Fluor 488 (1:200, Abcam, USA) was used to stain actin filaments. All samples were observed by inverted phase contrast microscopy. WB, FCM, and RT‐qPCR : Western blot was conducted as follows. Cells were lysed in RIPA lysis buffer to collect total proteins. SDS‐PAGE was performed, and the samples were transferred onto PVDF membranes. The samples were then incubated at 4 °C overnight with the following primary antibodies: anti‐GAP‐43 (1:10000, Abcam USA), anti‐GFAP (1:10000, Abcam, USA), anti‐Ki67 (1:5000, Abcam, USA), anti‐Tuj1 (1:1000, Abcam, USA), anti‐nestin (1:2000, Abcam, USA), anti‐S100 β (1:5000, Abcam, USA), anti‐N‐cadherin (1:5000, Abcam, USA), anti‐vinculin (1:10000, Abcam, USA), and anti‐integrin α5 (1:1000, Gibco, USA). The Ki67 expression was also evaluated using FCM. Total RNA was extracted from the cells using TRIzol reagent (Gibco, USA) according to the manufacturer's protocol. RNA was reverse‐transcribed into cDNA with PrimeScriptTM (Takara). According to the manufacturer's instructions, samples were run on a real‐time PCR biosystem. The primer sequences were N‐cadherin (Forward (5′‐3′) CAGGGCCCTTTGCATTTGAC), integrin (Forward (5′‐3′) TGTCCTACTGGTCCCGACAT), vinculin (Forward (5′‐3′) TGGTCTAGCAAGGGCAATGAC), Ki67 (Forward (5′‐3′) ACAGGGCTTAGGAAACAGTCC), nestin (Forward (5′‐3′) GGGGGTAGGAGATGCCTTTG), GFAP (Forward (5′‐3′) TGCATGTACGGAGTATCGCC), GAP‐43 (Forward (5′‐3′) ACCTAAGGAAAGTGCCCGAC), Tuj1 (Forward (5′‐3′) AGCTCACCCAGCAGATGTTC), S100 (Forward(5′‐3′), CGATGCCCCGGAAAGTTAGA), and GAPDH (Forward(5′‐3′), GGCAAGTTCAACGGCACAGT). Animal Surgery : 45 male Sprague Dawley rats (weighing 150–200 mg) were housed in a specific pathogen‐free atmosphere. The animals were randomly divided into three experimental groups: a GO/PCL group (15 rats), a PCL group (15 rats), and an autograft group (15 rats). Animals were anesthetized via an intraperitoneal injection of 40 mg kg −1 sodium pentobarbital. Under sterile conditions, a small incision was created in the right leg of the rat to expose the sciatic nerve located 4 mm below the skin. The surrounding muscles were detached with blunt dissection. Then, a 15 mm defect was created in the right thigh of each rat for the different implants. A GO/PCL conduit, a PCL conduit or an autologous nerve was sutured to the proximal and distal ends of the injured nerve with 6‐0 nylon sutures. The muscle soft tissue and skin were sutured accordingly with 3‐0 nylon sutures. Then, 800 000 units of penicillin were administered to each rat immediately after surgery to prevent infection. Subsequent postoperative observations were made at week 6, 12, and 18. In this study, all animal care and use were performed according to the guidelines approved by the Institutional Animal Care and Use Committee of Shanghai Jiao Tong University (SJTU, No. A2017072). Functional Analysis and Electrophysiological Assessment : Infection, edema, and surgical wound healing were assessed in each group. Walking track analysis was used to evaluate nerve regeneration based on the following formula: SFI = (−38. 3 × (EPL − NPL)/NPL) + (109. 5 × (ETS − NTS)/NTS) + (13. 3 × (EIT − NIT)/NIT) − 8. 8. The foot print length is the distance between the heel and the top of the third toe, and toe spread is the distance from the first to the fifth toe, and intermediary toe spread is the distance between the second and fourth toes. These metrics were determined for the foot on the unoperated side (normal foot print length, NPL; normal toe spread, NTS; normal intermediary toe spread, NIT), and evaluations of the experimental foot print length (EPL), experimental foot toe spread (ETS), and experimental foot intermediary toe spread (EIT) were made for the foot of the operated side. The SFI ranges from −100 to 0, with −100 corresponding to complete nerve dysfunction and 0 representing good repair. Electrophysiological assessments of the NCV and DCMAP, were conducted immediately after the walking track analysis at 6, 12, and 18 weeks postoperatively. Before starting the electrophysiological recordings, conduits that were not degraded were removed from the regenerated nerves under anesthesia. A monopolar recording device was used to record the NCV and DCMAP with digital electromyographs of the gastrocnemius muscle for all groups. Histological Analysis : Immediately after the electrophysiological evaluations, 15 mm sections of the right regenerated nerves of the experimental rats were dissected. The NGC was carefully removed with the regenerated sciatic nerve remaining inside. Thereafter, the conduit was opened with small scissors to expose the nerve. Nerves were cut into ultra‐thin 5 µm thick sections for further histological evaluation. Both 1% toluidine blue staining and TEM were performed. The regenerated nerve number and thickness, the diameters of myelinated fibers, and the thickness of the myelin sheath were calculated using Image‐Pro Plus, as described in the previous studies. The gastrocnemius muscles of the experimental leg was removed and stained with hematoxylin and eosin. Random fields of view were selected in the images to evaluate the muscle fibers using Image‐Pro Plus, and the muscle was assessed by the percentage of muscle fiber area (Pm) according to the equation: Pm = Am/At × 100%, where Am is the area of muscle fibers, and At is the total area of the field. Assessment of Angiogenesis : At 6, 12, and 18 weeks postoperatively, the middle sections of the regenerated sciatic nerves were dissected, and the transverse layer was used for CD34 immunofluorescence assays. In short, tissues were prepared on a thin slide. The primary antibody was rabbit anti‐CD34 antibody (1:100, Abcam, USA), and FITC conjugated goat anti‐rabbit antibody (1:200, Abcam, USA) was used as a secondary antibody. Quantification was performed by counting the number of microvessels in randomly selected views to evaluate the MVD. The CD31 staining procedure was as follows. In brief, rehydration of the sections in paraffin by a graded ethanol series was followed by antigen retrieval via heating in sodium citrate. The samples were then incubated in the primary rabbit anti‐human CD31 antibody (1:150, Abcam, USA) in 2% NDS at 4 °C overnight. Then, the slides were incubated with the secondary biotin‐conjugated donkey anti‐rabbit antibody (1:500, Abcam, USA) for 1 h at room temperature. Then, the sections were dehydrated and mounted. Quantification of the pro‐angiogenic characteristic of the GO/PCL scaffolds was based on the VLS area, VLS density, and average VLS area as percentages. Western Blot Analysis : Regenerated nerve tissues were lysed in NP40 lysis buffer supplemented with 1 × 10 −3 m phenylmethyl sulfonyl fluoride and 1 × 10 −3 m protease inhibitor cocktail. Samples were subjected to SDS‐PAGE and analyzed by Western blot using specific antibodies for VEGFR2 (1:2000, Abcam, USA), p‐VEGFR2 (1:1000, Abcam, USA), eNOS (1:1500, Abcam, USA), p‐eNOS (Ser1177) (1:1000, Abcam, USA), AKT (1:1500, Abcam, USA), p‐AKT (Ser474) (1:1000, Abcam, USA) and β‐actin (as a loading control). The test was repeated three times, Images were scanned and band density was analyzed using Image‐Pro Plus software, version 6. 0. The uncut blots are displayed in Figure S13 in the Supporting Information. Material Degradation Measurement : The nerves were dissected and the remaining GO/PCL nerve conduits were kept at 6, 12, and 18 weeks postoperatively. The weight of the conduits from 6, 12, and 18 weeks after surgery was subtracted from the original GO/PCL conduit to know the total amount of GO release and PCL degradation. In order to quantitatively show the GO release, the 6‐week, 12‐week, 18‐week, and preoperative GO/PCL nerve conduits in the dichloromethane solution (the same volume for each conduit) were dissolved separately. Then, the mixed solution received sonication. Afterward, the GO was extracted by centrifugation. GO sediment weight was measured by vacuum drying it for 24 h. The GO release and PCL degradation amount could be calculated according to the following formula: accumulation GO release amount (mg) = original GO weight (mg) − GO sediment amount (mg); PCL degradation amount (mg) = original conduit weight (mg) − remaining conduit weight (mg) − GO release amount (mg). Statistical Analysis : All tests were repeated three times, and the results are displayed as the mean ± standard deviation. Unpaired Student's t ‐tests were used for statistical analyses. A p value of 0. 05 was considered significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700513
| 2,018
|
Advanced Science
|
Polysaccharide‐Based Controlled Release Systems for Therapeutics Delivery and Tissue Engineering: From Bench to Bedside
|
Abstract Polysaccharides or polymeric carbohydrate molecules are long chains of monosaccharides that are linked by glycosidic bonds. The naturally based structural materials are widely applied in biomedical applications. This article covers four different types of polysaccharides (i. e. , alginate, chitosan, hyaluronic acid, and dextran) and emphasizes their chemical modification, preparation approaches, preclinical studies, and clinical translations. Different cargo fabrication techniques are also presented in the third section. Recent progresses in preclinical applications are then discussed, including tissue engineering and treatment of diseases in both therapeutic and monitoring aspects. Finally, clinical translational studies with ongoing clinical trials are summarized and reviewed. The promise of new development in nanotechnology and polysaccharide chemistry helps clinical translation of polysaccharide‐based drug delivery systems.
|
1 Introduction The clinical efficacy of low‐molecular‐weight chemotherapeutics and functional biological macromolecules (i. e. , proteins and oligonucleotides) is often limited by a number of obstacles, including unfavorable solubility, loss of bioactive structure prior to reaching the disease lesion site, inadequate cellular uptake, short plasma half‐lives due to rapid renal clearance or enzymatic degradation, drug resistance driven by overexpression of the efflux transporter, and unwanted side effects of nonspecific cytotoxic drugs caused by off‐target effect during chemotherapy. 1 The development of a smart nanoscaled drug delivery system (nanoDDS) has entered the mainstream to not only address these issues but also to aid the advancement of personalized nanomedicine for noninfectious diseases, especially cancer. 2 With the achievable and tunable size and structure, such nanovehicles can be properly designed to cross the smallest capillary wall while avoiding clearance by a mononuclear phagocyte system (MPS), resulting in a prolonged blood stream duration. Due to the enhanced permeability and retention (EPR) effect, 3 macromolecules and large nanoparticles can be more effectively trapped in tumor tissues than low‐molecular‐weight molecules and small nanoparticles. 4 On the other hand, high‐molecular‐weight bioactive molecules (e. g. , cytokines and growth factors) have limitations due to their instability in delivery both in vitro and in vivo as well as their immunogenicity and shorter half‐lives. To overcome these limitations, modern drug formulation technologies have facilitated researchers' abilities to create the commonly named “second generation” of protein drugs to overcome the above limitations. Moreover, based on the molecular weight, secondary structure, and availability of surface groups, polymer–protein or fusion protein conjugates have been created. However, protein folding may also be altered through and after the modification process. 6 Therefore, there is a great need to design delicate DDSs to fulfill the protection of protein therapeutics with enhanced half‐lives and reduced immunogenicity. This strategy can then be used in protein pharmaceutical areas. For decades, various nanotherapeutics have been developed for use in humans, most of which can be formulated with several main types of DDSs, such as liposomes, micelles, polymeric conjugates, inorganic nanoparticles, and others. Among them, polysaccharides are the most recognized biomaterials that are derived from natural carbohydrate polymers. They are generally regarded as safe (GRAS) and are broadly used in the food industry. Here the term “GRAS” is a general standard defined under sections 201(s) and 409 of the Federal Food, Drug and Cosmetic Act (the Act), meaning any substances that have been generally recognized, among qualified experts and adequately shown to be safe under the conditions of its intended use. 7 They have also been applied as an excipient for drug formulation by regulatory authorities in different countries. 8 Basically, polysaccharides are carbohydrates with more than two sugar molecules covalently bonded together by glycosidic linkage. In addition to polysaccharides, there are also monosaccharide and disaccharides within the definition of carbohydrates. They offer a wide range of functional versatility and structural diversity due to their variable molecular weight and abundant reactive groups (i. e. , amine, carboxyl, carbonyl, and hydroxyl groups) on the polysaccharide backbone. 9 Polysaccharides of natural origin commonly exist in various species, including plants (cellulose), animals (chitosan, nature origin obtained from chitin of exoskeletons crustaceans and insects), algae (alginate), and microorganisms (dextran). Compared to other types of synthetic hydrophobic polymers, polysaccharides hold a large number of hydroxyl groups or other hydrophilic groups, such as carboxyl groups in alginate and amino groups in chitosan, which afford additional aqueous solubility and reinforce bioadhesion and biorecognition characteristics via noncovalent bonding (e. g. , electrostatic interactions) between biological tissues and polysaccharides ( Figure 1 ). 10 For example, chitosan, the only natural positively charged polysaccharide, is capable of attaching to the negatively charged mucosal layers via electrostatic interactions. 11 Similarly, hyaluronic acid can recognize and bind to the glycoprotein CD44 antigen on cell surfaces. 12 Moreover, their intrinsic functional moieties can serve as attachment points for chemotherapeutics, imaging probes, and targeting agents using facile chemical modification, 13, 14, 15 such as PEGylation and antibody conjugation to provide prolonged circulation time and site‐specific accumulation activities. Furthermore, due to their parallel biochemical properties with human extracellular matrices, polysaccharides are easily recognized and metabolized by the body, 16 and they have been discovered to be involved in many biological processes including immune recognition and cell signaling, 17 which are responsible for activation of antimicrobial and anti‐inflammatory responses. 18 In addition, these biopolymers undergo enzymatic and/or hydrolytic degradation in vivo, leaving innocuous degradation products that can either be reused in biological systems or cleared by the immune system. 19 Thus, with the aforementioned features, polysaccharides have a promising future as an accessible therapeutic delivery system. Figure 1 Schematic illustration of unique physiochemical and biological properties of polysaccharides. As naturally based biomaterials, polysaccharides have been largely explored for their ability for targeted delivery and control released to improve the therapeutic index of drugs (e. g. , chemotherapeutics, antibiotics, proteins, peptides, and nucleic acids) using various routes of administration. 20, 21 The recent advances in polysaccharide‐based nanomaterials have driven apparent trends toward multifunctional and more complex controlled release systems (CRSs), which will take a big step forward in achieving theranostics and regenerative medicine with improved therapeutic efficacy, mechanical properties, and safety profiles. The aim of this review is to present the state‐of‐the‐art in identification, functionalization, characterization, and application of bioactive polysaccharides originated from natural sources. Various biomedical applications were emphasized including tissue engineering, regenerative medicine, and cancer theranostics. This review will help to explore and investigate novel chemical and biological strategies for functional materials, promoting the clinical translations of polysaccharide‐based materials in biomedical applications. 2 Polysaccharides and Chemical Modification for Controlled Release We have briefly introduced the importance of controlled drug throughout our introduction session. In general, the goal of controlled drug delivery, not only limited to polysaccharide‐based DDSs, are listed as follows. (1) To protect the drug from degradation. This is also being used in the protection of protein‐based biomolecules, such as cytokines and growth factors, which contain sophisticated secondary structure that can be degraded through delivery routes. 22 (2) To enhance the half‐life of certain drug. A common example is insulin delivery, which requires instant injection after each meal. Dextran–insulin nanoparticles have been created to meet this need. 23 (3) To maximize the therapeutic effects while reducing the side effects. This is commonly seen in cancer therapy, where chemotherapy/radiotherapy affects patients' body condition severely. We have a chapter in the later session to discuss how researchers have been creative to effective cure cancer with the novel technologies of theranostics. 24 (4) To take full advantages of existing drug in comparison with identifying a new drug molecular or potential intracellular pharmaceutical target. The research and economic burden to identify a novel drug molecule or to discover a novel signal pathway for drug target is huge. Therefore, researchers revisit some of existing drug molecules, which also have a comprehensive safety/therapeutic profiles as being in market already, and utilize novel drug delivery techniques to make them perform better or for some other type of disease, saving money on both the research and clinical trial stages. 25 There are many different principles as researchers' guidelines when designing novel DDSs. These principles are not limited to polysaccharide‐based DDSs, but being extensively studied and utilized throughout the design. These drug release mechanisms lead and trigger the design of versatile polysaccharide‐based DDSs with the help of various chemical modification as approaches. (1) The mesh size of the materials could control the diffusion and release of drug molecules. 26 When the porous structure of the diameter is larger than the drug over three times (γ mesh > 3 × γ drug ), diffusion is the dominant factor for drug release. Strokes–Einstein equation is usually used to determine the diffusivity ( D ), which depends on the size (radius) of the drug molecule (γ drug ) and the viscosity of the solution (η) ( R is the gas constant and T is the absolute temperature) 27 (1) D = R T 6 π η γ drug When the mesh size has similar radius with drug molecule (γ mesh ≈ γ drug ), drug diffusion will be stalled by steric hindrance. 26 Therefore, the approach to control the porous size becomes important in designing DDSs. Researchers have controlled the size of porous structure by adjusting the concentrations of polysaccharides or the cross‐linkers. (2) The particle degradation could control the release of drug molecules. Such design of DDSs usually contain degradable cross‐linker. 26 One common example is 3, 3′‐dithiobis(sulfosuccinimidyl propionate), which could be cleaved with reducing agents, such as glutathione (GSH) to achieve intracellular redox drug delivery. 28 Enzymatic degradation sequence could also be used in design novel degradable cross‐linker. Phelps et al. reported of a protease degradable peptide cross‐linker GCRDVPMSMRGGDRCG, which could be cleaved by matrix metallopeptidase and able to deliver growth factor in vivo. 29 (3) The material swelling behavior could control the release of drug molecules. 26 Polysaccharide‐based hydrogel particles will swell to absorb water and the size of the porous structure will be increased, releasing encapsulated drug. Various factors will contribute to the degree of swelling behavior, such as pH, 30 temperature, 31 ionic strength, 32 electric fields, 33 light, 33 and glucose, 34 which have been extensively studied in the field of drug delivery. (4) The mechanical deformation could also induce drug release from the matrix. This strategy is usually designed in cooperation with hydrogel system. By applying mechanical force, hydrogel matrix will deform, leaving enlarged mesh size and triggering convective flow within the network. 35 Such impulsive change usually create a pulsatile release profile in certain disease scenario, such as insulin delivery after each meal. 26, 36 Based on the chemical composition, structure, solubility, and derivative sources, there are many possible approaches to classify polysaccharides. Considering the chemical composition, we can divide polysaccharides into two categories: (1) homopolysaccharides or homoglycans, which contain a single type of monosaccharides, such as chitin and chitosan, starch, and cellulose; and (2) heteropolysaccharides or heteroglycans, which consist of multiple types of monosaccharides, such as alginate, glycosaminoglycan, hyaluronic acid, and pectin. Other strategy to categorize polysaccharides includes the electronic charge for each specific polysaccharide molecules. For example, chitosan has usually positive charge whereas alginate has negative charge in general. For this chapter, we discuss five different types of polysaccharides based on their origins: chitosan from shellfish, 36 alginate from algae, 37 hyaluronic acid from various mammals (human, pig, beef, etc. ) and bacteria, 38 dextran from bacteria, 39 and cyclodextrin, which are synthetic substances obtained from enzymatic degradation of starch. 40 We chose these types of polysaccharides as they have been extensively studied and applied in the field of CRS. Molecules with same side groups share similar chemical modifications. In addition, comparable preparation approaches could also be applied to form DDSs. Being as traditional polysaccharide‐based biomaterials, the industrial processing techniques are mature, providing good resources for researchers and clinicians to practice. All the following materials have been going or already gone for clinical trials for various applications. These polysaccharides are good representatives to demonstrate the great research potential and broad applications of polysaccharide materials: from bench to bedside. There are also other types of polysaccharides, which have also shown variety of modification and fabrication potential and biomedical applications, such as carrageenan, 41 pullulan, 42 pectin, 43 cellulose, 44 starch, 45 and some new polysaccharides from marine or bacterial origin. 46 Interested readers are referred to the reviews that we have listed above for more details. 2. 1 Chitosan‐Based CRS Second to cellulose, chitin is the second most abundant natural amino polysaccharide throughout the world, which is the typical component of shellfish exoskeletons and fungal cell walls. It is a linear cationic polymer of N ‐acetyl‐ d ‐glucosamine (2‐acetylamino‐2‐deoxy‐ d ‐glucose) units that are joined by β‐1, 4 linkages. 47 Chitosan is produced by the deacetylation of chitin, which consists of β‐(1, 4)‐linked deacetylated units of d ‐glucosamine and N ‐acetyl‐ d ‐glucosamine 47, 48 ( Figure 2 ). Chitosan has been studied and used within the pharmaceutical area for almost three decades. As is widely recognized, chitosan‐based nanoplatform is one of the most promising DDSs due to its positive attributes of superior biocompatibility, notable biodegradability (metabolized by lysozyme), low toxicity, and positively charged characteristics based on its primary amino groups (this property enables electrostatic interaction with negatively charged macromolecules, nucleic acids, proteins, mucosal surfaces, etc. ). 48, 49 In spite of these advantages, the aqueous solubility of chitosan is relatively poor at neutral pH in some cases. 50 However, molecular weight and residual acetyl groups in the chitosan may also play important roles in the solubility of chitosan. To overcome the solubility issue, acidic solution (pH <6. 5) has been used while introducing additional water solubilizing groups. Reducing the molecular weight and elevating the degree of deacetylation can also facilitate additional solubilization, but this can spontaneously affect the physicochemical properties of chitosan. 51 Figure 2 Structures of repeating units of some of the polysaccharides discussed in this review. Branching is not shown for dextran. The structure of alginate and hyaluronic acid are shown the two linkage types rather than a formal repeating unite. The chitin and chitosan structures shown represent extremes of a continuum of structures. Owing to the pH‐induced solubility of chitosan modified drug conjugates, Park et al. developed a pH sensitive glycol chitosan‐based drug conjugate for photodynamic therapy. 52 This DDS consisted of three functional moieties that grafted on amine groups along the chitosan chain. The photosensitive drug Chlorin e6 and polyethylene glycol (PEG) residues were crosslinked with glycol chitosan through dicyclohexyl carbodiimide (DCC)/ N ‐hydroxysulfosuccinimide (NHS)‐mediated amidation. 3‐Diethylaminopropyl isothiocyanate (DEAP) was then grafted onto the chitosan backbone via thiourea bond formation. DEAP was exploited as an endogenous stimulus for pH triggered drug release in acidic tumor tissue that lead to conformational changes of nanoparticles from a coiled (at pH 7. 4) into an uncoiled structure (at pH 6. 8). Moreover, the protonation of DEAP residues generates the additional singlet oxygen to provide higher phototoxicity for cancer cells. In another study, a facile and controlled graft polymerization of N ‐(2‐hydroxyethyl)prop‐2‐enamide and chitosan was achieved by using γ‐ray irradiation of bis(R, R′‐dimethyl‐R″‐acetic acid) trithiocarbonate. 53 The anticancer drug chromone‐3‐carboxaldehyde was then grafted on the amino groups of chitosan via Schiff‐base bond formation, which was a cleavable covalent bond that undergo hydrolysis at low pH conditions. Specifically, this amphiphilic copolymer conjugate could self‐assemble into micelle nanoparticles in a water solution. In addition to pH‐sensitive DDSs, Hu et al. described a selective redox‐responsive chitosan‐based glycolipid‐like micelles, which was able to control the drug release rate by GSH concentration in tumor cells. 54 The polymer was prepared via a two‐step synthesis. First, disulfide linker (bis‐2‐carboxyethyl disulfide) was conjugated with stearylamine through amide bond formation by applying DCC/4‐dimethylaminopyridine (DMAP). Second, the carboxyl‐terminated intermediate product was conjugated to amino groups on chitosan through the 1‐ ethyl‐3‐(3‐dimethylaminopropyl)carbodiimide hydrochloride (EDC a water soluble carbodiimide)/NHS‐mediated coupling reaction. The low pH reaction environment of the EDC‐catalyzed peptide formation also provided a good solubility environment for chitosan. This study demonstrated that the disulfide linkage could be designed as a cleavable linker to trigger the release of a conjugate payload under a certain level of GSH concentration, making the cargo drugs accumulated in tumor cells. Xu et al. reported an oxidation and pH‐responsive chitosan‐based nanoparticle that branched with ferrocene moieties for 5‐fluorouracil delivery. 55 The DDS was prepared by reductive alkylation of chitosan with ferrocenecarboxaldehyde in the presence of NaBH 4, 56 and spherical nanoparticle or vesicles were formed via self‐assembly of ferrocene–chitosan at different concentrations in an acid solution. The 5‐fluorouracil was encapsulated through ultrasonication, and this drug payload was control released in the presence of an oxidative agent at low pH conditions due to the layer‐by‐layer electrostatic repulsion and loss of ferrocene moieties (π–π stacking) of the nanoparticles. Of the available oral DDSs for treating digestive disorders, for example, Jing et al. prepared amoxicillin‐loaded chitosan nanoparticles to target the urea transport protein of Helicobacter pylori based on mild ionic gelation of ureido‐conjugated chitosan with sodium tripolyphosphate (TPP). 57 The ureido group was conjugated to the amino and hydroxyl position of chitosan through amidation and esterification with 12‐ureidododecanoic acid, respectively. The in vitro simulation experiment demonstrated that both of the DDSs were inactivated at pH 1. 2 but effectively released amoxicillin at pH 6. 0 and pH 7. 0, which resulted from the destabilization and weakened electrostatic interaction between chitosan derivatives and TPP at higher pH conditions. Moreover, the multiple amino groups on the backbone of chitosan would be able to perform any amine related conjugations with other molecules, including methacylation. 58 Methacylation reaction could enhance the durability of adhesive interfaces. Diolosà et al. reported of using methacrylic acid mixed with chitosan to enhance the adhesion durability with the restorative resin (hydrophobic layer) and the dentine (hydrophilic counterpart) for clinical dental restorations. Methacrylated glycol chitosan can also be cross‐linked using UV light with Irgacure 2959 photoinitiator. Carbodiimide chemistry could also be performed with the amine group on chitosan. Rafat et al. reported the use of EDC/NHS‐mediated cross‐linking reaction or hybrid polyethylene glycol dibutyraldehyde/EDC/NHS to combine collagen and chitosan molecules. These collagen–chitosan composite network scaffolds were verified to enhance the mechanical strength and elasticity for corneal tissue regeneration. 59 As is well established, unique properties of chitosan make it capable of therapeutic delivery for various application sites including oral, ocular, nasal, vaginal, buccal, parenteral, and intravesical drug delivery. 48 2. 2 Alginate‐Based CRS Alginate refers to linear anionic polysaccharides derived from brown algae and bacteria, consisting of repeating units of β‐1, 4‐linked d ‐mannuronic acid (M) and l ‐guluronic acid (G) in varying ratios. Their physicochemical properties (e. g. , mechanical flexibility, cross‐linking reactivity, and ionic binding types) have been found to be largely dependent on M/G proportion, the length of block segments, and molecular weight. 40, 60 High contents of G blocks of the alginate are able to form rigid hydrogels with divalent cations such as Ca 2+, each of which orderly binds to two opposing G blocks, resulting in so‐called egg‐box conformational arrangement ( Figure 3 ). 61 This is because calcium ions induce chain–chain association. A junction zone (the confirmation structure turn between two adjacent monosaccharide repeat units) was proposed by Grant et al. as confirmed by circular dichroism. 62 Such facile synthesis approach of these alginates makes them suitable for cell transplantation and tissue regeneration. The semipermeable membrane that alginate and calcium form allows the diffusion of nutrients and therapeutics, maintaining the transplanted cells to grow within the hydrogel, whereas transplanted cell population are protected from the immune systems, which are less immunogenic compared to free cell injection. 63 Oppositely, using high M block content of the alginates has been found to be less adhesive and exhibits immunostimulatory activity. 64 This was because M–alginate contained higher level of polyphenol, endotoxins, and proteins compared to G–alginates without additional purification steps before applying in vitro and in vivo. Orive et al. also suggested that purification of alginate by free‐flow electrophoresis would reduce the total impurity content, without provoking foreign body reactions. 38 This requires high processing and purification standard from alginate industry. The ionically cross‐linking reaction usually occurs by exchanging sodium ions from G blocks with multivalent cations (i. e. , Ca 2+ ). Therefore, a higher percentage of guluronic acid of the alginate type corresponds to a tighter ionically cross‐linking network, which results in a prolonged release profile. 65 The mild gelation method has enabled wide application of this type of particles in delivery drugs, plasmid DNAs (pDNAs), growth factors, or even live cells. 37 The formation of ionic cross‐linking is reversible—that is, adding chelating agent (i. e. , ethylenediaminetetraacetic acid) to the already formed nanoparticles will destabilize the cross‐link network, causing particle degradation. Figure 3 Molecular structure of calcium–alginate junction zone through ionic cross‐linking. 65 Alginates have been increasingly employed as a favorable delivery nanoplatform for biomacromolecules and a wide variety of other substances (e. g. , growth factors, 40 cytokines, 37 doxorubicin, 65 paclitaxel (PTX), 66 DNA, 67 RNA) 68 to achieve controlled release by varying pore size of alginate hydrogels 69 as well as by cleavable chemical conjugation using active hydroxyl and carboxyl groups in the polymer backbone. 40 This is also one of the most mucoadhesive polymers applied in tissue engineering applications. 70 The most popular routes of conjugation are either to form an amide bond by an EDC/sulfo‐NHS or DCC/DMAP reaction in either aqueous or organic solvent. 71, 72 The water‐based reaction is capable of directing bioconjugation without prior organic solvent dissolution, and the excess of reagent and byproducts can be easily removed by dialysis or gel‐filtration. In order to make alginate soluble in organic solvent, Pawar and Edgar reported a strategy to dissolve tetrabutylammonium salts of alginic acid in polar aprotic solvents containing tetrabutylammonium fluoride, 73 which is able to react with alginates homogeneously in organic solvents, such as dimethyl sulfoxide and dimethylformamide (DMF). In addition, it has also been reported that hydroxyl groups can be oxidized to form alginate aldehyde with more reactive groups and a faster degradation profile. 74, 75 Since the periodate oxidation of alginate cleaves the carbon–carbon bond of the cis‐diol group in the urinate residue ( Figure 4 ), which changes the chain conformation, it promotes the hydrolysis of alginate in aqueous solutions. 74 Numerous alginate conjugates have been reported; we now present some recent examples of alginate‐based CRSs. Figure 4 Schematic representation of the chemical synthesis of oxidized alginate. Feng et al. described redox‐sensitive alginate nanogels for intracellular delivery of doxorubicin (DOX). The nanogels were prepared via in situ cross‐linking of the alginate and the coupling agent cystamine through carbodiimide chemistry and a miniemulsion method. 76 DOX was encapsulated into nanogels by exploiting the electrostatic interactions between the cationic DOX and the anionic alginate. In one study, vascular endothelial growth factor A (VEFGA)‐encapsulated alginate microsphere with the incorporation of a cyclic arginylglycylaspartic acid (cRGD) peptide and PEG moieties was developed. 72 This alginate‐based microsphere was designed for receptor‐mediated intracellular delivery and release of the vascular endothelial growth factor A (VEGFA) in primary human mesenchymal stem cells (MSCs) to regulate osteogenic differentiation as a potential therapeutic application. In this study, amine‐terminated PEG oligomers were grafted on alginate through EDC/NHS activation, and cRGD–alginate conjugation included two steps. First, NH 2 –PEG–SH was reacted with 2, 2′‐dithiodipyridine to generate pyridine modified NH 2 –PEG. Then, the modified NH 2 –PEG–pyridine was coupled to alginate through carbodiimide chemistry. Another interesting example of alginate‐based oxidation‐responsive delivery system was formed by the conjugation of deferoxamine onto alginate aldehyde via a Schiff‐base reaction and followed by reduction. 77 These conjugates were explored for their removal of excess iron from the body. It was expected that the alginate–deferoxamine conjugates could protect deferoxamine from metabolism by globulin during circulation and release the active deferoxamine at target sites by local oxidative stress status. In addition, the hydroxyl group of alginate can be reacted with methacrylic anhydride via esterification, which can then be cross‐linked upon exposure to long‐wavelength UV light in the presence of a photoinitiator. 78 Jeon et al. developed a protocol using 2‐aminoethyl methacrylate to react with the carboxyl group on alginates ( Figure 5 ), providing alternatives to design alginate‐based photosensitive materials. 79 Jeon et al. also suggested that photo‐cross‐linked oxidized methacrylated‐alginate hydrogels can enhance cell adhesion and spreading compared to those prepared with nonoxidized alginate, 80 since the free aldehyde group can bind to amines present on cell surface proteins or extracellular matrices. 80, 81 These above strategies for alginate modification provide versatile functionalities for the delivery of therapeutics in a controllable manner and show potential with extensive implementation in the development of innovative DDSs. Figure 5 Schematic representation of the chemical synthesis of methacrylated alginate and photo‐cross‐linking of methacrylated alginate. Reproduced with permission. 79 Copyright 2009, RSC. 2. 3 Hyaluronic Acid (HA)‐Based CRS Hyaluronic acid, also known as hyaluronan, is a naturally occurring linear polysaccharide that consists of repeating disaccharide blocks of d ‐glucuronic acid and N ‐acetyl‐ d ‐glucosamine joined by a glycosidic linkage. It was first isolated from the vitreous humor of bovine eyes by Meyer and Palmer. 82 Later, it was found to ubiquitously exist in the extracellular matrix of most neural and connective tissues. 83 Due to the presence of the carboxyl group on each of the glucuronic acid units, HA is naturally negatively charged, which allows it to absorb a large amount of water and swell up to 1000 times its solid dimensions. 38 The swelling behavior allows the release of drug molecules in a controlled manner. 84 In the field of tissue engineering, the hydrophilic and viscoelastic properties of HA can not only reduce the friction of the joints but also provide a cushion effect for surrounding tissues. 83 In addition to its biodegradable and noncytotoxic features, HA has also been considered to be nonimmunogenic and have anti‐inflammatory properties that depend on its molecular weight. 85 HA is regularly involved in the regulation of angiogenesis, inflammatory, fibrosis, and cancer‐promoting processes. 86 Moreover, HA can also serve as a targeting molecule that specifically binds to some cell surface receptors including CD44 and receptor for HA‐mediated motility. 12, 87 Owing to its inherent bioactive nature, HA is widely applied as a targeting carrier for delivering therapeutics to tumor tissues, 88 and as important building blocks for tissue engineering and regenerative medicine, 89 as well as a common ingredient in cosmetic dermatology. 90 Similar to alginate, the available hydroxyl and carboxyl groups on HA are commonly used for conjugation via methacrylation 91 and carbodiimide reaction, 91 respectively. Methacrylated HA can be photo‐cross‐linked using either ultraviolet (UV) radiation 91 or visible light. 91 The distinctive acetamide group (—NHCOCH 3 ) of HA is available for deacetylation with the presence of hydrazine sulfate to restore the amine group, which can then undergo amidation reactions for further modification. 92 Due to the unique and valuable physicochemical property of HA, researchers have been able to design and modify the HA to obtain new specific features for therapeutic delivery in controllable manner. Some recent examples are now given. Hulsart‐Billström et al. demonstrated a two‐component hydrazide‐modified HA hydrogel‐based adhesive scaffold for bone regeneration through the enzymatic release of active bone morphogenetic protein (BMPs). 93 In this report, HA–aldehyde (HA–al), HA–hydrazide (HA–hy), and HA–bisphosphonate (BP)–hydrazide (HA–BP–hy) derivatives were used as starting materials, which were obtained from the carbodiimide‐mediated amide coupling of HA carboxyl groups at the carbazate terminal of the reagent. Subsequently, the HA–al solution was mixed with BMP‐2 containing solution of HA–BP–hy or HA–hy to formulate hydrogels encapsulating BMP‐2. The positively charged BMP‐2 can be electrostatically trapped in a negatively charged HA–BP hydrogel, and sustainably released through enzymatic digestion. Moreover, the BP functional group promoted the attachment of the cell to the surface of the HA hydrogel due to the additional Ca 2+ ‐mediated linkages. In another example, Baier et al. developed GSH responsive HA‐based nanocapsules by using Cu(I)‐catalyzed “click” reaction polymerization of azide‐functionalized HA and disulfide functionalized dialkyne at the oil‐in‐water miniemulsion droplet interface. 94 The encapsulated sample dye was released after cellular uptake through the cleavage of disulfide bridges with the presence of GSH in the polytriazole shell of HA nanocapsules. Fan et al. described a cationic liposome–HA–PEG hybrid nanopolyplexes (NPs) for intranasal vaccination with subunit antigens. 95 It was composed of a positively charged 1, 2‐dioleoyl‐3‐trimethylammonium propane liposomes with incorporation of negatively charged l ‐cysteine modified HA (HA–SH) and was further decorated with thiolated PEG via the thiolation of the HA–SH layer on the outer shell of the NPs. This study demonstrated that the F1‐V antigen and monophosphoryl lipid A (MPLA) encapsulated liposome–HA–PEG hybrid NPs could serve as a potential vaccine delivery platform with enhanced biocompatibility, stability, and controlled release for intranasal vaccination against infectious pathogens. Another example of liposome–HA hybrid NPs was reported by Li et al. 96 The HA–DOX encapsulated liposome was fabricated via two step electrostatic interactions. First, the hydrophobic core of DOX was formed with the presence of soybean oil, and then the HA‐based nanopolyplexes (HA‐NPs) were prepared by ion‐pairing between HA (negative charge) and DOX cores (positive charge). Second, the as‐synthesized HA‐NPs were further encapsulated in liposomal carriers to afford the sustained‐release of DOX by selectively targeting CD44‐positive tumor cells in vivo. Moreover, electrostatic interactions were also employed for targeted gene delivery. Liang et al. prepared a self‐assembled ternary complex consisting of pDNA, branched polyethylenimine (B‐PEI), and HA–epigallocatechin gallate (EGCG) for CD44‐targeted delivery of nucleotides. 97 This DDS was first stabilized by self‐assembly of pDNA and B‐PEI via electrostatic interactions, and the resulting positively charged pDNA/B‐PEI complexes were subsequently coated by HA–EGCG conjugates. These ternary complexes processed an efficient targeting cancer cell transfection due to the CD44‐targeting ability, and B‐PEI induced endosome escape and the strong nucleotide‐binding affinity of catechin. Recently, Zhong et al. formed endosomal pH‐activatable micelles via the self‐assembly of HA‐ b ‐dendritic oligoglycerol block copolymer (HA–dOG–PTX). 98 The azide‐terminated [G1. 0] dendritic oligoglycerol conjugate (N 3 –dOG) was first reacted with succinic anhydride to form N 3 –dOG–COOH, which was then converted to N 3 –dOG–vinyl ether by adding 2‐chloroethyl vinyl ether under a nitrogen atmosphere. PTX was grafted to N 3 –dOG–vinyl ether via an acetal linkage through the acid‐catalyzed reaction of the 2′‐hydroxyl group in PTX with vinyl ether terminates. HA–alkynyl was synthesized by the reductive amination of HA terminal aldehyde with propargylamine, and conjugated to N 3 –dOG–PTX through a click reaction. The resulting conjugate (HA–dOG–PTX) was then self‐assembled into prodrug micelles, which demonstrated higher payload, CD44 targetability and pH‐response capabilities than the free drug. The formation of bioreducible HA composites has also been reported by disulfide cross‐linking HA to other molecules. 99, 100 Han et al. described the fabrication of DOX‐loaded bioreducible HA nanoparticles (DOX–HA–ss‐NPs) with a redox‐responsive drug release profile and improved antitumor efficacy in the treatment of SCC7 tumor in a xenograft model. 100 First, alkyl‐terminated HA was prepared by reductive amination, as aforementioned, and 2‐(pyridyldithio)‐ethylamine (PDA) was conjugated with alkyne–HA through EDC‐mediated amidation. Second, another building block azide‐functionalized polycaprolactone (PCL–N 3 ) was synthesized via the ring‐opening polymerization of caprolactone, followed by tosylation of PCL–OH, and conversion into the azide group by a nucleophilic displacement reaction. 101 Finally, the PDA‐conjugated HA– b –PCL copolymer was obtained via Huisgen cycloaddition between PCL–N 3 and alkyne–HA–PDA. The resulting shell cross‐linked HA–ss‐NP was formed by the dithiothreitol (DTT) catalyzed cross‐linking of the PDA‐conjugated HA– b –PCL, and loaded with DOX through an emulsion method. Similarly, a bioreducible core‐cross‐linked HA micelle (CC‐HAM) for anticancer therapy was reported in another study of Han et al. 99 In this case, the building block of HA‐based core‐cross‐linked polymeric micelle was also prepared by Huisgen cycloaddition between alkyne–HA and azide terminated poly(pyridyl disulfide methacrylate) [P(PDSMA)]. The P(PDSMA) was synthesized by polymerization of the monomer 2‐(pyridine‐2‐yldisulfanyl) ethyl methacrylate with 2‐azidoethyl‐2‐bromo‐2‐methylpropanoate. The DOX‐loaded CC‐HAM was formed via the self‐assembly of amphiphilic HA– b –P(PDSMA) with the presence of DTT. In a related work, Zhong et al. described the redox‐sensitive HA– l ‐lysine methyl ester‐lipoic acid (HA–Lys–LA) conjugates for active targeting delivery of DOX to the drug resistant CD44 positive human breast tumor in vitro and in vivo. 102 The l ‐lysine methyl ester was first grafted to HA via EDC/NHS activation. The amino group of intermediate (HA–Lys) was then conjugated with the carboxylic group of lipoic acid via DCC/DMAP‐mediated amidation. The resulting cross‐linked NPs were obtained by self‐assembly of HA–Lys–LA conjugates with the presence of DOX and a catalytic amount of DTT. Among its applications, the broad spectrum of options for HA chemical modifications can be applied to achieve a specific targeting and long‐lasting delivery of various therapeutics, including protein, peptides, and small molecule drugs. 103 2. 4 Dextran‐Based CRS Dextran is a group of branched anhydroglucose polymer composed of alpha‐1‐6 glucose‐linked glucan with side chains alpha‐1‐3 linkages attached to the backbone units of dextran. They are produced from the fermentation of sucrose by certain lactic acid bacteria. Dextran holds several physicochemical advantages regarding its superb water solubility, surface resistance to nonspecific protein adsorption, 104 and ease of chemical derivatization, which make it suitable to be modified for therapeutic delivery. The abundant microbial enzyme dextranases, which could enzymatically digest dextran in colon tissue, making dextran suitable as potential nanovehicles for colon specific drug delivery. 105 Similar to other types of polysaccharides, the properties of dextrans are strongly dependent on their structure including molecular weight and degree of branching. According to the backbone structure of dextran, the hydroxyl groups of dextran units and the terminal aldehyde groups of dextran are two common reaction sites for chemical conjugation. Alternatively, two additional aldehyde groups can also be established from the periodate oxidation of dextran. 106 In one example, intracellular acidity‐sensitive dextran–DOX conjugates (Dex–O–DOX) were prepared by employing an oxime click reaction between the amino group in DOX and the terminal aldehyde group of dextran, 107 which afforded the pH‐triggered intercellular release of DOX via the breakdown of the Schiff‐base linkage. The in vitro and in vivo evaluation revealed that Dex–O–DOX increased antitumor activity and reduced toxicity compared with the reduction type Dex– b –DOX. Zhu et al. developed a lysosome‐targeted acidity‐responsive nanomicelles (Dex/Chol–PBA) through self‐assembling dextran and phenylboronic acid modified cholesterol. 108 First, phenylboronic acid was coupled with cholesterol by adding N ‐methylimidazole. 109 Then, Dex/Chol–PBA nanomicelles were prepared by dynamic self‐assembly between Dex and Chol–PBA via pH‐dependent phenylboronate linking. Through the in vivo/in vitro evaluation work, it was clearly confirmed that DOX‐loaded Dex/Chol–PBA nanomicelles exhibited an efficient cholesterol‐assisted cellular uptake, lysosome‐acidity induced drug liberation, and excellent safety profile. Recently, Cao et al. combined two types of stimuli‐sensitive dextran conjugated prodrugs for combinatory cancer therapy. 110 They first prepared a dextran propargyl carbonate (dex—C≡C) by activating propargyl alcohol with the presence of carbonyl diimidazole. 111 Later, the activated compounds were coupled with a dextran backbone through the formation of hydrolyzable carbonate esters. Subsequently, the dex—C≡C was conjugated with azide‐functionalized redox‐sensitive (disulfide bond) camptothecin derivative (CPT—ss—N 3 ) via Huisgen cycloaddition, which resulted in the formation of the prodrug Dex—ss—CPT. For another pH‐sensitive dextran–hydrazone–doxorubicin (Dex—hyd—DOX) prodrug, a similar synthetic approach was applied to couple the azide‐functionalized pH‐responsive (hydrazone bond) DOX derivative (DOX—hyd—N 3 ) with dex—C≡C. The preclinical evaluation of the combinatory therapy using Dex—ss—CPT and Dex—hyd—DOX micelles demonstrated significant anticancer activity by passively targeting tumor microenvironment and optimizing the synergistic effect molar ratio of DOX and CPT. Another interesting case for the redox‐responsive DOX carrier for triggered drug release using the GSH reducible dextran–Pt(IV) conjugate was reported by He et al. 112 They first prepared a carboxyl group functionalized platinum(IV) complex, 113 and then they synthesized the amphiphilic dextran–Pt(IV) conjugate based on the esterification between the carboxyl group of the Pt(IV) complex and the hydroxyl group of dextran. The DOX was further encapsulated into the hydrophobic center of dextran–Pt(IV) conjugate through self‐assembly. In cancer cells with presence of reductants including GSH and ascorbate, Pt(IV) moieties were reduced to the active Pt(II) form and cleaved from dextran side chains to induce the disruption of the conjugate structure, leading to the rapid liberation of dual drugs. 112 2. 5 Cyclodextrin (CD)‐Based CRS CDs are cyclic oligosaccharides constructed by 6 (α) or 7 (β) or 8 (γ) glucopyranose units through α‐1, 4‐glycosidic linkages and possess a cage‐like structure with a hydrophobic interior cavity and a hydrophilic exterior surface ( Figure 6 ). They are synthetic compounds obtained from the enzymatic hydrolysis of starch by Bacillus macerans. 114 The extraordinary trapping ability leads to a host–guest interaction between hydrophobic guest species and the interior cavity of CDs, given the modified physicochemical properties of the guest molecules in biological milieu. Since they are generally safe, inexpensive, water soluble, and easily functionalized, CDs have been intensively explored for therapeutic delivery. However, due to the relatively small cavity size of CDs, only limited number of molecules can be encapsulated for drug delivery; thus, the CD‐conjugated amphiphilic nanoformulation in the form of host–guest complexes has been increasingly developed in recent years. 115, 116 In addition, supramolecular hydrogels that utilize the interactions between a CD host and guest polymers to form inclusion complexes have attracted considerable attention to the tissue engineering field. 117 Figure 6 Schematic representations of molecular structure and geometric dimensions of α, β, and γ‐cyclodextrin. On the basis of construction for CD‐based assemblies, there are several exciting possibilities to design host/drug complex systems for therapeutic delivery ( Figure 7 ), including substrate–CD inclusion complexes (substrate/CD ratio: 1:1, 1:2, 2:1, and 2:2), amphiphilic CD conjugates, and CD‐based pseudo‐polyrotaxane (PPR). Recently, there is a great trend of employing synergistic interactions to constitute CD‐based self‐assemblies that result from the combination of various intermolecular forces, such as hydrophobic, electrostatic, covalent, and hydrogen binding. 118, 119 Besides, their formation and dissociation of CD‐based self‐assemblies are designed to be sensitive to biological milieu variations, for instance, pH value, temperature, redox, and enzyme. 116, 120 CDs are Food and Drug Administration approved cyclic macromolecules for application in food, cosmetics, and pharmaceuticals. By taking full advantage of these features, CDs can be utilized as molecular valves to control the conformational change of the supramolecular system for the release of therapeutic payloads. In this regard, we will highlight recent advances in the chemical modification and bioapplication of CD‐based CRS. Figure 7 Schematic illustration of three types of CD‐based assemblies. A) Substrate–CD inclusion complexes. B) Amphiphilic CD conjugates. C) CD‐based pseudo‐polyrotaxane (PPR). 2. 5. 1 Substrate/CD Inclusion Complexes Nobusawa et al. described a pH‐sensitive fullerene (C 60 )/6‐amino‐γ‐CD (ACD) inclusion complex for photodynamic therapy. 121 First, the protonatable primary amino moieties were grafted on the primary face of γ‐CDs through the reduction of intermediate azide modified γ‐CDs by employing triphenylphosphine (PPh 3 ) in DMF. 122 Subsequently, each C 60 was hydrophobically encapsulated in two γ‐CDs under neutral pH conditions. The in vitro evaluation of photodynamic therapy for HeLa cells demonstrated that the protonation of the amine groups of C 60 /ACDs at slightly acidic conditions led to the electrostatic repulsion of the wide rim, followed by the shrinkage of narrow rim, thus triggering the release and aggregation of C 60 surrounded by protonated ACDs. 121 Besides drug inclusion, CDs were also utilized as a key component for preparing stimuli‐induced supramolecular vesicles. Recently, Nayak and Gopidas designed and synthesized unusual supramolecular vesicles through the spontaneous self‐assembly of β‐CD/adamantane (AD)‐based bis‐inclusion complex (β‐CD∈AD–AD∋β‐CD). 123 Regarding the β‐CD∈AD–AD∋β‐CD system, the AD–AD molecule behaved as the amphiphilic bridge, which consisted of an ethylenedipyridine core and two adamantane moieties on both ends. The hydrophilicity is attributed to positively charged pyridinium residue, and adamantane moieties acted as hydrophobic head. In contrast, two β‐CDs served as the hydrophilic cap that accommodates the adamantane ends to form the bis‐inclusion complex. The AD–AD molecule was synthesized by the alkylation of amine at ethylenedipyridine with adamantyloxy ethyl bromide. 123 DOX loaded β‐CD/AD vesicles showed potential application for controlled release of drug by addition of a competitive inclusion binder such as adamantane carboxylate that simultaneously disrupted the vesicles. Similarly, Ma et al. developed a simple supramolecular self‐assembled binary vesicle based on tyrosine/β‐CD inclusion complexes. 124 Since the primary rim of the β‐CD molecule has a stronger hydrophilic properties than the carboxyl group of tyrosine exposed at the exterior part of secondary rim of β‐CD, the tyrosine/β‐CD amphiphiles self‐assembled into binary vesicles that were driven by hydrophobic–hydrophobic interaction. 124 The vesicles can respond to multiple exogenous stimuli, by disrupting the conformation of the vesicles via competitive guest molecules (1‐hydroxyadmantane) and copper ions. The 1‐hydroxyadmantane entered the cavity of β‐CD to replace tyrosine, or the copper ions coordinated with tyrosine molecules to form stable metal–organic complexes. 124 2. 5. 2 Multifunctional Amphiphilic CD Conjugates One of the prominent strategies in designing CD‐based therapeutic delivery systems is multifunctional conjugates that integrate stealth effects, active targeting, stimuli‐response, and imaging monitoring to provide greater therapeutic improvement. For example, the most remarkable CD‐based conjugate formulations for gene delivery is β‐cyclodextrin–polyethylene glycol copolymer (β‐CDP) polyplexes and its derivatives, which were developed by Davis and co‐workers. In their early study, pDNA encapsulated PEGylated β‐CDP polyplexes were designed and prepared through electrostatic interaction between positively charged β‐CDP polyplexes and negatively charged pDNA under physiological conditions 125 The β‐CDP was synthesized via the cross‐linking reaction of dimethyl suberimidate and amine‐functionalized β‐CD. Adamantane–PEG (AD–PEG) conjugates were self‐assembled with polyplexes to form Ada–PEG/β‐CDP inclusion complexes for prolonged circulation. In another study, they modified the AD–pep–PEG with galactose to obtain the active targeting ligand AD–pep–PEG–gal, which was decorated on pDNA loaded PEGylated β‐CDP polyplexes for selective binding to hepatocytes through overexpressed asialoglycoprotein receptors. 126 In 2007, they reported similar CDP polyplexes by replacing galactose with transferrin (AD–PEG–transferrin) for targeted delivery of siRNA to transferrin‐receptor‐upregulated HeLa cells 127 as well as to metastatic Ewing's sarcoma in a mouse model. 128 They further evaluated the safety profile with escalating intravenous doses of siRNA containing AD–PEG–transferrin polyplexes in nonhuman primates. 129 Ultimately, it entered a clinical trial (under the name CALAA‐01) for RNA interference (RNAi) in human tumors. 130, 131 Another prominent example of CDP nanoparticles is CRLX101, which was designed to address the poor drug solubility, insufficient chemical stability in physiological environments, and off‐target toxicity of CPT. 132 CPT is a topoisomerase I inhibitor with remarkable anticancer activity, which was interventionally linked to the repeating units of CD and PEG blocks via a glycine linker. Such arrangement could lead to the self‐assembly of the copolymer into 20–60 nm sized particles due to the host/drug interactions between adjacent CDP strands. 133 The resulting CRLX101 possessed neutral surface charge and with PEG blocks exposed to the outer layer. In addition, the CPT can be activated at target sites through the cleavage of the glycine linker that was medicated by both the base‐catalyzed and enzymatic hydrolysis of the ester group. 132 Both preclinical and clinical studies demonstrated improved solubility and an extended circulation time as well as reduced toxicity of CPT; these studies also exhibited enhanced therapeutic efficacy of CPT. 132, 133, 134, 135 Namgung et al. described self‐assembled polyplexes that were prepared by a multivalent inclusion complexation between a polymer–β‐CD conjugate (pCD) and a polymer–paclitaxel (pPTX) with active targeting and controlled release of PTX via in vivo enzyme‐degradation and the hydrolysis of ester linkages. 136 First, the β‐CD was grafted on poly(isobutylenealt‐MAnh) through esterification of maleic anhydride units with a single‐selectively deprotected hydroxyl group of β‐CDs. Second, the 2′‐hydroxyl group of PTX was preferentially reacted with anhydride groups of the poly(methyl vinyl ether‐alt‐MAnh) to form pPTX. 136 Next, the FCR‐675 fluorescent dye and targeting ligand, AP‐1 peptide were conjugated to pPTX by an amine–anhydride reaction and a PDA linkage, respectively. 136 The FCR‐675/AP‐1 grafted pPTX was then self‐assembled with pCD through a multivalent inclusion complexation, and the resulting polyplexes were found to have higher stability and solubility than that of the monovalent PTX‐β‐CDs. They also exhibited the stimuli‐responsive PTX release and potential tumor targeting through passive and active targeting mechanisms. Wajs et al. have recently reported stable redox or light responsive hollow nanocapsules based on ferrocene/β‐CD or azobenzene/α‐CD‐decorated dextran polymers. 137 Both kinds of nanocapsules were prepared through layer‐by‐layer self‐assembly of host/guest polymers that deposited on the surface of Au colloid templates. The Au nanoparticles were initially coated by thiolated β‐CD or α‐CD dextran polymers (host), followed by the deposition of ferrocene or azobenzene functionalized dextran polymers (guest) on the outer layer via host/guest interaction. Finally, they removed the oxidative core to obtain the hollow nanocapsules. 137 In this report, the authors demonstrated that Rhodamine B can be encapsulated and released via a reversible one‐electron redox process (ferrocene‐based nanocapsules) and UV‐light irradiation (azobenzene‐based nanocapsules) by the altering wall permeability of the inclusion complex. 137 2. 5. 3 CD‐Based PPR CD‐based PPRs are noncovalently interlocked supermolecular architectures that are comprised of linear polymer components (guests) and encircled by CD components (hosts), and they are advancing rapidly in the area of stimuli‐responsive materials due to their unique features. 138 Using similar mechanisms, Dandekar et al. developed a cationic α‐/β‐CD‐based polyrotaxane, which can condense nucleic acids into nanoplexes for in vitro gene delivery. 139 The CD polyrotaxane was obtained by subsequent incubation of amine functionalized β‐CD and α‐CD with ionene‐6, 10 polymer, as the CD rings were threaded over the polymer chain with temperature activated noncovalent interactions. 139, 140 The nanoplexes were then formulated with pDNA and siRNA via electrostatic interaction. The cellular investigations demonstrated that their nanoplexes could successfully overcome the endosome degradation with low cytotoxicity for intracellular gene delivery. In one study, we developed self‐healing, thermoresponsive host–guest inclusion complexes (i. e. , Pluronic F108 incorporated alginate‐ graft ‐β‐cyclodextrin) for cell transplantation and drug delivery. 71 To synthesize alginate‐ graft ‐β‐cyclodextrin, p‐toluenesulfonyl (TosCl) chloride was first reacted with β‐CD to yield β‐CD–TosCl. Then 1, 6‐hexanediamine (HDA) and ethylenediamine (EDA) were reacted with β‐CD–TosCl to obtain β‐CD–HDA and β‐CD–EDA, followed by amide bond formation with alginate via carbodiimide chemistry. Finally, the resulting product alginate‐ graft ‐β‐CD was self‐assembled with the difunctional guest molecule Pluronic F108 through a host/guest interaction. This because the hydrophobic moieties of Pluronic F108 is held within the cavity of β‐CD. Based on these unique intermolecular interactions, such supramolecular inclusion complex exhibits shear‐thinning properties and affords excellent thermal‐responsive behavior to the injectable hydrogel. Such shear‐thinning hydrogel flows similar to low‐viscosity fluids under shear stress during injection. However, as soon as the fluid comes out the needle, hydrogel recovers by itself without additional trigger factors, such as UV light. Shear‐thinning hydrogels have been extensity studied in various disease model and even 3D printing polysaccharides. 26, 141 Recently, Badwaik et al. reported three cholesterol terminated Pluronic (F‐127, L‐35, and L‐81) cationic polyrotaxanes (PR + ) threaded with N, N ‐dimethylaminoethylamine (DMEDA)‐functionalized 2‐hydroxypropyl (HP)–β‐CD for siRNA delivery. 142 DMEDA was conjugated to HP–β‐CD via a carbonyldiimidazole‐mediated coupling reaction. 143, 144 The HP–β‐CD units were first threaded onto the Pluronic copolymer backbone, followed by introducing tris(2‐aminoethyl)amine at both ends, and finally end‐capping the branched diamine termini with cholesteryl chloroformate. 142, 143 The resulting PR + :siRNA formulation was obtained through electrostatic interactions between the PR + and siRNA payload, which exhibited higher performance than Lipofectamine 2000, while maintaining low cytotoxicity and high in vitro stability. 143 Tamura et al. developed a novel acid‐responsive β‐CD‐based polyrotaxanes for the treatment of Niemann–Pick type C (NPC) disease. 145 NPC disease is a rare inherited lysosomal storage disorder with mutations in NPC1 and NPC2 genes. 146 The key feature of the disease mechanism is the accumulation of cholesterol within lysosomes, and it has been found that intracellular cholesterol can be effectively dissolved away through inclusion complexation with HP–β‐CD. However, excessive HP–β‐CD can induce various acute toxicities in animal models; 147 thus, to overcome the toxic issue, they designed and synthesized a pH‐sensitive polyrotaxanes system comprised of three different components: Pluronic P123 polymer, threading 2‐(2‐hydroxyethoxy)ethyl (HEE)‐functionalized β‐CDs and terminal N ‐triphenylmethyl (N‐Trt) blocks. 145 The in vitro evaluations demonstrated that the acid‐responsive β‐CD‐based polyrotaxanes can be internalized into cells through endocytosis and spontaneously dissociated the HEE–β‐CDs under acid environments. 145, 148 When polyrotaxane occupies the β‐CD cavity by the polyrotaxane structure, it will not only mask the cytotoxicity by preventing the extraction of cholesterol in membranes, but also provide improved therapeutic efficacy by three orders of magnitude over HP–β‐CD. 145 3 Preparation Approaches The advanced understanding of material chemistry and engineering techniques facilitates multiple strategies to fabricate polysaccharide‐based DDSs. In this section, we discuss the chemistry basics associated with different cross‐linking forces within polysaccharide systems and the engineering techniques used to fabricate polysaccharide‐based DDSs. 3. 1 Intra‐ and Intermolecular Forces in Polysaccharide Systems 3. 1. 1 Covalent Cross‐Linking To maintain the network of polysaccharide NPs that avoid dissolution of the hydrophilic polymer chains/segments into the aqueous phase, chemical cross‐linking is usually performed while maintaining the biodegradability of the materials ( Figure 8 A). In chemically cross‐linked NPs and gels, covalent bonds are established between functional groups of polymeric chains or are mediated by covalent cross‐linking molecules with at least two active moieties. 9 The chemical linkages in the matrix structure are usually designed either to be biodegradable or stimuli‐responsive under specific endogenous and exogenous conditions. 20, 21, 149 Although the covalent cross‐linkages are the major driving force, other noncovalent forces (e. g. , hydrogen bonding and hydrophobic interactions) could also be involved, depending on the types of polysaccharides and chemical modifications employed. In general, labile bonds including peptide bonds (carbodiimide‐mediated reactions), ester bonds (anhydride‐mediated esterification), and disulfide bonds (oxidation of the thiol groups) commonly facilitate the intramolecular cross‐linking of the polysaccharide network. 21, 150 In the previous paragraph, we discussed the methacylation reaction and its function in photocross‐linking reactions, which is an interesting example of covalent bonds being applied to design polysaccharide NPs. Figure 8 Schematic illustration of different intra‐ and intermolecular forces in polysaccharide systems. A) Covalent cross‐linking. B) Metal–polymer coordination. C) Electrostatic interactions. D) Hydrophobic interactions. 3. 1. 2 Metal–Polymer Coordination In contrast to covalent cross‐linking, metal–polymer coordination forms stronger bridges between polysaccharide chains through coordinate–covalent bonds (chelation) between metal cations (e. g. , calcium, copper, iron, zinc) and negatively charged ligand moieties of polysaccharides (Figure 8 B). 9, 151 This intramolecular force enables the reversible and facile formation of metal–polysaccharide nanocomposites, 152 such as hydrogels with variable physicochemical properties that depend upon the size and the valence of anionic metals, as well as degree of chemical modification and concentration of the polysaccharide. 152, 153, 154 In addition, metal–polymer coordinates are generally pH sensitive, which is favorable for controlled drug release, although this may also cause instability of the cross‐linked network. 65 To date, alginate is a well‐known example of polysaccharide that can be cross‐linked by metal–coordinate interactions by exchanging sodium ions from the guluronic units with divalent cations, mainly the Ca 2+ ions. 155 These calcium ions are coordinated to the hydroxyl and carboxyl groups of four α‐ l ‐guluronic acid units from two adjacent chains of the alginates, 156 and as a result, the hydrogel network with a so‐called “egg‐box” structure is formed. 157 The alginate gel beads can be prepared at room temperature and physiological pH; thus, they are widely used for the immobilization of living cells and the controlled release of a variety of proteins. 158 3. 1. 3 Electrostatic Interactions In addition to anionic polysaccharide being coordinately cross‐linked with metallic ions, polyelectrolyte complexes (PECs) can also be obtained by electrostatic interactions between oppositely charged polysaccharide and polyelectrolytes in solution (Figure 8 C) 154, 159 PECs provide a reversible and noncovalent physical linkage without using any reactive agents and catalysts for the immobilization of therapeutic payloads. PECs are any positively or negatively charged macromolecules like nucleic acids (e. g. , pDNA, siRNA), proteins (e. g. , albumin, collagen, gelatin), polysaccharides (e. g. , chitosan, hyaluronic acid, alginate), and synthetic polycation and polyanion polymers (e. g. , polyethylenimine, polyacrylic acid). 153, 160 The complexation, stability, and physical properties (e. g. , permeability, swelling) of PECs are determined by several factors, including the intrinsic properties of PECs (e. g. , ionic strength, charge density, molecular weight, flexibility) and physicochemical environment (e. g. , temperature and pH of the solution, type of solvent, degree of interaction between PECs and polysaccharides) as well as the order and duration of mixing PECs. 21, 149, 152, 160, 161 Among the existing polysaccharides, chitosan is the most commonly applied cationic polysaccharide to form PECs due to its biocompatible and water‐soluble features, 20, 149 whereas hyaluronic acid, 162 dextran sulfate, 163 alginate, 164 nucleic acids, 165 and some aspartic acid and glutamic acid‐rich peptides/proteins are used as anionic polyelectrolytes. 166, 167 In addition, anionic polysaccharides can also form PECs with positively charged peptides/proteins, such as polylysine, which is a positively charged peptide that electrostatically combines with alginate to form PEC nanoparticles. 166 3. 1. 4 Hydrophobic Interactions Upon introducing hydrophobic segments onto the hydrophilic polysaccharide chains, amphiphilic copolymers are produced. These copolymers tend to self‐assemble into stable conformations to minimize the free energy by spontaneous formation of hydrogen bonding between the hydrophilic backbone of the polysaccharide and water molecules. Hydrophobic blocks undergo self‐association to form a hydrophobic domain due to the unfavorable interaction with water (Figure 8 A). 168 The amino, carboxyl, and hydroxyl groups present on polysaccharide backbone are the most utilized functional pendant groups to conjugate a wide range of hydrophobic segments, such as cholesterols, fatty acids, bile acids, polyesters, pluronic polymers, poly(alkyl cyanoacrylate), and hydrophobic drugs. 20, 21 Various self‐aggregates that are based on hydrophobized polysaccharides (e. g. , hydrogel nanoparticles, micelles, polymersomes, oil in water (O/W) emulsions) can be formed for the controlled delivery of hydrophobic compounds via stimuli responses, such as pH, temperature, and enzyme‐degradation. 168, 169 For clinical translation of controlled drug delivery formulations, a number of parameters including size, solubility, loading capacity, surface charge, physiological stability, and drug release kinetics need to considered, which can be achieved by adjusting the functional group, molecular weight, and concentrations of the hydrophobic block and polysaccharides. 154, 170 Among those amphiphilic copolymers, amphiphilic CDs have gained significance in pharmaceutical formulations to encapsulate hydrophobic drug molecules through their hydrophobic cavity. Recent progress in the development of CD‐based complexation system has inspired the way of supramolecular self‐assembly for drug delivery, and examples have been discussed in the previous section of this review. 3. 2 Fabrication Methods and Techniques 3. 2. 1 Emulsification Method Emulsification is one of the most evolved methods for the preparation of polymeric nanoparticles for research and pharmacotherapy applications. The success of emulsion comes from several attributes, such as optical clarity, ease of preparation, thermodynamic stability, and increased surface area. Phase behavior studies have shown that the size of the droplets is determined by the surfactant phase structure (bicontinuous microemulsion or lamellar) at the inversion point that is induced by either material composition or temperature. The preparation of an emulsified system is generated by mixing two or more immiscible liquids and using mechanical processes, such as stirring or ultrasonication. Generally, depending on the type of liquid that is used for the dispersed and continuous phase, O/W or water in oil (W/O) emulsions can be formed, and multiple emulsions (e. g. , W/O/W and O/W/O) can also be achieved to enhance the efficacy of formation of emulsion droplets and to encapsulate drugs with different solubility in different phases. According to the size of the droplets, the emulsions are classified into three main types: a microemulsion is primarily referred to as a thermodynamically stable droplet with size ranging from 10 to 100 nm; a nanoemulsion is characterized by a thermodynamically unstable but kinetically stable feature with droplet sizes mostly between 20 and 500 nm; and a macroemulsion represents a classical emulsion system that often exhibits thermodynamically unstable and weakly kinetically stable behavior with droplet size greater than 1 µm. 171 Since polysaccharides are usually water soluble, W/O is mostly applicable for the fabrication of polysaccharide‐based nanoparticles. The emulsion‐cross‐linking method was initially applied to the preparation of chitosan nanoparticles for 5‐fluorouracil delivery. In this process, a chitosan aqueous solution was emulsified in toluene, followed by cross‐linking with glutaraldehyde to harden the droplets. 172 The principle of cross‐linking was based on a Schiff‐base reaction between the aldehydic group of the glutaraldehyde and the primary amines of chitosan, 173 which formed the inter‐ and intramolecular covalent network to firm up the structure of the chitosan particle. However, there are concerns over the toxicity of the glutaraldehyde used, which compromised the biocompatibility of chitosan‐based emulsions. Therefore, efforts had been made to ameliorate the cross‐linking method. One solution is to replace the glutaraldehyde with biocompatible cross‐linking agents such as glyceraldehyde and genipin. 174, 175 Recently, Song et al. prepared PEG‐modified ultrasmall chitosan nanoparticles as indocyanine green (ICG) carriers with the average size around 5 nm for tumor photothermal therapy in vivo. 175 An aqueous dispersion of chitosan was added into the microemulsion system consisting of cyclohexane, 1‐octanol, and Triton X‐100, and the mixture was stabilized using ultrasound. The microemulsions were hardened by genipin cross‐linking. 175 PEG‐modified chitosan–genipin nanoparticles were prepared via the conjugation of succinimidyl carboxymethyl ester (SCM–PEG) on the surface of the nanoparticles, and ICG molecules were subsequently loaded into the nanoparticles using electrostatic interactions. 175 When irradiated with a NIR laser, cells incubated with CG–PEG–ICG nanoparticles showed cell viability around 15%. The in vivo bioavailability and efficacy of the photothermal therapy effect on the treatment of U87 xenograft tumors by intravenous and intramuscular injection was evaluated, respectively, and the results demonstrated that CG–PEG–ICG nanoparticles exhibited prolonged retention time of ICG in the mice body as well as low toxicity with effective tumor phototherapy (tumor injected with CG–PEG–ICG nanoparticles containing ICG more than 100 µg mL −1 (100 µL)). 175 In addition to covalent approaches, the emulsion‐ionic cross‐linking interaction has also been applied to prepare chitosan microspheres. For instance, Zou et al. reported that sodium TPP, a biocompatible polyanion, was introduced to prepare cross‐linking chitosan microparticles (5–10 µm) for pH‐responsive release of bovine serum albumin (BSA). 176 The controlled release of BSA was mediated by diffusion via the swelling behavior of chitosan microspheres, which exhibited a higher swelling ratio and was more promising than glutaraldehyde cross‐linked microspheres. 176 Machado et al. described the preparation of W/O type nanoemulsions of aqueous alginate solutions through the phase inversion temperature emulsification method. 177 In this experiment, they employed nonionic ethylene oxide oligomers (C 12 E 4 ) as a temperature dependent surfactant, which exhibits increased hydrophobicity with rising temperatures. The structure of emulsions could change from O/W to W/O via temperature control. Ionic cross‐linking of the alginate was performed by introducing aqueous CaCl 2 to the emulsions under stirring, and prepared nanoparticles were collected through addition of excess oil. This method allows the preparation of finely dispersed calcium alginate nanoparticles in the sub‐200 nm range without a large input of mechanical process. 177 Recently, we have reported utilizing 5% Span 80 in mineral oil as the oil phase with addition of tween 80 as the surfactant and 1% alginate solution to form alginate microparticles cross‐linked by CaCl 2. The reaction was easily performed on bench top at room temperature, simplifying the previously stated method, but achieving evenly distributed nanoparticles ( Figure 9 ). In addition to the normal alginate solution, we also used a PEGylated alginate for multifunctional microparticles with a similar method. The polymer (alginate or PEGylated alginate)/drug solution was slowly added to biological‐grade mineral oil containing surfactants. CaCl 2 was added to the system while stirring to cross‐link alginate to from stable microparticles. After the reaction, the particles were washed several times to remove the mineral oil. The obtained particles were spherical in shape with an average diameter of 1–5 µm and can be lyophilized and stored for long‐term application. 22 Figure 9 Schematic representation of fabrication of micropsheres via water/oil emulsion. Reproduced with permission. 22 Copyright 2014, Elsevier. 3. 2. 2 Desolvation (Coacervation or Precipitation) Method The desolvation method is a facile synthetic approach that often involves coacervating or precipitating a polysaccharide matrix in an aqueous solution and forming polymeric micro‐/nanoparticles by addition of desolvating agents, such as salts or alcohols. This process is induced by the competitive binding of desolvating agents to water molecules in a previously formed polysaccharide solution. The surrounding water molecules are consequently dissociated from the polysaccharide micro‐/nanoparticles due to the higher affinity between water and the desolvating agents. 178, 179 A cross‐linking agent is commonly used for further stabilization and adapted for the controlled release of the therapeutic payload. One significant benefit of the desolvation approach is that usually no heated reaction condition is required, as some encapsulated drug molecules or bioactive agents are thermodynamically unstable. 180 However, the stabilization of the resulting micro‐/nanoparticles should be carefully controlled, since the cross‐linking reaction can lead to high polydispersity. 181 The utilization of cross‐linking agents (PEG–dialdehyde) for stabilization of the particle carrier was initially reported by Berthold et al. in 1996. 182 Since then, such procedure is widely used in preparation of polysaccharide‐based drug carriers, especially the chitosan micro‐/nanoparticles. For example, Mao et al. developed chitosan‐based nanocarriers (ranging from 100 to 250 nm) for in vitro and in vivo gene delivery. 183 In this approach, the chitosan–DNA complex was formed via electrostatic interaction, and sodium sulfate was facilitated as a desolvating reagent to separate nanoparticles from the solution. Glutaraldehyde was introduced for stabilizing the chitosan–DNA nanoparticles without damage of DNA. 183 The resulting nanoparticles were further conjugated by PEG and transferrin to reduce the aggregation and enhance the transfection efficiency, respectively. Agnihotri and Aminabhavi synthesized timolol maleate‐encapsulated chitosan nanoparticles for ophthalmic delivery. 184 The chitosan nanoparticles were formed via desolvation with the dropwise addition of acetone in the aqueous acetic acid solution containing the mixture of chitosan and timolol maleate, followed by cross‐linking with glutaraldehyde. 184 The resulting nanoparticles had sizes ranging from 118 to 203 nm, and the drug release rate was dependent on the level of cross‐linking and the molecular weight of chitosan. Al‐Ghananeem et al. prepared hyaluronan nanoparticles for intratumoral delivery of paclitaxel. 185 Nanoparticles were obtained from the desolvation of HA in a Tween 20 aqueous solution using sodium sulfate as desolvating agents and cross‐linked with glutaraldehyde after paclitaxel loaded into HA coacervates. Although the desolvation method simplified the purification process, the introduction of toxic glutaraldehyde would potentially impede the in vivo application if the purification of the product did not meet regulatory requirements. Moreover, the experimental optimization is always required, since various parameters such as initial molecular weight and concentration of polysaccharide, amount of desolvating agent, agitation speed, as well as molar ratio of polysaccharide/therapeutic payload can greatly influence the resulting characteristics of the nanoparticles. 3. 2. 3 Polyelectrolyte Complexation and Ionotropic Gelation The use of electrostatic interactions and metal–polymer coordination between polysaccharides and counterions or polyelectrolytes has drawn considerable attention. This facile and mild approach offers several unique advantages, including a nontoxic process, reversible cross‐linking, an organic solvent‐free process, and easy scaling. The materials applied in the fabrication of polyelectrolyte nanocomplexes can be divided into two main categories: (1) Small counterions or molecules, such as divalent chloride salts (e. g. , CaCl 2, MgCl 2, CuCl 2 ), pyrophosphate, citrate, sulfate; and (2) Oppositely charged macromolecules, including polyphosphates, polylysine, alkyl sulfates, polyglutamic acid, and polysaccharides. One of the classic early studies of polyelectrolyte complexation was reported by Calvo et al. 186 In this method, various amounts of chitosan and BSA were dissolved in aqueous solutions that contained acetic acid. Then, sodium TPP in water was subsequently mixed with chitosan solution under agitation, spontaneously producing chitosan nanoparticles. The TPP/chitosan mole ratio, stirring rate, and the degree of deacetylation of chitosan can crucially influence the particle size and surface charge. In addition, the nanoparticle size can also be affected by the molecular weight of oppositely charged cross‐linking agents, that is, employing small counterions or molecules results in smaller particle sizes than using oppositely charged macromolecules. 178 Recently, polyion nanocomplexes based on the layer‐by‐layer deposition of sodium alginate and chitosan has been applied for improving the lipid membrane stability of nanoliposomes in the gastrointestinal tract. 187 This study used different concentrations of chitosan and sodium alginate in aqueous solutions with pH adjusted to 5. 5. The first layer was formed by addition of negatively charged nanoliposomes into chitosan solution under constant stirring for 1 h, followed by adding chitosan coated nanoliposomes into sodium alginate solution via same procedure, and resulting in the formation of alginate–chitosan coated nanoliposomes. Interactions between the ternary polysaccharide systems have been applied to develop injectable nanonetworks for controlled insulin delivery. For example, Gu et al. developed a glucose‐responsive nanoparticle‐based polymeric network 188 that was composed of four components, including an acid‐degradable acetal‐incorporated m‐dextran, chitosan‐ and alginate‐based surface coatings, and bioactive encapsulations (i. e. , glucose oxidase, GOx; catalase, CAT; and human recombinant insulin). The preparation of the nanoparticle‐based nanonetwork started by the formation of m‐dextran nanoparticles via a double emulsion (water‐in‐oil‐in‐water)‐based solvent evaporation/extraction method. 188 A certain amount of m‐dextran in dichloromethane (DCM) was emulsified with an aqueous mixture of insulin, GOx, and CAT in specific ratios by sonication. The obtained primary emulsion was added into the chitosan and alginate aqueous solution with sustained sonication separately. The double emulsion was then transferred into chitosan and alginate aqueous solution and eliminated the DCM through agitation, followed by centrifugation. The nanonetwork was then prepared through polyelectrolyte complexation by mixing the aqueous solution of chitosan‐ and alginate‐coated nanoparticles together under constant stirring and was collected by centrifugation. 188 In the ionotropic gelation technique, polysaccharide‐based polyelectrolytes can be used, such as the widely investigated alginate and chitosan, which can chelate with counterions to induce the gelation and form a particulate or meshwork structure. Alginate is one of the most well‐known examples and has been extensively reported. In the case of the formation of calcium alginate hydrogels, three general approaches can be used. One is the diffusion or external gelation method, where alginate solution is added dropwise into a bath of calcium chloride solution. The hydrogel matrix is formed through the diffusion of the calcium cations from the external continuous phase into the interior structure of alginate droplets. 40, 65, 189, 190 The second method is the in situ gelation or so‐called internal gelation. In this approach, the insoluble calcium source (e. g. , calcium salt) is mixed with an alginate solution, and the release of the calcium ions is triggered by altering the pH of the system or by increasing the solubility of calcium source, which subsequently leads to the formation of the Ca–alginate gel. 65 The third method is the hot‐made preparation through the controlled cooling from high‐temperature hydration of a medium that consists of alginate, salt, and a sequestrant. 191 Comparing these methods, the diffusion method is a rapid and high yield gelation process that produces an inhomogeneous Ca–alginate gel, in which the concentration of Ca–alginate gelation is dependent on the thickness of the gel. 192 While in situ gelation provides a homogeneous ionotropic gel with a uniform distribution of calcium ions, 193 the hot‐made preparation of the Ca–alginate gel is plainly limited to the incompatible use of heat‐labile substances. Alginate ionotropic gels prepared by different methods can exhibit distinct properties (e. g. , stiffness, strength, permeability, pore size). Externally cross‐linked alginate matrix usually possesses greater matrix strength than internally alginate cross‐linked matrix, despite matrix strength can be balanced between two types of alginate matrix by adjusting the amount of cross‐linker used. Matrix flexibility can also be altered by controlling the amount and size of CaCO 3 used in internal gelation method, but little impact on the strength of matrix. Both approaches are potentially applicable as a coating or delivery system. 189 High molecular weights of alginate and the presence of nongelling ions can improve the uniformity of the Ca–alginate gel created with the diffusion method. 193 3. 2. 4 Self‐Assembly Method Self‐assembly is a method that involves the self‐ruling organization of polysaccharide compounds into nanostructures without human interference. The joint use of self‐assembly and other methods is commonly applied for the preparation of novel supramolecular assemblies in drug delivery applications. CDs are the most widely used cyclic oligosaccharides in the drug delivery field to enhance the solubility, stability, and bioavailability of drugs. As previously mentioned, there are mainly three types of inclusion‐complex formations between substrate (drug) and CD (host), and several techniques have been used to prepare CD‐based inclusion complexes, such as the coprecipitation technique, the kneading technique, the neutralization precipitation technique, the coevaporation technique, and the microwave irradiation technique. 119, 194 In the coprecipitation technique, CD is initially dissolved in an aqueous solution, and the substrate is introduced when stirring the CD solution. The solubility of CD can be increased up to 20% with elevated temperature if the substrate molecule is thermally unstable at higher temperatures. The precipitate of inclusion complexes is formed during the continuous cooling and agitation, which is then collected by centrifugation or filtration, and may be washed with a water‐miscible solvent. 195 However, this technique is limited in its scaling‐up production ability due to the large amount of water that is required for poor solubility of CDs, as well as the massive amount of energy used for heating and cooling. Besides, some organic additives can influence the complexation efficiency of the substrate (drugs), 194, 195 which is needed to take into account a particular case. The kneading technique is one of the widely used methods for inclusion complexation. 119 In the course of its preparation, the CD is mixed with a specific amount of water or hydroalcoholic solutions to form a paste. The substrate is subsequently added to the paste and homogenized for a certain amount of time, which is then dried by vacuum desiccators. 194 The kneading method was successfully utilized for encapsulation of various drugs in both small‐ and large‐scale production, including azomethine, sulfamethoxazole, linalool, and difluorinatedcurcumin. 196 Neutralization precipitation is a technique used for the precipitation of ionizable inclusion complexes, which are prepared by dissolving the substrate in an alkaline solution and mixing with an aqueous CD solution. The pH of the resultant mixture is neutralized by adding a hydrochloric acid solution while stirring; then, the precipitate is formed and collected by filtration, followed by desiccation. 197 However, this method is limited to encapsulate acid‐ and alkaline‐labile substrates. 194 The coevaporation technique is a simple and economic method that involves the mixing of two different miscible solutions (for instance, an aqueous CD solution and an alcoholic solution of a substrate) to form an emulsion of inclusion complexes. Then, the solvent is evaporated and dried under vacuum to obtain the pulverized product. 197 Microwave irradiation is an effective and convenient technique for the rapid complexation of CD and a substrate. In this process, the CD and substrate are dissolved in a solvent and reacted for a short period of time using a microwave oven. When the reaction is completed, the free substrate, cyclodextrin, and residual are removed by a solvent mixture, and the resultant precipitate is dried in a vacuum oven. 197, 198 3. 2. 5 Microfluidic Methods To ensure that the sizes of the nanoparticles are evenly distributed, a homogenizer is often used in the emulsion process to reduce the sizes of the droplets in liquid–liquid dispersions, generating stable homogenized particles. However, the inherent random process makes it a nonideal strategy to fabricate polysaccharide nanoparticles in industry. Microfluidics has shown unparalleled advantages for the synthesis of polymer particles and have been utilized to produce hydrogel particles with a well‐defined size, shape, and morphology. Most importantly, during the encapsulation process, microfluidics can control the number of cells per particle and the overall encapsulation efficiency. Therefore, microfluidics is becoming a powerful approach for cell microencapsulation and the construction of cell‐based drug delivery systems. 199 An example of Ca 2+ ‐cross‐linked alginate microspheres were generated from a microfluidic device by Chen et al. They reported a versatile method of droplet microfluidics to fabricate alginate microspheres while simultaneously immobilizing an anti‐ Mycobacterium tuberculosis complex Immunoglobulin Y (IgY) and anti‐ Escherichia coli IgG antibodies primarily on the porous alginate carriers for specific binding and binding affinity tests. 200 They actually presented the shape and surface structure of calcium‐cross‐linked alginate microspheres under microscopy. They were generally round with an undulating membrane. Tiny porous structures were shown in zoomed in pictures of microsphere surfaces. Microfluidic devices utilize the science of manipulating and controlling fluids and particles at micrometer or sub‐micrometer dimensions to exploit a wide range of biological applications such as high‐throughput drug screening of single cell or molecular analysis and manipulation, drug delivery and advanced therapeutics, biosensing, and point of care diagnostics, among others. 201 Fluid flow in microchannels is diffusion‐based laminar flow due to the low Reynolds numbers. 202 Several materials have been casted to make microfluidic devices, including polymer (including polydimethylsiloxane, polymethylmethacrylate, polycarbonate, cyclic olefin copolymer), 203 silicon, 204 and metal. 205 Typically, syringe pumps or microfabricated pumps provide pressure‐driven flow in the microchannels, and electrokinetic devices provide other choices for pumping liquids. Reagent solutions are manipulated inside microfluidic devices. A T‐junction type of channels is usually designed to generate droplets alternatively and fuse tow reagent droplets in a tapered chamber. In the long switch back channel, particles with nano‐ or microsizes can then be synthesized in each droplet reactor and collected at the end of device. 206 In our group, we have designed a microfluidic‐flow‐focusing device which is consistently reproducible, readily characterized, and easy to test and use to produce homogeneous alginate microparticles ( Figure 10 ). Microparticles with the same size were pumped out of the T‐junction and then collected at 1 m CaCl 2 solution. High speed camera recording helped to identify the process of formation of a single droplet in the microfluidic devices. 207 Microfluidic devices allow researchers to control the physical conditions and behavior of fluids in a micro‐/nanoscaled domain to fabricate polysaccharide biomaterials, offering versatile solutions for fabrication, manufacturing, and research in the fields of cell biology, pharmacology, and tissue engineering. We believe the continued enhancements of technology of microfluidic devices will produce much smaller and uniform polysaccharide nanoparticles while maintaining portable and cheap solutions for large‐scale industrial manufacturing applications. Figure 10 A) Photograph of microfluidic device. B, C) Microscopic photographs of alginate/oil droplets pumped out T‐junction inside the microfluidic device. Reproduced with permission. 207 Copyright 2015. 4 Preclinical Advancements The goal of biomaterials is to assist the body's self‐healing process with the engagement of different cells/tissues as well as drug molecules. Drug delivery systems are tailor‐designed to promote the therapeutic efficacy of existing drug molecules in controlled manner. Our discussion focuses on two major categories of biomedical applications: (1) tissue engineering with regenerative medicine and (2) targeted delivery and theranostic applications in the field of treatment of diseases. 4. 1 Tissue Engineering and Regenerative Medicine Polysaccharides are able to form hydrogels and micro‐/nanoparticles after certain reactions, which can encapsulate drug for therapeutic application. Tissue engineering is an emerging biomedical field that aims to assist and enhance the regeneration of body tissue defects that are too large to self‐repair or to substitute for the biological functions of damages/injured organs. To promote tissue regeneration or wound healing, many protein growth factors are required. For example, some growth factors are able to induce angiogenesis, which then supplies oxygen and nutrients to cells transplanted for organ substitution to maintain their biological functions. 208 Some growth factors are also shown to stimulate the proliferation and differentiation activity of stem cells via certain cellular signal pathways. 209 However, the biological effects of growth factors cannot always be expected because of their poor in vivo stability, unless a drug delivery system is contrived. 208 Various growth factors have shown to affect the proliferation and survival of multipotential stromal cells, including transforming growth factor beta (TGF‐β), the fibroblast growth factor (FGF), the VEGF, the platelet‐derived growth factor, the epidermal growth factor, the hepatocyte growth factor, and the Wnt family. 210 Almubarak et al. summarized the role of commonly used growth factors in angiogenesis and osteogenesis and highlighted the current status of preclinical and clinical trials. 211 Our group has reported using alginate microparticles to stealth‐deliver VEGF intracellularly to mesenchymal stem cells (hMSCs), inducing osteogenesis differentiation of hMSCs. The alginate microparticle prevented the delivered VEGF to interact with the VEGF surface receptor (VEGFR), which could potentially direct hMSCs into the osteoblast linage rather than adipocytes. 22 Liu et al. showed that the stealth delivery of VEGF effectively contributed in the differentiation signal pathway; they found that the intracellular expression of VEGFA but not external application of the growth factor could cure osteoporosis. 212 As expected, hMSCs could endocytose VEGFA–microparticles within 48 h coculturing and differentiate into osteoblast after 14 d ( Figure 11 ). The utilization of alginate microparticles provides a possible solution to activate the intracrine mechanism, which may be different from the paracrine mechanism with respect to directing cell fates. Table 1 lists examples of using different polysaccharides to control delivery of certain bioactive agents for various applications. Figure 11 A) Protein results of human MSCs after 14 d in culture with differentiation growth medium. Sample aliquots were collected and osteoprotegerin concentration (pg mL −1 ) in the supernatant medium was measured using an enzyme‐linked immunosrobent assay (ELISA) assay ( n = 3). * indicates significantly higher result, p < 0. 0001. B) Reverse transcription polymerase chain reaction (RT‐PCR) results verify production of osteocalcin and RUNX2 in MSCs after 14 d in culture with osteogenic differentiation growth medium. Collagen type I was upregulated for MSCs pretreated with VEGFA‐encapsulated Alg– g –PEG microspheres prior to differentiation. Glyceraldehyde‐3‐Phosphate Dehydrogenase (GADPH) was used as an internal control. C) Schematic representation of hMSCs endocytosis alginate‐based microparticles (loading with VEGFA), inducing osteogenesis differentiation. Reproduced with permission. 22 Copyright 2014, Elsevier. Table 1 Examples of different polysaccharides cargo system to control delivery certain bioactive agents for various applications Polysaccharide type Bioactive agents Application Role in tissue engineering Ref. Alginate Amidated pectin hosting doxycycline (Antibiotics) Wound healing Inhibit bacterial‐infection‐caused necrosis in wound healing process 256 Alginate VEGF Osteoporosis Promote osteogenesis differentiation of hMSCs rather than adipogenesis 22 Alginate Human fibroblast growth factor 1 (FGF‐1) Cartilage defects Promote the in vitro development of mature adipocytes 257 Human bone morphogenetic protein 4 (BMP‐4) Alginate Human fibroblast growth factor 1 (FGF‐2) Peripheral artery disease and coronary artery disease Promote neovascularization and restore blood flow and tissue function of heart muscle 258 Alginate FGF‐1 Hypoxia Enhancement of graft neovascularization in a retrievable rat tomentum pouch 259 Alginate Transforming growth factor‐beta (TGF‐β) Articular cartilage defects Controlled delivery of TGF‐β selectively to the injury site and improve the repair of articular cartilage defects in rabbit model 260 Alginate Insulin‐like growth factor‐1 (IGF‐1) Nervous system disorders such as stroke Enhanced the proliferation of the encapsulated NSCs 261 Hyaluronic acid None Atherosclerosis Reach the atherosclerotic lesion after systemic administration with high potential as carrier for diagnosis and therapy of atherosclerosis 261 Hyaluronic acid VEGF Wound healing Promote angiogenesis and accelerate healing 262 Chitosan Hyaluronic acid VEGF Development of vascular network during implantation Promote angiogenesis 263 Chitosan PGDF–BB Poly( l ‐lactide‐ co ‐glycolide) (PLGA)‐grafted hyaluronic acid Bone morphogenetic protein‐2 (BMP‐2) and IGF‐1 Bone regeneration Promote the attachment, proliferation, spreading, and alkaline phosphatase (ALP) activity of human adipose‐derived stem cells (hADSCs) 264 Alginate microspheres within hyaluronic acid hydrogels TGF‐β3 Cartilage repair Promote neo cartilage formation 265 Parathyroid hormone related protein (PTHrP) Hyaluronic acid/chitosan nanoparticles embedded in porous chitosan scaffold DNA encoding TGF‐β1 Cartilage tissue engineering Promotion of chondrocyte proliferation 266 Glycidyl methacrylated dextran BMP‐2 Wound healing Periodontium drug delivery 267 Acetalated dextran Hepatocyte growth factor fragment Myocardial infarction (MI) Largest arterioles, fewest apoptotic cardiomyocytes bordering the infarct, and the smallest infarcts 268 Methacrylated dextran VEGF Ischemia Increase blood perfusion and angiogenesis 269 Chitosan–polyethylenimine BMP‐2 gene Repair of bone defect Affect cell differentiation through a BMP‐2 signal pathway and promote new bone formation at the defect area 270 John Wiley & Sons, Ltd. Polysaccharide‐based micro‐/nanoparticles provide protection to the protein‐based growth factors, offering a versatile release profile in a controlled manner while reducing the risk of having site effects, and are able to deliver the bioactive agent to target cells. In addition, stimulation can form different types of protein cargos, and the cellular signaling will then influence the cellular process, such as attachment, proliferation, migration, and differentiation, demonstrating the potential for applying these strategies for promoting tissue regeneration. 213 4. 2 Targeted Delivery and Theranostic Applications Cancer remains one of the worlds' major causes of death 214 and the improvement of effective therapies continues to challenge researchers. The great biocompatibility and the availability of multifunctional conjugation make polysaccharide nanoparticles arguably one of the best drug delivery vehicles for cancer treatment. With optimal size and surface properties, polysaccharide nanoparticles can be designed and engineered to increase the bloodstream circulation time and reach the target tumor lesion. Due to the enhanced permeability and retention effect to the ligands conjugation, nanoparticles are accumulated in the tumor tissue while delivery anticancer therapeutics are entrapped inside the particles, providing a higher targeting efficacy compared to traditional drug delivery methods. Conjugation target moieties also facilitate the precise delivery of chemotherapeutics, resulting in higher treatment efficiency and lower side effects. 215 Table 2 lists some selected examples of polysaccharide‐based drug delivery systems encapsulated with therapeutic/diagnostic agents for cancer therapy. Although different types of polysaccharides have been assessed to develop a suitable anticancer/theranostic nanosystems (Table 2 ), only negligible amount of development can reach clinical trials. It should be noted that complex chemical conjugation may result in unexpected toxicity after systemic administration because of impurity of products. The future study of polysaccharide‐based CRS is seen to be centered on the improvement of sensitivity and specificity of stimuli‐responsive triggers, as well as safety profile after systemic delivery. Table 2 Examples of polysaccharide‐based drug delivery systems for controlled delivery anticancer agents Polysaccharide type Anticancer agents Imaging agents Cancer type Result and application Ref. Hyaluronic acid (HA) None Cy5. 5 Xenograft subcutaneous dorsa of athymic nude mice To visualize the biodistribution of HA nanoparticles accumulating into the tumor with a combination of passible and active targeting mechanism 271 Liposome–protamine–hyaluronic acid TGF‐β siRNA None Melanoma Induction of antigen‐specific immune response and target modification of tumor microenvironment; powerful tool for immunotherapy 272 Chitosan siRNA for VEGRA, VEGFR1, and VEGFR2 None Breast cancer Suppressive effect on VEGF expression and tumor volume 273 Chitosan/alginate Doxorubicin None HepG2 hepatoma cells xenografts Induce the apoptosis of HepG2 tumor cells both in vitro and in vivo 274 Alginate Doxorubicin None Liver tumor Tumor necrosis; heart cells and healthy liver cells surrounding the tumor were not affected 275 Glycyrrhetinic acid‐modified alginate Doxorubicin None Hepatoma carcinoma Tumor inhabitation rate reach 79. 3% 282 Alginate– g –poly( N ‐isopropylacrylamide) (PNIPAAm) Doxorubicin FCR‐675 Squamous cell carcinoma DOX‐loaded alginate– g –PNIPAAm micelles showed excellent anticancer therapeutic efficacy in a mouse model without any significant side effects 277 Alginate Cisplatin(CDDP) Cy5. 5 Human caucasian ovary adenocarcinoma Enhance delivery of CDDP into ovarian tumor tissues and improved the antitumor efficacy of CDDP, while reducing nephrotoxicity and body weight loss in mice 278 N ‐trimethyl chitosan Cisplatin–alginate complex None Human ovarian and lung carcinoma Induce apoptosis 279 Hyaluronic acid Cisplatin None Human malignant gliomas Induce apoptosis 280 Chitosan Destran–doxorubicin None Various cancer types Induce apoptosis and shrink tumor size 281 Hyaluronic acid Cisplatin siRNA that downregulate antiapopotic genes overexpressed in cisplatin resistant tumor Indocyanine green Lung cancer Overcome the Multidrug resistance effect of lung cancer in xenograft model and induce apopotosis 282 John Wiley & Sons, Ltd. One attractive strategy for intracellular controlled release of anticancer agents is the exploitation of the redox‐responsive system, which contains disulfide bonds that can be cleaved by overexpressed glutathione in tumor cells. For example, Hu et al. reported chitosan‐based glycolipid‐like CSO—ss—SA (CSO: chitosan; SA: stearylamine) micelles for selective release of DOX/PTX by responding to the reducing environment in tumor cells. 54 CSO—ss—SA micelles exhibited a desired reduction‐sensitivity as they were able to promote fast degradation and release of the drug in 10 × 10 −3 m of GSH. An in vitro drug release study indicated that CSO—ss—SAs could quickly deliver the drug into the human ovarian cancer cells (SKOV‐3) and human normal liver cells (L‐02) through endocytosis pathway with significant higher delivery efficiency in SKOV‐3 compared to L‐02. Besides, the cellular inhibition rate of PTX‐loaded CSO—ss—SA micelles was positively correlated with the intracellular GSH concentration in SKOV‐3 cells. A mouse xenograft model study showed that CSO—ss—SAs could effectively deliver the drug into tumor tissue via the EPR effect ( Figure 12 ). Although there was high deposition of CSO—ss—SAs in the liver and spleen, the drug release mainly existed in the tumor. Compared with Taxol at the same doses, PTX‐loaded CSO—ss—SA micelles provided a distinguished antitumor effect with a rather low dose of PTX. Overall, this study emphasizes that the rational design of a selective redox‐responsive system could serve as a smart platform for drug delivery with the least toxicity and rapid intracellular drug liberation in tumor cells. Figure 12 A) Transmission electron microscopy (TEM) images of CSO—ss—SA (up) and PTX loaded CSO—ss—SA) (down). B) In vivo whole body images and C) average fluorescent signal. D) Fluorescent images of organs of tumor bearing nude mice after injection DiR‐labeled CSO—ss—SA for 12, 24, 48, and 72 h. Reproduced with permission. 54 Copyright 2015, Elsevier. In addition to redox‐triggered drug release, the pH difference between tumor and normal tissues, as well as between the cytoplasm and endosomes can also be harnessed for controlled release of the chemotherapeutics. For instance, Feng et al. described the surface coating of DOX‐loaded mesoporous silica nanoparticles (MSNs) with multilayers of alginate/chitosan to impart pH responsiveness of the nanocarriers (DOX@PEM–MSNs; i. e. , DOX‐loaded polyelectrolyte multilayer (PEM)–green fluorescence FITC‐labeled MSNs (FMSNs)). 216 The release of DOX was triggered by acidic intracellular or extracellular environments. An in vitro study on HeLa cells showed that the intracellular release of DOX from nanocarriers was pH dependent (lowering pH increased the release rate), and a sustained DOX accumulation in the nucleus led to prolonged therapeutic efficacy ( Figure 13 ). Moreover, an in vivo evaluation in healthy rats demonstrated that these DOX@PEM–MSNs carriers exhibited longer systemic circulation time and slower plasma elimination rate than free DOX. Compared with unmodified MSNs, the PEM–FMSNs showed superior hemocompatibility in terms of low hemolytic and cytotoxic effects against human red blood cells (RBCs), which endorses them as potential candidates for systemic delivery. Figure 13 A1, A3) Scanning electron microscopy (SEM) and A2, A4) TEM images of (A1/A2) MSN and (A3/A4) PEM–MSNs with adjacent photographs showing good dispersity in water. SEM images of human red blood cells (RBCs) cultured on B1, B2) MSNs and B3, B4) PEM–MSNs. C) TEM images of intracellular uptake of PEM–MSNs in HeLa Cells; (C2) was the zoomed in image of (C1). D) Confocal laser scanning microscopy (CLSM) images of HeLa cell coincubation with free DOX, DOX encapsulated PEM–MSNs, and PEM–FMSNs for 10 min, 0. 5 h, 1 h, 3 h, 6 h, and 12 h. Reproduced with permission. 216 Copyright 2014, ACS. Dysregulated enzyme expression is often associated with numerous diseases, particularly cancer, inflammatory, and infectious diseases. Certain upregulated enzymes (e. g. , matrix metalloproteinases, cathepsins, caspases, thrombins, glucuronidase) could be considered as specific endogenous triggers for the release of therapeutic and diagnostic agents. HA‐coated MSNs loaded with DOX were also reported by Zhang et al. They grafted the biotin–HA on desthiobiotin decorated MSNs via a streptavidin‐mediated cross‐linkage, which prevented the DOX release from the pores of the MSNs. 217 Once the MSN–HA/DOX was specifically taken up by CD44‐positive cancer cells by receptor‐mediated endocytosis, DOX was released from the pore of MSN by competitive binding of cytoplasmic biotin and desthiobiotin to streptavidin. In vitro examinations showed that MSN–HA could be internalized by HT‐29 and Colon‐26 cells (CD44 positive), and the release of DOX was promoted significantly in the presence of hyaluronidase (HAase) and/or biotin. HA‐coated MSNs displayed higher cell viability than bare MSNs. An in vivo safety evaluation demonstrated that despite that MSN–HA showed little nonspecific interaction with proteins, blood cells, and macrophages, MSN–HA could significantly improve the biocompatibility of MSNs by surface coating HA. Evaluation of MSN–HA/DOX on a colon‐26 xenograft tumor model showed that MSN–HA/DOX had better antitumor effect than free DOX, owing to the presence of extracellular matrix‐localized HAase and intracellular biotin in the tumor site that triggered the disintegration of biotin–HA from MSNs, which thus enhanced its antiproliferative activity in a solid tumor. Zhang et al. reported using glycyrrhetinic acid (GA)‐modified alginate nanoparticle to target delivery of DOX to kunming mice for curing liver cancer. GA is a commonly used bioactive ligand for modification of DDS and results in additional accumulation of drug molecules in the liver sites. Passive targeting with enhanced permeability was also a leading cause for liver cancer accumulation. Instead of focusing on the therapeutic effect of shrinking the tumor, the authors also evaluated the side effects with regards to the DOX chemotherapy. The in vivo study results suggested that after a single tail‐vein injection of 7 mg kg −1 body weight, the concentration of DOX in the liver reached 67. 8 ± 4. 9 µg g −1, which was 2. 8‐fold and 4. 7‐fold higher compared to non‐GA modified alginate nanoparticles and free DOX HCl. A histological examination showed tumor necrosis in both experimental groups. Most importantly, the heart cells and the liver cells surrounding the tumor were not affected by administration of DOX/GA–ALG NPs, whereas myocardial necrosis and apparent liver cell swelling were observed after DOX·HCl administration. 217 The RNAi technique has open a new route for cancer therapy and several candidates are being clinically tested. In the development of RNAi‐based techniques, imaging methods provide a visible and quantitative solution to investigate the therapeutic effect at anatomical, cellular, and molecular levels and they are able to noninvasively trace the distribution and study the biological processes in preclinical and clinical stages. 218 Nanocarrier‐mediated delivery of RNAi therapeutics usually encounters different biological barriers, including reaching the circulation, crossing the vascular barrier, cellular uptake, and endosomal escape. With advancements in chemical modification and nanotechnology, polysaccharide nanoparticles are diverse in size and charge and are widely applied as platforms for simultaneous gene/drug delivery and imaging. 14 Yoon et al. reported a novel type of biodegradable hyaluronic acid‐ graft ‐poly(dimethylaminoethyl methacrylate) (HPD) conjugates that can form complexes with siRNA and that can be chemically cross‐linked via the formation of the disulfide bonds under facile conditions to exhibit high stability in 5% serum solution over the un‐cross‐linked ones. The in vivo study, which was performed using FPR675‐labeled HPD with siRNA complexes, showed the efficacy of selective accumulation of the complexes at the tumor site after intravenous injection into tumor‐bearing mice, achieving a successful gene silencing effect while being able to be monitored with a whole‐body near infrared flurescence (NIRF) imager. 219 While the application of polysaccharide coated particles show promising result in research, there are still obstacles before more clinical trials are tested. One problem is to target sites that are located farther from the magnetic source. Future research should focus on designing multimodality imaging probes with polysaccharide coatings to enhance the use of particle‐based imaging‐based contrasts, offering versatile solutions for early cancer detection and monitoring. 220 Aside from cancer therapeutics, CD‐based supramolecular assemblies have received attentions in confronting genetic and rare diseases. 148, 221 A series of stimuli‐cleavable β‐CD‐based polyrotaxanes (PRXs) have been investigated by Tamura and Yui. 222 According to a recent study, a redox‐responsive β‐CD‐threaded PRX has been developed for the treatment of NPC disease, 221 which was achieved by the controlled released of β‐CDs from HEE group‐modified Pluronic P123 via intracellular disulfide bonds cleavage. In this report, they compared the efficacy of PRXs and hydroxypropyl‐β‐CD (HP‐β‐CD) for treatment of the autophagy failure in NPC disease. Usually, an increased number of LC3‐positive puncta can be observed from NPC patient‐derived fibroblasts (NPC1 fibroblasts). When treated with HP‐β‐CD, the autophagic degradation activity was further disturbed by the increasing amount of LC3‐positive puncta and levels of p62 in the NPC1 fibroblasts, whereas the PRX‐based treatment diminished both the amount of LC3‐positive puncta and levels of p62 in the NPC1 fibroblasts via the mammalian target of rapamycin (mTOR)‐independent pathway. The evaluation of the mRFP‐GFP‐LC3 reporter gene expression that demonstrated the redox‐responsive PRXs mediated the generation of autolysosomes to approach for autophagic protein degradation ( Figure 14 ). In this regard, the developed β‐CD‐threaded bioresponsive PRXs offered a promising treatment for NPC disease based on simultaneously improving the cholesterol accumulation in lysosomes and damaged autophagy functionality in NPC1 fibroblasts. Figure 14 A) CLSM images of normal and NPC1 fibroblast cells expression mRFP‐GFP‐LC3 coincubated with HP–β‐CD and HEE—ss—PRX nanoparticles at the concentration of 10 × 10 −3 and 1 × 10 −3 m. B) Quantification of cell population with production of autophagosomes (mRFP + and GRP + ) and autolysosome (mRFP + and GRP − ). Reproduced with permission. 221 Copyright 2015, American Society for Biochemistry and Molecular Biology. Natural polysaccharides also play important roles in the diagnosis and therapy of cardiovascular diseases due to their unique features including binding affinities to atherothrombotic sites, immunomodulation and therapeutic effects as well as their use as a platform for therapeutics delivery. 223 For example, the recognition of stabilin‐2 and CD44 receptors by HA during the pathogenic process of atherosclerosis has been explored for active targeting theranostics. 224, 225 Recently, Lee et al developed HA‐NPs as therapeutic carriers for active targeting atherosclerosis, 224 which was prepared through self‐assembly of HA‐5b‐cholanic acid‐Cy5. 5 conjugates. The evaluation of cellular internalization of HA‐NPs demonstrated the stabilin‐2 or CD44 receptor‐mediated endocytosis mechanism, as cellular uptake of HA‐NPs was significantly inhibited by the pretreatment of an excess amount of free HA. In vivo fluorescence imaging of atherosclerotic lesion by tail vein injection of Cy5. 5‐labeled HA‐NPs into ApoE‐deficient mice revealed that 24 h postinjected HA‐NPs successfully highlighted the atherosclerotic lesion with a stronger signal than the normal aorta. The confocal microscopy imaging showed colocalization of the stabilin‐2/CD44 antibody and HA‐NPs in the atherosclerotic plaque ( Figure 15 A, B). Besides, in vivo fluorescence imaging also demonstrated superior targeting efficiency of HA‐NPs compared to passively targeted HGC‐NPs (Figure 15 C–E). Overall, the study described the potential theranostic application of HA‐based nanopolyplexes for atherosclerosis. Figure 15 A) Fluorescent images of Cy5. 5‐labeled HA‐NPs in atherosclerotic plaque in ApoE KO and normal mice. B) Zoom in fluorescent images of HA‐NPs in isolated plaques with immunostaining of Stabilin‐2 (STAB2) and CD44 with insetted images show the 4′, 6‐diamidino‐2‐ phenylindole (DAPI) nuclei stain. C) In vivo live image of HA‐NP in atherosclerotic lesion in ApoE KO mice. D) Fluorescent images of isolated aorta after sacrificing the ApoE KO mice. E) Quantitatively analysis of the fluorescent intensity of HGC‐NPs and HA‐NPs form atherosclerotic lesion images. Reproduced with permission. 224 Copyright 2015, RSC. 5 Clinical Translations: Progress and Challenges Despite the great potential of polysaccharide‐based DDSs in various preclinical studies of disease treatment, they are still elusive to the market and only limited amounts of products have entered clinical trials. We have listed some ongoing and completed clinical trials for polysaccharide‐based nanoproducts that are not limited to particulate DDSs but can be used for other therapeutic applications as well ( Table 3 ). There are several types of polysaccharide products based on a drug‐conjugated delivery system, which can be modulated to be stimuli‐responsive or receptor‐mediated targeting. 13 Five of the known polysaccharide‐based conjugates for anticancer treatment in clinical tests are AD‐70, DE‐310, Delimotecan, ONCOFID‐P‐B, and CRLX101. Table 3 Examples of polysaccharide‐based DDSs in clinical trials. IVI, intravenous infusion; CDP, cationic cyclodextrin polymer; AD–PEG, adamantane polyethylene glycol; hTf, human transferrin protein Polysaccharide Product name Composition Delivery route Diseases or conditions Development stage Country, year [Ref. ] Dextran AD‐70 Doxorubicin, dextran IVI Refractory solid tumors Phase I discontinued Germany, 1993 226 DE‐310 Exatecan mesylate, carboxymethyl‐dextran IVI Advanced solid tumors Phase I Netherlands, UK, Canada, 2005 231 Delimotecan (MEN 4901/T‐0128) Camptothecin (T‐2513), carboxymethyl‐dextran IVI Solid tumors Phase I Netherlands, Italy, 2008 283 Chitosan Milican Holmium‐166, chitosan Percutaneous injection Small hepatocellular carcinoma Phase II South Korea, 2006 236 Hyaluronic acid RadiaPlex Sodium hyaluronate Topical skin Radiation dermatitis Phase III USA, 2007 284 ONCOFID‐P‐B PTX, hyaluronic acid Intravesical instillation Bladder cancer Phases I, II Italy, 2011 233 Cyclodextrin CALAA‐01 siRNA (RRM2), CDP, AD–PEG, hTf IVI Solid tumors Phase I USA, 2008 285 CRLX101/IT‐101 β‐Cyclodextrin, PEG copolymer–camptothecin IVI Ovarian/tubal/peritoneal cancer Phases I, II USA, 2012 286 Rectal cancer Phases I, II USA, 2013 287 Advanced solid tumors Phase I USA, 2015 134, 288 Lung cancer Phases I, II USA, 2016 289 Alginate DIABECELL Neonatal porcine islets, poly‐ l ‐ornithine, alginate mixture Xenotransplantation Type I diabetes Phases I, II, III New Zealand, 2009 240 IK‐5001 Calcium gluconate, sodium alginate Intracoronary injection Acute/ST‐elevation myocardial infarction, congestive heart failure Phase I USA, Germany, Israel, 2010 241 OligoG CF‐5/20 Alginate oligosaccharide Inhalation Cystic fibrosis Phases I, II UK, Norway, 2014 242 John Wiley & Sons, Ltd. AD‐70 : AD‐70 is a dextran–anthracycline conjugate that is consists of a oxidized dextran polymer with molecular weight of 70 000 Da and doxorubicin conjugations through the Schiff‐base reaction with glycine attached on dextran. 226 The principle of this conjugation approach was based on the concept that hypoxic tumor milieu are expected to promote the liberation of active drugs. In phase I of the study, AD‐70 was administered via intravenous infusion in 13 patients. 226 Dose‐limiting toxicities (DLTs) including significant thrombocytopenia and hepatotoxicity were observed in several patients; they were attributed to specific uptake of dextran–doxorubicin conjugates by MPS, since the dextran (glucose polymer) can be recognized by glucose transporters on macrophages. Besides, the Schiff‐base formation can certainly yield aldehyde residues that could induce toxicity. 227 Despite its promising results for Schiff‐base‐mediated tumor selectivity in an animal model, the progress to the next clinical phase was discontinued. During the phase I clinical trial, AD‐70 showed unexpected toxicity due to the immunogenic effect caused by non‐biodegradable nanoformulation that consists of modified side chain of oxidized dextran. 228 DE‐310 : DE‐310 is a macromolecular DDS that was discovered by Daiichi Pharmaceutical Co. , Ltd. It is composed of a topoisomerase I inhibitor (DX‐8951f, a camptothecin analogue), and a biodegradable carboxymethyl–dextran polyalcohol polymer that are covalently attached via a Gly–Gly–Phe–Gly peptidyl linker. 229 The design rationale of this macromolecular carrier was intended to afford passive targeting based on the EPR effect and the controlled release of the parent drug DX‐8951f using enzymatic cleavage of the peptidyl spacer by lysosomal proteases (cathepsins). 230 A phase I clinical trial with DE‐310 revealed that in a total of 27 patients, one patient with metastatic adenocarcinoma achieved complete remission. Another patient with metastatic pancreatic cancer achieved partial remission. And a total of 14 patients had stabilized disease progression. 231 Neutropenia, thrombocytopenia, and hepatotoxicity were the main DLTs. The study concluded that DX‐8951 was sustainably released from DE‐310 over a prolonged period, yet there was no detectable drug concentration in red blood cells, skin, and saliva, which supportively implies that DE‐310 could improve the therapeutic index of drug DX‐8951f. However, the insufficient sample size and data prevent clear conclusions from being drawn. Delimotecan : Delimotecan is another carboxymethyl–dextran conjugate containing the camptothecin analogue 10‐(3′‐amino‐propyloxy)‐7‐ethyl‐(20S)‐camptothecin (T‐2513). T‐2513 is bound to the polymer via the triglycine linker, which can be specifically cleaved by cathepsin B and subsequently release the active drug. 232 Cathepsin B is a lysosomal cysteine protease that is upregulated in a wide variety of human tumors; hence, the presence of the triglycine linker is important for enhancing tumor selectivity and reducing toxicity. In a phase‐I study, 22 patients received the Delimotecan treatment, and two partial remissions were observed in patients with head and neck cancer. However, adverse hematological effects such as leukopenia hematologic leukocytopenia and neutropenia and nonhematologic symptoms including skin rash, fatigue, and diarrhea had occurred after Delimotecan therapy. The clinical trial confirmed that Delimotecan had a prolonged circulation half‐life and enhanced Delimotecan retention in tumor tissues (especially when the tumor is enriched in tumor‐associated macrophages) as well as the ability to increase the release of T‐2513 via enzymatic cleavage. ONCOFID‐P‐B : ONCOFID‐P‐B is a PTX–HA bioconjugate supplied by Fidia Farmaceutici S. p. A. that has entered the phase‐I evaluation. 233 The covalent conjugation between PTX and HA was intended to improve the hydrophilicity of the active pharmacophore. 234 A total of 16 patients with bladder cancer were treated by intravesical instillation of ONCOFID‐P‐B acid solution. No DLT was observed during the treatment, and drug concentrations were always under the detectable level; in addition, 9 patients achieved complete remission. However, 50% of the total patients experienced recurrence or progression. 233 The study results demonstrated that ONCOFID‐P‐B was safe for treatment of nonmuscle invasive bladder cancer. Moreover, HA‐based bioconjugates can certainly improve the bioavailability of active drug PTX for intravesical chemotherapy. CRLX101 : CRLX101 is a polysaccharide‐based compound discovered by Cerulean Pharma Inc. It consists of active pharmacophore camptothecin that covalently attached to a CDP. The rationale of this approach was to encapsulate the hydrophobic camptothecin via intra‐ and intermolecular interactions between adjacent polymer units (cyclodextrin and polyethylene glycol) and lead to self‐assembles in aqueous solution. 133 The resulting nanoparticles varied in size from 20 to 60 nm, due to its neutral surface charge and presence of PEG blocks that together provide a stealth effect to avoid nonspecific uptake by mononuclear phagocyte system. 235 Moreover, the physical encapsulation of camptothecin in CRLX101 nanoparticles prevents camptothecin from enzymatic degradations in circulation. Previous clinical data demonstrated that CDP‐based DDSs can address not only plasma solubility and toxicity, but also the therapeutic index of camptothecin. 132, 134 Milican : Unlike the above‐mentioned polysaccharide conjugates, Milican consists of a radioisotope holmium‐166 ( 166 Ho(NO 3 ) 3 ·5H 2 O), which complexed with chitosan (2‐deoxy‐2‐amino‐ d ‐glucose) polymer as an embolic platform for ablative radiotherapy. The phase‐IIb clinical study shows the outstanding efficacy and long‐term safety of Milican for the treatment of small hepatocellular carcinoma (<3 cm in size) through intratumoral injection with ultrasonographic guidance. 236 However, the effectiveness for treatment of larger tumors is currently undergoing for further evaluation. CALAA‐01 : CALAA‐01 was a pioneering targeted siRNA nanotherapeutic that was developed by Davis in 1996. 237 This delivery system consists of an anti‐RRM2 (ribonucleotide reductase subunit 2) siRNA payload and cyclodextrin‐containing polymer particle core attached with AD–PEG group and some AD–PEG covalently linked to human transferrin (Tf–PEG–AD) for tumor targeting and cellular internalization. The imidazole residues are also present to promote the endosomal escape, which exploit the protonation of amines in an acidic environment and induce the influx of protons and chloride ions into endosome, elevating the osmotic pressure and resulting in the disruption of endosomal membrane. 238 The human phase Ia/Ib clinical data from 24 patients with different cancers showed that CALAA‐01 was well tolerated during the initial dose escalation. 131 However, two patients were experiences with DLTs after the trials were reopened. Although the delivery system of CALAA‐01 has been proved effective for targeted delivery, the full ability of CALAA‐01 failed to meet its primary end point in this trial. 239 In addition to the anticancer therapeutics, a number of polysaccharide‐based DDSs were also used for treatment of other types of diseases, such as type I diabetes (DIABECELL), 240 heart disorders (IK‐5001), 241 and cystic fibrosis (OligoG CF‐5/20). 242 These examples of polysaccharide‐based nanotherapeutics undergoing clinical trials certainly played an important role in the future development of polymeric nanomedicine. However, opportunities have always been accompanied by challenges. Translational nanomedicine is a relatively new interdisciplinary field that has revolutionized the traditional knowledge of disease and therapy through cutting edge bionanotechnology. 243 Therefore, it is challenged by limited previous experience with addressing various concerns, such as nanoparticle formulation, delivery mechanisms, toxicity investigation, and revealing the biochemical basis of the interactions between NPs and complex biological systems. 244 First, one of the major considerations is the design and formulation of polysaccharide‐based DDSs. Natural‐based polysaccharides are not a single discrete chemical system, as they vary in number and distribution of repeating building blocks along the polymer backbone. 20, 245 Molecular weight and composition are therefore important influences on the solubility, chain flexibility, intra‐ and intermolecular forces, carrier size/shape, loading capacity, and surface charge. These physicochemical properties can subsequently determine the biophysiological behavior, such as plasma solubility, aggregation states, and immunogenicity. In this regard, regulatory control including the bench‐top synthesis and characterizations must be taken into account for successful clinical translation. The dose of polysaccharide‐based products may need to be scaled up from animal models to human trials, since statistical analyses showed that only less than 1% of the injected dose can reach the desired target sites for most published researches. 246 In this case, reproducibility is another issue for manufacturing, since the structure of polysaccharide‐based polymers varies from bench to bench and time to time in terms of the source purchased, molecular weight variation, functional group modification, and purification. 13 Second, knowledge of the degradation profile of different polysaccharides will also be very important for the design of the polysaccharide‐based DDSs. Breaking up the noncovalent cross‐linking (intra‐ and intermolecular forces) networks is usually the initial step. For the nondegradable cross‐link, the degradation of the polysaccharide backbone will become the leading process of polysaccharide‐based DDSs in vivo. In a biological environment, polysaccharides usually undergo enzymatic or hydrolytic degradation into nontoxic byproducts. The degradation process usually begins with the random breaking up of β‐1, 4‐ glycosidic bonds followed by the observation of N ‐acetyl linkage deacetylation degree. As the average molecular weight decreases, the degree of deacetylation increases, resulting in a polysaccharide backbone scission and destruction of side functional groups, including carbonyl, amine, and hydroxyl. 247 It is difficult to control the polysaccharide backbone degradation in vivo and predict the clear out mechanism of polysaccharide‐based DDSs, as the incomplete degradation will result in a burst release and compromised mechanical properties. Recent studies have designed polysaccharide‐based stimuli‐responsive (pH, thermal, mechanical force) materials to control or tune the drug release profile during in vivo application. 248 However, more animal studies should focus on monitoring the clearance and degradation of polysaccharide‐based materials in the long run. Another challenge is to consider the heavy reliance on the EPR effect for passive targeting in oncology. In principle, nanoparticles with sizes less than 500 nm can cross the tumor blood vessels due to the irregular gaps developed by less tightly formed endothelial cells. However, the size of the gaps varies with the type and stage of tumor. 246 Hence, the heterogeneity of the EPR effect and limited relevant experimental information from patients can certainly result in high levels of uncertainty for delivery efficiency. 249 In this respect, development of specific biomarkers or imaging agents to determine the strong EPR effects in patients for preselecting appropriate patients could be an option. 244, 250 Once the nanoparticles have reached the target sites, they have to deal with the intratumoral microenvironment to across the tumor vascular barrier. However, nanoparticles and even smaller chemotherapeutics can only insufficiently diffuse into deep tumor space. This is due to the abnormalities of the vasculature development and the rising interstitial fluid pressure inside the tumor; together, this forms a barrier for transportation of chemotherapeutics, nanoparticles, imaging agents, etc. 251 These obstacles have been associated with drug resistance to certain chemotherapeutics. 252 One possible solution is to revert the abnormalities and function of tumor vessels to a relatively normal state by using antiangiogenic therapy. 253 This in turn can ultimately improve the response to therapeutic treatment and control the tumor progression. Concerns have also expressed that the in vitro cellular study and in vivo animal model do not fully simulate the physiology and pathophysiology in humans, in particular, the xenografted human tumors in immunodeficient mice, which may not be able to mimic the true tumor microenvironment and predict therapeutic response in human patients. 135, 254, 255 Therefore, this could be a grand challenge in estimating possible outcomes before human trials start using preclinical results. Nevertheless, genetically engineered mice models can be useful to enhance our understanding of translational nanomedicine with more realistic approaches. 254 In summary, like other types of nanoparticle, budding success of clinical translation for polysaccharide‐based nontherapeutic medicines relies on the integration of multidisciplinary knowledge (involving life science, clinical medicine, material science, chemistry, and engineering) and collaboration among regulatory authorities, pharmaceutical companies, academics, and governments. 6 Concluding Remarks Polysaccharide‐based DDSs have emerged as one of the major naturally based polymers for biomedical application due to their excellent biocompatibility and biodegradability, structural stability, broad source, and versatile chemical compositions. Various chemical modifications of chemistry have been explored to increase the functionalities of the polysaccharide polymers. Meanwhile, novel engineering techniques and devices have been developed for DDS fabrications. These have generally made it possible for encapsulating different types of drug molecules (e. g. , protein, oligonucleotides, small molecules) with a desirable release profile to target tissue and great pharmacokinetic/pharmacodynamic (PK/PD) properties. The preclinical and clinical studies represent the possibility of utilizing polysaccharide‐based DDSs to enhance the therapeutic efficacy of biopharmaceutics. Despite the largely evolving knowledge and techniques, few of the polysaccharide‐DDSs have been translated into clinical studies due to limit knowledge regarding their drug release properties, targeting and therapeutic efficacy, and degradation profile. Therefore, a better understanding of material/tissue interactions is greatly needed in the field of polysaccharide‐DDSs. While compiling more convincing characterizations both in vitro and in vivo would be helpful, utilizing additional engineering modeling and monitoring techniques will also be useful for predicting the therapeutic response for clinical applications. Furthermore, the continuing development of de novo material fabrication techniques will produce better, stable, and evenly distributed polysaccharide‐based drug carriers that can be used to tailor disease targeting models. We foresee more clinical translations studies with polysaccharide‐based materials in the near future. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201700527
| 2,018
|
Advanced Science
|
Antibacterial Hydrogels
|
Abstract Antibacterial materials are recognized as important biomaterials due to their effective inhibition of bacterial infections. Hydrogels are 3D polymer networks crosslinked by either physical interactions or covalent bonds. Currently, hydrogels with an antibacterial function are a main focus in biomedical research. Many advanced antibacterial hydrogels are developed, each possessing unique qualities, namely high water swellability, high oxygen permeability, improved biocompatibility, ease of loading and releasing drugs, and structural diversity. Here, an overview of the structures, performances, mechanisms of action, loading and release behaviors, and applications of various antibacterial hydrogel formulations is provided. Furthermore, the prospects in biomedical research and clinical applications are predicted.
|
1 Introduction Since the first discovery of penicillin in 1928, 1 antibiotics have been widely used in the antibacterial field. With the development of public hygiene and biomedical technology, many infections have been effectively suppressed or even conquered, and the quality of life for human beings has been significantly improved. However, a serious issue that still remains is that the use of antibiotics has led to the emergence of multidrug resistant microorganisms, which are very difficult to combat. 2 This has led to over 13 million people dying per year from infectious diseases worldwide. 3 What was the most disappointing was that the corresponding antibiotic‐resistant bacteria emerged almost immediately after the advanced antibiotics were approved, e. g. , the fidaxomicin‐resistant Enterococci (K‐1476) and the methicillin‐resistant Staphylococcus aureus ( S. aureus )(MRSA). 4, 5, 6 Figure 1 shows the history of the development of antibacterial agents followed by the acquisition of resistance by microorganism. Synthetic antibacterial agents, such as salicylate (SAL), chlorhexidine (CHX), isothiazolinone (ITZ), thiosemicarbazone (TSC), octenidine (OCT), and quaternary ammonium (QA) compounds, also face constant threats because of the drug resistance acquired by microorganisms. 7 Additionally, conventional antibiotics also face other problems, such as solubility, overdose, and cytotoxicity. Therefore, an efficient and safe drug delivery system, which can reduce the risk of bacterial drug‐resistance and regulate the toxicity of antibacterial drugs, is in high demand. Figure 1 History of antibacterial agents and acquisition of resistance by microorganism. Challenged by the ever‐growing threats from drug‐resistant pathogenic microorganisms, researchers have been studying various advanced antibacterial materials. Among them, heavy metal ions and natural extracts were discovered and applied in the antibacterial field. However, these materials can inhibit and kill not only the pathogenic microbes, but also normal cells in the human body, which limits the potential applications for these materials. Hydrogels are a form of 3D porous materials, which consist of polymer chains with either physical or chemical crosslinking. 8, 9, 10, 11 Hydrogels have been extensively studied as an alternative material for antibacterial applications. By carefully selecting monomers and crosslinkers, the desired abilities of hydrogels, such as the hydrophilicity and porosity, can be developed for antibacterial applications. Moreover, some types of hydrogels also have an inherent antibacterial property. According to the classification of hydrogel matrices and the antibacterial agents, the antibacterial hydrogels are divided into three types: (i) inorganic nanoparticle‐containing hydrogels, (ii) antibacterial agent‐containing hydrogels, and (iii) hydrogels with inherent antibacterial capabilities. This article will describe the syntheses, performances, action mechanisms, loading and release behaviors, and applications of antibacterial hydrogels, as depicted in Scheme 1. Scheme 1 Compositions, performances, and applications of antibacterial hydrogels. 2 Inorganic Antibacterial Agent‐Incorporated Hydrogels Inorganic antibacterial materials mainly include metal ions and metallic oxide nanoparticles. Commonly used metal/metal ions include, but are not limited to, silver (Ag), gold (Au), and copper (Cu). Metallic oxide metal nanoparticles that are utilized include zinc oxide (ZnO), titanium dioxide (TiO 2 ), and nickel oxide. Currently, the most widely used inorganic antibacterial materials are silver nanoparticles (Ag NPs) and ZnO NPs. Inorganic antibacterial material‐loaded hydrogels can not only enhance the antibacterial properties, but can also maintain antibacterial activity for a long period of time, which reduces the likelihood of bacterial resistance arising. Figure 2 illustrates the possible antibacterial mechanisms of the metal and metallic oxide nanoparticles. 12 To summarize, the nanoparticles cause damage to bacterial cell membranes or detrimental alterations to organelles. It should be emphasized, however, that some of these mechanisms are speculative and require further discussion and demonstration. Figure 2 Antibacterial mechanisms of metal and metallic oxide nanoparticles. Reproduced with permission. 12 Copyright 2013, Elsevier. 2. 1 Metal Nanoparticle‐Loaded Hydrogels 2. 1. 1 Silver Nanoparticle‐Loaded Hydrogels Since thousands of years ago, even before the word “microorganism” was established, Ag was already regarded as an antibacterial agent. Ag powder was first documented in medical history to be applied in wound healing for the treatment of ulcers by Hippocrates. Ag continues to play an important role in biomedical applications, such as for wound dressings, textiles, and bone implants. With the development of nanoscience and nanotechnology, the recent applications of Ag are mainly in the form of nanoparticles. 13 Ag NPs are emerging as an efficient antibacterial agent, although the mechanisms remain unclear. The most accepted hypothesis is that the silver ion (Ag + ) can bind to the bacterial cell membrane through the interaction between Ag + and the thiol group in proteins on the cell membrane, thus affecting the bacterial cell's viability by inhibiting the replication of DNA (Figure 2 ). Hydrogels containing Ag NPs include two types of matrices: (i) the natural polymers or modified natural polymers and (ii) synthetic polymers. The polysaccharides play an important role in the natural hydrogel matrix. Alginate is one of the linear natural polysaccharides that can form hydrogels via ionic interactions with Ca 2+. Stojkovska et al. incorporated Ag NPs into sodium alginate (SA) microbeads through an electrochemical procedure, which efficiently released Ag NPs and/or Ag + and showed antibacterial activity against S. aureus. 14 More specifically, the maximal concentration of released Ag from SA microbeads was about 0. 3 × 10 −3 m, which killed 95. 8% of the bacteria after 1 h of coincubation. These results showed that SA was successfully utilized for preparation of SA microbeads incorporated with Ag NPs as antimicrobial agents against S. aureus. Madhusudana Rao et al. further contributed to this research by creating SA‐based semi‐interpenetrating polymer network hydrogels for the incorporation of Ag NPs. The research showed that Ag nanocomposite hydrogels could be used for biomedical applications, such as wound dressings and even water purification. Furthermore, Neibert et al. described a method to enhance the mechanical strength of SA hydrogel loaded with Ag NPs. The calcium‐ or N, N ‐methylenebisacrylamide‐crosslinked SA fibers were loaded with Ag NPs, which could be applied to wound dressings or utilized for healing purposes. 15, 16 All the hydrogels loaded with Ag NPs showed good antibacterial activity against Gram‐negative (G−) bacterium Escherichia coli ( E. coli ). 17, 18 The natural and biodegradable SA nanocomposite hydrogels showed a sustained release of Ag and a long‐term antibacterial activity. Chitosan (CS) and chitin (CT) have inherent antibacterial and metal‐binding properties. CS‐ or CT‐based hydrogels like CS/2‐GP/nanosilver hydrogels (GP, glycerophosphate) 19 and silver molybdate NANOPARTICLEs/CT matrix, are also commonly used for antibacterial applications. Ag 2 Mo 2 O 7 /CT hydrogels provide green synthesis processes and excellent antibacterial abilities against E. coli. 20 With the help of CS or CT, nanosilver hydrogels had enhanced the efficacy and reduced the toxicity. Reddy et al. demonstrated that the natural and nontoxic gelatin contributed to anchoring and stabilizing Ag NPs. Thus, they prepared poly(gelatin acrylamide) silver nanocomposite hydrogels for the inactivation of bacteria. 21 In addition to utilizing natural polymer for natural polymer‐ or modified natural polymer‐based hydrogels as antibacterial matrices, many synthetic polymers have also been applied to fabricate the Ag NP‐loaded hydrogels, such as polyacrylamide (PAAm), poly(acrylic acid) (PAA), poly(ethylene glycol) (PEG), poly(vinyl alcohol) (PVA), poly( N ‐vinyl‐2‐pyrrolidone) (PVP), as well as short peptides and their derivatives. The main advantage of using these hydrogels as a matrix is that the morphologies and sizes can be easily controlled by changing the amount of crosslinkers and monomers in the hydrogel network. 22, 23, 24 For example, compared with Ag NPs alone and Ag + ‐bonded hydrogels, the Ag NP‐loaded PAAm/PVA hydrogels fabricated by Varaprasad and co‐workers exhibited a higher antibacterial activity toward E. coli. This was because the Ag NPs in the hydrogels had good dispersion capability throughout the hydrogel network. Styrene sulfonic acid sodium salt was incorporated into the hydrogels to form the Ag NP‐loaded hydrogel composed of poly(acrylamide‐ co ‐styrene sulfonic acid sodium salt) and CS, which could combat the most sensitive strains of Bacillus subtilis ( B. subtilis ). 23 In order to increase the stability and dispersity of metal nanoparticles in aqueous media and control the nanostructure, semi‐interpenetrating network hydrogels composed of Pluronic and PAAm were simultaneously prepared by free radical crosslinking polymerization and served as nanoreactors for the synthesis of Ag NPs. 22 The Ag NP‐loaded hydrogels formed by mixing of PAAm with itaconic acid (IA) 25 or starch (ST) 26 were also reported to possess good antibacterial properties while providing a green synthesis process. Boonkaew et al. synthesized 2‐acrylamido‐2‐methylpropane sulfonic acid sodium salt hydrogels containing Ag NPs. The hydrogel with 5. 0 × 10 −3 m Ag NPs displayed the highest antimicrobial activity for wound infection prevention without cytotoxicity. 27 Simon et al. synthesized a N ‐terminally 2‐(naphthalen‐6‐yl)acetic acid‐protected Phe‐Phe‐Cys peptide (Nap‐FFC) hydrogel, which incorporated Ag NPs and showed inhibition against both Gram‐positive (G+; MRSA) and G− (i. e. , Acinetobacter baumannii ) bacteria. 28 It is important to note that the hydrogels had excellent biocompatibilities compared to human cervical carcinoma HeLa cells ( Figure 3 ). All these hydrogels demonstrated noticeable antibacterial properties, which gave researchers more confidence on the exploitation of Ag NP hydrogels. Figure 3 Biocompatible Ag NP‐derived tripeptide supramolecular hydrogel for antibacterial wound dressings. A, B) Schematic illustration of antibacterial tests of Nap‐FFC and Ag NP@Nap‐FFC hydrogels. C) Cytotoxicity assay of Ag NP@Nap‐FFC nanocomposites toward HeLa cells. Reproduced with permission. 28 Copyright 2016, Royal Society of Chemistry. Hydrogel matrices obtained from different synthesis processes possess differing characteristics. P(AA‐ co ‐PEGMA)/Ag NP composite hydrogels were developed by Lee and Tsao, offering a promising bioadhesive patch or wound dressing material (PEGMA, poly(ethylene glycol) methyl ether acrylate). 29 Ag NP‐coated CS wafer‐loaded PVA hydrogels (PVA/Ag–CHW hydrogels) were formulated by a sonication technique and then used as a wound dressing. The PVA/Ag–CHW hydrogels improved the re‐epithelialization, increased angiogenesis, and enhanced wound healing without any undesirable inflammatory response. 30 A thermoplastic hydrogel was synthesized from multiblock PEG–POSS (POSS; poly(hedral oligosilsesquioxane)) polyurethanes by Wu et al. 31 Without Ag, the hydrogel exhibited the most rapid and extensive biofilm formation. Meanwhile, the Ag‐containing nanofibrous hydrogel possessed outstanding biofilm resistance and antibacterial property that lasted over 14 days. PVA/PVP‐based hydrogels fabricated by Eid et al. containing Ag NPs were reported to be uniformly distributed and highly stable. 32 The pH‐sensitive poly(methyl methacrylate‐ co ‐methacrylic acid)/Ag NP hydrogels synthesized by a free radical crosslinking copolymerization approach have the potential to be utilized as an antibacterial biomaterial. 33 All these hydrogels mentioned above displayed antibacterial ability against E. coli, S. aureus, Pseudomonas aeruginosa ( P. aeruginosa ), and B. subtilis. Additionally, an antibacterial coating made of poly( l ‐lysine)/hyaluronic acid multilayer films and liposomes loaded with Ag + was also explored. 34 The strong antibacterial effect was attributed to the diffusion of Ag + from the AgNO 3 coating, which resulted in Ag + aggregation around the membranes of bacteria. Moreover, other small antibacterial molecules like antibiotics could be loaded into hydrogels using this method to accomplish the goal of delayed drug delivery. Furthermore, there are mussel‐inspired Ag NP hydrogels synthesized with water‐soluble PEG, which contain reactive catechol moieties inspired by mussel adhesive proteins. Mussels possess these adhesive proteins because it is crucial for mussels to adhere to almost any surface in an aqueous environment. This application of biomimicry is a highly promising antibacterial biomaterial coating and tissue adhesion. 35 Although Ag NP‐based hydrogels have so many advantages, their applications are still far from what is expected. They are less effective in G+ bacteria compared to G− bacteria due to the high resistance from the peptidoglycan within the cell walls of G+ bacteria. 36 Furthermore, the development of nanoparticles was largely restricted because of their physical and chemical instability, therefore stabilization of metallic nanosystems will become a promising area of research within nanoscience and nanotechnology. Although Ag ions are efficient bactericides at a concentration of as low as ≈0. 001–0. 05 ppm, their tissue toxicity and cytotoxicity should be discussed. The serum albumin in human blood can reduce the antibacterial effect of Ag NP‐embedded hydrogels as well. 37 In addition, it is reported that Ag NPs resulted in several negative impacts on genes. The balance between anti‐reactive oxygen species (ROS) response and DNA damage and the balance between mitosis inhibition and chromosome instability may play significant roles in Ag‐induced toxicity. 38 Thus, there is a preference to minimize the toxicity and reduce the influence of serum albumin when designing Ag NP‐based hydrogels. Additionally, more nontoxic and environmental‐friendly synthesis strategies of Ag NP‐based hydrogels, such as the size‐controllable synthesis of Ag NPs with tobacco mosaic virus as a biomediator without any external reducing agents should be developed. 39 2. 1. 2 Gold Nanoparticle‐Loaded Hydrogels Although Au is universally considered to be biologically inert, gold nanoparticles (Au NPs) have diverse biological functions. Au NPs play a significant role in biological applications, such as cell imaging, photothermal therapy, sensing, and antimicrobial activities. 40 Au NPs can be designed to be various sizes and functionalized with desired polymers, thus recognized as biocompatible materials. Daniel‐da‐Silva et al. developed Au/gelatin hydrogel nanocomposites, which were crosslinked with genipin. When triggered by thermal stimuli, the composites had the potential for release of the encapsulated Au NPs. 41 Au NPs possess antibacterial capability by attaching to bacterial membranes, thus leading to the leakage of bacterial contents or the penetration of the outer membrane and peptidoglycan layers, resulting in bacterial death. Au NPs also reverse bacterial resistance to some extent when combined with non‐antibiotic or antibiotic molecules. 42 However, compared with Ag NP‐loaded hydrogels, the antibacterial Au NP‐loaded hydrogels remain insufficiently explored. The N ‐isopropylacrylamide‐based hydrogels containing Au NPs 43 and the pH‐responsive poly(methacrylic acid) hydrogel microcapsules as Au NP nanoreactors 44 have been reported, but their antibacterial properties remain unstudied. Gao et al. demonstrated that hydrogels containing Au NP‐stabilized liposomes displayed excellent antibacterial properties on S. aureus without skin toxicity to mice ( Figure 4 ). 45 In their research, the carboxyl‐modified Au NPs were absorbed onto the outer surfaces of cationic liposomes as stabilizers. The hydrogel formulation allowed for controllable viscoelasticity and tunable liposome release rate. The released Au NPs subsequently fuse with bacterial membranes in a pH‐dependent manner. In summary, the hydrogel formulation exhibited great promise for applications against various microbial infections. Furthermore, in order to obtain better antibacterial properties, some researchers fabricated bimetallic (i. e. , Ag and Au) hydrogel nanocomposites, which achieved the desired antibacterial activity. Varaprasad et al. prepared the dual‐metallic (Ag 0 –Au 0 ) nanoparticle‐loaded hydrogels through a green process with mint leaf extracts as the hydrogel networks, which exhibited significant antibacterial activity against Bacillus and E. coli. 46 Figure 4 Hydrogel containing nanoparticle‐stabilized liposomes for topical antimicrobial delivery. A) Schematic illustration of hydrogel containing nanoparticle‐stabilized liposomes for topical antimicrobial delivery. B) Fluorescence study of fusion interaction between AuC–liposome hydrogel and S. aureus bacteria. C) The toxicity evaluation of AuC–liposome hydrogel using a mouse skin model. Reproduced with permission. 45 Copyright 2014, American Chemical Society. 2. 1. 3 Other Metal Nanoparticle‐Loaded Hydrogels Apart from these commonly used metal nanoparticles, the antibacterial cobalt‐exchanged natural zeolite (ZEO)/PVA hydrogels were proved to possess antibacterial activity against E. coli. 47, 48 ZEO/PVA hydrogel with 0. 48 wt% and higher cobalt‐exchanged ZEO contents showed efficient antibacterial activities against G− bacteria (i. e. , E. coli and S. aureus ). Cu–SA hydrogels prepared through electrostatic extrusion were bactericidally effective against E. coli and MRSA ( Figure 5 ). 49 Cu–SA hydrogels, with higher Cu(II) loading (≈100 × 10 −6 m ), were produced by electrostatic extrusion using gelling solutions with Cu(II). The Cu–SA hydrogels exhibited immediate bactericidal effects against S. aureus and E. coli. Figure 5 Study and potential biomedical application of Cu–SA. Reproduced with permission. 49 Copyright 2016, IOP. Generally, metallic nanoparticles can attach to and destroy the integrity of bacterial membranes, leading to the leakage of bacterial contents, such as nucleic acids, through the outer membrane and peptidoglycan layer, resulting in the inhibition of protein synthesis. However, the mechanisms behind the antibacterial effects of metallic nanoparticles have not been confirmed. Despite being used in low concentrations, the toxicities of the metal‐based materials remain a major concern. Further studies are required to investigate the effects of particle size, morphologies, surface properties, associated signal transduction mechanisms, and applied concentration of metallic nanoparticles on antibacterial properties. Many of the aforementioned metals, including their alloys and the metal nanoparticles that are applied in modern medical biomaterials, need to be further explored. However, despite the previously discussed questions and challenges, hydrogels provide a convenient and controllable platform for the production of biocompatible functionalized metal nanoparticles. 2. 2 Metallic Oxide Nanoparticle‐Loaded Hydrogels In addition to the metal nanoparticle‐loaded hydrogels, metallic oxide nanoparticle‐loaded hydrogels also possess good antibacterial properties. The antibacterial mechanism of metallic oxide nanoparticles differs from metal nanoparticles. The photocatalysis is the main antibacterial mechanism of metallic oxide nanoparticles. 50 Under the ultraviolet irradiation of sunlight, large amounts of free radicals, namely hydroxyl radicals and oxygen radicals, are produced at the surface of metallic oxide nanoparticles. When the free radicals are exposed to microorganisms, the organic matter of the microorganisms are oxidized into carbon dioxide, therefore the metallic oxide nanoparticles can kill microorganisms in a relatively short amount of time. Among the various metallic oxides, ZnO is the most popular antibacterial agent. 51, 52, 53 ZnO NPs are widely used in many cosmetic materials because they exhibit antibacterial activity and non‐cytotoxicity at the appropriate concentrations. Sudheesh Kumar et al. developed CT hydrogel/ZnO composite bandages for wound healing and collagen deposition. 54 They are effective against both G+ and G− bacteria as well as high‐temperature resistant and high‐pressure resistant bacterial spores. 55 Similar to Ag NP‐loaded hydrogels, a composite bandage of SA hydrogel loading ZnO NPs prepared by Mohandas et al. exhibited enhanced swelling, blood clotting, and antibacterial activity. The hydrogel/ZnO NP composite bandage exhibited excellent antimicrobial activities against various strains of bacteria (e. g. , E. coli, S. aureus, Candida albicans, and methicillin‐resistant S. aureus ). When utilized on human dermal fibroblast cells, the composite bandage was nontoxic at low concentrations of ZnO. 56 The antibacterial hydrogel coatings made from ZnO NP‐incorporated poly( N ‐isopropylacrylamide) (PNIPAM) were demonstrated to be effective alternatives for biomedical device coatings. This composition of hydrogel exhibited bactericidal activity against E. coli. 57 Furthermore, the hydrogel coatings showed no cytotoxicity toward the mammalian cell line (3T3) over one week. Yadollahi et al. fabricated CMC/CuO (CMC, carboxymethyl cellulose) nanocomposite hydrogels via in situ formation of CuO NPs within swollen CMC hydrogels. The resultant hydrogels exhibited excellent antibacterial effects against both G+ and G− bacteria. 58 Archana et al. reported that the TiO 2 NP‐loaded CS–pectin composite hydrogel generated wound dressings with photoactive property, excellent biocompatibility, good antibacterial ability, and enhanced wound closure rate. 59 Inorganic antibacterial agent‐loaded hydrogels have relatively stable antibacterial properties and high temperature resistance. Unfortunately, their biocompatibilities are unsatisfactory for human implantation. As an alternate, the organic antibacterial agents are used in synthetic antibacterial hydrogel. Organic antibacterial agent is generally classified into small molecule antibacterial agents and polymer antibacterial agents. The hydrogel matrix can be composed of natural polymers and their derivatives, in particular ST, gelatin, CS, CMC, SA, as well as synthetic polymers including PVA and PVP. 3 Antibiotic‐Loaded Hydrogels Although discovered after antibacterial metal agents, antibiotics are undoubtedly the most common and effective antibacterial agents. 60 However, the drug‐resistant effect that bacteria possess has been the biggest obstacle in the development and applications of antibiotics. To overcome this, it is more promising and practical to minimize the dosage of conventional antibiotics rather than to explore new antibiotics. 61 Local antibiotic administration, by delivering the adequate bactericidal dose of antibiotics directly into the infected site without significantly overtaking the systemic toxicity level, has drawn increasing attention in recent years. 62 In biomedical research, fibers, beads, gels, and many other materials are used to deliver antibiotics. Hydrogels, a form of local administration matrix, offer a high surface area to volume ratio and structural controllability, such as porosity to mimic natural tissues. As a result, it is easy for hydrogels to selectively release their loaded drugs at desirable sites, 63, 64 while maintaining high water content and biocompatibility. 65 Some of the antibiotic‐loaded hydrogels are summarized in the following sections. 3. 1 Ciprofloxacin‐Loaded Hydrogels Ciprofloxacin (CIP) is a fluoroquinolone‐based antibacterial agent, with a broad antibacterial spectrum against both G+ and G− bacteria. The partition coefficient (log P, octanol–Tris model) of CIP is −1. 31 5. 66 This is the gold standard for various topical applications, such as for eye and skin infections. 66, 67 The antibacterial mechanism of CIP relies upon the blockage of bacterial DNA duplication by binding to the DNA gyrases and causing double‐stranded ruptures in bacterial chromosomes. Thus, the drug resistance to this antibiotic develops slowly. 68 The toxicity of CIP is dosage‐related and excessive doses can cause damage to tissues. Utilizing hydrogels as a local delivery system can sufficiently resolve this issue. CIP can be self‐assembled with a tripeptide ( d ‐Leu‐Phe‐Phe) and incorporated into antibacterial nanostructured hydrogels with high drug loading efficiency (DLE) and a prolonged release. 64 This CIP–peptide self‐assembled hydrogel showed high antimicrobial activity against S. aureus, E. coli, and Klebsiella pneumoniae. Furthermore, no cytotoxicity was found in hemolysis assays of red blood cells or in cultures of fibroblast cells. Two electrosynthesized polyacrylate hydrogels loaded with CIP prevented the Ti implant‐associated infections. The antibiotic‐modified hydrogel coatings had a long‐term release property, which exhibited antimicrobial activity against MRSA and good biocompatibility with MG63 human osteoblast‐like cells. 62 A hydrogel generated by the polymerization of 3‐aminophenylboronic acid with PVA for CIP loading was reported to facilitate wound healing in diabetes patients. 69 The hydrogel composite exhibited an ability to bind glucose and release CIP, which demonstrates the possibility of using it for wounds, particularly in diabetic patients. Colon‐associated diseases like constipation were reported to be treated with hydrogels containing laxative psyllium and CIP. The hydrogel with laxative action of psyllium and slow release of CIP exhibited a satisfactory therapeutic effect for treatment of diverticulitis. 70 A liposomal hydrogel containing CIP was reported to improve the maximum ocular availability in the cornea of albino rabbits. 71 Shi et al. conjugated CIP to the hydrogel network structure and obtained a composite hydrogel with ultraviolet‐triggered CIP release behavior. The composite hydrogel showed excellent antibacterial effects against MRSA. 72 3. 2 Gentamicin‐Loaded Hydrogels Gentamicin (GEN) is a traditional broad‐spectrum antibiotic used for the treatment of infections of the skin, soft tissues, and wounds, but its systemic toxicity (e. g. , kidney) and low plasma concentration remain a problem, which hinders its applications. Local administration of functional GEN hydrogels offers an efficient solution. Posadowska et al. fabricated an injectable drug delivery system, which consists of GEN‐loaded poly(lactide‐ co ‐glycolide) (PLGA) NPs embedded in the gellan gum hydrogel. The system was suitable for injection and was antibacterially active against Staphylococcus saprophyticus without affecting the bone forming cells. 73 Sa et al. developed a class of thermosensitive CS–GP hydrogels incorporating nanosized hydroxyapatite (nHA)/antibiotic GEN. The thermosensitive hydrogels were introduced into polymethylmethacrylate (PMMA) bone cement, resulting in an increased mineralization capacity and an enhanced antibacterial activity of the cement ( Figure 6 ). 74 Figure 6 Beneficial effects of biomimetic nHA/GEN‐enriched CS–GP hydrogel on performance of injectable PMMA. A) Synthetic process of PMMA‐based cements. B) Morphology of the GS–GP hydrogel. C) Antibacterial activity of samples by zone of inhibition test. Reproduced with permission. 74 Copyright 2015, Royal Society of Chemistry. GEN‐loaded PVA and PVA–AAm hydrogels crosslinked by Sterculia can be a form of potent antibacterial wound dressings due to their good biomedical properties, specifically blood compatibility, tensile strength, burst strength, water vapor permeability, and oxygen diffusion. 75, 76 Superabsorbent polysaccharide GEN hydrogels based on pullulan derivatives also present a broadened view about antibacterial hydrogels. The ability to expand to 4000% of its initial volume provides the hydrogels with a quick hemostatic ability and a capacity to prevent the wound bed from accumulation of exudates. 77 Phospholipid‐modified solid lipid microparticles encapsulating GEN were loaded into different polymer hydrogels. Among them, poloxamer 407 microgels displayed the most desirable properties specifically rapid antibacterial activity, in vitro diffusion‐dependent permeation, ability to spread, and appropriate viscosity. 78 These results indicated that the same drug can achieve different diffusion speeds on hydrogels due to the different matrices being employed. 3. 3 Vancomycin‐Loaded Hydrogels Clinically, vancomycin (VAN) is an antibiotic that is considered as the last form of defense against an infection. However, VAN‐resistant Enterococcus was recently discovered. 79 As mentioned above, utilizing hydrogels as a drug delivery system protects and enhances the effectiveness of VAN. Gustafson et al. developed a charged hydrogel as a carrier. The charged hydrogel, which was loaded with VAN over 500 µg mg −1 hydrogel, was able to control the VAN delivery and was used to combat the surgical site infections against MRSA ( Figure 7 ). 80 Development of an injectable gellan gum‐based PLGA NP‐loaded system, 81 injectable Pluronic–α‐CD supramolecular gels (CD, cyclodextrin), 82 and hydrogels consisting of thiolated chitosan crosslinked with maleic acid‐grafted dextran 83 provided new opportunities for antimicrobial research. The photo‐crosslinked methacrylated dextran and poly( l ‐glutamic acid)‐ graft ‐hydroxyethyl methacrylate (PGA‐ g ‐HEMA) hydrogels were studied 84 and both exhibited excellent antibacterial properties and desirable release capabilities. Figure 7 Controlled delivery of VAN via charged hydrogels. A) VAN release from charged hydrogels. B) Best fit of data as calculated by phenomenological mathematical model described in text. C) Comparison of fitted model to obtained data points for VAN‐loaded and ‐unloaded (0% and 50% sodium methacrylate (SMA)) hydrogels. D) Zone of clearing assay comparing 0% and 50% SMA. Reproduced from ref. 80. 3. 4 Other Antibiotic‐Loaded Hydrogels In addition to the aforementioned antibiotic‐loaded hydrogels, other antibiotic‐loaded hydrogels were developed as well. Ampicillin sodium‐loaded PVA–SA hydrogel exhibited strong antibacterial property to both G+ and G− bacteria and improved hemolysis. 85 Cephalosporin is a widely used neutrapen‐resistant and broad‐spectrum β‐lactamase‐based antibiotic. 86 Hydrogels containing cefditoren pivoxil achieved gastroretentive effect 87 and methoxy poly(ethylene glycol)‐ co ‐poly(lactic acid‐ co ‐aromatic anhydride) hydrogels containing cefazolin offered a stable release profile without an initial burst release and effective antibacterial properties against E. coli. 88 Levofloxacin‐loaded hyaluronic acid hydrogels were reported to be able to attack bacteria within the cells for both S. aureus and P. aeruginosa strains. 61 A hydrogel based on (−)‐menthol, which is a traditional cooling compound followed by an amino acid derivative through an alkyl chain, provided an innocuous environment to living cells and was able to deliver lincomycin to the local infection site. 89 Furthermore, the hydrogel composites were completely innoxious to HeLa cells. Doxycycline (DOX) was also loaded in situ into a thermosensitive hydroxypropyl‐β‐cyclodextrin (HP‐β‐CD) hydrogel for ophthalmic delivery. 90 The release of DOX from hydrogel followed a zero order equation, suggesting that it occurred due to corrosion of the poloxamer hydrogel. The liposomes‐in‐hydrogel delivery systems can control and prolong the release of mupirocin (MIP). MIP is a promising antibiotic that is well tolerated in topical administration with minimized side effects and leads to improved burn therapy. 91 4 Biological Extract‐Loaded Hydrogels Biological extracts include extracts from plants and animals. 92 Seaweed extract‐based hydrogel was reported as an antibacterial wound dressing. 93 PVA composite hydrogels based on combinations of agar and carrageenan have been proved to be useful as wound dressings in the treatments of burns, nonhealing ulcers of diabetes, and other external wounds. 94 Although some studies have stated that SA does not display antibacterial properties, it can be an ideal material for wound dressings due to its morphology, fiber size, porosity, degradation, and swelling ratio. 93, 95 Allicin–CS complexes were proved to be active against bacteria involved in spoilage and can be used as an antibacterial agent in foods. 96 Curcumin (CUR), a nontoxic and bioactive agent found in turmeric, has been applied for centuries as a remedy to various ailments. 97 However, its applications were limited by its low aqueous solubility and poor bioavailability. As a result, hydrogels incorporated by CUR nanoparticles were developed. Ag NP–CUR loaded hydrogels utilized for wound dressings were reported to exhibit good antibacterial property and sustained release, which indicated enormous therapeutic values. 98, 99 SA hydrogels encapsulated with essential oils, such as lavender, thyme oil, peppermint, tea tree, rosemary, cinnamon eucalyptus, and lemongrass, were reported to be qualified as disposable wound dressings due to the distinctive antibacterial properties of essential oils. 100 Among the biological extracts from animals, honey was the most easily acquired. Honey showed an antimicrobial activity in the management of various wounds. 101 CMC hydrogels incorporated with propolis honey were prepared by gamma radiation to produce a functional wound dressing. 102 Hydrogel contact lenses incorporated with lysozymes derived from normal tears exhibited remarkable antibacterial activity due to the inherent antibacterial property of lysozymes. 103 Vitamin E (VitE) is an important antioxidant and biodegradable extract. Hydrogels of VitE‐functionalized polycarbonates for antibacterial applications displayed an excellent compatibility with human dermal fibroblast. It can be loaded with cationic polymers and/or fluconazole at minimum biocidal concentrations to kill bacteria and fungi. 104 Polysaccharides with antibacterial ability are often natural macromolecules or their derivatives, such as ST and CS, which are frequently used for the preparation of hydrogels because of their nontoxicity, biodegradability, biocompatibility, and abundance in nature. 105, 106 Some of these polysaccharides have inherent antibacterial activity. Among them, CS is the most popular polysaccharide. CS has a wide antibacterial spectrum and high killing rate against G+ and G− bacteria while displaying a low toxicity toward mammalian cells. 107 CS can dissolve in weakly acidic solution and release NH 2 +, then bind with negatively charged macromolecules on the microbial cell surface to achieve bacterial stasis. 108 The polymers mainly composed of CS and semi‐interpenetrating carboxymethyl chitosan (CMCS)/polyacrylonitrile hydrogels were reported to present good antibacterial activity when the CMCS content was increased. 109 Hydrogel coatings prepared by electrophoretic codeposition of CS/alkynyl CS exhibited better antibacterial activities than pure CS hydrogel. 110 CS‐grafted polymer‐based hydrogels containing mica nanocomposite produced a rougher surface while maintaining antibacterial activity. 111 These biological extracts are easy to obtain, handle, possess excellent biocompatibilities, and good antibacterial properties, making them promising antibacterial biomaterials. 5 Synthetic Antibacterial Drug‐Loaded Hydrogels Synthetic antibacterial drugs discussed here refer to the nitroimidazoles, sulfanilamide groups, and other frequently used drugs, but do not include semisynthetic antibiotics nor biological extracts. Although the special chemical structures benefit synthetic drugs significantly, they carry risks and damages to the normal tissues as well. A stable and safe delivery system for them is necessary. Hydrogel composed of dextrin and PAA was utilized for the delivery of ornidazole, which is a nitroimidazole‐derived antibacterial drug used for the digestive system. It showed effects on anaerobic bacteria and amoeba 112 with pH‐ and temperature‐controllable release profiles. 113 Moreover, the hydrogel with degradable characteristics showed no cytotoxic behavior toward human mesenchymal stem cells. Hydrogels based on dextrin grafted with poly(2‐hydroxyethyl methacrylate) (PHEMA) were also good candidates for the orally administered drug delivery system in the colon region. 65 CS/gelatin/β‐GP hydrogel containing metronidazole was tested as an injectable form for periodontal infection. It was able to maintain the release of metronidazole in concentrations that were effective for killing pathogenic bacteria of Clostridium sporogenes. 114 PAA–CS composite hydrogels containing tinidazole (TIN) and theophylline also have been studied to control and sustain TIN and theophylline delivery. 115 Simply put, in the presence of CS, the acrylic acid and N′ ‐methylene bis‐acrylamide were crosslinked by radical copolymerization to synthesize the composite hydrogels. CHX is considered to be a promising antibacterial agent that possesses a broad antibacterial spectrum including both G+ and G− bacteria. 116 CHX‐contained poly(ethylene glycol)‐ block ‐poly( l ‐lactide) nanoparticles were loaded in hydroxyethyl cellulose hydrogel, allowing the hydrogel system to enhance antibacterial activity against Enterococcus faecalis for root canal system disinfection. 117 CS‐HTCC/GP‐0. 1% CHX (CS, quaternized CS, and α, β‐GP loading with 0. 1% CHX (w/v)) thermoresponsive hydrogels showed an excellent antibacterial effect against Porphyromonas gingivalis, Prevotella intermedia, and Actinobacillus actinomycetemcomitans. 118 Chlorhexidine diacetate‐contained poly(2‐hydroxyhexyl methacrylate‐ co‐N‐ isopropylacrylamide) hydrogels are a promising thermoresponsive and antibacterial biomaterial (e. g. , Staphylococcus epidermidis ( S. epidermidis )). 119 Wound dressings composed of OCT‐loaded nanocellulose were proven to possess antibacterial activity with minimized side effects. 120 This OCT‐loaded nanocellulose exhibited a slower OCT release rate of up to 96 h, which demonstrated high biocompatibility in human HaCaT keratinocytes and antimicrobial activity against S. aureus. PHEMA‐conjugated β‐CD or directly crosslinked HP‐β‐CD hydrogels were applied to load TSC, an antibacterial drug used in ophthalmic diseases for fabricating antibacterial soft contact lenses. 121 Cetylpyridinium chloride‐immobilized PVA hydrogel offered a sustained release profile for wound therapy. 122 Chloramine‐T and sulfadiazine sodium coloaded hydrogels composed of PVA, PVP, and glycerin showed an excellent swelling capacity, which accelerated the wound healing with an antibacterial effect. 123 A PVP–iodine hydrogel was found to enhance the epithelialization and reduce the loss of skin grafts in wound therapy. 124 Poly( N ‐hydroxyethyl acrylamide) (PHEAAm)/SAL hydrogels provided both antibacterial and antifouling functions. 125 This research showed that SA‐treated PHEAAm hydrogels could inhibit both G+ E. coli RP437 and G− Staphylococcus epidermidis. Alginate hydrogel spheres releasing ITZ achieved long‐term antibacterial activity by improving the alkali and heat resistance abilities. 126 Evidence indicates that the synthetic drug‐loaded hydrogels could achieve desirable drug delivery as well as avoid risks and minimize side effects. It was of equal importance that hydrogels offer potential for widespread application of antimicrobial and antiviral agents. 6 Carbon Material‐Loaded Hydrogels Some carbon materials combined with hydrogels were developed for inhibition of bacteria. Venkatesan et al. prepared CT–carbon nanotube hydrogels by freeze‐lyophilization method, which exhibited antimicrobial activity against S. aureus, E. coli, and Candidatropicalis. 127 Composite CT/active carbon hydrogels prepared by ammonia vapor treatment showed an potential application to be used as wound dressings. 128 Graphene oxide (GO) also has immense potential in the antibacterial field. A facile one‐pot method was used to synthesize GO‐based hydrogels (i. e. , benzalkonium bromide/GO hydrogel and benzalkonium bromide/polydopamine/reduced GO hydrogel), which exhibited strong antibacterial activity against G+ and G− bacteria. 129 Zeng et al. prepared an Ag/reduced GO hydrogel by a facile hydrothermal reaction, which consisted of two parts: (i) a controlled porous reduced GO network and (ii) well‐dispersed Ag NPs. 130 The antibacterial hydrogels were generated by crosslinking the Ag/graphene composites with acrylic acid and N, N′ ‐methylene bisacrylamide, which exhibited good antibacterial abilities against E. coli and S. aureus. The excellent biocompatibility, high swelling ratio, and good extensibility were also found in this hydrogel system. 131 7 Hydrogels with Inherent Antibacterial Activity Hydrogels with inherent antibacterial activity discussed here refer to the hydrogels that contain antibacterial components. 132 These hydrogels, with inherent antibacterial activity, were developed in recent years as effective antibacterial agents with little or even no side effects compared to the traditional ones. The main forms of these hydrogels are discussed below. 7. 1 Hydrogels with Antibacterial Polymers Antibacterial polymers include nonstimulated antibacterial polymers and potential antibacterial polymers. The most common nonstimulated antibacterial polymers have certain components in their structures that are important for antibacterium. The hydrogels composed of thermoresponsive PNIPAM and redox‐responsive polyferrocenylsilane macromolecules exhibited strong antibacterial activities while maintaining high biocompatibilities. 133 The redox‐induced formation of hydrogel–Ag composites showed a good antimicrobial activity against E. coli. pH‐sensitive and thermal‐sensitive hydrogels based on HEMA and IA copolymers possess potential biomedical applications, especially for skin treatments and wound dressings. 134 P(HEMA/IA) could block the entry of S. aureus and E. coli into hydrogel dressing. In addition, no evidence of cell toxicity or considerable hemolytic activity was observed in an in vitro study of P(HEMA/IA) biocompatibility. Hydrogels prepared by the photopolymerization of PEG diacrylate and a monomer containing ammonium salt (RNH 3 Cl) demonstrated both antibacterial and antifouling properties. 135 The potential antibacterial polymers are a class of polymers that could be converted to become antibacterial under certain conditions, such as exposure to light. The photodynamic poly(2‐hydroxyethyl methacrylate‐ co ‐methyl methacrylate) (P(HEMA‐ co ‐MAA)) copolymers crosslinked by porphyrin were reported to be promising for the prevention of intraocular lens‐associated infectious endophthalmitis. 136 Another photodynamic PHEMA‐based hydrogel also exhibited light‐induced bactericidal effect through the release of nitric oxide. 137 These antibacterial polymers provided not only antibacterial materials but also responsive delivery and release methods. 7. 2 Hydrogels with Antibacterial Peptides Antibacterial peptides (AMPs) are an abundant and diverse group of molecules produced by many types of tissues and cells in plant and animal species. 4 They are recognized as a possible source of panacea for the treatment of antibiotic‐resistant bacterial infections. 138, 139 AMPs have strong antibacterial activities against a very broad spectrum of microorganisms, including G+ and G− bacteria, fungi, and even viruses. 140 It is generally accepted that the antibacterial mechanism of AMPs is that they associate with the membrane leading to disruption of the bacterium ( Figure 8 ). 138 Bai et al. designed an amphiphilic peptide A 9 K 2 that could effectively inhibit both G+ and G− bacterial strains. 141 The enzymatic A 9 K 2 hydrogel possessed good biocompatibility and showed excellent selectivity by favoring the adherence and spreading of mammalian cells. Baral et al. prepared an antibacterial dipeptide, which showed excellent antibacterial activity against G− bacteria ( E. coli and P. aeruginosa ), as well as high biocompatibility with human red blood cells and human fibroblast cells. 142 Wang et al. fabricated enzymatically crosslinked ε‐poly‐ l ‐lysine hydrogels, which exhibited efficient antibacterial activity against both G+ and G− bacteria. 143 Peptide‐based β‐hairpin hydrogels were reported with MAX1 peptides by Salick et al. in 2007 144 and with arginine‐rich peptides by Veiga et al. in 2012. 145 Both of them are self‐assembly peptides that exhibited potent antibacterial activity. Figure 8 Antimicrobial peptides: pore formers or metabolic inhibitors in bacteria. A) Barrel‐stave model, B) toroidal model, and C) carpet model of antimicrobial peptide‐induced killing. D) Mode of action for intracellular antimicrobial peptide activity. Reproduced with permission. 138 Copyright 2005, Nature. The self‐assembled peptides comprised of two antibacterial peptides (KIGAKI) 3 —NH 2 and a central tetrapeptide linker can maintain a stable β‐hairpin structure. 146 Polylysine, a popular AMP reported by Zhou et al. , has been applied in photopolymerized antibacterial hydrogels, which generated promising coatings for medical devices and implants. 147 In addition to antibacterial peptide maximin‐4‐loaded PHEMA hydrogels, 148 l ‐cysteine‐ and silver nitrate (AgNO 3 )‐loaded hydrogels were proved to possess antibacterial activity against Staphylococci, Bacilli, Escherichia, and P. aeruginosa strains. 149 Although the hydrogels with AMPs displayed some disadvantages, such as tissue toxicity and hemolysis, 150 they are still attractive due to their increased antibacterial ability and biocompatibility when compared to synthetic drugs with similar structures ( Figure 9 ). 151 Extensive research is being carried out to improve the biocompatibility. Specifically, a hydrogel of cell adhesive polypeptides and PEG with inherent antibacterial activity was developed by Song et al. as a potential scaffold for cutaneous wound healing. 152 The hydrogel formed by crosslinking poly(Lys‐Ala) with 6‐arm poly(ethylene glycol)‐amide succinimidyl gluta exhibited significant antibacterial activity against S. aureus and E. coli. Moreover, a protein anchor developed to immobilize functional protein to poly(ethylene glycol) diacrylate microspheres in 2013 proved to be a fascinating method to maintain therapeutic efficacy without toxicity. 153 Figure 9 Comparative surface antimicrobial properties of synthetic biocides and human apolipoprotein E‐derived antimicrobial peptides. A) Dual‐polarization interferometer associated with mass uptake of ApoEdpL‐W (a peptide derivative of human apolipoprotein E) and polyhexamethylene biguanide (PHMB). B) CLSM images showing the growth of P. aeruginosa and L929 cell line exposed to a PHEMA hydrogel that has been previously exposed to (a) phosphate‐buffered saline (PBS), (b) ApoEdpL‐W, (c) CHX, (d) PHMB, and (e) triclosan. Reproduced with permission. 151 Copyright 2013, Elsevier. 7. 3 Amphoteric Ion Hydrogels Amphoteric ion hydrogel works similarly with AMPs. The electrostatic interactions facilitate the bindings between the polymers and anionic bacterial membranes, resulting in the physical destruction of membrane structures and cell death. 154 QA compound is one of the most famous antibacterial materials. Antibacterial hydrogels containing QA groups synthesized by Zhou et al. through a facile thiol–ene “click” reaction exhibited excellent antibacterial efficacy against MRSA. 155 A cellulose (CEL)‐based hydrogel containing QA groups were prepared by Peng et al. via a simple chemical crosslinking and showed strong antibacterial ability against Saccharomyces cerevisiae. 156 He and co‐workers designed photo‐crosslinked polymer ionic hydrogel films incorporating QA chloride groups, which exhibited antibacterial ability against E. coli with almost 100% killing efficiency. 157 PEG hydrogel networks incorporated by polycarbonate via Michael addition by Liu et al. were reported to have more than 99. 99% efficiency against MRSA. 158 It is worth mentioning that antimicrobial and nonfouling hydrogel did possess significant skin toxicity or hemolytic activity. When combined with hydrogels, the amphiphiles perform the same as the AMPs. Polyampholytic hydrogels with high antibacterial activity can exhibit high water absorbency. 157 Potent antibacterial hydrogels based on anti‐inflammatory N ‐fluorenyl‐9‐methoxycarbonyl amino acid/peptide‐functionalized cationic amphiphiles exhibited efficient antibacterial activity against both G+ and G− bacteria. 159 To achieve a bifunctional hydrogel with both antibacterial and antifouling capacities, a zwitterionic hydrogel was conjugated with an antibacterial agent, SAL. The resultant hydrogel can reach one‐SAL‐per‐monomer DLE while maintaining the nonfouling property at protein and bacteria levels. 160 To improve the biocompatibilities of amphiphiles, Dutta et al. developed cholesterol‐based amino acid containing hydrogels. Ag NPs were synthesized in situ and the amphiphile–Ag NPs soft nanocomposite exhibited notable antibacterial property. 161 In addition to the hydrogels developed against normal G+ and G− bacteria, an anti‐mycobacterial supramolecular hydrogel based on amphiphiles was developed by Bernet et al. It retains specific and chain‐length dependent antibacterial and anti‐mycobacterial activity while showing no antiproliferative and negligible cytotoxic effects. 162 8 Hydrogels with Synergetic Effects Hydrogels with synergetic effects refers to hydrogels containing two or more antibacterial agents, which can enhance antibacterial effects. Metal nanoparticles and antibiotics are commonly reported to be incorporated into hydrogels together to obtain synergetic effects. In addition, as described above, the combined utilization of Ag NPs and reduced GO is also a common mechanism to enhance antibacterial effects. 130, 131 8. 1 Synergetic Effective Hydrogels Containing Metal Nanoparticles Metal nanoparticles in synergetic effective hydrogels were mainly Ag NPs. Ag NPs can be loaded into synthetic amphiphiles, amino acids, and even biological extract‐based hydrogels. 163 Ag NP composite systems are more suitable for biomedical applications because of their good biocompatibility with biological molecules, cells, and tissues. 164 Amphiphilic hydrogels with in situ‐synthesized Ag NPs reported by Dutta et al. exhibited improved biocompatibility and antibacterial efficacy, which are promising in applications of biomedicine and tissue engineering. 165 They also reported self‐assembly amino acid‐based amphiphilic hydrogels containing in situ‐synthesized Ag NPs, which exhibit lethal bactericidal activity toward both G+ and G− bacteria while maintaining the growth of mammalian cells, which remained unaffected on the surface. 166 As reported in 2011, the hydrogels based on l ‐cysteine and AgNO 3 were used to prepare bactericidal fibers and fabrics, and the Ag + glutathione hydrogel exhibited improved cytocompatibility. 167 Furthermore, the composite hydrogel offered more possibilities in potential biomedical applications like wound dressings for burn victims. For other combinations, the antibacterial efficacy of these hydrogel nanocomposites was largely enhanced by the incorporation of both Ag NPs and CUR. The entrapped Ag NPs and CUR molecules were constantly released, thus the hydrogel nanocomposites could be applied in enormous prolonged severe infection therapies. 99 8. 2 Synergetic Effective Hydrogels Containing Antibiotics Hydrogels containing antibiotics exhibited more potent antibacterial property and biocompatibility when combined with other antibacterial materials. A ZnO/GEN–CS composite gel with a controlled release profile was reported to be promising in the treatment of wounds. The composite gel of ZnO, GEN, and CS significantly improved the minimum inhibitory concentrations (MICs) against G+ and G− bacteria compared with the GEN control group ( Figure 10 ). 168 Bacterial CEL polymers grafted by RGDC (R: arginine; G: glycine; D: aspartic acid; C: cysteine) and GEN offered an inspiring and effective antibacterial composite. 169 CIP was combined with different materials, from metal nanoparticles to amphiphiles, to develop synergetic effective antibacterial hydrogels. Antibacterial nanostructure hydrogels containing self‐assembled CIP and tripeptide were reported by Marchesan et al. and played a significant role in the design of cost‐effective nanomaterials for prolonged drug release. 64 The magnetically mediated release of CIP‐loaded super‐paramagnetic nanocomposites provided the synergetic effective hydrogels with an effective drug release approach. 68 The quaternized gellan–CS particles were demonstrated to be potent in sustained release applications of CIP. 170 In addition, tetracycline hydrochloride/Ag NP composite hydrogels were developed and inhibited bacteria in a simulated colon environment. 171 All these synergistically effective composite hydrogels offer possible approaches to minimize the dosage of antibiotics required. Figure 10 A controlled release ZnO/GEN–CS composite with potential applications in wound treatment. A) (a) Zn–CS gel (12:1) cut in shapes, (b) close view of a ZnO–CS cube, and (c) ZnO–CS cube after three months in water. B) Stability of ZnO–CS gel in laboratory atmosphere. C) GEN release from ZnO/GEN nanopowder and ZnO/GEN–CS gel. D, E) Graphic representation of MICs and inhibition diameters of GEN and ZnO/GEN–CS. Reproduced with permission. 168 Copyright 2014, Elsevier. 9 Summary and Prospects Hydrogels as antibacterial biomaterials can be an alternative and amenable solution to traditional antibiotic treatments. Controlled and prolonged release, local administration, stimulated switch on–off release, enhanced mechanical strength, and improved biocompatibility are all important advantages that a broad diversity of hydrogels can provide and that is exactly what antibacterial biomaterials currently require. Antibacterial hydrogels can be widely applied in the field of wound dressings, urinary tract coatings, catheter‐associated infections, gastrointestinal infections, osteomyelitis, and contact lens. Based on current research regarding the development and application of antibacterial hydrogels, most researchers have been investigating hydrogels composed of polysaccharides, PEG, or other hydrophilic polymers in combination with a variety of bactericidal substances. For hydrogels to be utilized therapeutically, biocompatibility and biodegradability are the utmost important requirements. Furthermore, as a drug carrier, hydrogels should have high DLE. In regards to the side effects, there was no inflammation in the adjacent connective tissue after biodegradation of the hydrogels. Based on the above factors, intelligent hydrogel platforms should be exploited to overcome the challenges of local antibacterial drugs. Although inorganic antibacterial agents like Ag NPs have good antibacterial properties, the unsatisfactory biocompatibility and dosage dependency limit their applications. Moreover, many drug‐resistant bacteria have evolved because of the misuse of traditional antibiotics and other antibacterial drugs. The special antibacterial mechanism of antibacterial peptides provides a solution to the issue of bacterial resistance. Therefore, fabricating antibacterial peptide hydrogels through the incorporation of antibacterial peptides with hydrogels will be the key to overcome these limitations. Antibacterial hydrogels will finally be able to conquer the vast issues of traditional therapies. Antibacterial biomaterials, their unique combinations, and the approaches currently being developed will provide a promising future for anti‐infection treatment. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201700550
| 2,017
|
Advanced Science
|
3D‐Bioprinted Osteoblast‐Laden Nanocomposite Hydrogel Constructs with Induced Microenvironments Promote Cell Viability, Differentiation, and Osteogenesis both In Vitro and In Vivo
|
Abstract An osteoblast‐laden nanocomposite hydrogel construct, based on polyethylene glycol diacrylate (PEGDA)/laponite XLG nanoclay ([Mg 5. 34 Li 0. 66 Si 8 O 20 (OH) 4 ]Na 0. 66, clay )/hyaluronic acid sodium salt (HA) bio‐inks, is developed by a two‐channel 3D bioprinting method. The novel biodegradable bio‐ink A, comprised of a poly(ethylene glycol) (PEG)–clay nanocomposite crosslinked hydrogel, is used to facilitate 3D‐bioprinting and enables the efficient delivery of oxygen and nutrients to growing cells. HA with encapsulated primary rat osteoblasts (ROBs) is applied as bio‐ink B with a view to improving cell viability, distribution uniformity, and deposition efficiency. The cell‐laden PEG–clay constructs not only encapsulated osteoblasts with more than 95% viability in the short term but also exhibited excellent osteogenic ability in the long term, due to the release of bioactive ions (magnesium ions, Mg 2+ and silicon ions, Si 4+ ), which induces the suitable microenvironment to promote the differentiation of the loaded exogenous ROBs, both in vitro and in vivo. This 3D‐bioprinting method holds much promise for bone tissue regeneration in terms of cell engraftment, survival, and ultimately long‐term function.
|
1 Introduction Traditional strategies for bone tissue engineering are based on the facilitation of cell growth into engineered interconnecting scaffolds to generate a functional tissue construct for the reestablishment of structure and function in damaged bone tissues. 1 However, the realization of the desired levels of cell deposition and cell distribution in 3D scaffolds remains a great challenge. Recent developments in the tempospatial specific 3D‐bioprinting of cells and inks show promise for a new approach in bone tissue engineering. 2 With this technology, delicately tailored cell‐laden tissue constructs have been reportedly constructed to regenerate bone tissue. Zhang and co‐workers recently reported on bone mesenchymal stem cells (BMSCs)‐laden gelatin/sodium alginate/carboxymethyl chitosan hydrogel scaffolds prepared through a 3D‐bioprinting method. The prepared scaffolds exhibited good mechanical properties and favorable cytocompatibility, with cell viability of over 85% postprinting. 3 Neufurth et al. encapsulated bone‐related SaOS‐2 cells into a biologically inert sodium alginate matrix and used a second layer containing polyP·Ca 2+ to cover the bioprinted cell laden scaffold. The encapsulated cells exhibited a good proliferation rate and mineralization ability. 4 These previous studies were mainly devoted to imitating the native structure of bone tissues and improving the viability of 3D‐bioprinted cells in the short term. However, whether or not the 3D‐bioprinted cells can realize the functional features of bone tissue after in vivo implantation still remains unknown. Indeed, the long‐term in vivo evaluation of most of the previously reported cell‐laden scaffolds for bone regeneration is deficient and limited, due to the lack of ideal bio‐inks for 3D‐bioprinting to favorably support cell growth and development both in the short and long terms. In extrusion‐based printing, the materials which can be used as bio‐inks for bone tissue regeneration should satisfy the following basic requirements: (1) extrusion properties—suitable rheology for 3D‐bioprinting; (2) stability—the printed scaffolds should not collapse before solidification or crosslinking; (3) good biocompatibility and porosity—high cell viability after printing, with optimal avenues for the adequate diffusion of nutrients and oxygen to facilitate cell proliferation or differentiation in the short term; (4) appropriate mechanical properties; and (5) osteogenic capability to promote differentiation and new bone formation in the long term. Frequently used bio‐inks such as gelatin, 5 alginate, 6 gelatin/alginate, 4 gelatin/alginate/chitosan, 3 and poly(ethylene glycol) (PEG) dimethacrylate/gelatin 7 that have been used in the fabrication of cell‐laden scaffolds have shown high biocompatibility for cell viability in bone repair. However, they lack osteogenic capability to promote cell differentiation and new bone formation in the absence of growth factors, which represents a clear limitation for their successful application. Pati et al. first attempted to develop novel decellularized extracellular matrix (dECM) bio‐inks which could provide an optimized microenvironment and was conducive to the growth of 3D structured tissue in the long term. Their potential application for adipose, cartilage, and heart tissue regeneration was explored, but bone tissue was not mentioned. 8 Moreover, the inducement properties of this bio‐ink were only validated in vitro and the tedious preparation process and expensive cost further limit its application prospects. We report herein on the development of a novel biodegradable bio‐ink comprised of polyethylene glycol diacrylate (PEGDA) and laponite XLG nanoclay, which was photo‐crosslinked to form a stable gel (PEG–Clay) to support the printing process, facilitate optimal cell growth and function through the delivery of nutrients and oxygen, and promote osteogenesis due to the induced microenvironment forming by the released magnesium ions (Mg 2+ ) and silicon ions (Si 4+ ). Simultaneously, another bio‐ink, hyaluronic acid sodium salt (HA) with encapsulated cells, was designed to improve cell viability, distribution uniformity, and deposition efficiency. The two bio‐inks, via a two‐channel 3D‐bioprinting method, were alternately extruded to fabricate primary rat osteoblast‐laden (ROB‐laden) nanocomposite hydrogel constructs for bone regeneration. We evaluated the proliferation and differentiation behavior of the loaded cells and the osteogenic capability of ROB‐laden nanocomposite hydrogel constructs both in vitro and in vivo. 2 Results 2. 1 Characterization of Bio‐Inks A and B The 3D‐bioprinting of the cell‐laden constructs using two bio‐inks was performed over a few steps ( Figure 1 ). The two bio‐inks, denoted bio‐ink A and bio‐ink B, were prepared first. A PEG–Clay prehydrogel solution was designed for bio‐ink A containing photo‐crosslinkable PEGDA and the nanoclay, while HA was used to encapsulate cells as bio‐ink B. We synthesized biodegradable photo‐crosslinkable PEGDA with two classical molecular weights, M w = 4K and 10K, using acryloyl chloride according to a method previously reported. 9 1 H NMR spectra (Figure S1, Supporting Information) confirmed the successful formation of PEGDA4K and PEGDA10K. Then, the nanoclay was added to a 20% PEGDA (w/v) aqueous solution, resulting in a thickening of the viscosity from liquid form to sticky, rendering it suitable for room temperature 3D‐bioprinting ( Figure 2 A). As no obvious difference could be observed between the viscosities of the PEGDA4K and PEGDA10K solutions at 20% concentration, rheological analysis of 20%PEG4K–Clay prehydrogels with different nanoclay concentrations was performed. When less than 7% nanoclay (w/v) was added to the 20% PEGDA solution, the printed scaffold could not maintain a fixed shape with fused microstructure. When the nanoclay concentration exceeded 7%, the viscosity of the PEG–Clay prehydrogel solution was too high to be printed under lower pressure at room temperature. Moreover, the viscosity of the 20%PEG–7%Clay prehydrogel solution exhibited similar rheological behavior to 20% HA (w/v). Therefore, 7% nanoclay was selected as the optimal concentration to add to the 20% PEGDA aqueous solutions. UV light exposure led to the gelation of the PEG–Clay prehydrogel solution and Fourier transform infrared (FTIR) spectra confirmed the successful formation of the PEG–Clay hydrogels (Figure S2, Supporting Information). Thermogravimetric analysis (TGA) (Figure S3, Supporting Information) indicated that approximately 25% nanoclay was incorporated into the PEG–Clay hydrogels, in both 20%PEG4K–7%Clay and 20%PEG10K–7%Clay, which is very close to the original clay concentration (26%, relative to the total mass of PEGDA and nanoclay), and further confirmed the successful incorporation of nanoclay into the hydrogel matrix. Transmission electron microscopy (TEM) images highlight the good exfoliation and dispersion of nanoclay within the PEG4K–Clay hydrogel, with no obvious agglomeration observed, Figure S4 in the Supporting Information. The well dispersed nanoclay deposits contribute to the attractive mechanical properties of the PEG–Clay hydrogels over pure PEG hydrogels. As presented in Figure 2 B, C, the compression strength values recorded for the pure 20%PEG4K and 20%PEG10K hydrogels were 0. 096 and 0. 332 MPa, respectively. The compression strength increased upon incorporation of the nanoclay, with values of 0. 468 and 0. 976 MPa recorded for the 20%PEG4K–7%Clay and 20%PEG10K–7%Clay hydrogels, respectively. Moreover, the compression modulus increased from 9 and 49 to 22 and 90 kPa for the pure 4K and 10K PEG hydrogels and their nanoclay incorporating cousins, respectively. These results confirm that adding nanoclay to the hydrogel systems not only imparted printability to the PEG–Clay prehydrogel solutions but also enhanced the mechanical properties of the hydrogels, thus increasing the potential of the PEG–Clay hydrogels as component supporting materials for bone tissue regeneration. Figure 1 Stepwise illustration depicting the two‐channel bio‐ink system and method used for 3D‐bioprinting and the in vivo experiments. The composition of bio‐ink A and bio‐ink B (A), 3D‐bioprinting process of the ROB‐laden constructs using a two‐channel method (B) and the diagram for the in vivo experiments (C). Figure 2 Rheology characteristics of 20%PEG4K–Clay prehydrogel solutions with different nanoclay concentrations and the 20% HA solution A). Compression strength B) and compression modulus C) of PEG hydrogels and PEG–Clay hydrogels. The effect of UV exposure on ROBs in α‐MEM medium and encapsulated in 20% HA solution D). Asterisks (*) denote significant differences (* p < 0. 05). HA, a highly biocompatible material, was used without further crosslinking as a sacrificial material, to facilitate the gradual and uniform release of encapsulated cells within the scaffold 24 h after printing. Furthermore, the HA layer protects cells from UV damage during the PEG–Clay hydrogel crosslinking procedure. Aqueous 20% HA is similar in viscosity to the 20%PEG–7%Clay prehydrogel solution (Figure 2 A) and was suitable for the 3D‐bioprinting of cells. The choice of concentration is important, not only for printability but also the primary hydrogel structure should not be impacted by the addition of the second material. 10 Moreover, when the HA concentration exceeds 20%, the printing pressure needed for successful printing would surpass 100 kPa, which would negatively impact on the viability of the printed cells. 8, 11 Due to the susceptibility of cells to UV‐related damage, 12 we evaluated the ability of HA to act as a protective barrier against UV‐damage (Figure 2 D). ROBs directly dispensed in alpha minimum essential medium (α‐MEM) medium were acutely damaged by UV‐irradiation over time. However, when encapsulated in 20% HA and exposed to UV‐irradiation, no obvious damage was observed, even when the irradiation time was as long as 30 min. These results indicate that bio‐ink B (HA) was an ideal carrier for loading cells within the scaffold. To summarize, 20%PEG–7%Clay and 20% HA encapsulated cells were chosen as bio‐ink A and bio‐ink B, respectively, to construct cell‐laden scaffolds for further investigation in this study. 2. 2 3D‐Bioprinting of ROB‐Laden Constructs Using a Two‐Channel Method The ROB‐laden constructs were layer by layer fabricated through a two‐channel method, as depicted in Figure 1. The first layer as a stable base was built with only bio‐ink A. From the second layer to the end layer, bio‐ink A and bio‐ink B were alternately extruded to build a cell‐laden construct. Movie S1 in the Supporting Information shows the actual printing process and Figure 3 A shows the final structure of the scaffold. As the PEG–Clay prehydrogel solution could maintain its structure at room temperature, it was possible to first print the scaffolds individually and then photopolymerize them together. As photo‐crosslinking time increases, the degree of PEG–Clay scaffold crosslinking naturally increases. 13 In order to minimize possible damage and maintain the viability of the ROBs and ensure the formation of the PEG–Clay hydrogels in our constructs, 10 min UV‐irradiation was chosen. Figure 3 The final structure of 3D‐bioprinted scaffolds prepared by the two‐channel method, PEG–Clay was stained with gentian violet and HA was stained with Rhodamine A). Live/dead assay of ROBs 1 d after printing within PEG4K–Clay scaffolds B), I) viable ROBs (green, calcein AM), II) dead ROBs (red, EthD‐1), and III) merged. The status of ROBs after 3D‐bioprinting (PEG4K–Clay‐P) or traditional seeding (PEG4K–Clay‐S) on PEG4K–Clay scaffolds after culturing for 0, 1, 3, 5, and 7 d C). SEM images of ROBs 3D‐bioprinted within PEG4K–Clay scaffolds after 7 d of culture (D, ROBs in the red rectangular were marked with yellow color). Live/dead assay E, I) and cytoskeleton staining E, II) of ROBs 3D‐bioprinted within PEG4K–Clay scaffolds after 7 d of culture. Cell counting Kit‐8 (CCK‐8) analysis of ROBs after 3D‐bioprinting within PEG–Clay scaffolds (PEG–Clay‐P) or traditionally seeded on PEG–Clay scaffolds (PEG–Clay‐S) on days 1, 3, 5, and 7 d of culture F). ALP analysis of ROBs 3D‐bioprinted within PEG–Clay scaffolds after culturing for 4, 7, 14, and 21 d G), the insert images show ALP staining of ROBs on PEG4K–Clay‐P I) and PEG10K–Clay‐P II). Asterisks (*) denote significant differences (* p < 0. 05, ** p < 0. 01). After UV‐irradiation, 3D‐bioprinted nanocomposite hydrogel constructs with well‐distributed cells for bone regeneration were formed. As shown in Figure 3 C, when cells are seeded on pure 3D‐printed PEG–Clay scaffolds by the traditional cell‐seeding method, even if enough time is given for cell adhesion, most of the cells drop through the holes of the scaffold or just adhere to the surface, and ultimately are distributed nonuniformly. 14 In contrast, the number and distribution of cells within a scaffold can be precisely regulated in the proposed 3D‐bioprinting system. After one‐day of culture, most of the cells adhered within the scaffolds and the viability of the printed cells was higher than 95% (Figure 3 B). The results show that, our 3D‐bioprinted nanocomposite hydrogel constructs resulted in uniform cell distribution, effective deposition, and excellent viability. 2. 3 In Vitro Proliferation and Cell Morphology We further evaluated and compared the cell proliferation ability of PEG–Clay cell matrices prepared by both the 3D‐bioprinting method and the traditional cell‐seeding method (Figure 3 C, F). Regardless of the cell seeding protocol, the proliferation of ROBs on the PEG4K–Clay scaffolds was better than the PEG10K–Clay scaffolds. In the PEG10K–Clay scaffolds, cells maintained their spherical morphology and poorly spread, and while the cells were still alive they did not proliferate well. For the PEG4K–Clay scaffolds, the highest rate of proliferation was observed on days 3–5 of culture, in both the cell printing and traditional seeding methods. In the traditional cell‐seeding approach, only a few cells had adhered and spread out on the PEG4K–Clay scaffolds after 1 d of culture, and even after 3 d no obvious cell–cell contact could be observed. Due to the nonuniform distribution of cells, single cells and cell aggregation could be observed after 7 d of culture (Figure 3 C and Figure S5, Supporting Information). In contrast, for the 3D‐bioprinted cells, though few cells retained their spherical morphology after 3 d of culture, the cells were well distributed and spread after 5 d (Figure S5, Supporting Information). After 7 d of culture, almost complete ROBs coverage of the scaffolds was realized with almost no dead cells observed, as visualized by F‐actin staining of the cytoskeleton and live/dead experiment (Figure 3 E). Furthermore, since the cell‐containing component (HA) was printed between each of the PEG–Clay prehydrogel lines, the cells not only grew on the surface of the matrices but also grow within the scaffolds themselves, as confirmed by scanning electron microscope (SEM) imaging (Figure 3 D and Figure S5, Supporting Information). 2. 4 Osteogenic Differentiation of ROBs after 3D‐Bioprinting Alkaline phosphatase (ALP) activity assays and ALP staining were used to evaluate printed ROBs differentiation after 4, 7, 14, and 21 d of culture (Figure 3 G). After printing within both PEG4K–Clay and PEG10K–Clay scaffolds, the cells always demonstrated ALP secretion ability, with the largest increase observed at 14 d. The ALP activity of ROBs within the PEG4K–Clay scaffolds was significantly higher than within the PEG10K–Clay scaffolds on days 7, 14, and 21 d of culture. ALP staining was used to confirm these results after 21 d of culture (Figure 3 G). The images show that the ROBs within the PEG10K–Clay scaffolds were almost aggregated compared with the well spread conformation of cells growing within the PEG4K–Clay scaffolds. Moreover, the bluish intensity, which is directly proportional to the ALP activity, was much higher for the PEG4K–Clay scaffolds than the PEG10K–Clay scaffolds, which is consistent with the cell proliferation and differentiation results outlined. This phenomenon can partly be attributed to the incorporation of nanoclay into our 3D‐bioprinting systems; as PEG hydrogels have no osteogenic properties, 15 cells cannot grow or spread out on pure PEG hydrogels regardless of the PEG molecular weight. 16 The mechanical properties of PEG alone are too weak for cells to adhere, by dispersing nanoclay within the system the mechanical properties are significantly improved upon, as described (Figure 2 B, C). Moreover, our PEG–Clay scaffolds can gradually release bioactive ions (magnesium ions, Mg 2+ and silicon ions, Si 4+ ), which act to promote osteogenic differentiation. 17 Figure S6 in the Supporting Information presents the release curve of Mg 2+ and Si 4+ from PEG–Clay scaffolds. At all time points, the ion concentrations released from the PEG10K–Clay scaffolds were marginally higher than the PEG4K–Clay scaffolds, probably due to the lower crosslinking density of the PEG10K–Clay scaffolds. The measured ion concentrations were in the range considered effective for eliciting bone regeneration. 18, 19 The release concentration increased steadily for both sample types, from 44 and 56 µg mL −1 after 1 d of immersion to 106 and 115 µg mL −1 after 21 d of immersion, for Mg 2+ released from PEG4K–Clay and PEG10K–Clay scaffolds, respectively. The release concentration of Si 4+ increased from 108 and 128 µg mL −1 after 1 d of immersion to 245 and 263 µg mL −1 after 21 d of immersion, for PEG4K–Clay and PEG10K–Clay scaffolds, respectively. Notably, after 7 d of immersion, the release rate reduced, most likely because the surface ions are released more easily and faster than ions inside of the scaffolds, which are restricted and slow to release due to the slow degradation rate of the PEG‐based hydrogel. 20 2. 5 In Vivo Tibia Repair Experiments Taking into account the superior performance of the 3D‐bioprinted ROB‐laden PEG4K–Clay scaffolds in the in vitro studies over their higher molecular weight cousins, they and pure PEG4K–Clay scaffolds were assessed in tibia repair and ectopic osteoinduction experiments in vivo. Sequential fluorescent labeling 21 was used to monitor new bone formation around the injuries by applying three types of fluorochromes ( Figure 4 A): tetracycline (yellow) for 2–4 weeks, alizarin red (red) for 4–6 weeks and calcein (green) for 6–8 weeks. Since fluorochrome labels can bind calcium ions of newly formed bone and get incorporated into the site of mineralization, the new bone formed at different periods can be discriminated by different fluorochrome labels. Comparing the pure PEG4K–Clay scaffolds and the blank group with the ROB‐laden PEG4K–Clay scaffolds, more continuous and more abundant fluorescent emission was observed for the 3D‐bioprinted PEG4K–Clay scaffolds. The percentage area occupied by the three fluorochromes in the labeled bone was calculated and is shown graphically in Figure 4 C. The fluorochrome area of the ROB‐laden PEG4K–Clay scaffold was 8. 44%, clearly larger than the equivalent areas of the pure PEG4K–Clay scaffold (5. 30%) and the blank group (3. 22%). Furthermore, different from all three colors can be observed for both the pure PEG4K–Clay scaffolds and the ROB‐laden PEG4K–Clay scaffolds, the yellow color can hardly be observed in the blank group, which means new bone formed very slowly without promotion. Micro–computed tomography (Micro‐CT) 3D‐images showed that the highest amount of new bone was found in the marrow cavity around or inside the ROB‐laden PEG4K–Clay scaffolds (Figure 4 B). The bone volume of the ROB‐laden PEG4K–Clay scaffolds (10. 24%) was significantly larger than that of the pure PEG4K–Clay scaffolds (7. 20%) and the blank group (3. 78%; Figure 4 C). Figure 4 Sequential fluorescent labeling of blank, PEG4K–Clay scaffolds without ROBs, and PEG4K–Clay scaffolds with 3D‐bioprinted ROBs A). Yellow, red, and green represent tetracycline hydrochloride, alizarin red S, and calcein labeling, respectively. Characterization of implants and new bone formation by reconstructed 3D models after micro‐CT analysis B). Percentage of the area of fluorochromes‐stained bone and new bone volume/total volume (BV/TV) ratio C) (* p < 0. 05, ** p < 0. 01, blank as a control). As shown in Figure 5, the histological sections (8 weeks postimplantation), processed by Giemsa staining, 18 gave a full view of bone formation in the three groups. The new bone formed (pink part) around the defect area in the blank group was intermittent. The newly formed bone in both of the PEG4K–Clay groups was far more continuous, especially in the ROB‐laden PEG4K–Clay scaffolds. For pure PEG4K–Clay scaffolds, although new bone can only be observed around the scaffolds, osteoblasts and fibroblasts (purple and blue) can be found within the scaffolds, which means that cells had grown into the scaffold and indicates that new bone may form within the scaffold later. While osteoblasts and fibroblasts can be observed within the ROB‐laden PEG4K–Clay scaffolds, new bone can also be clearly observed around and within the scaffolds. H&E staining 22, 23 was used to further validate the Giemsa staining results (Figure 5 ). The results were consistent with the data collected from micro‐CT and new bone labeling analyses, verifying that PEG4K–Clay nanocomposite hydrogel scaffolds can stimulate new bone formation, and that the loading of exogenous cells into the scaffolds through 3D‐bioprinting has further positive implications for bone regeneration. Figure 5 Histological staining observations of the whole bone around the defects with low magnification (40×) and high magnification (100×), marked with the yellow rectangle. Giemsa staining and H&E staining were used to stain new bone tissue in the blank group, pure PEG4K–Clay group, and the cell‐laden PEG4K–Clay group. The red arrows show the formation of new bone and the green arrows show the fibrous tissues around the scaffolds 8 weeks after surgery. 2. 6 Ectopic Osteoinduction Experiments To further investigate the growth behavior of printed allogenic ROBs after in vivo implantation, PEG4K–Clay printed with green fluorescent protein (GFP)‐labeled ROBs scaffolds (GROB‐laden PEG4K–Clay scaffolds) were implanted into the muscles of Sprague–Dawley rats (SD rats) for ectopic osteoinduction. After 1, 3, 5, 7, 14, and 21 d of implantation, the implanted GROB‐laden PEG4K–Clay scaffolds were removed from the muscles, and the fibrous tissue growing around the scaffolds was removed for the observation of cell status within the scaffolds using fluorescence microscopy. As time progressed, the implants increasingly integrated with the surrounding tissues, so that it became increasingly difficult for the scaffolds to be exfoliated. It is obvious from Figure 6 A that GFP‐labeled ROBs still persisted even 21 d after implantation, indicating that no obvious immunoreaction occurred after implantation. This result was supported by H&E and Goldner's staining, 24 after GROB‐laden PEG4K–Clay scaffolds were implanted for 3 and 7 d (Figure 6 B). For both the 3 and 7 d implantations, no significant immunoreaction was observed. Compared with the 3 d implants, the 7 d implant scaffolds were filled with fibrous tissue and even osteoids were identified. We further comparitively explored the osteogenic properties of both the ROB‐laden PEG4K–Clay scaffolds and the pure PEG4K–Clay scaffolds as the controls, through H&E and Goldner's staining ( Figure 7 ) after 2, 4, and 8 weeks of implantation. After the first 2 weeks, a small quantity of new bone had already formed in the ROB‐laden PEG4K–Clay scaffolds, while new bone in pure PEG4K–Clay scaffolds was not obvious. After 4 weeks of implantation, numerous osteoblast cells and new bone can be observed around and even inside the ROB‐laden PEG4K–Clay scaffolds, and the amount of new bone formed around the ROB‐laden PEG4K–Clay scaffolds was larger than that of pure PEG4K–Clay scaffolds. After 8 weeks of implantation, both new bones and numerous osteoblast cells can be observed around the ROB‐laden PEG4K–Clay scaffolds and the pure PEG4K–Clay scaffolds. However, both the new bone volume and thickness around the ROB‐laden PEG4K–Clay scaffolds were much larger than the pure PEG4K–Clay scaffolds. These results further confirm that our PEG–Clay nanocomposite hydrogels could promote new bone formation and that the ROB loaded 3D‐printed PEG–Clay scaffolds showed better osteogenic properties. Figure 6 GFP‐labeled ROBs within 3D‐bioprinted PEG4K–Clay scaffolds after the in vivo ectopic osteoinduction experiment at 1, 3, 5, 7, 14, and 21 d time points A). H&E staining and Goldner's staining of 3D‐bioprinted PEG4K–Clay scaffolds after the in vivo ectopic osteoinduction experiment at 3 and 7 d under different magnifications (40×, 100×, and 200×) B). The green arrow shows the fibrous tissues and the yellow arrow shows the osteoid part. Figure 7 Histological staining observations of the in vivo ectopic osteoinduction experiment after implantation for 2, 4, and 8 weeks with low magnification (40×) and high magnification (100×), marked with the yellow rectangle. H&E staining A) and Goldner's staining B) were used to stain new bone tissue in the cell‐laden PEG4K–Clay group and pure PEG4K–Clay group. The yellow arrows show the formation of new bone after surgery. The reconstructed 3D models C) and micro‐CT picture D) were used to further confirm the formation of new bone 8 weeks after surgery. 3 Discussion In this study, ROB‐laden nanocomposite hydrogel constructs were fabricated by a two‐channel 3D‐bioprinting method by alternately extruding two bio‐inks (A and B). Bio‐ink A, PEG–Clay prehydrogel solution, offered a suitable viscosity for facilitating the process of 3D bioprinting, oxygen and nutrient delivery, and ultimately cell growth after crosslinking. The uniform distribution of the nanoclay in the PEG matrix not only enhanced the hydrogel mechanical properties but also rendered them more conducive to cell adhesion and proliferation than pure PEG hydrogels. Furthermore, the release of Mg 2+ and Si 4+ bioactive ions from the PEG–Clay scaffolds formed an induced microenvironment which stimulated the osteogenic differentiation of ROBs, to benefit bone regeneration. Simultaneously, bio‐ink B, based on HA, was applied as a vector to accurately and uniformly deposit an ROB load into the 3D‐printed scaffolds. The inclusion of HA not only guaranteed cell viability during 3D‐bioprinting with good distribution within the scaffolds, but its slow dissolution allowed for the gradual release of cells. Moreover, compared with the one‐channel method of 3D‐bioprinting, our two‐channel 3D‐bioprinting approach not only enhanced cell viability but also encouraged better cell spreading and proliferation. 25 Importantly, this study also provides an efficient alternative to load cells within scaffolds for tissue engineering. Compared with the traditional cell‐seeded scaffolds, our ROB‐laden scaffolds showed better cell distribution and effective cell deposition in vitro. Although ROBs survived and grew on the PEG–Clay scaffolds whether traditionally seeded or printed, the distribution of cells in bioprinted scaffolds was much better than in the cell‐seeded scaffolds. In the seeded scaffolds, most of the cells dropped through the scaffold holes leading to nonuniform distribution, while in the ROB‐laden scaffolds, cell encapsulation in HA ensured initial uniform cell deposition followed by slow cell release to allow for optimal cell adhesion within the scaffolds. Even though the same cell numbers were used for both protocols, a significantly higher number of cells could be observed in the cell‐laden bio‐printed scaffolds. After 7 d of culture, ROBs almost entirely covered the bio‐printed scaffold surfaces, with practically no dead cells observed; only a few dispersed ROBs were found on the traditionally seeded scaffolds. Moreover, the ROB‐laden PEG–Clay constructs also exhibited excellent osteogenic capability both in vitro and in vivo. In vitro experiments showed that ROBs successfully proliferated and differentiated, especially within the cell‐laden bio‐printed PEG4K–Clay scaffolds. For the tibia repair experiments, sequential fluorescent labeling results reveal the new bone formation process. Compared with the slow bone formation in the blank group, the new bone formation for both the pure PEG4K–Clay scaffolds and the ROB‐laden PEG4K–Clay scaffolds can be observed throughout the entire examination period, encouraged in‐part by the release of bioactive magnesium ions and silicon ions, which are well recognized promotion factors for osteogenic differentiation. 18, 19 Furthermore, loading cells within PEG–Clay scaffolds by our 3D‐bioprinting system further improved the bone formation ability in vivo, suggesting that exogenous bone‐related cells can also play a key factor in the promotion of bone regeneration. Additionally, the favorable integration of exogenous cells was further highlighted during the in vivo ectopic osteoinduction study. The GFP‐labeled ROBs, from allogenic childhood SD rat calvarial chips, within PEG4K–Clay constructs were tracked after implantation in the muscles of adult SD rats for 21 d, and no significant immune response was observed in the surrounding tissues, which points to the potential of replacing bone‐related stem cells with exogenous allogenic osteoblasts for bone regeneration. Prolonging the implantation time, new bone formation in muscles was found in both pure PEG4K–Clay scaffolds and ROB‐laden PEG4K–Clay scaffolds. Additionally, the osteogenic capability of ROB‐laden PEG4K–Clay scaffolds was significantly better than that of pure PEG4K–Clay scaffolds. The 3D‐bioprinting cell loaded PEG–Clay system described herein holds significant promise for therapeutic application in bone regeneration, as evidenced by the in vitro and in vivo studies described, the results of which point to not only the excellent osteoblast scaffold distribution and viability in the short term but also the ability to promote new bone formation in the long term. 4 Conclusion In summary, we successfully fabricated an osteoblast‐laden nanocomposite hydrogel construct via a two‐channel 3D‐bioprinting method. One channel carried bio‐ink A, PEG–Clay prehydrogel solution, which was suitably viscous to facilitate the 3D‐bioprinting process and was conducive to the delivery of oxygen and nutrients and cell growth after crosslinking. The other channel guided the accurate delivery of cells into the 3D scaffolds, using ROBs encapsulated in 20% HA solution. The HA component served to not only protect the ROBs from UV damage during the crosslinking process but also guaranteed uniform distribution and cell viability (more than 95% after 1 d). Furthermore, ROBs within the bioprinted scaffold showed better proliferation and differentiation than the same number of ROBs seeded on 3D PEG–Clay scaffolds. In tibia repair and ectopic osteoinduction experiments, ROB‐laden PEG–Clay scaffolds showed excellent osteogenic potential, due to the induced environment formed around the PEG–Clay scaffolds which was conducive to ROBs differentiation. This study offers a viable new approach for 3D‐bioprinting for the construction of bone substitutes in tissue regeneration. 5 Experimental Section Materials : PEG ( M w = 4K and 10K, Sigma‐Aldrich, St. Louis, USA), acryloyl chloride (98%, TCI, Shanghai, China), triethylamine (99%, TCI, Shanghai, China), diethyl ether (Lingfeng Chemical Reagent Company, Shanghai, China), 2‐hydroxy‐2‐methyl‐1‐phenyl‐1‐propanone (IRGACURE 1173, 98%, Sigma‐Aldrich, St. Louis, USA), Laponite XLG ([Mg 5. 34 Li 0. 66 Si 8 O 20 (OH) 4 ]Na 0. 66 ; BKY, Wesel, Germany), and HA ( M w = 350K, TCI, Shanghai, China) were all used as received. All other chemicals and solvents were analytical reagents and were purchased from Lingfeng Chemical Reagent Company (Shanghai, China) and used as received. Cell Culture : Primary ROBs were isolated from minced SD rats (born within 3 d) calvarial chips, as described previously; 26 this procedure was conducted in accordance with the guidelines set by the Ethics Committee for Animal Research, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. ROBs, between the third to fifth passage, were used in the 3D‐bioprinting system to evaluate cell viability, proliferation, and differentiation ability both in vitro and in vivo. ROBs were cultured in α‐MEM (Hyclone, Utah, USA) with 10% (v/v) fetal bovine serum (Corning, New York, USA), supplemented with antibiotics (100 U mL −1 penicillin, 100 µg mL −1 streptomycin) and incubated at 37 °C with 5% carbon dioxide (CO 2 ). Prehydrogel Solution Preparation for 3D‐Bioprinting : Crosslinkable PEGDA of different molecular weights ( M w = 4K and 10K) were synthesized as previously described. 9 The recipes for preparing prehydrogel solutions are listed in Table S1 in the Supporting Information. Briefly, PEGDA was dissolved in deionized water at first, and then 3 wt% of the photoinitiator IRGACURE 1173 (relative to the mass of PEGDA crosslinker) was added into the solution and stirred thoroughly under a nitrogen atmosphere until completely dissolved. The solution was then sterilized via filtration through a 0. 22 µm filter. Subsequently, ultraviolet sterilized nanoclay (Laponite XLG) was added into the solution and vigorously stirred on a clean bench for about 2 h, to obtain a homogeneous prehydrogel solution. Bubbles were removed from the prehydrogel solution by centrifugation (3000 rpm, 10 min) before being used. Sterilized HA powder (20%) was dissolved in sterile phosphate buffer solution (PBS, 100 mL) and stirred thoroughly on a clean bench until dissolved. For 3D‐bioprinting preparation, a cell suspension (50 µL) with 4 × 10 6 ROBs cells was added into 20% HA (1. 0 g) solution and gently stirred to achieve uniformity for printing. 3D‐Bioprinting Process : The printing device (3D scaffold printer) used for our experiments is a precision three‐axis positioning system (Bioscaffolder 3. 1, GeSiM, Grosserkmannsdorf, Germany). The system was placed on a clean bench and was sterilized using 75% alcohol and UV light before use. In our study, a two‐channel printing method was used, one channel for supporting matrix printing (PEG–Clay) and another for cell printing (HA); the printing process is shown in Figure 1 and Movie S1 in the Supporting Information. The dosing pressure of the syringe pump for the PEG–Clay prehydrogel solution and the cell containing HA solution was controlled between 55 and 65 and 70 and 80 kPa, respectively. The moving speed of the dispensing unit was set to 15 mm s −1. The cell containing scaffolds were printed in 60 mm diameter cell culture dishes and placed in a crosslinking oven (XL‐1000 UV Crosslinker, Spectronics Corporation, NY, USA) for 10 min polymerization under UV. Then, α‐MEM medium (8 mL) was added into each dish. The medium was changed the next day to allow time for HA cell release and adhesion within the PEG–Clay scaffold. Pure PEG–Clay scaffolds without ROBs, with the same structure as the 3D‐bioprinting scaffolds, were also printed by this system. Furthermore, PEG hydrogels and PEG–Clay hydrogels were prepared without printing, for basic characterization, and compression tests. Basic Characterizations : PEG (15 mg) or pure PEGDA (15 mg) was dissolved in D 2 O (0. 5 mL), and their respective 1 H NMR spectra were recorded on an AVANCE III 400 spectrometer (BRUKER, Madison, USA). FTIR spectroscopy (BRUKER VERTEX 70, Madison, USA), and TGA (Q600 SDT, TA Instruments, New Castle, USA) were used to characterize the nanoclay, PEGDA, PEG hydrogels, and PEG–Clay nanocomposite hydrogels. TGA was performed from room temperature to 1090 °C, at a heating rate of 10 °C min −1 in nitrogen. 27 The microstructure of the PEG–Clay scaffolds was investigated by TEM (Philips CM100, Massachusetts, USA). 28 For TEM imaging, lyophilized PEG–Clay scaffolds were embedded in epoxy resin for microtoming at −40 °C with a glass knife to ≈50 nm thick sections, which were then deposited on copper grids for imaging at 200 kV. Compression Test : To assess if the incorporation of the clay component into the hydrogel system improved the mechanical properties of the hydrogels, the compression properties of the PEG hydrogels and PEG–Clay hydrogels were tested on an Instron 5697 (Instron, Grove City, USA) universal material testing system at room temperature. All hydrogels were tested directly after polymerization. As pure PEG hydrogels are unprintable, all samples used in the compression test were prepared in a cylindrical shape, 4 mm in diameter, and 4 mm in height. The crosshead speed was set to 10 mm min −1. At least five samples were used for each test for statistical significance. Ion Leaching Analysis : Immersion tests were carried over different time points (from 1 to 21 d) to assess the quantities of magnesium and silicon ions released from the 3D‐printed pure PEG–Clay scaffolds with different PEGDA molecular weights. Printed scaffolds, 15 mm in width and five layers high, were immersed in PBS (8 mL) separately. Nine different periods of time (1, 2, 3, 4, 5, 6, 7, 14, and 21 d) were analyzed by inductively coupled plasma optical emission spectrometry (Perkin Elmer, Optima 7000DV, Massachusetts, USA) to determine the concentration levels of Mg 2+ and Si 4+ released and establish the relationship between ion release and time. Rheology Test : Dynamic rheological experiments with 20%PEG4K–Clay prehydrogel solutions of different nanoclay concentrations and 20% HA (20 g HA powder dissolved in 100 mL PBS) were carried out using a rheometer (MCR302, Anton Paar, Austria). Plate‐plate geometry with a plate diameter of 25 mm was used. The shear rates were controlled between 1 and 100 s −1 at room temperature. Proliferation Assay and Cell Morphology : Cell proliferation ability was determined after 1, 3, 5, and, 7 d of culture through the CCK‐8 assay, after 3D‐bioprinting. 29 After incubation with 10% CCK‐8 solution at 37 °C for 4 h, cell proliferation was quantified by measuring the optical density of the CCK‐8 solution at 450 nm, using a Multiskan spectrum reader (Bio Tek Synergy4, Winooski, USA). For testing, each 15 × 15 mm 2 printed scaffold was equally divided into four parts both for the proliferation and differentiation assays and the culture medium was 2 mL per part. In order to verify the advantages of the 3D‐bioprinting method over the traditional cell‐seeding method, the ROBs were seeded onto the top of the sterilized PEG–Clay scaffolds (7. 5 × 7. 5 × 2–7. 5 × 7. 5 × 3 mm 3 ) placed in 24‐well cell culture plates (Corning, New York, USA). The same cell density was used for cell seeding as the cell‐printing method, and the culture conditions were the same. The live/dead viability assay 30 was performed according to the manufacturer's instructions to test the viability of ROBs one day after printing and after 7 d culture. The samples were observed through laser scanning confocal microscopy (Leica SD AF, Hamburg, Germany). Excitation of 488 nm was used to detect the live (green) cells stained by calcein AM, and 561 nm excitation was used to observe dead (red) cells stained by EthD‐1. For cell morphology observations after 3D‐bioprinting and 7 d culture, ROBs cytoskeleton was stained with 50 µg mL −1 phalloidin‐rhodamine (Sigma‐Aldrich, St. Louis, USA) after fixation with 4% paraformaldehyde (PFA) and permeation by 0. 1% Triton. 31 SEM (Hitachi S4800 FEG, Tokyo, Japan) was used to detect the growth status of the ROBs within the PEG4K–Clay scaffolds after 3D‐bioprinting, fixation, gradient dehydration, and critical point drying. Differentiation Assay : The quantification of ALP activity and ALP staining were used to evaluate the osteoblast phenotype of ROBs grown within the scaffolds after 3D‐bioprinting. 32 After 4, 7, 14, and 21 d culture, the scaffolds printed with ROBs were rinsed three times with PBS, and then lysed in lysis buffer (300 µL, radio immunoprecipitation assay (RIPA) buffer, Beyotime Biotechnology, Shanghai, China). Cell debris was removed by centrifugation at 13 rpm, at 4 °C for 5 min, and then the supernatant (50 µL) was added to chromogenic substrate (50 µL) in a 96‐well plate and incubated at 37 °C for 2 h. Then, stop buffer (100 µL) was added to stop the reaction. Absorbance was measured at 405 nm using a microplate reader. Analysis of each sample was performed in triplicate and the total protein content was used to normalize the ALP activity by a commercially available protein assay kit (Pierce TM BCA Protein Assay Kit, ThermoFisher Scientific, Massachusetts, USA). In order to visualize the differentiation ability of ROBs printed on each scaffold, ALP staining for each scaffold was performed using a 5‐bromo‐4‐chloro‐3‐indolyl phosphate (BCIP)/nitrobluetetrazolium chloride (NBT) Alkaline Phosphatase Color Development Kit (Beyotime Biotechnology, Shanghai, China). In Vivo Evaluation : All the animal procedures and experiments were approved by the Ethics Committee for Animal Research, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. A rat tibia model was used for the bone defect repair experiment and the surgical procedures were all conducted under sterile conditions. The scaffolds both printed with ROBs and without ROBs were cut into sections of 2. 5 mm diameter and 5 mm length. These sections were implanted into tibia bone defects of 12‐week old male SD rats. First, the rats were anesthetized with pentobarbital sodium (40 mg kg −1 ) by intraperitoneal injection. Then, the critical size defects (2. 5 mm diameter and 5 mm length) were created close to the tibia plateau and in the middle of the tibia shaft of an SD rat leg by a 2. 5 mm drill. The implants were placed bilaterally resulting in two implants per rat, and the wound was closed carefully. Twelve SD rats were randomly divided into three groups corresponding to 3D‐bioprinting PEG4K–Clay scaffolds, pure PEG4K–Clay scaffolds without ROBs, and blanks as the controls. All rats were sacrificed 8 weeks after implantation, and the implants were harvested and fixed in 4% PFA for the micro‐CT assay and histological analysis later. In order to characterize the new bone formation and mineralization, a polychrome sequential fluorescent labeling method was used. 21 Three different fluorochromes were sequentially administered intramuscularly, 25 mg kg −1 tetracycline hydrochloride, 30 mL kg −1 alizarin red S, and 20 mg kg −1 calcein 2, 4, and 6 weeks after the operation. For the ectopic osteoinduction experiment, 24 SD rats (12‐weeks old) were used. Following disinfection of the operative region, the 3D‐bioprinting PEG4K–Clay scaffolds (using GFP‐labeled ROBs) were implanted into muscle pouches created in the left thigh of each rat. The control group pure PEG4K–Clay scaffolds without ROBs were implanted into the muscle pouches in the right thigh of each rat. The wounds were sutured carefully. At every time point (1, 3, 5, 7, 14, and 21 d, 4, and 8 weeks after implantation), three rats were euthanized. For the first 1–21 d time points, cell‐laden PEG4K–Clay scaffolds were exfoliated from the tissues to confirm the cell viability after implantation. For the 2, 4, and 8 week time points, the implanted region including the surrounding tissues was excised. The excised fragments were fixed in 4% PFA for histological analysis and the 8 week old implant scaffolds were analyzed by micro‐CT. New bone formation was determined by micro‐CT (SkyScan 1176, Bruker, Madison, USA) for the fixed samples. The scanning parameters were set at 60 kV, Al 1 mm filter, and 18 µm resolution. After scanning, the NRecon software (Skyscan, USA) and CTvol program (SkyScan) were used to reconstruct 2D and 3D models of the samples, and DataViewer software (SkyScan) and CTAn program (SkyScan) were used to determine the bone volume around the implants. After micro‐CT, 50 µm undecalcified sections were prepared using a Exakt system (model 310 CP bnad system, Exakt, Oklahoma City, OK, USA) for fluorescence labeling observation under a confocal laser scanning microscope (Leica TCS SP8, Hamburg, Germany). The excitation/emission wavelengths used to observe the chelating fluorochromes were 405/575, 543/620, and 488/520 nm for tetracycline hydrochloride (yellow), alizarin red S (red), and calcein (green), respectively. After fluorescence microscopy, Giemsa staining was used to visualize the mineralized bone tissue (pink) in the same sections. 18 The images were captured by a fluorescence microscope (Olympus, BX53, Tokyo, Japan). For paraffin sections, all samples (both tibias and excised fragments) were decalcified in 10% ethylene diamine tetraacetic acid (EDTA) for 6 weeks after fixation, embedded in paraffin and sectioned into 5–7 µm sections (Leica RM2235, Hamburg, Germany). Hematoxylin (Sigma‐Aldrich, St. Louis, USA) and eosin (C0105, Beyotime Biotechnology, Shanghai, China) (H&E) staining 22 and Goldner's staining 24 were used to detect the specific tissue response to the implanted materials. Images were taken under the fluorescence microscope (Olympus, BX53, Tokyo, Japan). Statistical Analysis : All the experiments were analyzed by one‐way analysis of variance (ANOVA) with Tukey' post hoc test and expressed as means ± standard deviations (SD). A p ‐value < 0. 05 was considered to be statistically significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201700666
| 2,018
|
Advanced Science
|
Engineering a Tumor Microenvironment‐Mimetic Niche for Tissue Regeneration with Xenogeneic Cancer Cells
|
Abstract The insufficient number of cells suitable for transplantation is a long‐standing problem to cell‐based therapies aimed at tissue regeneration. Xenogeneic cancer cells (XCC) may be an alternative source of therapeutic cells, but their transplantation risks both immune rejection and unwanted spreading. In this study, a strategy to facilitate XCC transplantation is reported and their spreading in vivo is confined by constructing an engineering matrix that mimics the characteristics of tumor microenvironment. The data show that this matrix, a tumor homogenate‐containing hydrogel (THAG), successfully creates an immunosuppressive enclave after transplantation into immunocompetent mice. XCC of different species and tissue origins seeded into THAG survive well, integrated with the host and developed the intrinsic morphology of the native tissue, without being eliminated or spreading out of the enclave. Most strikingly, immortalized human hepatocyte cells and rat β‐cells loaded into THAG exert the physiological functions of the human liver and rat pancreas islets, respectively, in the mouse body. This study demonstrates a novel and feasible approach to harness the unique features of tumor development for tissue transplantation and regenerative medicine.
|
1 Introduction Cell transplantation aimed at tissue reconstruction and regeneration holds promise to solve the fundamental challenges with organ transplantation, including an extreme shortage of whole organs suitable for medical implantation as well as its associated medical, social, and ethical issues. 1, 2, 3 Ideally, the therapeutic cells are collected from the patient's own tissue to avoid immunogenic rejection, expanded ex vivo and delivered back to the patient with or without the use of scaffolds (“autograft”). 4, 5 However, the number of obtainable cells is usually far insufficient to construct a functional tissue or organ. 3 One microliter of human tissue typically contains about 10 9 cells; and the liver, as an example of a whole organ, comprises over 10 11 hepatocytes. 6 It is unlikely to obtain extra billions of primary cells through isolation and ex vivo expansion because many somatic cells proliferate slowly or do not proliferate at all. This insufficiency is a long‐existing bottleneck hampering the clinical applications of cell transplantation. Besides, although theoretically stem cells have unlimited self‐renewal capacity and may provide more cells after expansion, 7 their differentiation into desirable lineage and generation of functional tissues are hard to control. 8 As such, cells that can rapidly proliferate and readily constitute a new tissue remain highly demanded in regenerative medicine. Cancer cells are not normally linked with tissue regeneration, but they indeed fulfil the above requirements for transplanted cells. First, cancer cells grow fast, proliferating 3–5 folds faster than normal cell lines. 9, 10 Second, many cancer cells can preserve the core functions of their tissue of origin. In tissue engineering research, cancer cells are commonly employed to test the performance of biomaterials scaffolds in vitro, 11 as their phenotype can simulate that of the target tissue. In some artificial livers (a type of ex vivo device), hepatoma (liver cancer cells) were even used to substitute normal hepatocytes and functioned well. 12 Third, a solid tumor has a complete structure of the organ where it arises, comprising parenchymal (neoplasm) and stromal cells, rich extracellular matrix (ECM) components, independent circulation and an immunosuppressive microenvironment. 13, 14 These features of tumor would precisely favor the regeneration of many tissue types. But certainly, if cancer cells were serving as therapeutic cells, their localization is important because their migration out of the implanted sites may risk tumorigenesis in other healthy organs and must be strictly prohibited. Nevertheless, if we transplant cancer cells for regenerative purposes, how can we promote their growth locally while preventing their migration globally at the same time? We proposed to harness the power of immunity to realize this goal. Specifically, we aimed to use xenogeneic cancer cells (XCC) and create an immunosuppressive microenvironment (an “enclave”) for these trans‐species cells to grow. However, their migration out of this niche would be naturally inhibited by the robust cross‐species immune response. First, it is common to create tumor models in cancer research by transplanting xenogeneic (human) cancer cells into immunocompromised mice—but not immunocompetent mice, where promptly triggered immunogenic rejection and could immediately eliminate the transplanted cells. 15 Likewise, this mechanism could be utilized to confine the implanted XCC in the engineered immunogenic enclave. Second, this enclave should mimic the characteristics of tumor microenvironment (TME). Like transplanted cells, the tumor is also recognized as “foreign” by the immune system but can effectively escape immunosurveillance to flourish. 16, 17, 18, 19 The key reason is the formation of its unique microenvironment, in which multiple biochemical and physical signals dynamically coordinate to educate immunocytes, including macrophages, dendritic cells (DCs) and lymphocytes—which continue infiltrating into the tumor as the blood vessels invade—into an immunosuppressive phenotype. 20, 21, 22, 23 The educated immune cells, in turn, help the tumor escape immune attack, promote neoplasm growth, encourage angiogenesis, and thereby shape TME into a better platform to support cancer cell growth. 24, 25 Therefore, cultivating and transplanting XCC in an engineered, TME‐mimetic matrix niche would exert the maximum potency of cancer cells for tissue regeneration while preventing their spreading in the host body. To prove this concept, in the present study, shown as Scheme 1, we aimed to design a TME‐mimicking matrix and evaluated its capability in supporting the growth of XCC in immunocompetent mice. The matrix comprised the soluble extract of a murine sarcoma (which are supposed to preserve the biological factors for TME formation), a synthesized hydrogel (which offers mechanical support and injectability) as well as recombinant proteins of basic fibroblast growth factor (bFGF) aimed at enhancing angiogenesis. 26 The prepared gel, carrying different types of XCC, was injected subcutaneously into mice to create the immunogenic enclave. The confinement of XCC growth within the enclave and meanwhile the development of various types of xenograft tissue were evaluated. Our results demonstrated the successful formation of a TME‐mimetic niche in the hydrogel implants that could support trans‐species cell growth into specific tissue structures in the back of immunocompetent mice, with no signs of cell migration out of the created enclave. Most encouragingly, two xenograft models, originating from human hepatic‐transformed cell line (THELE‐3) and rat insulin‐producing cell lines (INS‐1), exerted normal functions of the liver and pancreas islets, respectively, in the mice body. 2 Results 2. 1 Tumor Homogenate (TH) Creates a TME‐Mimetic Niche Our goal was to fabricate a TME‐mimetic matrix, made of soluble TH and an injectable hydrogel, for the culture of xenogeneic cells. We assumed that TH contained the essential ingredients to modulate the stromal cells into a pro‐tumor phenotype which would favor the growth of transplanted cells and creation of TME‐like microenvironment. To validate this assumption, we obtained TH from the soluble extract of murine sarcoma (derived by the transplantation of S180 cell line), profiled its protein composition by liquid chromatography mass spectrometry (LC‐MS) (Data File 1, Figure S1, Table S1, Supporting Information) and tested its effects on the behavior of primary fibroblasts and macrophages. We chose these two cells for testing because cancer‐associated fibroblasts (CAF) and tumor‐associated macrophages (TAM) are the two major stromal cell populations in TME that promote tumor development. 27, 28 The results from immunofluorescent (IF) staining ( Figure 1 a), Western blotting (WB, Figure 1 b) and quantitative polymerase chain reaction (PCR) (Figure 1 c) consistently suggested that treatment with TH for 48 h significantly stimulated the expression of α‐SMA and SDF‐1—two representative markers for CAFs—in primary mammary fibroblasts. 29 The treatment also upregulated the levels of Type I and IV collagen, which are also highly expressed in CAF (Figure 1 b and c). Notably, an antibody assay examining 78 proteins (Table S2, Supporting Information) revealed that the TH‐treated fibroblasts expressed abundant pro‐angiogenic and pro‐growth factors, which are usually secreted by CAF to promote the shaping of TME. The expression of angiogenesis‐related cytokines (100%), colony‐stimulating factors (CSF, 75%), and growth factors (50%) were highly up‐regulated (Figure 1 d). Further analysis of these data with gene ontology revealed the high enrichment of pathways associated with angiogenesis and anti‐inflammation, as well as the involvement of signals relating to several growth factors (Figure 1 e). These data indicated that TH could induce fibroblasts to acquire the phenotype of CAF. Scheme 1 Schematic diagram of the concept that an engineered TME‐mimetic niche formed by active soluble factors in tumor extract (TH), basic fibroblast growth factor (bFGF), and injectable hydrogel facilitates the xenogeneic cancer cells (XCC) to develop into a functional tissue in immunocompetent mice. Figure 1 Effects of tumor homogenate (TH) on the phenotype of primary fibroblasts and macrophages. a) Representative immunostaining images illustrating the expression of α‐SMA and SDF‐1 in primary fibroblasts treated with TH or PBS (Scale bar = 50 µm); b) Western blotting, and c) RT‐PCR analysis of the expression of fibroblast makers—α‐SMA, SDF‐1 and collagen type I and IV in primary fibroblasts treated with TH or PBS; d) Hierarchical cluster analysis of 78 kinds of soluble proteins in the culture medium of primary fibroblasts treated with TH relative to that with PBS according to molecular functions; e) Pathways enriched by the up‐regulated proteins shown in the panel d with Gene Ontology Pathway analysis; f) FACS analysis and g. immunofluorescent staining of CD206 and F4/80 in primary macrophages treated with TH or PBS, with IL‐4 as positive control (Scale bar = 50 µm); h) RT‐PCR analysis of the expression of macrophage makers—CCR7, CCR2, iNOS‐2, and Arg‐1 in primary macrophages cells treated with TH, PBS, or IL‐4; i) Hierarchical cluster analysis of 78 kinds of soluble proteins in the culture medium of primary macrophages treated with TH relative to that with PBS according to molecular function; j) Pathways enriched by the up‐regulated proteins shown in the panel (i) with Gene Ontology Pathway analysis. Images are representative of three independent experiments. Results are shown as mean ± SD. * P < 0. 05 after ANOVA with Dunnett's tests. We next analyzed the influence of TH on the primary bone marrow derived macrophages (BMDM). First, both flow cytometry analysis (Figure 1 f) and IF staining (Figure 1 g) indicated that the TH treatment up‐regulated the expression of CD206 in the macrophage population. Meanwhile, quantitative PCR analysis revealed increased levels of CCR2 and Arg‐1 and decreased expression of CCR7 and iNOS‐2 in the TH‐treated BMDM (Figure 1 h). As CCR7 and iNOS‐2 are M1 markers, while CD206, CCR2, and Arg‐1 are typical M2 markers, 30 the data suggested that TH triggered a M2‐way polarization of BMDM. Next, as revealed by the antibody assay and follow‐up ontology analysis, TH treatment upregulated the levels of CSF (25%), anti‐inflammatory cytokines (50%), pro‐angiogenic factors (60%) and growth factors (78. 57%) in BMDM (Figure 1 i), enriching the pathways associated with anti‐inflammation, angiogenesis and EGF receptor (Figure 1 j). Thus, as it endowed the primary fibroblasts with the phenotypes and functions of CAF, TH could also transform the primary macrophages into an M2 phenotype functionally similar to TAM. These TH‐educated cells switched to secrete cytokines that were typically produced by CAF and TAM in shaping up TME. These findings validated that TH could be used in engineering scaffolds to create a TME‐mimetic niche for cancer cell growth. 2. 2 Creation of TME‐Mimicking Microenvironment In Vivo by Implantation of THAG—A TH‐Containing Hydrogel Having validated the effect of TH on remodeling stromal cells, we speculated whether the TME‐like niche in vivo could be constructed by physically mixing TH with an injectable hydrogel and subcutaneously implanting the mixture into mice. We prepared the hydrogel by chemically crosslinking agarose and gelatin (ACG), according to previously reported methods, 31 and characterized it with scanning electron microscope (SEM) and IR spectrum (Figure S2a and b, Supporting Information). It also had a tunable phase transition temperature, mechanical strength, and flexibility (Figure S2c–e, Supporting Information). Then, TH (≈2–3 mg protein) was mixed with the liquid ACG (1% in PBS) at 42 °C, and the mixture solidified and became TH‐containing ACG (THAG) when the temperature decreased to 37 °C (Figure S2d, Supporting Information). THAG demonstrated excellent support of cell growth, as both fibroblasts and macrophages adhered well to its surface and proliferated both on its surface and inside its matrix (Figure S2f–h, Supporting Information). Additionally, fibroblasts and macrophages encapsulated in THAG expressed the markers of CAF (α‐SMA high /SDF‐1 high ) and M2‐polarization (CD206), respectively (Figure S2h, Supporting Information). We then injected THAG subcutaneously into the back of C57BL/6J mice every third day for four times (500 µL each time, Figure 2 a), with the same ACG gel with PBS as control. To further enhance angiogenesis, we added excessive bFGF (1000 U mL −1 ) in THAG (Figure S3, Supporting Information). A series of histological and cellular analyses indicated that THAG facilitated angiogenesis to a greater extent, as evidenced by a remarkably higher density of new blood vessels (gross view, Figure 2 b; H&E staining, Figure 2 c), more hemoglobin (Figure 2 d) and CD144 + cells (Figure 2 e), and elevated expressions of CD31 (an endothelial marker) and α‐SMA (a pericyte marker, Figure 2 f). Intriguingly, the implanted THAG appeared more transparent than the ACG gels (Figure S3a, Supporting Information), indicating a mild foreign body reaction to THAG. Further histological analysis showed less foreign body granuloma formation around THAG and fewer inflammatory cells infiltrated into THAG, as compared with the ACG group (Figure 2 g). On the contrary, significantly more CAFs invaded into THAG than ACG, as evidenced by IF staining (Figure 2 h) and flow cytometry analysis for Vimentin and SDF‐1 (Figure 2 i). Fibroblasts are adept at remodeling tissue microstructures by producing ECMs. Indeed, both Type I and IV collagens, which are crucial for the initial stage of tissue morphogenesis, 32 were found much more abundant in THAG than in ACG (Figure 2 j and k). Figure 2 Effects of THAG, an engineered matrix mimicking tumor microenvironment, in supporting neo‐tissue formation in vivo. a) Schematic diagram of subcutaneous injection of hydrogels incorporating tumor homogenate (THAG) or saline (ACG) into the back of C57BL/6J for different times; b) Gross view of in situ neo‐tissue formation in THAG or ACG at indicated time points, with regions from Day 12 samples magnified in the right panel; c) H&E analysis of neo‐tissues in THAG or ACG implanted into the back of mice (Scale bar = 100 µm); d) Hemoglobin content in the neo‐tissue in THAG or ACG 9 days after the first injection ( n = 7 per group); e) FACS analysis of CD144, a typical marker of endothelial cell, in neo‐tissue in THAG or ACG at Day 9; f) Representative images of immunofluorescent staining for CD31/α‐SMA in two groups at Day 9. Blood vessel density was calculated based on the confocal microscopical images shown on the right panel; g) Representative images of H&E staining illustrating foreign body reactions to THAG or ACG at Day 9(Scale bar = 250 µm); h) Representative images of immunofluorescent staining for vimentin and SDF‐1, typical fibroblast markers, in neo‐tissue in THAG or ACG (Scale bar = 100 µm); i) FACS analysis of cells expressing vimentin and SDF‐1 in neo‐tissue in THAG or ACG; j) Representative images of immunofluorescent staining for collagens type I and IV in neo‐tissue in THAG or ACG (Scale bar = 100 µm); k) Quantitative measurements of collagen contents in neo‐tissue in THAG or ACG ( n = 7–9 per group); l) The hardness of the neo‐tissue in THAG or ACG compared with that of heart, lung, spleen, lung, and kidney ( n = 7 per group); m) Scanning electron microscope of neo‐tissues in the two groups at indicated time points (red arrows indicate cells) (Scale bar = 20 µm). Images are representative for three independent experiments. Results are shown as mean ± SD. * P < 0. 05 after ANOVA with Dunnett's tests. In addition to facilitating tissue growth, THAG also exhibited proper mechanical strength. The hardness of THAG (≈10 HA) was much comparable to that of most organs and lower than that of ACG (over 20 HA) (Figure 2 l). Further SEM examination indicated that the softness of THAG was possibly owing to its degradation in vivo in response to an increase in cell density (Figure 2 m). The above data demonstrated that the implanted THAG created a desirable niche for tissue growth, with high degrees of angiogenesis and fibroblasts invasion as well as low degrees of foreign body reactions—the latter of which also suggested a successful suppression of the body's immune response against THAG and would be further investigated in the following experiments. 2. 3 Establishment of Immunosuppressive Niche in Implanted THAG We continued to evaluate whether THAG could create an immunosuppressive environment within its matrix after implantation in mice. This feature was crucial to our hypothesis as it served to protect the growth of XCC inside the immunologic enclave. The earlier in vitro tests showed that TH successfully educated macrophages into an immunosuppressive phenotype. Consistently, the neo‐tissue niches formed in THAG in vivo exhibited high immunosuppressive activities. First, a significantly higher proportion of macrophages (F4/80 + ) exhibited an M2‐like phenotype in THAG than in ACG, though the total numbers of macrophages were similar in these two groups ( Figure 3 a and Figure S4, Supporting Information). Next, the number and composition of lymph cells were dramatically changed by TH—the percentage of CD4 + T cells was much lower in THAG than in ACG (Figure 3 b). More importantly, the CD25 + Foxp3 + Treg cells, which play an indispensable role in maintaining immunological tolerance, 33 was significantly higher in the CD4 + proportion, five times of that in ACG (Figure 3 c). Besides, CD8 + T cells were barely detectable in THAG group (Figure 3 d), where CD19 + B memory cells were also dramatically fewer than in ACG (Figure 3 e). Further profiling of the soluble factors demonstrated up‐regulated CSF (100%), anti‐inflammatory cytokines (75%), growth factors (92. 86%), angiogenesis proteins (69. 23%) as well as ECM components (80%) in THAG (Figure 3 f), resulting in the enriched pathways relating to anti‐inflammation, EGF signaling and angiogenesis, which are all essential for neo‐tissue formation (Figure 3 g). Among the soluble factors, the levels of TGF‐beta1, IL‐10, VEGF, and EGF were determined by enzyme linked immunosorbent assay (ELISA) to be up‐regulated in THAG (Figure 3 h). These cytokines are well‐known immunosuppressive and mitogenic factors. Figure 3 Formation of immunosuppressive environment in THAG. a) (left) Representative FACS analysis of CD206/F4/80 and CD86/F4/80 and (right) M1/M2 proportion (calculated by the number of CD86 + /F4/80 + cells divided by that of CD206 + /F4/80 + cells) of the cells in neo‐tissue at Day 9 after first injection of THAG or ACG; b–e). Representative FACS analysis of b. CD4 + T; c. CD25 + Foxp3 + Treg (CD4 + gated); d. CD8 + T and e. CD19 + B cells in neo‐tissue as described above; f) Hierarchical cluster analysis of 78 kinds of soluble proteins in the supernatant of neo‐tissue homogenate in the animals treated with THAG relative to that with ACG according to molecular function; g) Pathways enriched by the up‐regulated proteins shown in the panel f with Gene Ontology Pathway analysis. h) The levels of growth factors (TGF‐β1, EGF and VEGF) and inflammatory cytokines (IL‐10, IL‐6, MCP‐1, IFN‐γ, TNF‐α and IL‐12p70) in neo‐tissue as in panel a. Results are shown as mean ± SD ( n = 8–10 mice per group; * P < 0. 05 versus ACG; ns = not significant). Besides, we assessed the possible effects of TH on the normal functions of the immune system. We compared the numbers of CD4 + /CD8 + T cells and CD19 + B cells in the peripheral blood around ACG and THAG implants against the untreated mice; we found no significant difference among these three samples (Figure S5a, Supporting Information). Consistently, we also found that the expression of IL‐10, IL‐6, MCP‐1, IFN‐γ and TNF‐α, the typical inflammatory factors, were also similar between THAG and sham‐operated control (Figure S5b, Supporting Information). These results indicated that THAG had little influence on the global immune system of the mice. As such, THAG was established as a TME‐mimetic, immunosuppressive enclave while having little influence on the global immune system, which was highly desirable according to our hypothesis. 2. 4 Growth of XCC in Implanted THAG We next investigated the growth of XCC in the implanted THAG. Our hypothesis was that this artificial TME‐mimetic niche could support xenograft cells to grow in immunocompetent mice. Human embryonic kidney epithelial cell 293T transfected with lentivirus labeled with GFP (GFP‐293T) were injected into THAG or ACG implanted in mice ( Figure 4 a). Observation with a bioluminescent imaging system through a period of 9 d demonstrated that the cells continued proliferating in THAG but did not grow in ACG (Figure 4 b). Further flow cytometry analysis confirmed that the cells were proliferating and reached nearly a quarter of the total population in THAG; while they failed to grow in ACG (Figure 4 c). Quantification of the above data showed that the number of 293T cells had increased by 20 folds since injection (Figure 4 d). Figure 4 Growth of xenogeneic cell lines in THAG. a) Schematic diagram of subcutaneous injection of trans‐species 293T cells infected with lentivirus encoding GFP (GFP‐293T) into neo‐tissue in THAG or ACG. b) Bioluminescence imaging of the mice injected with GFP‐293T cells in two groups. c) FACS analysis of GFP‐positive cells in two groups treated as in panel (a). d) Quantitative analysis of GFP‐positive 293T cells calculated by multiplying the total cell number and the percentage of GFP‐293T‐positive cells ( n = 7 per group). e) Confocal microscopical view and f) histological analysis (H&E) of GFP‐293T cells in THAG or ACG treated as in panel a (Scale bar = 100 µm); g) Images of immunofluorescent staining for human histone and h. type I collagen in THAG or ACG group treated as in panel (a) (Scale bar = 100 µm). Images are representative for three independent experiments. Results are shown as mean ± SD. * P < 0. 05 after ANOVA with Dunnett's tests. Interestingly, during the 9‐day culture in THAG, the 293T cells underwent a typical epithelial morphogenesis and developed into a regularly‐shaped tubular structure analogous to epithelial organoids (Figure 4 e and f). 34 We further stained the tissue sections with an antibody against human histone and confirmed that these epithelium‐like tissues were of human origin, i. e. , derived from the implanted 293T cells (Figure 4 g). Another notable result was the abundant production of type I collagen in THAG (Figure 4 h); this ECM component is required for both the growth of the implanted cells and tissue morphogenesis. Meanwhile, we examined whether the introduction of human‐origin 293T cells would change the immunosuppressive microenvironment already established within THAG. We determined the levels of a series of inflammatory cytokines both (i) inside the THAG/ACG (Figure S6a, Supporting Information) and (ii) in the serum and different organs (Figure S6b, Supporting Information) with ELISA. The outcomes suggested that, in agreement with the earlier findings, the immune response within the THAG implants was kept low in ACG, while the global immune system was unaffected by either implants. We also found that the implanted 293T cells did not migrate out of THAG and spread to any normal tissue in the mice body (Figure S6c, Supporting Information), which was crucial for the safe use of this method. These data suggested that THAG could provide a desirable environment for trans‐species tissue growth, by not only supporting the proliferation of xenograft cancer cells but also facilitating the reconstitution of their native morphologies. These advantages may favor the development of xenograft cells into specific tissue structures in vivo. 2. 5 Formation of Specific Tissue Structures by Xenograft Cells in Implanted THAG The above findings that 293T cells could undergo epithelial morphogenesis in THAG inspired us to ask whether other types of xenograft cells would do so when implanted into THAG. We selected four representative cell lines—including two epithelial cells Caco2 and Hep G2, one neural cell line SH‐SY5Y and one endothelial cell line HUVEC—and implanted them into THAG or ACG in the back of immunocompetent mice. All of them grew normally in THAG but not in ACG (Figure S7, Supporting Information). Remarkable epithelial morphogenesis was recapitulated in the neo‐tissues formed in the cell‐laden THAG, where Caco2 and Hep G2 cells developed into tubal organoids ( Figure 5 a, c, Figures S8 and S9, Supporting Information). 35, 36 IF co‐staining for human retinol‐binding protein II (RBP 2) or human hepatic nuclear factor 4 alpha (HNF 4α)—markers of Caco2 and Hep G2 cells, 34, 37 respectively—and E‐cadherin further testified that the tubular structures were constituted by the formed epithelium organoids (Figure 5 b and d). Figure 5 Morphology of specific tissue structures formed by human cell lines in THAG. a) Histological examination (H&E staining) of epithelioid structures and b) immunofluorescent staining for retinol‐binding protein II (RBP 2, marker of Caco2 cells) and E‐cadherin (marker of epithelium) in Caco2 cells‐laden THAG implanted in mice; c) Histological examination (H&E staining) of epithelioid structure and d) immunofluorescent staining for hepatic nuclear factor 4 alpha (HNF 4 α, marker of Hep G2 cells) and E‐cadherin in Hep G2 cells‐laden THAG implanted in mice; e) Histological examination (H&E staining) of neonatal neuron tissue and f) immunofluorescent staining for Tuj 1 and NeuN (both markers of neuron cells) in human neuroblastoma epithelial cell SH‐SY5Y‐laden THAG implanted in mice; g) Histological examination (H&E staining) of vascular structure and h) immunofluorescent staining for human CD31 (marker of endothelium), mouse α‐SMA and vimentin (markers of tunica media and external) in primary human umbilical vein endothelial cell (HUVEC)‐laden THAG implanted in mice; i) RT‐PCR analysis of specific markers of the epithelial, neural and vascular structures ( n = 7–9 per group); j) Immunofluorescent staining for human histone expressed by different xenograft cells in THAG (those in ACG as control shown in the inset figures). Images are representative for three independent experiments. Scale bar = 100 µm. Results are shown as mean ± SD. In the human neuroblastoma (epithelial cell SH‐SY5Y)‐laden THAG, the neonatal neuron tissue successfully developed during the period of the experiment (Figure 5 e and Figure S10, Supporting Information). At day 3 after the first injection, neuroepithelial cysts, one of the early stage structure of neuroepithelium, developed in the THAG group and significantly expressed E‐cadherin (Figure S11a, Supporting Information). 38 At day 8, which was 5 d after the second injection, neurons started to grow basally. Increasing numbers of cysts were observed, with up‐regulated N‐cadherin and down‐regulated E‐cadherin (Figure S11b, Supporting Information). At day 12, or 6 days after the third injection, the cells with cysts continued neural differentiation by extending dendrites and forming network, as evidenced by the staining for neurogenesis markers NeuN and βIII‐tubulin (Tuj 1) (Figure 5 f). In addition to these cancer cells, HUVEC, which are normal primary endothelial cells of human origin, also formed tubular structures after implantation into THAG (Figure S12, Supporting Information and Figure 5 g). Moreover, the implanted human cells successfully recruited tunica external and media cells from the mouse body to form a mature vasculature structure, 39 as evidenced by IF staining for human CD31 (endothelial layer, internal), anti‐mouse α‐SMA (smooth muscle cells, middle), and anti‐mouse vimentin (connective tissue, external; Figure 5 h). Besides, the transcriptional levels of both Lgr5+/PRMT and Sox9/GATA4, which are important genes during intestinal and hepatic epithelium generation, 40, 41 were up‐regulated as Caco2 and Hep G2 grew. Similarly, early neurogenic markers, such as Fgf5 for primitive ectoderm and Sox1 for neuroectoderm, 38 were expressed along the cultivation of SH‐SY5Y in THAG. And the levels of Flk1 and Tie2, two markers of endothelial development, 39 also increased as the humanized vessels developed in THAG (Figure 5 i). Further staining for species‐specific human histone confirmed that these well‐developed tissues with morphological features were derived from the seeded human cells (Figure 5 j). In summary, these results suggested that the microenvironment in THAG not only accommodated the growth of different xenograft human cells but also supported their development into specific tissues. Remarkably, these human tissues de novo formed in the mouse body underwent normal differentiation, organized into correct structures and exhibited typical morphologies. 2. 6 Physiological Functions of the Xenograft Tissues in THAG To test whether the xenograft tissues could exert normal physiological functions, we implanted a rat β‐cell line (INS‐1) and an immortalized human hepatocyte cell line (THLE‐3) into THAG created on the back of type I diabetic and normal mice, respectively. We induced diabetes in THAG‐ or ACG‐implanted C57BL/6J mice by intraperitoneally injecting streptozotocin (STZ) for consecutively five days at a dose of 45 mg kg −1. Blood glucose levels were monitored every third day; only mice with blood glucose levels above 200 mg dL −1 for two consecutive days were considered diabetic and received subsequent INS‐1 cell injection ( Figure 6 a). Figure 6 Function of INS‐1 cells in the rat ‘pancreas islet‐like’ tissue created in THAG in mice. a) Schematic diagram of diabetes reversal by subcutaneous injection of INS‐1 cells into neo‐tissue in THAG implanted in the back of mice; b) The levels of blood glucose monitored every third day following the diabetic mice were generated. The arrows named as a, b and c, respectively, indicate the corresponding treatment described as in panel (a); c) Histological examination (H&E staining) and d) immunofluorescent staining for insulin and nuclear protein Nkx6. 1 of neo‐tissue 27 days after the first injection of INS‐1 into the THAG or ACG implanted in mice; e) Blood glucose levels monitored at 0, 15, 30, 60, 90 and 120 min after glucose stimulation (2 g kg −1 ) during the intraperitoneal glucose tolerance tests (IPGTT) 27 d after the injection of INS‐1 cells; f) Rat and mouse insulin in the serum of different groups measured at 30 min during IPGTT treated as in panel e. Images are representative for three independent experiments. Scale bar = 100 µm. Results are shown as mean ± SD ( n = 7–9 per group). Implantation of INS‐1 cells efficiently lowered the level of blood glucose in the THAG‐bearing mice after three days, and this effect maintained through a period of 27 d, during which THAG was replenished every 6 d. We intentionally stopped replenishing THAG at day 27, and the animals returned to hyperglycaemia after another 18 d. On the contrary, these cells failed to change the condition at all in the ACG‐bearing mice throughout the 45 d (Figure 6 b). A set of histological examinations coupled with IF staining revealed a clear, Langerhans‐like structure expressing insulin and the nuclear protein Nkx6. 1 in the new tissue within THAG (Figure 6 c, d and Figure S13, Supporting Information). Furthermore, we tested the response of this islet‐like xenograft tissue to glucose challenge—the intraperitoneal glucose tolerance test (IPGTT). The mice were fasted for overnight and intraperitoneally administrated with glucose (2 g kg −1 body weight) before their blood glucose levels were monitored at the indicated time points. The IPGTT outcomes indicated that the level of blood sugar was controlled well in the THAG‐bearing mice, suggesting that the “pancreas islet‐like” tissue developed in THAG secreted insulin sensitively and timely in response to glucose (Figure 6 e). Interestingly, we confirmed that the insulin presented in the serum of the THAG‐bearing mice was of rat origin, i. e. produced by the xenograft cells (Figure 6 f). Having validated the function of xenograft pancreas‐like tissue growing in THAG, we generated a “humanized liver‐like tissue” using the same approach and examined its function in mice. The injected human liver epithelial cells (THLE‐3) demonstrated a typical large, polygonal morphology of the liver cells ( Figure 7 a and Figure S14, Supporting Information), expressing human histone, CYP2D6 and HNF 4α (both are markers of liver cells) in the newly formed tissue in THAG (Figure 7 b–d). These humanized, liver‐like tissues faithfully exerted the function of the liver, producing human serum albumin at 6. 053 ng mL −1 in serum and up to 18. 862 ng mL −1 in the new tissue growing from THAG, starting at around day 10 (4 d after the third injection of THAG). However, the transplanted cells into ACG failed to produce albumin (Figure 7 e). Moreover, after administrating the mice with debrisoquine (DB), which human and mice metabolized differently, 42 we detected 4‐hyroxydebrisoqune (4OH‐DB)—a human‐specific metabolite—in the urine collected from mice implanted with THLE‐3‐laden THAG. Whereas, this agent could not be metabolized by the mice with ACG (Figure 7 f). Thus, the THAG‐supported liver‐like tissue could secrete human albumin and performed drug metabolism as the human liver does. These findings suggested that THAG created in the mouse body not only supported the growth of human or rat cell lines into specific tissues, but also facilitated these trans‐species tissues to exert their typical physiological functions in the mouse body. Figure 7 Function of THELE‐3 cells in the human “liver‐like tissue” developed in THAG in mice. a) Histological examination (H&E staining) of THLE‐3 cells‐laden THAG or ACG implanted in mice; the outlined regions highlight the large sized cells in polygonal shape, which is a typical morphology of liver cells; b) Immunofluorescent staining for human histone and c) CYP2D6 (specifically expressed by human liver cells) and d) Nuclear protein HNF 4α in the samples as in panel (a); e) The levels of human albumin detected in serum and local tissues of the mice implanted with THLE‐3 cells‐laden THAG or ACG; f) Metabolic ratios determined by dividing the AUC0‐8h (the area under the curve from time 0 until 8 h) ratio of 4‐hyroxydebrisoqune (4‐OHDB) to that of debrisoquine (DB) in untreated mice as well as mice implanted with THLE‐3 cells‐laden THAG or ACG. Images are representative for three independent experiments. Scale bar = 100 µm. Results are shown as mean ± SD ( n = 7–9 per group). * P < 0. 05 after ANOVA with Dunnett's tests. 3 Discussion One of the most exciting goals in regenerative medicine is to develop functional tissue/organs in the body using transplanted cells. To date, attempts towards this aim have been substantially hampered by several major obstacles, notably including the lack of sufficient cells suitable for transplantation and immune rejection against the “foreign” cells after transplantation. 3, 43 In this study, we showed that xenogeneic cancer cells, delivered in an engineered matrix mimicking the TME, could successfully generate new functional tissues in vivo without being eliminated by the host immunity. These cells demonstrated an unparalleled (and underappreciated) potential in serving as a new, ample source of therapeutic cells for transplantation, while the TME‐mimetic niche played an indispensable role in supporting the survival and function of these cells. A key inspiration to our investigation was the similarity between the tumor and transplanted tissues—both are recognized as “foreign” by the host immunity while both require blood supply (which inevitably brings immunocytes) from the host. 44, 45 The tumor could tactically solve this dilemma thanks to its unique TME with two distinctive features—pro‐tumor growth and immunosuppressive. 17, 18, 46, 47 Thus, we assumed that re‐creation of analogous TME would provide the proper ground to support xenograft cells to grow in the host body. Since little was known about the detailed mechanisms of TME formation, 48, 49 the most efficient and straightforward way to establish a TME‐mimetic niche would be implanting a tumor homogenate‐containing hydrogel matrix (THAG) in vivo, as we eventually performed. Our strategy proved successful. A comprehensive set of data showed that THAG preserved the biological characteristics of TME and effectively developed into an immunoprotective, vascularized niche. Such niche, as expected, effectively facilitated the settlement of xenogeneic cells of various tissue origins—from human kidney, liver, intestine, endothelium, neuron to rat pancreas—and promote their development into specific, fully functional tissues in immune competent mice. Notably, in agreement with our hypothesis, the engineered THAG not only created optimal conditions for tissue growth but also prevented immunogenic rejection from the host—which is the major cause of failure for xenografts and allografts. 15 Although some recent studies devised “immuno‐isolating” materials to prevent the infiltration of immunocytes into the implants, 50 such protection also cut off blood vessels and exclude the entry of other stromal cells that are essential for the neo‐tissue formation, leading to low vascularization and poor host integration that eventually cause implant failure. By creating the TME‐mimicking niche, our approach successfully translated the tumor's strategy to escape immune attack into the growth of xenogeneic tissues. Our data showed that THAG was well integrated with the circulation system of the mice body while received minimal foreign body reactions. Instead of being barred from the implants, macrophages and fibroblasts—the two major stromal cells of high plasticity—abundantly infiltrated into THAG, vastly changed their phenotypes and actively shaped up the immunosuppressive enclave by further regulating the functions and numbers of the other stromal cells, such as the endothelial cells and lymph cells. As a result, inside the enclave, the transplanted “human” or “rat” tissues were formed, without being eliminated by the host immune rejection; out of the enclave, neither the spreading of xenogeneic cells nor the compromise in the host immunity was found. These findings collectively highlighted the efficacy and safety of THAG as an engineering matrix for tissue regeneration. There are at least two interesting directions for furthering our exploration. First, though we have validated the effect of tumor homogenate in creating TME and analyzed its components with proteomic tools, it is also a critically deficient aspect for its possibly uncontrollable effects on the body, especially the immune system. We still need to investigate the mechanism in this process and identify the key components out of the many, to make the process more controllable. Second, though we had foreseen the growth of xenogeneic cancer/immortalized cells in THAG, we did not predict the intriguing outcomes of morphogenesis, which indicated a considerable extent of differentiation of these cell lines in THAG. Further investigations are in demand to explore the mechanisms underlying their morphogenesis. Besides, according to the properties of the engineered tissues demonstrated in our study, we prospect the technology is more suitable to construct endocrine organs, such as the pancreas, than to engineer other types of organs. In the future, it is challenging and inspiring to design more complex tissues or organs with elaborated improvement of the present method. In summary, in the present study, we demonstrated the creation of an engineered matrix that exerts the unique features of TME to support the growth of xenogeneic tissues. The designed THAG matrix, an injectable hydrogel system incorporating tumor extract as an active ingredient, could enable the survival, proliferation and function of various xenogeneic cell lines in immunocompetent mice. Using this approach, pancreas and liver tissues respectively from rat and human cells were successfully developed in THAG implanted into the back of mice. Our findings suggest that application of a TME‐like, immunosuppressive niche, in combination with employment of xenogeneic cells, may potentially solve two fundamental challenges in cell transplantation—i. e. , low cell availability and immune rejection. The novelty, feasibility, and openness of our approach may inspire the design of new strategies for tissue engineering and other cell‐based therapies. 4 Experimental Section Reagents : Agarose, gelatin, N, N ′‐Carbonyldiimidazole (CDI), dimethyl sulfoxide (DMSO), STZ, and all other chemicals used in this study were purchased from Sigma‐Aldrich (St. Louis, MO, USA) unless otherwise stated. Interleukin‐4 (IL‐4) and bFGF were purchased from PeproTech (New Jersey, USA). Debrisoquine and 4‐hyroxydebrisoqune were obtained from Toronto Research Chemicals (TRC, Toronto, Canada). Synthesis and Characterization of Agarose‐CDI‐Gelatin Conjugate Hydrogel : ACG were synthesized according to a previously reported method. 31 Briefly, agarose powder (4%; w/v) was suspended in DMSO, heated at 80 °C to dissolve and cooled down to room temperature. Next, CDI (1%) was added to activate the hydroxyl groups on the sugar chain of agarose. Subsequently, gelatin (6%) was mixed with the solution and stirred overnight at room temperature. The resulting solution was placed in a dialysis tube, dialyzed against distilled water to remove DMSO and remaining CDI and lyophilized to obtain ACG. ACG was characterized by SEM (SFEG Leo 1550, AMO GmbH, Aachen, Germany) and Fourier's transform infrared spectroscopy spectra (Shimadzu Corp. , Kyoto, Japan) with the 4000–400 cm −1 scanning range. The mechanical properties of the hydrogel were tested using a table‐top material tester (EZ‐Test‐500 N; Shimadzu, Kyoto). Briefly, the solutions were poured into cylindrical tubes and cooled in an ice bath. The obtained cylindrical gels were compressed at a crosshead speed of 3. 0 mm min −1 for the stress–strain profiles test. Rheological measurements were performed in a TA Instruments AR1000 Rheometer using the parallel plate shear mode. Dynamic viscoelastic measurements were performed to measure the storage modulus, G′, the loss modulus, G″, and the loss angle tangent, tan (delta). ACG hydrogel solutions of different concentration were examined with the temperature sweeps between 65 and 25 °C. Extracts from TH and Proteomics Analysis for LC‐MS Profile of TH : To generate the heterotopic tumor model, mouse sarcoma cell line S180 cells (1 × 10 6 ) were injected subcutaneously into the left arm pits of the animals. Mice bearing implanted tumors were sacrificed when the sizes of the implanted tumors reached about 0. 5 cm. The tumors were removed, immersed in ice‐cold phosphate buffered saline pH 7. 4 (PBS), minced, and washed with the same solution. The mince was homogenized with a Teflon/glass homogenizer. The homogenate was centrifuged at 12 000 rpm for 10 min at 4 °C. The pellets were discarded and the supernatant was collected as TH. The protein in the supernatant was quantified using Bradford assay (Biorad, CA, USA) and then stored at −80 °C. The composition of TH was analyzed by label‐free LC‐MS adhered to a method described previously. 51 Briefly, total protein (100 µg) was reduced by adding dl ‐dithiothreitol (Sigma‐Aldrich) (1 m, 60 °C, 1 h), and free cysteines were alkylated with 1 m iodoacetamide (Sigma‐Aldrich) (room temperature, 10 min in the dark). The alkylated proteins were further washed with 100 × 10 −3 m tetratehylammonium bromide for three times at 4 °C for 20 min by centrifugation at 12000 rpm. Then the protein was digested with porcine sequencing grade trypsin (LC‐MS Grade, Sigma‐Aldrich) overnight at 37 °C. The samples were then subjected to LC‐MS analysis on a Shimadzu UFLC 20ADXR HPLC system in‐line with an AB Sciex 5600 Triple TOF mass spectrometer (AB SCIEX, Framingham, Massachusetts State, USA). Samples were analyzed in three technical replicates. Identification of peptides and proteins from continuum LC‐MS data was performed with the ProteinPilot 4. 5 software (AB SCIEX), based on the Paragon database search algorithm. Proteins were analyzed by searching the mouse taxon of the UniProtKB/ SwissProt database (release 2011_11) using the precursor and fragmentation data provided by the LC‐MS acquisition method. Then, LC‐MS profile of TH was analyzed by Gene Ontology ( http://geneontology. org/ ) according to biological process, molecular function, and cellular component or with gene ontology pathway database. Both “GO analysis” and “Pathway analysis” were analyzed in the standard enrichment computation method. Cells Preparation, Isolation, and Treatment—Cell Preparation : Human embryo kidney epithelial cell 293T, human colon epithelial cells Caco2, human liver epithelial cells Hep G2, human neuroblastoma cells SH‐SY5Y, primary human umbilical vein endothelial cells (HUVEC), rat insulinoma beta cell INS‐1, and immortalized human liver epithelial cells THLE‐3 were obtained by Stem Cell Bank, Chinese Academy of Sciences (Shanghai, China). GFP stably expressing 293T cells (GFP‐293T) were sorted in the presence of puromycin (10 µg mL −1 ; Sigma‐Aldrich) after the cells were transfected with GFP‐labeled lentivirus carrying puromycin‐resistant marker. Cells were cultured in DMEM or RPMI 1640 medium containing 10% fetal bovine serum (Thermo Scientific, MA, USA), harvested at ≈80% confluence, washed twice with phosphate buffer saline (PBS), counted and re‐suspended in PBS before injection. Isolation and Treatment of Primary Mammary Fibroblast : The primary fibroblasts were isolated form mammary glands of female C57BL/6J mice according to a reported protocol. 52 Briefly, glands were digested with collagenase I and hyaluronidase in DMEM/F12 (Thermo Scientific). After digestion, the tissues were washed and cultured in DMEM/F12 media supplemented with 10% FBS and 1% penicillin/streptomycin at 37 °C. The cells were sub‐cultured after the cultures reached confluency. After 2–3 passages, stromal fibroblasts in high homogeneity were obtained. All the stromal fibroblasts used in the experiments were at less than ten passages to maintain the closest phenotype to the primary tissues. To analyze the effects of TH on the primary fibroblast, the cells were treated with TH (30 µg mL −1 ) for 48 h. Isolation and Treatment of Bone Marrow Derived Macrophages : BMDM were collected and cultured according to a published protocol. 53 Briefly, bone marrow cells from the femurs of C57BL/6J mice were harvested by flushing with Hanks' balanced salt solution without Ca 2+ /Mg 2+. A single‐cell suspension was created by passing through a 21‐gauge needle. Non‐adherent cells were removed after culture for 4 h and purified monocytes were incubated for 7 d in RPMI 1640 supplemented with FBS (10%) and M‐CSF (50 ng mL −1 ) to obtain macrophages. Activation of BMDM was carried out by the addition of TH (30 µg mL −1 ) for 48 h. M2 macrophages induced by IL‐4 (20 ng mL −1, overnight) were set as positive control. Cell Growth on or in TH‐Incorporated ACG In Vitro : The gel plates were prepared by pouring of 1% ACG (2. 0 mL) or THAG (ACG mixed with TH) into each well of the 6‐well culture plates. Primary fibroblast or macrophage cells were seeded on the gel plates (5 × 10 5 cells/well) and cultured for 1, 3, and 5 d. CCK‐8 kit (Dojindo Laboratories, Kumamoto, Japan) was used to examine the proliferation of these two cells at indicated time points, with cells cultured in monolayer on tissue culture polystyrenes (TCPS) as control. After 3 d, the live cells were stained with Calcein‐AM (Thermo Scientific) and their morphology was assessed under a TE2000‐U inverted phase‐contrast microscope (Nikon, Tokyo, Japan). Furthermore, to analyze the morphology of these cells in THAG in vitro, primary fibroblast cells or macrophage cells were mixed into the pre‐sterilized THAG (1%) and deposited into each well to reach a final cell density of 5 × 10 6 cells mL −1. The mixture was incubated at 37 °C for gelation. Then, cell culture medium was added and was changed every other day. Xenograft Cell Implantation Model : Male or female C57BL/6J mice (20 ± 2 g) of the same ground were obtained from Model Animal Research Centre of Nanjing University (Nanjing, China). All animals had free access to rodent chow and water, and were treated in strict accordance with the institutional ethical regulation on animal experiments. Animal protocols were reviewed and approved by the Animal Care and Use Committee of Nanjing University, and conformed to the Guidelines for the Care and Use of Laboratory Animals published by the National Institutes of Health. In Vivo Gelation Model : To generate the gelation model in vivo, the hydrogel solution was subcutaneously injected into the back of mice at 42 °C, and the injection site was immediately cooled by ice compress to quickly solidify the gel, based on the thermo‐reversal of 1% ACG from liquid form at 42 °C to solidified state at 37 °C (or lower temperature). ACG or ACG incorporated with TH (≈2–3 mg proteins; THAG) in sterile physiological saline mixed with or without bFGF (1000 U mL −1 ) in total 500 µL volume was injected into C57BL/6J mice continuously for different times every third day. At the indicated time points, the neo‐tissue niches formed in the injection site were extracted and subjected to histological analysis. Xenograft Cell Implantation : After the gel formation in vivo, different kinds of xenograft cell lines (5 × 10 7 ; GFP‐293T, Caco2, Hep G2, SH‐SY5Y, HUVEC and THLE‐3) mixed into 1% ACG or THAG (both incorporating bFGF) were injected into the site of gels for three times. Intravital Imaging : Mice implanted with xenograft GFP‐293T cells were monitored by intravital imaging. Briefly, 3 d after each injection, the animals were anesthetized with isoflurane and the hair at the site of injection was removed. GFP bioluminescence in the dorsum of mice was imaged by IVIS Lumina XR system (PerkinElmer, Waltham, MA, USA) at an excitation wavelength of 488 nm. The obtained images were analyzed using Velocity 3D Image Analysis Software (PerkinElmer, Waltham, MA). Determination of Proteins by Enzyme Linked Immunosorbent Assay : Serum or supernatant of tissue homogenate was collected and frozen at −80 °C before use. The levels of human albumin (ALB), rat insulin, mouse insulin, collagen type I, collagen IV, growth factors (TGF‐β 1, EGF, VEGF), and typical inflammatory factors (IL‐6, IL‐10, MCP‐1, IFN‐γ, TNF‐α, and IL‐12 p70) were measured using corresponding ELISA Quantitation Kits (Abcam, UK) according to the manufacturer's instructions. Profiling of Soluble Factors by Mouse Cytokine Array Kit and Mouse Angiogenesis Array Kit : To investigate the effect of TH on the tissue‐forming niches and the functional change of fibroblasts and macrophages, the supernatant of tissue homogenate and these two kinds of cells treated with TH were collected. Their soluble components were analyzed by membrane‐based antibody assay kit for 78 kinds of different soluble factors (R&D, USA; including Proteome Profiler Mouse Cytokine Array kit and Mouse Angiogenesis Array Kit) according to the manufacturers' protocols. The expression profiles of these 78 factors underwent hierarchical cluster analysis for molecular functions by Gene Ontology. The up‐regulated proteins were further investigated with Gene Ontology Pathway database, with pathways ranking within top 10 listed. Quantitative Hemoglobin Assay : The hemoglobin in tissue was detected with a quantitative hemoglobin assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturers' instructions. Briefly, the supernatant of neo‐tissue niches homogenate was mixed with potassium ferricyanide and potassium cyanide for 5 min. Absorbance was determined at 540 nm on a microplate reader. The results were referenced to a standard curve made by cyanmethemoglobin in different concentrations. The Hardness Test : The hardness of neo‐tissue niches and other tissues were measured using an elasticity‐measuring instrument equipped with a coil spring (Type A Durometer) (CL‐150SL, ASKER, Japan). A constant force was applied to each tissue sample and the mean value representing three individual tests was recorded. Flow Cytometry Analysis : Cells or tissues were digested to generate a single‐cell suspension, which was then blocked with 1% bovine serum albumin and incubated with the fluorescence‐conjugated monoclonal antibodies specific for the following cell surface markers in the dark for 30 min at 4 °C: CD206, CD86, F4/80, CD4, CD8, CD19, CD25, Foxp 3, CD144, vimentin, and SDF‐1 (eBioscience, MA, USA). The samples were centrifuged at 400–500 × g for 5 min at 4 °C to remove unbound antibody. After rinsing for three times, each sample was resuspended for analysis using a BD fluorescence activated cell sorter (FACS) Calibur (BD Biosciences, San Jose, CA, USA). Unconjugated antibodies and IgG controls were run in parallel to set the background. Besides, to analyze the growth of GFP‐293T cells as xenografts, the neo‐tissue was digested and subjected to FACS. The number of GFP‐positive cells was calculated by multiplying the total cell number in the injection site and the percentage of positive cells. Western Blotting : Proteins were separated by SDS‐PAGE and transferred onto the polyvinylidine difluoride membranes. The membranes were blocked with skim milk and incubated with diluted primary antibody—SDF‐1, α‐SMA, collagen type I (Col I), collagen type IV (Col IV), and glyceraldehyde‐3‐phosphate dehydrogenase (Abcam, Cambridge, MA) at 4 °C with gentle shaking overnight. After five times of washing with PBST (PBS with 0. 1% Tween‐20), the membranes were probed with horseradish peroxidase‐conjugated anti‐rabbit, anti‐mouse, or anti‐goat IgG (Life Technologies, Grand Island, NY, USA) for 1 h at room temperature. After rinsing, bands were visualized with fluorography using an enhanced chemiluminescence system (Cell Signaling Technology). RNA Isolation and Quantitative Real‐Time PCR : Total RNA from cells or tissues was extracted by using Trizol (Life Technologies). Real‐time PCR was performed by using a SYBR Prime Script RT‐PCR Kit (Takara Bio, Shiga, Japan) in an ABI 7300 Fast Real‐time PCR System (Applied Biosystems, FosterCity, CA). Each sample was analyzed in triplicates and repeated for three or four independent assays. The level of each gene was normalized to that of β‐actin. Primers used are listed below (Invitrogen, Carlsbad, CA, USA): Name Forward (5′‐3′) Reverse (5′‐3′) α‐SMA GAGCGTGAGATTGTCCGTGA GGTGCTGGGTGCGAGG SDF‐1 CTCTGCATCAGTGACGGTAA CTCTTCTTCTGTCGCTTCTT Col IV α1 TATGTCCAAGGCAACGAGC AACCGCACACCTGCTAATG Col I α CAACAGTCGCTTCACCTACAGC GTGGAGGGAGTTTACACGAAGC CCR2 CTCTACATTCACTCCTTCCACT TACAAACTGCTCCCTCCTT CCR7 TTCAACATCACCAATAGCAG GAAGGCATACAAGAAAGGG Arg‐1 AACACGGCAGTGGCTTTAACC GGTTTTCATGTGGCGCATTC iNos‐2 CAGCTGGGCTGTACAAACCTT CATTGGAAGTGAAGCGTTTCG GATA4 TGGCGTCTTAGATTTATTCAGGTTC TGTGCCAACTGCCAGACTACC PRMT GGAACACTCAATCCCAATAACC CTACTTTGACTCCTATGCCCACT Lgr5+ GTCAGTGTTCTTAGTTCAGGCAAAT CGTTCGTAGGCAACCCTTCTC Sox4 CTTTTCCCCTTTCTCCTTCTA TCTAACCTGGTCTTCACCTACTG Fgf5 TCTCCTTTTATCTGCCCCCT GAGCAGATGCACTCATTCCA Sox1 CGAGCCCTTCTCACTTGTT TTGATGTTGGGGGTAT Sox9 AGGAAGCTGGCAGACCAGTA TCCACGAAGGGTCTCTTCTC Flk1 GGCTCTTTCGCTTACTGTTCT CCTGCCTACCTCACCTGTTTC Tie2 AGGGAGTCCGATAGACGCTGT GGACCCATCAAATCCAAGAAG β‐actin GGTGTGATGGTGGGAATGGG ACGGTTGGCCTTAGGGTTCAG John Wiley & Sons, Ltd. The Diabetic Mice Model : The insulin‐dependent diabetic mice were prepared after the formation of neo‐tissues by intraperitoneal injection of STZ (45 mg kg −1 ; in acetate phosphate buffer, pH 4. 5) to C57BL/6J male mice for consecutive 5 days. The mice were considered diabetic when their blood glucose levels exceeded 200 mg dL −1 for two consecutive days. The Pancreas Islet‐Like Xenograft Tissue Model Formed by INS‐1 : After the diabetic mice were generated, the INS‐1 cells were injected into the neo‐tissue niches of diabetic mice every third day for three times. Next, the injection site was replenished with TH every 6 d until day 27. During the process, the levels of blood glucose were monitored every third day following the first injection of INS‐1 with glucose meters (Roche, Basel, Switzerland). Monitoring continued until the end of experiment, when the mice were euthanized and tissues retrieved. IPGTT : To detect the pancreas islet‐like xenograft tissue's responses to glucose challenge, IPGTT was conducted 27 d after the first injection of INS‐1 into ACG or THAG group. This assay could further assess the tissue's metabolic capacity in response to a glucose bolus. Briefly, animals were fasted overnight before receiving an intraperitoneal glucose bolus (2 g kg −1 ). The levels of blood glucose were monitored at 0, 15, 30, 60, 90, and 120 min after injection. Blood samples were also collected to measure the glucose‐stimulated secretion of rat insulin. In these experiments, blood was centrifuged for 10 min at 12 000 rpm and serum was stored at −80 °C until use. Assessment of Drug Metabolism Activity : To analyze the drug metabolism activity, the humanized “liver‐like tissue” model formed by THLE‐3 cells in ACG or THAG group was challenged with debrisoquine, a reagent known to be metabolized differently by mice and humans. Sham‐operated C57/B6J mouse was used as control. Briefly, after debrisoquine (2 mg kg −1 ) was orally administrated, urine (0–8 h) was collected in acetate buffer (0. 5 m, pH 5. 0). KOH (1 n ) was added to urine samples and incubated at 80 °C for 3 h, before being neutralized by equivalent volume of HCl (1 n ). Acetonitrile containing 1% acetic acid was added and centrifuged (15 000 rpm, 4 °C, 5 min). The supernatant was subjected to liquid chromatography‐tandem mass spectrometry (LC/MS/MS). The area under the curve from time 0 until the last measurable urine concentration (AUC 0‐8 ) was calculated using the linear trapezoidal rule. Metabolic ratios were determined by dividing AUC 0‐8 of 4‐hyroxydebrisoqune by AUC 0‐8 of debrisoquine. Histological Analyses : The tissue samples were collected, frozen at optimal cutting temperature medium and cut into sections for H&E staining according to the manufacturer's instructions with slight modifications. The stained sections were photographed at different magnification times under a microscope. Under blindfold conditions under a standard light microscopy, the tissue was randomly examined. Meanwhile, frozen tissue sections for immunofluorescent staining were fixed with 4% paraformaldehyde and stained with primary antibody at 4 °C overnight. The primary antibodies included anti‐human CD31, anti‐mouse CD31, anti‐human histone, anti‐human RBP 2, anti‐human HNF 4 α, anti‐human Tuj1, anti‐human neuron (NeuN), anti‐human CYP 2D6, anti‐mouse SDF‐1, anti‐mouse Col I, anti‐mouse Col IV, anti‐mouse F4/80, anti‐mouse CD206, anti‐mouse E‐cadherin, anti‐mouse N‐cadherin, anti‐mouse smooth muscle actin, anti‐mouse vimentin, anti‐rat insulin, and anti‐rat Nkx6. 1. Next, the sections were incubated with secondary antibody Alexa Fluor (Life Technologies) for 1 h at room temperature, followed by 4, 6‐diamidino‐2‐phenylindole (DAPI) for nuclear staining. All fluorescence including GFP bioluminescence were captured with a Nikon confocal microscope (C2+, Nikon, Tokyo, Japan) and analyzed using Nis‐element advanced research software (Nikon). Statistical Analysis : The results are expressed as mean ± standard deviation (SD). Data were statistically analyzed using Prism software (GraphPad) and assessed for normality or homogeneity of variance with D‐test and Levene test. Differences between multiple groups were compared using one‐way analysis of variance (ANOVA) with Dunnett's tests or, if appropriate, repeated measures ANOVA test with post hoc Bonferroni correction. Differences between two groups were evaluated using the unpaired Student's t ‐test. A value of P < 0. 05 was considered significant and “ns” stands for “not significant. ” Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201700678
| 2,017
|
Advanced Science
|
Valence State Manipulation of Cerium Oxide Nanoparticles on a Titanium Surface for Modulating Cell Fate and Bone Formation
|
Abstract Understanding cell–biomaterial interactions is critical for the control of cell fate for tissue engineering and regenerative medicine. Here, cerium oxide nanoparticles (CeONPs) are applied at different Ce 4+ /Ce 3+ ratios (i. e. , 0. 46, 1. 23, and 3. 23) to titanium substrate surfaces by magnetron sputtering and vacuum annealing. Evaluation of the cytotoxicity of the modified surface to cultured rat bone marrow mesenchymal stem cells (BMSCs) reveals that the cytocompatibility and cell proliferation are proportional to the increases in Ce 4+ /Ce 3+ ratio on titanium surface. The bone formation capability induced by these surface modified titanium alloys is evaluated by implanting various CeONP samples into the intramedullary cavity of rat femur for 8 weeks. New bone formation adjacent to the implant shows a close relationship to the surface Ce 4+ /Ce 3+ ratio; higher Ce 4+ /Ce 3+ ratio achieves better osseointegration. The mechanism of this in vivo outcome is explored by culturing rat BMSCs and RAW264. 7 murine macrophages on CeONP samples for different durations. The improvement in osteogenic differentiation capability of BMSCs is directly proportional to the increased Ce 4+ /Ce 3+ ratio on the titanium surface. Increases in the Ce 4+ /Ce 3+ ratio also elevate the polarization of the M2 phenotype of RAW264. 7 murine macrophages, particularly with respect to the healing‐associated M2 percentage and anti‐inflammatory cytokine secretion. The manipulation of valence states of CeONPs appears to provide an effective modulation of the osteogenic capability of stem cells and the M2 polarization of macrophages, resulting in favorable outcomes of new bone formation and osseointegration.
|
1 Introduction Understanding the interactions between stem cells and biomaterials surface is essential for controlling stem cell fates, including adhesion, proliferation, and osteogenic differentiation, in bone tissue engineering and regenerative medicine. 1 Tremendous advancements have been made in tailoring bone biomaterials to modulate stem cell fates, based on two general strategies: (1) modulation of the physical and chemical properties of biomaterial surfaces, 2 and (2) incorporation of bioactive cues onto biomaterial surfaces. 3 For instance, electroactive biomaterials have been designed to stimulate the osteogenesis of stem cell, thereby inducing subsequent bone regeneration. As inspired by the piezoelectric property of bones, Yu et al. designed the microscale piezoelectric zones (MPZs) with the use of K 0. 5 Na 0. 5 NbO 3 ceramics to guide the stem cell osteogenic differentiation in vitro and in vivo. 4 Indeed, when an inverse piezoelectric effect was introduced to bone, the strain induced in bone matrix could stimulate the cellular activity of bone cells and then improved bone regeneration. 5 When low‐intensity pulsed ultrasound (LIPUS) was applied to porous titanium alloy scaffold, the micromechanical strain induced by LIPUS would therefore initiate osteoblastic differentiation, bone formation and maturation as well as bony ingrowth in the porous scaffold. 6 A recent study by Ning et al. proposed the use of periodic microscale electric field (MEF) to induce bone‐implant integration. Two parallel semiconducted anatase and rutile TiO 2 zones were established on titanium implant and this electrical cue was able to direct stem cell osteogenic differentiation and bone regeneration on implant surface. 7 However, identification of the influences of specific surface features of biomaterials on stem cell functions remains difficult due to the substrate complexity, and the mechanism underlying their interaction is still not well understood. Nonetheless, macrophages appear to play a critical role in host reactions in the early stage of bone biomaterial insertion, and their initial response to a biomaterial can determine the success of biomaterial‐associated osteogenesis. [[qv: 1b, 8]] In response to various tissue microenvironmental cues, highly plastic macrophages can alter their phenotypes and functions to display the classically activated proinflammatory M1 and alternatively activated antiinflammatory M2 forms. 9 The prohealing M2‐polarized macrophage phenotype can elicit positive outcomes on osteogenesis, angiogenesis, and osseointegration, [[qv: 8a, 10]] and macrophage polarization can be modulated by the surface physics and chemistry cues of bone biomaterials. Therefore, modulation of the macrophage polarization response to biomaterials may be a promising strategy for eliciting favorable outcomes for bone tissue engineering and regenerative medicine. 11 In fact, as a foreign body, a biomaterial implant is recognized by the host immune system and triggers an immune reaction that may eventually determine the in vivo fate of the bone biomaterials. [[qv: 11f, 12]] Introduction of an implant into the body initiates an inflammatory cascade due to cell and tissue damage, and this cascade persists for roughly 4 d. 13 This early inflammatory response is beneficial to the host, but termination of subsequent inflammation is critical for blocking tissue damage and promoting tissue regeneration. 14 Therefore, an in‐depth understanding of the mechanism underlying the immune response mediated by a bone biomaterial would aid in the development of novel bone biomaterials that can create a favorable local immune environment for optimal osteogenesis and osseointegration. Recently, cerium oxide nanoparticles (CeONPs) have attracted increasing interest for biological applications because of their excellent catalytic activities, which arise from quick and expedient switches in oxidation state between Ce 4+ (CeO 2 ) and Ce 3+ (Ce 2 O 3 ) during redox reactions. 15 In particular, their capability to quench free radicals (i. e. , antioxidant activity) give CeONPs great potential for use in therapy of diseases caused by reactive oxygen species (ROS), such as retinal degeneration, 16 cardiovascular pathology, 17 neurodegenerative disorders, 18 and spinal cord injury. 19 Most of the CeONP characteristics are endowed by the co‐existence of both Ce 4+ and Ce 3+ oxidation states. In addition, nano‐CeO 2 regions promote cell proliferation on a functionalized polymer scaffold surface, whereas nano‐Ce 2 O 3 regions have an inhibitory effect on cell proliferation. 20 CeO 2 nanoparticles can also induce aligned growth of stem cells and improve the bioactivity of polymer scaffold surfaces. 21 They also show favorable biocompatibility and exert a protective effect on normal cells even at levels that exhibit antitumor effects on cancer cells. 22 This mixed valence state of CeONPs has therapeutic potential due to the possibility of scavenging ROS in cells. 16, 23 Previous in vitro studies have shown that CeONPs can react catalytically with superoxide and hydrogen peroxide to mimic the biological actions of superoxide dismutase (SOD)[[qv: 23b, 24]] and catalase. 25 At physiological pH, in particular, Ce 3+ and Ce 4+ sites on CeONP surfaces show this SOD and catalase mimetic activity, respectively. 15 For these reasons, CeONPs supplied to murine J774A. 1 macrophages can scavenge ROS and inhibit inflammatory mediator production. 26 These significant advantages of CeONPs support the speculation that CeONPs could be useful in modifying bone biomaterials to impart immunomodulatory characteristics that will allow regulation of macrophage behavior and promotion of stem cell osteogenic differentiation and bone tissue regeneration. Nevertheless, few designs that have immobilized CeONPs onto bone biomaterials have also demonstrated immunomodulatory and osteogenic‐inducing activity. Titanium‐based biomaterials are widely applied in orthopedic and dental implants due to their desirable biocompatibility and mechanical properties. 27 However, their bioactivity is insufficient in terms of bone‐implant osseointegration and fibrotic scarring mediated by macrophages is usually found during foreign body response. 28 Therefore, the aim of this study was to design a functional titanium surface that is able to modulate the macrophage immune response and the osteogenic capability of stem cells, followed by stimulation of bone tissue regeneration. A promising strategy to achieve this goal would appear to be modulation of the Ce valence states of extracellular CeONPs. The biofunction of Ce 4+ and Ce 3+ particles is size dependent, and the fraction of Ce 4+ in particles generally increases with increases in the particle size. 29 Hence, a magnetron sputtering technique was adopted to control the particle size and the Ce 4+ /Ce 3+ ratio of CeONPs was optimized by deposition time. We hypothesized that this functional surface with its higher Ce 4+ /Ce 3+ ratio can effectively modulate macrophage behavior, thereby promoting new bone formation. Following characterization of the physical and chemical properties of the CeONP‐modified titanium surface, the cytocompatibility of this surface was tested by culturing rat bone marrow mesenchymal stem cells (BMSCs), with subsequent in vivo animal implantation for osteogenic capability assessment. The detailed biological responses and underlying mechanism of new bone formation were investigated by culturing, rat BMSCs and RAW264. 7 macrophages, respectively, onto the CeONP‐modified surface. 2 Results 2. 1 Materials Characterization Figure 1 a–d shows the surface morphologies of various samples. Acid cleaning left a relatively flat topography on the CeONP‐0 surface (Figure 1 a). Magnetron sputtering (2 min) and vacuum annealing then revealed tiny homogeneously distributed nanoparticles on the CeONP‐1 surface (Figure 1 b). Increasing the deposition time to 3 min (Figure 1 c) and 5 min (Figure 1 d) caused these nanoparticles to develop gradually into relatively larger ones. The X‐ray diffractometry (XRD) patterns (Figure 1 e) and the diffraction peaks of the titanium substrate revealed a diffraction peak for the CeO 2 (111) facet. 21 In addition, a peak of TiO 2 emerged due to the crystallization of a natural oxide layer on the titanium surface. 30 Therefore, these nanoparticles had the characteristics of CeO 2 nanoparticles. Figure 1 Surface morphologies of samples CeONP‐0 a), CeONP‐1 b), CeONP‐2 c), and CeONP‐3 d). XRD patterns e) and XPS full spectra f) of the samples. Development of Ce 3d XPS spectra f) and the fitted Ce 3d spectra of CeONP‐1 h), CeONP‐2 i), and CeONP‐3 j), with the corresponding analysis results k) of valence concentration of Ce(IV) and Ce(III). UV–vis absorbance spectra l) of the samples and the corresponding conversion curves using the Kubelka–Munk function m). Fitted high resolution XPS spectra of Ti2p core levels for the samples CeONP‐0 n), CeONP‐1 o), CeONP‐2 p), and CeONP‐3 q). Figure 1 f shows the surface X‐ray photoelectron spectroscopy (XPS) survey spectra of various samples. The corresponding surface elemental compositions are listed in Table 1. Increases in the deposition time from 2 to 5 min increased the content of the Ce element on the surface from an initial 0 at% for CeONP‐0, 3. 57 at% for CeONP‐1, and 5. 82 at% for CeONP‐2 to 7. 58 at% for CeONP‐3. Accordingly, the content of the Ti element on the surface decreased from 25. 53 at% for CeONP‐0, 25. 50 at% for CeONP‐1, and 21. 31 at% for CeONP‐2 to 17. 13 at% for CeONP‐3. Interestingly, the content of the surface O element dropped from 74. 47 at% for CeONP‐0 to 70. 93 at% for CeONP‐1, and then increased to 72. 87 at% for CeONP‐2 and 75. 29 at% for CeONP‐3. Table 1 Elemental components of CeONP‐0, CeONP‐1, CeONP‐2, and CeONP‐3 analyzed by XPS. Analyzing area: 2. 0 mm × 2. 0 mm Sample Elemental concentration [at%] Ce Ti O CeONP‐0 0 25. 53 74. 47 CeONP‐1 3. 57 25. 50 70. 93 CeONP‐2 5. 82 21. 31 72. 87 CeONP‐3 7. 58 17. 13 75. 29 John Wiley & Sons, Ltd. The high resolution XPS spectra of Ce 3d were acquired from the surfaces of various CeONP samples, as shown in Figure 1 g; this consisted of two parts: (i) 916. 7, 907. 3, and 901. 0 eV for Ce 3d 3/2 and 898. 3, 888. 8, and 882. 4 eV for Ce 3d 5/2 of CeO 2 (solid line); (ii) 904. 0 and 899. 7 eV for Ce 3d 3/2 and 885. 8 and 880. 7 eV for Ce 3d 5/2 of Ce 2 O 3 (dashed line). 20 The evolution of the Ce 3d XPS spectra was analyzed by further fitting of these high resolution spectra, as shown in Figure 1 h–j. A detailed description of the Ce 3d spectra fitting procedure was given previously. 31 Figure 1 h–j shows qualitatively the relative changes in the Ce 3d XPS spectra for various CeONP samples. The corresponding quantitative results of the valence concentrations shown in Figure 1 k confirm that, as deposition time increased, the concentrations of surface Ce 4+ (CeO 2 ) increased from 0% for CeONP‐0, 31. 5% for CeONP‐1 and 55. 1% for CeONP‐2 to 76. 3% for CeONP‐3. Accordingly, the concentrations of surface Ce 3+ (Ce 2 O 3 ) decreased from 68. 5% for CeONP‐1 to 23. 7% for CeONP‐3. Figure 1 i depicts the UV–vis absorption spectra of the various samples. A strong absorption at ≈400 nm for CeONP‐0 was ascribed to the absorption edge of TiO 2. Interestingly, the increase in deposition time from 2 to 5 min shifted the absorption edge to a longer wavelength range. The diffuse reflectance spectra were converted into equivalent absorption coefficients using the Kubelka–Munk function: 32 α = (1 − R ) 2 /2 R, αhν = C 1 ( hν – E g ) 2, hν = 1240/λ, where α is the optical absorption coefficient near the absorption edge for the indirect interband transition, R is the reflectance of the semiconductor, C 1 is a constant for the indirect transition, hν is the photon energy, E g is the indirect bandgap energy (eV), and λ is the wavelength (nm). Figure 1 m depicts the ( αhν ) 1/2 plot versus hν ; here, the vertical segments of the spectra were extended to intersect with the hν axis to obtain the E g for the CeONP samples. The narrowed bandgap agreed with the red shift of the absorption edge, which corresponded to the increase in nanoparticle size and surface Ce 4+ content for the CeONP samples from 2 to 5 min. 33 In addition, the high resolution XPS spectra of Ti2p core levels from the various samples were fitted in Figure 1 n–q. The doublet peak at 458. 8 eV and 464. 4 eV belongs to the Ti2p 3/2 and Ti2p 1/2 in TiO 2, respectively. 30 This indicates that the TiO 2 remained on titanium substrate after depositing cerium oxides. 2. 2 Cell Viability Figure 2 shows the proliferation and viability of rat BMSCs after culturing on various samples for 1, 4, and 7 d. At day 1, no significant difference was noted among the groups. However, after 4 d of culture, obvious differences emerged between the CeONP samples and the control group. The CeONP samples significantly promoted cell proliferation on the surface in a manner dependent on the Ce 4+ /Ce 3+ ratio (0. 46 for CeONP‐1, 1. 23 for CeONP‐2, and 3. 23 for CeONP‐3). Furthermore, the increase in the surface Ce 4+ content resulted in a better promotion effect on cell proliferation for the CeONP‐2 and CeONP‐3 samples than for the CeONP‐1 samples ( P < 0. 05). At day 7, the cell growth on all samples maintained an upward trend, with an apparent difference between the testing groups and the control group. CeONP‐3 induced the highest cell proliferation and viability. Thus, an increase in the surface Ce 4+ content can promote the proliferation of rat BMSCs on CeONP samples. Figure 2 CCK‐8 assay results of proliferation and viability of rat BMSCs after 1, 4, and 7 d of culture on samples CeONP‐0, CeONP‐1, CeONP‐2, and CeONP‐3. Note: * P < 0. 05, ** P < 0. 01, *** P < 0. 001 versus CeONP‐0; # P < 0. 05, ## P < 0. 01 versus CeONP‐1; $$ P < 0. 01 versus CeONP‐2; ns, not significant. 2. 3 In Vivo Osseointegration 2. 3. 1 Micro‐CT Evaluation of Bone Formation All the samples were implanted in rat femur bones for 8 weeks to enable the in vivo evaluation of osseointegration. Figure 3 a shows the reconstructed micro‐CT images along the central axis of the implants, with the implants marked in pink and the newly formed bones marked in grey. This figure shows that the bone volume is visually larger around the surface of the CeONP‐3 implant than around the other implants. As shown in Figure 3 d, the BV/TV of CeONP‐3 was 14. 1%, which was significantly higher than the 7. 8% obtained with CeONP‐0 ( P < 0. 05) and 2. 1% with CeONP‐1 ( P < 0. 01). CeONP‐2 also had a better outcome of 10. 8% compared with CeONP‐1 ( P < 0. 01). A similar trend was observed for Tb. Th (Figure 3 e) and bone mineral density (BMD) (Figure 3 f). Both CeONP‐3 (0. 46 mm −1 ) and CeONP‐2 (0. 44 mm −1 ) produced higher outcomes than CeONP‐1 (0. 14 mm −1, P < 0. 01) for Tb. N (Figure 3 g). For these parameters, worse outcomes were obtained with CeONP‐1 than with CeONP‐0 ( P < 0. 05). Consequently, the new bone formation around the implants displayed an obvious correlation with the surface Ce 4+ content in the CeONPs. Figure 3 Micro‐CT images of reconstructed 3D models of surrounding bones in transverse and longitudinal views, with or without implants a), accompanied by the corresponding quantitative analysis results of BV/TV d), Tb. Th e), BMD f), and Tb. N g). Sequential fluorescent labeling observation b), accompanied by the corresponding analysis results h) of the area of bone stained with the three fluorochromes. (Note: Red labeling, Alizarin Red S, week 2; yellow labeling, tetracycline, week 4; green labeling, calcein, week 6; * P < 0. 05, ** P < 0. 01 versus CeONP‐0; ## P < 0. 01 versus CeONP‐1; $ P < 0. 05 versus CeONP‐2. ) Histological observation c) of the CeONP‐0, CeONP‐1, CeONP‐2, and CeONP‐3 sections stained with Van Gieson's picrofuchsin, accompanied by the corresponding analysis results i) of bone‐implant contact from the histomorphometric measurement at ×10 magnification. (Note: * P < 0. 05 versus CeONP‐0; ## P < 0. 01 versus CeONP‐1; $ P < 0. 05 versus CeONP‐2. ) 2. 3. 2 Sequential Fluorescent Labeling The process of new bone formation and mineralization around the implants was recorded using three types of fluorochrome at different time points and the results are shown in Figure 3 b. This figure shows that CeONP‐3 and CeONP‐2 had better outcomes in terms of promoting new bone formation when compared with CeONP‐0 and CeONP‐1. The percentages of fluorochrome labeled area are shown in Figure 3 h. At week 2, the percentage of Alizarin Red S labeled area (red) was significantly higher for CeONP‐2 and CeONP‐3 than for CeONP‐0 and CeONP‐1 ( P < 0. 01). CeONP‐1 had a lower value than CeONP‐0 ( P < 0. 05) and CeONP‐3 had a better outcome than CeONP‐2 ( P < 0. 05). At week 4 and week 6, the percentages of tetracycline hydrochloride labeled area (yellow) and calcein labeled area (green) exhibited a similar tendency. Therefore, these results confirmed that the increase in surface Ce 4+ content in the CeONPs could promote new bone formation and mineralization at the bone‐implant interface. 2. 3. 3 Histological Observation The van Gieson's picrofuchsin staining is a central histological test for establishing whether implants are in direct contact with the surrounding bones. The histological staining results for the sections in the present study are shown in Figure 3 c. The corticocancellous site shows a close apposition of bone to implant for CeONP‐3. The implant surface bonds tightly and directly with the newly formed bone, virtually without fibrous or connective tissue that would prevent its direct contact with new bone. Similarly, CeONP‐2 also produces a good outcome regarding new bone apposition. By contrast, only a small amount of new bone appeared around the CeONP‐1 implant within the corticocancellous bone, and an interspace was apparent between the newly formed bone and the implant surface. The percentages of bone‐implant contact (BIC) in the osseointegration region, observed at ×10 magnification, are given in Figure 3 i. In agreement with the micro‐CT results, significantly higher BIC percentages were obtained with CeONP‐3 and CeONP‐2 than with CeONP‐0 ( P < 0. 05) and CeONP‐1 ( P < 0. 01). The BIC percentage was also lower for CeONP‐1 than for CeONP‐0 ( P < 0. 05) and the BIC percentage was higher for CeONP‐3 than for CeONP‐2 ( P < 0. 05). Therefore, an increase in the surface Ce 4+ content in the CeONPs can promote new bone formation around the implant and enhance osseointegration. 2. 3. 4 Biomechanical Push‐Out Test The biomechanical push‐out test was utilized to evaluate the quality of osseointegration of the implants. Figure 4 shows that, among the various groups, CeONP‐3 had the largest failure load of 175. 7 N whereas CeONP‐1 had the lowest failure load of 80. 6 N. Both CeONP‐3 and CeONP‐2 (164. 7 N) had larger failure loads than CeONP‐0 (135. 9 N, P < 0. 05) and CeONP‐1 ( P < 0. 01). The failure load was lower for CeONP‐1 than for CeONP‐0 ( P < 0. 01). Therefore, an increase in the surface Ce 4+ content on CeONPs could enhance the osseointegration quality of the implant. Figure 4 Results of biomechanical testing. Note: * P < 0. 05, ** P < 0. 01 versus CeONP‐0; ## P < 0. 01 versus CeONP‐1; $ P < 0. 05 versus CeONP‐2. 2. 3. 5 Scanning Electron Microscopy (SEM) Observation The surface morphologies of the pushed‐out implants were observed by SEM and the element distributions in the implants were mapped by energy dispersive X‐ray spectroscopy (EDS) for further examination of the new bone formation on the surface of the implants in the bone marrow cavity. The SEM images in Figure 5 show the typical interface bonding status between the implants and new bone tissues. The CeONP‐0, at low magnification, shows some large blocks of new bone tissues remaining on the implant surface, indicating a strong interface bonding strength between the implant and new bone tissues. At high magnification, the implant showed a rough surface due to the adhesion of new bone tissues (Figure S1, Supporting Information). This rough structure was more apparent at the higher magnification in Figure 5. By contrast, CeONP‐1, shows a relatively bare implant surface, indicating poor interface bonding strength between the implant and new bone tissues. At high magnification, the implant displayed a relatively smooth surface morphology (Figure S1, Supporting Information) and this became more obvious at higher magnification, as shown in Figure 5. CeONP‐2 and CeONP‐3 show larger blocks of new bone tissues remaining on the implant surface, indicating a very strong interface bonding strength between the implant and new bone tissues. This was further supported by the rough morphology seen under high magnification. Therefore, the SEM results were agreed well with the push‐out results in Figure 4. Figure 5 Surface morphologies of the implants of various samples, accompanied by the corresponding EDS elemental mapping and distribution of C, O, P, Ca, Ti, and Ce. EDS mapping showed that the newly formed bone tissues were characterized by intense distribution maps of calcium (green) and phosphorus (gray). Figure 5 shows the mapping results that were very consistent with the SEM results. The weak distribution maps for titanium (blue), as a counterpart, can be used to estimate the coverage and thickness of new bone tissues on the implants. Therefore, the SEM and EDS results agreed well with the results for new bone formation obtained with the other methods. 2. 4 Osteogenic Mechanism 2. 4. 1 In Vitro Osteogenic Regulation Further understanding of the outcome of new bone formation and mineralization and their regulation by CeONPs was obtained using quantitative real‐time polymerase chain reaction (PCR) assays. The key osteogenic‐related markers, including alkaline phosphate (ALP) ( Figure 6 a, f), collagen type I (Col‐I) (Figure 6 b, g), osteocalcin (OCN) (Figure 6 c, h), osteopontin (OPN) (Figure 6 d, i), and runt‐related transcription factor 2 (Runx‐2) (Figure 6 e, j), were investigated to determine the influence of the surface Ce 4+ content of the CeONPs on the osteogenic differentiation of rat BMSCs at the molecular level. The cells were cultured on CeONPs for 1 and 7 d. At day 1 (Figure 6 a–e), a significant promoting effect was observed for the expression of the ALP, Col‐I, OCN, OPN, and Runx‐2 genes in response to CeONP‐3 and CeONP‐2 but not in response to CeONP‐1 and CeONO‐0 ( P < 0. 05). CeONP‐1 induced an obvious down‐regulation of these osteogenesis genes when compared with CeONP‐0 ( P < 0. 05). At day 7 (Figure 6 f–j), the difference in the expression of these genes between CeONP‐3 and other groups became more significant ( P < 0. 01). Figure 6 Expression levels of osteogenic‐related genes ALP a, f), Col‐I b, g), OCN c, h), OPN d, i), and Runx‐2 e, j) by real time PCR after culturing rat BMSCs on various samples for 1 and 7 d. ALP activity of rat BMSCs after 7 d of culture on various samples k). ARS activity of rat BMSCs after 14 d of culture on various samples l). Detection of OPN expression after culturing rat BMSCs on the samples for 7 d by immunofluorescence assay m). Note: * P < 0. 05, ** P < 0. 01, *** P < 0. 001 versus CeONP‐0; # P < 0. 05, ## P < 0. 01, ### P < 0. 001 versus CeONP‐1; $ P < 0. 05, $$ P < 0. 01, $$$ P < 0. 001 versus CeONP‐2. ALP is an early marker of BMMSC differentiation, so its measurement at day 7 was used to evaluate the osteogenic differentiation potential of the CeONP samples. As shown in Figure 6 k, significantly higher ALP activity was observed for both CeONP‐3 and CeONP‐2 than for CeONP‐1 ( P < 0. 01) and CeONP‐0 ( P < 0. 05). ALP activity was also lower for CeONP‐1 than for CeONP‐0 ( P < 0. 05). Similarly, Alizarin Red S (ARS) was determined to evaluate calcium nodule formation in the samples. As shown in Figure 6 l, ARS activity was apparently higher for CeONP‐3 and CeONP‐2 than for CeONP‐0 and CeONP‐1 ( P < 0. 05). OPN expression in the rat BMSCs was detected by an immunofluorescence assay with DyLight 488. As shown in Figure 6 m, CeONP‐3 induced the strongest immunofluorescence labeling of OPN expression, whereas CeONP‐1 produced the weakest green fluorescence. More BMSCs that expressed the relevant specific protein were detected for CeONP‐3 than for the other groups. Therefore, an increase in the surface Ce 4+ content in the CeONPs can promote in vitro osteogenic differentiation of rat BMSCs, thereby accounting, at least in part, for the enhanced new bone formation around the implant and for osseointegration. 2. 4. 2 Macrophage Response Figure 7 shows the analysis results for flow cytometry assay and ELISA of RAW264. 7 macrophages. The percentage of M1 cells that expressed the surface marker CCR7 showed the following trend (Figure 7 a–d): CeONP‐1 (87. 64%) > CeONP‐0 (66. 52%) > CeONP‐2 (58. 64%) > CeONP‐3 (46. 46%). By contrast, the percentage of M2 cells expressing the surface marker CD206 showed the following trend (Figure 7 e–h): CeONP‐1 (30. 18%) < CeONP‐0 (45. 40%) < CeONP‐2 (51. 80%) < CeONP‐3 (69. 70%). The production of the pro‐inflammatory cytokine tumor necrosis factor‐α (TNF‐α) in the supernatants recovered from various samples (Figure 7 i) indicated a significantly higher TNF‐α concentration for CeONP‐1 than for CeONP‐3 ( P < 0. 01) and CeONP‐2 ( P < 0. 05) at day 1. The TNF‐α concentration was also lower in response to CeONP‐3 than in response to CeONP‐0 ( P < 0. 05). The same trend was observed after 4 d of culture. A statistically significant difference was observed between the responses to CeONP‐3 and CeONP‐2 ( P < 0. 05), and the concentration was higher for CeONP‐1 than for CeONP‐0 ( P < 0. 05). Figure 7 Expression of cell surface markers on RAW264. 7 macrophages determined by flow cytometry, showing the percentages of M1 phenotype (CCR7, a–d) and M2 phenotype (CD206, e–h). Production of cytokines TNF‐α i) and IL‐10 j) secreted by RAW264. 7 macrophages cultured on various samples determined by ELISA. Note: * P < 0. 05, ** P < 0. 01 versus CeONP‐0; # P < 0. 05, ## P < 0. 01, ### P < 0. 001 versus CeONP‐1; $ P < 0. 05 versus CeONP‐2. Production of the anti‐inflammatory cytokine interleukin‐10 (IL‐10) in the medium retrieved from various samples (Figure 7 j) indicated higher expression of IL‐10 in response to CeONP‐3 and CeONP‐2 when compared with CeONP‐0 and CeONP‐1, and this difference was statistically significant at both day 1 ( P < 0. 05) and day 4 ( P < 0. 01). At day 1, a higher IL‐10 level was induced in response to CeONP‐3 than to CeONP‐2 ( P < 0. 05), and at day 4, a lower IL‐10 level was obtained with CeONP‐1 than with CeONP‐0 ( P < 0. 05). Therefore, an increase in surface Ce 4+ content in the CeONPs suppresses the production of a proinflammatory cytokine but induces higher levels of an antiinflammatory cytokine; these are both characteristic responses of M2 macrophages. 3 Discussion During magnetron sputtering using highly pure CeO 2 target ( Scheme 1 a), subtle fluctuations in deposition may cause a dynamic departure from stoichiometry, which then induces a localized cerium or oxygen excess and creates oxygen or cerium vacancy point defects. 34 In the case of localized oxygen excess, the charge neutrality demands the creation of cerium vacancies (defect reaction in Kröger–Vink notation: O ↔ V Ce 4 ′ + 4h • + O O × ). The holes (h • ) are trapped at the nearest‐neighbor Ce 4+ sites to create Ce 5+. However, this happens only with difficulty since the highest oxidation state is 4 + for cerium. In the case of localized cerium excess, charge neutrality demands the creation of oxygen vacancies (Kröger–Vink notation: O O × ↔ V O •• + 2e′ + 1/2O 2 ). The electrostatic attractive forces can trap the electrons (e′) at the Ce 4+ sites to create Ce 3+. As a result, the heterogeneous deposition of cerium and oxygen atoms can create an oxygen‐deficient nonstoichiometric phase (CeO 2− x ) with oxygen vacancies. Scheme 1 Schematic illustration of CeONPs immobilization on titanium implant biomaterials for bone tissue engineering and regenerative medicine: a) a layer of CeONPs surface fabricated on titanium surface by using magnetron sputtering is able to induce osteogenic differentiation of rat BMSCs and M2 phenotype polarization of macrophages, thereby resulting to the boost of new bone regeneration in vivo; b) the interactions of CeONPs with the superoxide anion radicals ( • O 2 − ) produced by mitochondria, including SOD mimetic and catalase mimetic activities; c) the correlations of valence state of CeONPs (Ce 4+ /Ce 3+ molar ratio) layer with the cell fates of BMSCs and macrophages on osteogenesis. The relative content of Ce 4+ or Ce 3+ is a function of the particle size of CeONPs. In general, the fraction of Ce 3+ in particles increases as the particle size decreases; i. e. , the Ce 4+ /Ce 3+ ratio increases with increases in the particle size. 29 Here, by tuning the deposition time of the magnetron sputtering, the particle size was controlled and a high Ce 4+ /Ce 3+ ratio was obtained for the CeONPs (Scheme 1 a). XPS analysis revealed the changes in the surface element compositions of the various samples, as shown in Table 1. The development of surface oxygen content serves as an indicator of the change in the surface oxygen vacancy concentration in the various samples. The apparent drop in the surface oxygen content from 74. 47 at% for CeONP‐0 to 70. 93 at% for CeONP‐1 indicated the presence of oxygen vacancies on the surface. Subsequent increases in the surface oxygen content revealed a decrease in the number of oxygen vacancies on the surface, which corresponded to the increase in surface Ce 4+ content and the decrease in surface Ce 3+ content in the CeONPs. The sampling depth of XPS analysis is ≈10 nm, 35 so the measured Ce 4+ and Ce 3+ contents were derived from the outermost surface of the CeONPs. With regard to the bonding between CeNPs and titanium substrate, only the diffraction peaks of CeO 2, TiO 2, and Ti phases are detected in XRD (Figure 1 e). Also, all the XPS spectra of Ce3d core levels in Figure 1 g–j belong to the cerium oxide, i. e. , CeO 2 or Ce 2 O 3 and all the XPS spectra of Ti2p core levels in Figure 1 n–q are labeled to TiO 2. Hence, no interfacial reaction is found between cerium oxide and titanium oxide on the titanium substrate. 36 Therefore, we believe that the cerium oxides are physically bonded to the titanium surface during magnetron sputtering. The Ce 4+ state is the preferentially formed (CeO 2 ) oxide of cerium. Nevertheless, intrinsic defects are usually present, so that cerium will be present, in part, in the Ce 3+ state (Ce 2 O 3 ) containing oxygen vacancies. 37 The relative contents of Ce 3+ and Ce 4+ are a function of the particle size of the CeONPs, 15, 29 such that the proportion of Ce 3+ in the particles increases with decreasing particle size. The particle size of CeONPs can be regulated by magnetron sputtering and vacuum annealing, and the vacuum annealing can cause the deposited cerium oxide to grow into nanoparticles via Ostwald ripening. An increase in deposition time can further contribute to an increase in the particle size of the CeONPs. 38 The relative contents of surface Ce 3+ and Ce 4+ in the CeONPs were manipulated in this way in the present study, as shown in Figure 1 k. The calculated surface Ce 4+ /Ce 3+ ratios for the CeONPs were 0. 46, 1. 23, and 3. 23 for CeONP‐1, CeONP‐2, and CeONP‐3, respectively. Therefore, the prepared CeONPs were in a mixed valence state with an increasing surface Ce 4+ /Ce 3+ ratio. As shown in Figure 2, an increase in the surface Ce 4+ /Ce 3+ ratio can promote the proliferation of rat BMSCs on CeONPs. This agrees with previous work showing that the nano‐CeO 2 region on a polymer scaffold promoted cell proliferation, while the nano‐Ce 2 O 3 region inhibited cell proliferation. 20 Figure 2 shows that the prepared CeONPs samples had favorable cytocompatibility. The in vivo animal test performed to investigate the influence of surface Ce 4+ /Ce 3+ ratios on bone formation and mineralization around the CeONP‐immobilized implants confirmed that the extent of new bone tissue formation and mineralization depended on the surface Ce 4+ /Ce 3+ ratio of the CeONPs (Scheme 1 a). The micro‐CT, sequential fluorescent labeling, and histological analysis results presented in Figure 3 further confirm that an increase in the surface Ce 4+ /Ce 3+ ratio can promote new bone formation and mineralization around a CeONP‐immobilized implant. The better outcome for the biomechanical push‐out test, as shown in Figure 4, also demonstrates the increasing osseointegration quality of the implant with the increasing surface Ce 4+ /Ce 3+ ratio. These results were further supported by the SEM morphology observations and EDS element mapping of the newly formed bone tissues around the CeONPs‐immobilized implants, as shown in Figure 5 and Figure S1 (Supporting Information). The mechanism by which the Ce 4+ /Ce 3+ ratio on the CeONP surface modulates new bone formation and mineralization was investigated by analyzing the expression levels of the osteogenesis‐related genes, ALP, Col‐I, OCN, OPN, and Runx‐2, by quantitative real‐time PCR in in vitro cultured rat BMSCs treated with various CeONP samples. The expression of these markers, as shown in Figure 6 a–j, indicated an apparent relationship between the osteogenic differentiation of rat BMSCs and the surface Ce 4+ /Ce 3+ ratio of the CeONPs. An increase in the surface Ce 4+ /Ce 3+ ratio promoted the in vitro osteogenic differentiation of rat BMSCs cultured on CeONPs. This response was also evident at the protein level, as shown in Figure 6 k–m. SOD is an enzyme that catalyzes the disproportionation of superoxide, the most common free radical in body, into H 2 O 2 and O 2. [[qv: 23b]] The Ce 4+ /Ce 3+ valence switch capability can endow CeONPs with SOD mimetic activity. The CeONPs with lower Ce 4+ /Ce 3+ ratio have a higher SOD mimetic activity, and the superoxide dismutation by CeONPs is expressed as 24 (1) O • 2 − + Ce 3 + + 2 H + → H 2 O 2 + Ce 4 + ; O • 2 − + Ce 4 + → O 2 + Ce 3 + Catalase is a protective enzyme found within almost all living organisms exposed to oxygen. This enzyme catalyzes the degradation of H 2 O 2, a powerful and potentially harmful oxidizing agent, into H 2 O and O 2. 39 CeONPs can possess catalase mimetic activity that appears in a redox‐state‐dependent manner, and a higher Ce 4+ /Ce 3+ ratio gives a higher catalase mimetic activity. 25 Interestingly, the SOD mimetic activity and catalase mimetic activity of CeONPs are the opposite of that expected based on the Ce 4+ /Ce 3+ ratio (Scheme 1 b). CeONPs showing SOD mimetic activity will generate H 2 O 2. Both in vitro and in vivo, excess H 2 O 2 is believed to be more toxic than superoxide because it is the substrate for the Fenton reaction that creates the hydroxyl radical (•OH), the most destructive of the ROS. 40 Fortunately, CeONPs have both SOD and catalase mimetic activities, so the H 2 O 2 created during the SOD mimetic process can enter into the catalase mimetic dismutation cycle and be decomposed into innocuous H 2 O and O 2 ; these reactions make CeONPs an excellent antioxidant. 21, 26 Nevertheless, the antioxidative function of the CeONPs is only effective when the two enzyme mimetic activities are coordinated; in other words, the H 2 O 2 decomposition rate should be equal to or higher than the production rate. The SOD mimetic and catalase mimetic activities of the CeONPs should have a close relationship with the osteogenic differentiation of rat BMSCs, since these enzyme mimetic activities can regulate the production of ROS (Scheme 1 c). The overproduction of ROS will decrease the osteogenesis of stem cells, but exogenous antioxidant treatment will promote this osteogenesis. 41 This dichotomy may account for the promoting effect observed for an increase in the surface Ce 4+ /Ce 3+ ratio with respect to the in vitro osteogenic differentiation of rat BMSCs cultured on CeONPs. A recent study using cerium oxide nanoparticles for functional recovery of spinal cord injury (SCI) indicated that the nanoparticles had a higher +4 oxidation status than +3 (Ce 4+ /Ce 3+ = 2. 9) and these CeONPs possibly carried out an effective consumption of ROS to give rise to SCI recovery at the cost of changing their status primarily from Ce 4+ to Ce 3+. 19 The host immune response is an essential component of the biomaterial‐mediated osteogenic effect. [[qv: 1b]] Therefore, murine RAW264. 7 macrophages were cocultured with CeONP samples to obtain a further understanding of the relationship between the surface Ce 4+ /Ce 3+ ratio of the CeONPs and new bone formation and mineralization. As shown in Figure 7, an increase in the surface Ce 4+ /Ce 3+ ratio of the CeONPs promoted macrophage polarization to the M2 phenotype and an increase in the production of the anti‐inflammatory cytokine IL‐10. Similarly, a decrease in the surface Ce 4+ /Ce 3+ ratio promoted macrophage polarization to the M1 phenotype and increased the production of the proinflammatory cytokine TNF‐α. Taken together, these findings indicate that manipulation of the surface Ce 4+ /Ce 3+ ratio of the CeONPs can modulate the balance of anti‐inflammatory and proinflammatory cytokines in macrophages and create an anti‐inflammatory microenvironment (Scheme 1 c). Interestingly, previous work that investigated the capability of CeONPs to scavenge ROS and inhibit inflammatory mediator production in murine J774A. 1 macrophages indicated that oxidative stress and proinflammatory iNOS protein expression were abated by CeONP stimulation. 26 More importantly, recent work has shown that the prohealing M2‐polarized macrophage phenotype can elicit positive outcomes, both in vitro and in vivo, on osteogenesis, angiogenesis, and osseointegration. [[qv: 8a, 10]] The CeONPs appear to show a valence‐dependent modulatory effect on macrophage polarization and cytokine production. In this context, in the present study, CeONP‐1 had a negative effect on the balance between antiinflammatory and proinflammatory macrophage polarization and cytokine secretion, and its outcome was poorer even than that of CeONP‐0. The long‐term survival and integration of biomaterials are largely determined by activation of the host immune system. [[qv: 1b, 42]] Once a biomaterial is implanted into the host body, a sequence of immune responses and healing processes will occur in the surrounding tissues. [[qv: 1b, 42a]] Macrophages play a vanguard role in the recognition of and adhesion to the foreign biomaterial. [[qv: 42b, 43]] The close relationship between the immune and skeletal systems means that activated macrophages contribute to both the success and the failure of transplantation of a foreign biomaterial. The macrophages exert this dichotomous effect by secreting cytokines that modulate either osteogenesis or inflammation, thereby promoting or inhibiting new bone formation. [[qv: 42b, 44]] A favorable macrophage polarization can create an osteogenic microenvironment that improves osteogenesis, whereas an unfavorable macrophage polarization may exacerbate inflammation and destroy the tissue–biomaterial integration. 12, 45 In general, the M2 macrophage phenotype accounts for antiinflammation and tissue regeneration, whereas the M1 macrophage phenotype is proinflammatory and causes tissue destruction. The results presented here indicate that an increase in the surface Ce 4+ /Ce 3+ ratio of the CeONPs can promote the polarization of the healing‐associated M2 macrophage phenotype and increase the secretion of antiinflammatory cytokine IL‐10. The antiinflammatory cytokines secreted by M2 macrophages can resolve inflammation and promote wound healing. IL‐10 can inhibit pro‐inflammatory cytokine secretion and activity and the secretion of granulocyte‐macrophage colony stimulating factor and nitric oxide (NO) in macrophages. 46 By contrast, the secretion of pro‐inflammatory cytokines by M1 lead to delayed bone healing and pathogenic bone loss. [[qv: 42b, 43]] As a result, the inhibition of TNF‐α secretion by an increased surface Ce 4+ /Ce 3+ ratio would benefit new bone formation and wound healing around a CeONP implant. In summary, the mechanism of surface Ce 4+ /Ce 3+ ratio of CeONPs on titanium substrate to induce bone‐material integration is illustrated by a schematic diagram (Scheme 1 ). Manipulating the surface Ce 4+ /Ce 3+ ratio of CeONPs can modulate macrophage polarization and cytokine secretion, and facilitate appropriate immune reactions that balance antiinflammation and pro‐inflammation, which lead to the satisfactory outcomes of new bone formation and bone–biomaterial integration. The introduction of a biomaterial into the body initiates an inflammatory cascade due to cell and tissue damage, and this cascade persists for roughly 4 d. 13 This early inflammatory response is highly beneficial to the host, it must subsequently be terminated to avoid tissue damage and to promote tissue regeneration. 14 Any unrestrained pro‐inflammatory M1 macrophage polarization induced by biomaterials will therefore impair new bone formation and osseointegration[[qv: 8b]] and prevent wound healing. 47 Therefore, CeONPs biomaterials must have appropriate modulatory effects on the balance between antiinflammation and pro‐inflammation immune reactions in order to elicit the desired outcomes of bone regeneration and osseointegration. 4 Conclusion A customized magnetron sputtering and vacuum annealing protocol is applied to establish a layer of cerium oxide nanoparticles (CeONPs) with different surface Ce 4+ /Ce 3+ ratios (0. 46, 1. 23, 3. 23) on titanium surface serving as an experimental platform to examine the regulatory effect of CeONPs on cell fate and bone formation. The CeONPs with a mixed valence state and a high surface Ce 4+ /Ce 3+ ratio exhibit better cytocompatibility with rat BMSCs. Moreover, when implanted into the rat femurs, it is found that new bone formation and bone–implant integration is highly correlated with the surface Ce 4+ /Ce 3+ molar ratio. The in vivo outcomes are supported by the in vitro studies of rat BMSCs cultured with CeONP samples. The results confirmed that the osteogenic related gene and protein expressions were significantly upregulated, when the cells cultured with the titanium surface with higher Ce 4+ /Ce 3+ ratio. Furthermore, while culturing with RAW264. 7 murine macrophages, the polarization of macrophages to the M2 phenotype is highly expressed on the surface with increased Ce 4+ /Ce 3+ ratio in which the gain of prohealing M2 percentage can increase antiinflammatory cytokine production. Thus, the mixed valence state of CeONPs has the potential to induce bone regeneration without the need for any exogenous osteogenic inducer. These results suggest that manipulation of the valence state of CeONPs can exert a desirable modulatory effect on stem cell and macrophage fates to elicit the beneficial outcomes of CeONPs on new bone formation and osseointegration. 5 Experimental Section Sample Fabrication : The cerium oxide nanoparticles were deposited with a magnetron sputtering apparatus (ULVAC Corp. , Model ACS‐4000‐C4) using a high purity CeO 2 target at a radio frequency power of 80 W. Acid cleaned metal titanium plates or rods (99. 95 wt%) were chosen as the deposition substrate. In brief, titanium plates (10 mm × 10 mm × 1 mm, 20 mm × 20 mm × 1 mm, and 20 mm × 10 mm × 1 mm) and rods (Ø 2 mm × 7 mm) were first pickled in oxalic acid solution (5 wt%) at 100 °C for 2 h to clean the surfaces, followed by thorough washing with ethanol and fresh water. Before CeO 2 deposition, the deposition chamber was first pumped down to ≈10 −4 Pa and then pure Ar gas (99. 999%) was introduced at 50 sccm. During deposition, the substrate was rotated along the vertical axis at a speed of 10 rpm to improve homogeneity. The samples were fabricated by depositing CeO 2 on the substrate target by sputtering for 2, 3, and 5 min. The samples were then vacuum annealed at 450 °C. For simplicity, the obtained cerium oxide samples were designated as “CeONP‐1, ” “CeONP‐2, ” and “CeONP‐3, ” respectively. The acid‐cleaned and vacuum‐annealed metal titanium acted as the control material and was designated as “CeONP‐0. ” Sample Characterization : The surface morphologies were studied using field‐emission SEM (LEO 1530, Germany); the instrument was capable of EDS. The crystallinity was studied by XRD (Rigaku Ultima IV, Japan) using a Cu Kα (λ = 1. 541 Å) source in the range of 2θ = 20°–90° with a glancing angle of 1°. Phase identification was done with the help of the JCPDS database. The surface chemical composition and chemical state were determined by XPS (Physical Electronics PHI 5802) using an Al Kα source (1486. 6 eV). The optical diffuse reflectance spectra were recorded on UV–vis–NIR spectrophotometer (Model UV‐4100, Hitachi Corp. ). Cytocompatibility Evaluation: Cell Culture : The rat BMSCs were purchased from the Cell Bank of Chinese Academy of Sciences and cultured in Dulbecco's modified Eagle's medium (DMEM; HyClone) supplemented with 10% fetal bovine serum (FBS; Gibco) and a 1% antibiotic/antimyotic solution. The BMSCs were cultured at 37 °C in a humidified 5% CO 2 incubator and passaged every 3 d at ≈80% confluence. Only the confluent BMSCs at passages 3–5 were harvested for further study. Cell Proliferation Assay : The proliferation activities of rat BMSCs on various samples were determined using the CCK8 assay. Initially, 5 × 10 4 cells were seeded into the wells of a 24 well plate. After 1, 4, and 7 d of culture, CCK8 solution, at a volume of ≈10% of the culture medium, and incubated for 1 h at 37 °C to react with cells. The medium was refreshed and the absorbance was measured at 450 nm on an ELX ultra microplate reader (BioTek, Winooski, VT). Four samples were used for each group and all tests were repeated three times. In Vivo Animal Experiment: Surgical Procedures : All experimental protocols concerning animals were approved by the Animal Committee of the Ninth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. A rat femoral model was used in this study. Thirty‐two six‐month‐old female Sprague–Dawley (SD) rats were randomly divided into the following four groups (8 rats/group): (i) CeONP‐0 group ( n = 8), (ii) CeONP‐1 group ( n = 8), (iii) CeONP‐2 group ( n = 8), and (iv) CeONP‐3 group ( n = 8). Surgical procedures were performed under sterile conditions, as described previously. 48 Briefly, the rats were first anesthetized by intraperitoneal injection of ketamine. After the hind limb was shaved, an ≈0 mm longitudinal incision was made across the knee joint along the lateral side of the extensor mechanism. A pilot hole was drilled along the long axis of the femur through the intercondylar notch and the distal femoral metaphysis, and the implants were inserted. After surgery, all rats received antibiotic and analgesic injections intramuscularly for three postoperative days. Sequential Fluorescent Labeling : A polychrome sequential fluorescent labeling method was used to assess the process of new bone formation and mineralization. 48 At 2, 4, and 6 weeks after surgery, different fluorochromes were intraperitoneally administered at a sequence of 30 mg kg −1 Alizarin Red S (Sigma), 25 mg kg −1 tetracycline hydrochloride (Sigma), and 20 mg kg −1 calcein (Sigma), respectively. Sample Preparation : All rats were sacrificed at weeks 8 after surgery. Left femurs, with four groups of implants, were harvested and trimmed into smaller samples for subsequent use (i. e. , 8 left femurs/group, n = 8). Micro‐CT Evaluation : The presence of newly formed bone around the inserted implants was determined using micro‐CT (GE explore Locus SP Micro‐CT, USA). The scanning parameters were set at 80 kV and 80 µA, with 3000 ms of exposure time and 15 µm of resolution. After scanning, 3D images were reconstructed using NRecon software (SkyScan, USA) and the CTvol program (SkyScan). The BMD, bone volume fraction (bone volume/total volume, BV/TV), trabecular number (Tb. N), and trabecular thickness (Tb. Th) were analyzed for newly grown bone tissues using DataViewer software (SkyScan) and the CTAn program (SkyScan). Histomorphometric and Histological Observation : After micro‐CT scanning, the femur samples of each group were dehydrated in a graded series of alcohol solutions from 75% (v/v) to 100% (v/v), and embedded in polymethylmethacrylate. The samples were cut into 150 µm thick sections using a Leica SP1600 saw microtome (Leica, Nusseloch, Germany). These sections were then ground and polished to a final thickness of ≈40 µm for fluorescence labeling observation with a confocal laser scanning microscope (CLSM; Leica TCS Sp2 AOBS, Germany). Excitation/emission wavelengths for the chelating fluorochromes were 405/580 nm for tetracycline hydrochloride (yellow), 543/617 nm for Alizarin Red S (red), and 488/517 nm for calcein (green). The percentage of each single fluorochrome staining area, indicating the new bone formation and mineralization at 2, 4, and 6 weeks after surgery, was calculated with Image‐Pro Plus software (Media Cybernetics, Silver Spring, MD, USA). After fluorescence analysis, these sections were stained with Van Gieson's picrofuchsin for histological observation and histomorphometric analysis. Push‐Out Test : This biomechanical test was conducted on a universal material testing system (Instron, High Wycombe, UK). A special custom‐designed holder was used for the test samples to ensure the test force was along the long axis of the implants, which were trimmed to fit into the holder. All tests were conducted at a loading rate of 5 mm min −1. The load–displacement curves were recorded during the pushing period. The failure load was defined as the peak load value of the load‐displacement curve. SEM Observation : Rat femurs with four groups of implants were randomly selected for SEM observation. Briefly, after the push out tests, the implants were fixed with 2. 5% glutaraldehyde solution at 4 °C overnight, and then sequentially dehydrated in a graded series of ethanol solutions (30, 50, 75, 90, 95, and 100%, v/v). Prior to SEM observation, the samples were sputter coated with platinum. The corresponding elemental distribution on the implant surfaces was detected by EDS mapping. In Vitro Osteogenic Evaluation: ALP Activity Assay : Rat BMSCs were seeded in 24 well plates at density of 5 × 10 4 cells per well. After 7 d of culture, the cells were stained using an ALP kit (Beyotime, China), according to the manufacturer's instructions. For ALP quantitative assay, the cells were incubated with p‐nitrophenyl phosphate (Sigma) at 37 °C for 30 min. ALP activities were determined by recording optical density (OD) values at 405 nm. Total protein contents were calculated using a Bio‐Rad protein assay kit (Bio‐Rad, USA) and normalized with a range of bovine serum albumin (BSA) (Sigma) standards at 630 nm. ALP activities were expressed as OD values at 405 nm per mg of total protein. Calcium Deposition Assay : Calcium deposition was evaluated at 14 d by ARS staining. Cells were washed twice with phosphate buffered saline (PBS) and fixed in 95% alcohol for 15 min. The cells were stained with 0. 1% ARS solution and then desorbed with 10% cetylpyridinium chloride (Sigma) for quantification. The OD values for absorbances of the eluent were recorded at 590 nm. Total protein contents were measured using the Bio‐Rad protein assay kit at 630 nm. Results were normalized and expressed as OD values per mg of total protein. Real‐Time PCR Analysis : Total cellular RNA was extracted at 1 and 7 d with TRIzol reagent (Invitrogen) according to the manufacturer's instruction. Two micrograms of total RNA was used as the template for reverse transcription with Prime‐Script RT reagent kit (Takara Bio, Shiga, Japan). The expression of osteogenic genes, including Runx‐2, OCN, OPN, Col‐I, and ALP, was determined using a real‐time PCR system (Bio‐Rad) with SYBR GREEN PCR Master Mix. The primer sequences for these genes are listed in Table S1 (Supporting Information). Gene‐specific primers were synthesized commercially (Shengong Co. , Ltd. Shanghai, China). The housekeeping gene, glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH), was used for normalization. CeONP‐0 acted as the control for relative gene expression. All assays were carried out in triplicate. Immunofluorescence Observation : OPN expression was detected by seeding rat BMSCs on samples at a density of 5 × 10 5 cells per well and cultured for 7 d. The samples were then washed with PBS three times and fixed in 4% paraformaldehyde at 4 °C for 30 min. Subsequently, the cells were permeabilized with 0. 1% Triton X‐100 for 30 min and blocked in 10% goat serum for 1 h at room temperature. A specific primary antibody targeting rat OPN (Abcam) was added at 1:100 dilution for overnight coincubation at 4 °C. DyLight 488‐conjugated anti‐mouse IgG antibody (Boster, China) at a 1:100 dilution was added and incubated for 30 min at room temperature in the dark. These samples were observed by CLSM after the cell nuclei were contrast‐labeled with DAPI (Sigma, USA). Macrophage Study: Cell Culture : The RAW264. 7 murine‐derived macrophage cell line was purchased from the Cell Bank of Chinese Academy of Sciences and maintained in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin at 37 °C in humidified 5% CO 2 incubator. The culture medium was exchanged every 48 h. The growing cells were passaged at ≈80% confluence by scraping, and only passages 3–5 were used in the study. Flow Cytometry Analysis : Flow cytometry was used for quantitative analysis of the expression of CCR7 (M1 marker) and CD206 (M2 marker). In total, 5 × 10 5 cells were seeded onto various samples (20 mm × 20 mm × 1 mm). After culturing for 4 d, the cells were trypsinized and scraped from the sample surfaces, centrifuged, and resuspended in 1% BSA for 30 min at ambient temperature to block nonspecific antigens. The cells were then incubated with fluorescein isothiocyanate (FITC)‐conjugated anti‐mouse F4/80, allophycocyanin (APC)‐conjugated CCR7, and phycoerythrin (PE)‐conjugated CD206 for 1 h in the dark at ambient temperature in a final volume of 100 µL. FITC‐conjugated rat IgG2a, κ, APC‐conjugated rat IgG2a, κ, and PE‐conjugated rat IgG2a, κ acted as isotype controls. All flow cytometry antibodies were purchased from eBioscience. After washing twice in 1% BSA, the cells were resuspended in 1% BSA and analyzed with a Guava flow cytometer (Millipore, USA); 5 × 10 3 cells were analyzed in each test. Data were analyzed using guavaSoft 3. 1. 1 software. Cytokine Measurement : The production of relevant cytokines was measured using enzyme‐linked immunosorbent assays (ELISAs). After 1 and 4 d of culture, the culture medium was aspirated and centrifuged, and the supernatant was utilized for analyses. The concentrations of TNF‐α and IL‐10 were measured with ELISA kits (Anogen, Canada), following the manufacturer's instructions. Statistical Analysis : All data were expressed as mean ± standard deviation. Statistically significant differences ( P ) among groups were measured by one‐way analysis of variance (ANOVA) and SNK post hoc analysis, based on normal distribution and equal variance assumption test. All statistical analyses were performed using SPSS v. 10. 1 software (IBM SPSS, Armonk, New York, USA). A value of P < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201700817
| 2,018
|
Advanced Science
|
A Multimaterial Scaffold With Tunable Properties: Toward Bone Tissue Repair
|
Abstract Polyetheretherketone (PEEK)/β‐tricalcium phosphate (β‐TCP) scaffolds are expected to be able to combine the excellent mechanical strength of PEEK and the good bioactivity and biodegradability of β‐TCP. While PEEK acts as a closed membrane in which β‐TCP is completely wrapped after the melting/solidifying processing, the PEEK membrane degrades very little, hence the scaffolds cannot display bioactivity and biodegradability. The strategy reported here is to blend a biodegradable polymer with PEEK and β‐TCP to fabricate multi‐material scaffolds via selective laser sintering (SLS). The biodegradable polymer first degrades and leaves caverns on the closed membrane, and then the wrapped β‐TCP is exposed to body fluid. In this study, poly( l ‐lactide) (PLLA) is adopted as the biodegradable polymer. The results show that large numbers of caverns form on the membrane with the degradation of PLLA, enabling direct contact between β‐TCP and body fluid, and allowing for their ion‐exchange. As a consequence, the scaffolds display the bioactivity, biodegradability and cytocompatibility. Moreover, bone defect repair studies reveal that new bone tissues grow from the margin towards the center of the scaffolds from the histological analysis. The bone defect region is completely connected to the host bone end after 8 weeks of implantation.
|
1 Introduction Polyetheretherketone (PEEK) is a promising biopolymer, as its mechanical strength and elastic modulus are close to those of human cortical bone. 1 β‐tricalcium phosphate (β‐TCP), a bioceramic material, has good bioactivity as well as biodegradability. 2 The binary scaffold composed of PEEK and β‐TCP may theoretically combine the appropriate mechanical properties of PEEK and the good bioactivity and biodegradability of β‐TCP. While in fact PEEK acts as a closed membrane in which β‐TCP is completely wrapped after the melting/solidifying processing, plus PEEK membrane almost does not degrade, 3 making it difficult for the binary scaffold to exhibit the bioactivity and biodegradability. How to exert the bioactivity and biodegradability of the binary scaffold consisted of undegradable biopolymer and bioactivity ceramic still remains a challenge. A few researches indicated that blending biodegradable polymer into undegradable polymer could exhibit the biodegradability of scaffolds. 4 Subramanian et al. had fabricated polyhexylthiophene/poly(lactide‐co‐glycolide) (PLGA) scaffolds via electrospinning and found that the scaffolds possessed good biodegradability due to the addition of biodegradable PLGA. [[qv: 4b]] Some researches had been conducted by blending of biodegradable polymer with bioceramic to display the bioactivity of scaffolds. 5 Diba et al. had fabricated the polycaprolactone/forsterite scaffolds via salt leaching/solvent casting method and found that the scaffolds exhibited good bioactivity. [[qv: 5a]] The above studies indicated that it may be an effective method to solve the problem by blending a biodegradable polymer with PEEK and β‐TCP. When the scaffold formed by blending a biodegradable polymer with PEEK and β‐TCP was immersed in body fluid, the biodegradable polymer will first degrade, leaving a lot of caverns on the closed membrane. And then the wrapped β‐TCP may expose from the membrane and exchange ions with body fluid. Among various biodegradable polymers, poly( l ‐lactide) (PLLA) has been approved by the United States Food and Drug Administration for use in tissue engineering owing to its good biodegradability and biocompatibility. 6 It can degrade through hydrolysis of ester groups (COO − ) in the polymer backbone to lactic acid, which is a normal metabolite of the human body. The lactic acid can finally metabolize to carbon dioxide and water through tricarboxylic acid cycle. 7 Previous studies have also reported that PLLA possessed good biocompatibility for cell attachment, proliferation and differentiation, and bone regeneration. 8 In the present study, biodegradable PLLA powders were blended with the PEEK and β‐TCP powders, and PEEK/β‐TCP/PLLA multimaterial scaffolds were fabricated by selective laser sintering (SLS). The biodegradability of the scaffolds was characterized in the form of morphological change and weight loss in phosphate buffered saline (PBS). The bioactivity was assessed by examining apatite formation on scaffold surface in simulated body fluid (SBF). Besides, cytocompatibility was determined by studying the cell behaviors including adhesion, proliferation, and differentiation. Moreover, in vivo bone defect repair capacity of the scaffolds was investigated by radiological examination and histological analysis. 2 Results and Discussion The PEEK/β‐TCP/PLLA mixture powders were prepared by adjusting the PEEK and PLLA content while the total content of PEEK and PLLA was kept at 80 wt%. The four diffraction peaks (2θ = 18. 7°, 20. 7°, 22. 8°, and 28. 7°) were corresponded to PEEK, 9 and the characteristic peaks for both PLLA (2θ = 16. 3°) 10 and β‐TCP (2θ = 25. 7°, 27. 7°, 31. 1°, and 34. 3°) 11 were presented, suggesting the presence of both PLLA and β‐TCP in the mixture powders. The PLLA powders and β‐TCP powders were uniformly dispersed throughout the PEEK matrix (Figure S1, Supporting Information). The scaffolds possessed 3D porous structures with dimensions of Φ15 mm × 21 mm fabricated by SLS technology ( Figure 1 A). The pore sizes and strut sizes of the scaffolds were about 450 and 500 µm, respectively. The mapping distribution of C, O, Ca, and P elements indicated the β‐TCP powders were uniformly distributed in the scaffolds (Figure 1 B). The Ca/P ratio was 1. 48, which was close to the ratio of 1. 5 for β‐TCP. 12 It has been reported that an average pore size in the range of 300–500 µm of scaffolds is essential for new bone formation. 13 The pore structure can be fabricated via SLS, 14 electrospinning, 15 gas foaming, 16 particulate leaching, 17 and so on. Among these, SLS has received considerable attentions because it can easily control the pore size. 18 In this study, the pore size of the PEEK/β‐TCP/PLLA scaffolds was fabricated at around 450 µm by controlling laser spot diameter and scanning line interval. Figure 1 The characteristics of scaffolds. A) Optical graphs. B) EDS mapping images. The biodegradation behaviors of the scaffolds were characterized in terms of the morphological change and weight loss in PBS at 37 °C. After being immersed for 28 d, the surface morphologies of the scaffolds with 0–50 wt% of PLLA content changed significantly due to PLLA degradation ( Figure 2 A). The 0PLLA scaffolds had a compact structure (Figure 2 A1), and caverns were not found on the surface there. Some small caverns were formed on the 10PLLA scaffolds (Figure 2 A2). The cavern size and depth increased rapidly as PLLA content increased. For the 30PLLA scaffolds, some large caverns appeared on the scaffold surface due to the fusion of smaller ones, and the caverns grew deeper and more irregular (Figure 2 A4). Most of the smaller caverns disappeared and the larger caverns formed on the 50PLLA scaffolds (Figure 2 A6). The scaffold structure was relatively stable as PLLA content increased from 0 to 30 wt%, while parts of the scaffold structure were collapsed as PLLA content further. The weight loss of the scaffolds with 0–50 wt% of PLLA content over a 28‐day period was obtained by gravimetric analysis (Figure 2 B). There was no weight loss for the 0PLLA scaffolds, indicating no degradation of the scaffolds. The weight loss of the scaffolds increased with increasing PLLA content. Figure 2 The degradation behaviors of the scaffolds after PBS immersion. A) SEM micrographs of the scaffolds with 0–50 wt% of PLLA content (A1–A6) after PBS immersion for 28 d. B) Weight loss of the scaffolds with 0–50 wt% of PLLA content as a function of degradation time. The surface morphologies and the corresponding energy dispersive X‐ray spectroscopy (EDS) mapping images of the scaffolds without and with PLLA before and after PBS immersion for 28 d were characterized using scanning electron microscopy (SEM) and EDS analysis ( Figure 3 ), respectively. For the 0PLLA scaffolds, the β‐TCP particles were completely wrapped in the PEEK membrane. The main chemical elements including C, O, Ca, and P were observed from the EDS mapping images. The C element was derived from PEEK, while the Ca and P elements were derived from β‐TCP. The Ca element concentrations in the scaffolds before and after PBS immersion were 6. 14% and 6. 66%, respectively. For the PEEK/β‐TCP/PLLA scaffolds with PLLA, the surface morphologies changed markedly with lots of caverns after PBS immersion for 28 d. Some of the β‐TCP particles were exposed on the cavern wall due to the PLLA degradation. As a result, the Ca element concentration in the scaffolds increased significantly to 9. 74%, which was much higher than that in the scaffolds before PBS immersion. Figure 3 The surface morphologies and the corresponding elemental mapping images of the 0PLLA scaffolds and the scaffolds with PLLA before and after PBS immersion. The bioactivity was investigated by evaluating the apatite formation on their surfaces in SBF for 28 d, and the corresponding morphologies were observed by SEM ( Figure 4 A). It could be observed that there are no deposits on the 0PLLA scaffold surface (Figure 4 A1), while the number of deposits increased significantly with increasing PLLA content (Figure 4 A1–A6). The deposits grew in size and formed a continuous layer on the surface of scaffolds when PLLA content increased to 30 wt% or above. To identify phase compositions of the deposits, X‐ray diffraction (XRD) was conducted on the scaffolds after SBF immersion for 28 d, while the scaffolds without SBF immersion were served as control. Before SBF immersion, the characteristic peaks corresponding to both PEEK and β‐TCP were detected in the 0PLLA scaffolds (Figure 4 B). The characteristic peak at 2θ = 16. 3° belonging to PLLA appeared and the relative intensity increased with the increasing of PLLA content. There were no other peaks formed in the scaffolds beside the characteristic peaks of PLLA, β‐TCP, and PEEK. These results indicated that no chemical interaction between PLLA, PEEK, and β‐TCP occurred. After SBF immersion, a characteristic peak at 2θ = 31. 8° was ascribed to the diffracting plane (211) of hydroxyapatite (HAP), 19 and the relative intensity increased with increasing PLLA content in the scaffolds (Figure 4 C). The results demonstrated that the scaffolds possessed the capability to induce apatite layer formation, and the formation capability could be enhanced as PLLA content increased from 0 to 30 wt% while changed little as PLLA content further increased. Figure 4 The bioactivity of the scaffolds with 0–50 wt% of PLLA content before and after SBF immersion for 28 d. A) SEM micrographs of the scaffolds (A1–A6) after SBF immersion. B) XRD patterns before SBF immersion. C) XRD patterns after SBF immersion. Ideally, scaffold materials should be biodegradable while new bone tissue is growing, and have good bioactivity to promote the bone‐binding ability and osteocompatibility for bone regeneration. Some recent studies have used surface functionalization methods 20 (such as noncovalent functionalization and covalent functionalization) and biodegradation behavior approaches 21 (such as copolymerization and blending) to synthesize scaffold materials for tissue engineering applications. Abueva et al. [[qv: 20a]] had fabricated a multichannel hydroxyapatite scaffold, which was functionalized with phosphonic groups using poly(vinyl phosphonic acid) to allow adsorption of a chitosan layer. This modification improved the biocompatibility of the scaffold and also served as a buffer between the implant scaffold and bone tissue. Kundys et al. [[qv: 21a]] had synthesized polylactide (PLA)–poly(1, 4‐butylene adipate) (PBA) copolymers and found that the copolymerization of PBA with PLA could adjust the degradation rate of copolymers. It is well known that PEEK is a bioinert material and thereby has limited ability to bind with natural bone tissue. 22 Introduction of bioceramic (such as TCP, HAP, bioglass, and so on) is an effective method for improving the bioactivity. 23 In fact, PEEK acts as a closed membrane in which bioceramic is completely wrapped after the melting/solidifying processing, plus PEEK almost does not degrade, hence the scaffolds cannot display the bioactivity and biodegradability. Therefore, in this present study, biodegradable PLLA was introduced to exert the bioactivity of β‐TCP. The results from the biodegradability and bioactivity tests showed that a lot of caverns formed on the closed membrane due to PLLA degradation, and the caverns became larger and deeper with PLLA content increasing. As a result, the wrapped β‐TCP particles exposed from the membrane into body fluid environment. The β‐TCP particles degraded and exchanged ions with body fluid, and hence induced the deposition of apatite on scaffolds surface. Compression tests were conducted to investigate the compressive strength and modulus at different PLLA content ( Figure 5 A). It could be seen that the 0PLLA scaffolds had the strong ability to resist compressive deformation with the maximum compressive strength (33. 64 ± 1. 62 MPa) and modulus (2. 62 ± 0. 24 MPa). Conversely, the 50PLLA scaffolds had the lowest compressive strength (16. 89 ± 1. 29 MPa) and modulus (1. 28 ± 0. 23 MPa). The strength and modulus of other types of scaffolds were intermediate between those of the 0PPLA scaffolds and 50PLLA scaffolds, which exhibited a trend of decline with the increase in PLLA content. Moreover, it was interesting to observe that the strength and modulus of the scaffolds changed relatively stable as PLLA content increased from 0 to 30 wt%, while decreased dramatically as PLLA content further increased. The strength and modulus of the 30PLLA scaffolds were considerably higher than that of the 40PLLA scaffolds and 50PLLA scaffolds ( P < 0. 01). In fact, PEEK possessed excellent strength and modulus, which were much higher than that of PLLA. 24 As a result, it was reasonable to make a conclusion that incorporation of PLLA into PEEK decreased the mechanical properties of scaffolds, and the mechanical properties were mainly dominated by the continuous phase PEEK in the case of PLLA content less than 30 wt%. Figure 5 The mechanical and thermal properties of the scaffolds with 0–50 wt% of PLLA content. A) Compressive strength and modulus (** P < 0. 01). B) TGA and DSC profiles. Thermogravimetric analysis (TGA) and differential scanning calorimeter (DSC) analysis were carried out to investigate the thermal properties of the scaffolds at a flow rate of 10 °C min −1 (Figure 5 B). It could be seen that the scaffolds without PLLA revealed two‐stage thermal decomposition, while the scaffolds with PLLA exhibited three‐stage thermal decomposition from TGA curves. The thermal decomposition of PLLA mainly occurred within a temperature range between 180 and 380 °C. 25 The stage ranging from 550 to 630 °C was due to the thermal decomposition of PEEK. 26 The results from TGA analysis indicated that the addition of PLLA in PEEK decreased the thermal stability of scaffolds, as the starting decomposition temperature shifted to lower temperatures. In DSC curves, the endothermic peak at about 334 °C was attributed to the melting temperature of PEEK. 27 Two endothermic peaks located at about 174 and 355 °C were attributed to the melting temperature and decomposition temperature of PLLA, 28 respectively. The relative peak intensity increased with increasing PLLA content. The results indicated that both PEEK and PLLA existed in the scaffolds. Cell adhesion assay was carried out by seeding MG‐63 cells onto scaffolds and culturing in culture medium for 7 d ( Figure 6 A). Cells attached tightly on scaffold surface with well‐flattened and expanded morphology. Cells cultured on the PEEK/β‐TCP/PLLA scaffolds with PLLA (Figure 6 A2–A6) had well‐spread morphology compared with that on the 0PLLA scaffolds (Figure 6 A1). They were fully spread and formed a confluent layer on the surface of scaffolds when PLLA content was 30 wt% or above. The proliferation of MG‐63 cells was investigated by CCK‐8 assay (Figure 6 B). It could be seen that all the scaffolds possessed the capability for cell proliferation, and the optical density increased with culture time. Compared with the 0PLLA scaffolds, the PEEK/β‐TCP/PLLA scaffolds with PLLA significantly up‐regulated cell proliferation ( P < 0. 01). The cell proliferation on the scaffolds with 30 wt% or above PLLA content was higher than that on the scaffolds with lower PLLA content ( P < 0. 01 or P < 0. 05). The differentiation of MG‐63 cells cultured on the scaffolds with 0–50 wt% of PLLA content for 7 d was assessed in terms of alkaline phosphate (ALP) activity (Figure S2, Supporting Information). The ALP activity of MG‐63 cells on the 30PLLA scaffolds was higher than that on the 0PLLA scaffolds, indicating the significant up‐regulated osteogenic differentiation of the cells. The enhanced cell adhesion and accelerated proliferation and differentiation might be related to the improved ions exchange ability between β‐TCP and culture medium due to the PLLA degradation. Previous studies have demonstrated that the released Ca and P ions from bioceramic to culture medium could stimulate cell adhesion, proliferation, and differentiation. 29 Figure 6 The adhesion and proliferation of cells on the scaffolds with 0–50 wt% of PLLA content. A) SEM micrographs of cells after 7 d of culture. B) Cell proliferation on the scaffolds after cell culture using MTT assay (* P < 0. 05, ** P < 0. 01). The viability of MG‐63 cells was assayed using Calcein‐AM staining after 7 d of cell culture, where green fluorescence indicated live cells ( Figure 7 A). It could be observed that cells spread very well, and a large proportion of live cells adhered on the scaffold surface. The number of live cells increased significantly with increasing in PLLA content in the scaffolds. The percentage of spread cells on the scaffolds was obtained using Photo Shop software (Figure 7 B). The number of cells on scaffold surface increased with culture time prolonging. The percentage of spread cells on the 30PLLA scaffolds was significantly higher than that on the scaffolds with 20 wt% PLLA content or below ( P < 0. 01), indicating an improved cell proliferation. The results were consistent with that of the cell adhesion and proliferation assays. Figure 7 Viability of cells on the scaffolds with 0–50 wt% of PLLA content. A) Calcein‐AM staining images of cells after 7 d of cell culture. B) Percentage of spread cells after 1 and 7 d of cell culture (** P < 0. 01). Considering the application for bone regeneration, scaffolds should be biocompatible, mechanically competent, biodegradable, and bioactive. From the above analysis, it was clear that a higher PLLA content in scaffolds led to a higher degradation rate, bioactivity, and cytocompatibility. The mechanical properties of the scaffolds changed relatively stable as PLLA content increased from 0 to 30 wt%, but decreased dramatically as PLLA content further increased due to the collapse of scaffolds. The scaffolds possessed good degradability, bioactivity, and cytocompatibility at PLLA content of 30 wt%. The comprehensive performances of the scaffolds were undoubtedly optimal when PLLA content was 30 wt%. Therefore, the 30PLLA scaffolds were adopt to evaluate the cell viability and differentiation as well as the bone defect repair capacity, while the 0PPLA scaffolds served as control. Live MG‐63 cells were stained with calcein acetoxymethylester (Calcein‐AM, green), dead cells were stained with propidium iodide (PI, red), while cell nucleuses were stained with 4, 6‐diamidino‐2‐phenylindole (DAPI, blue) ( Figure 8 A). Live cell density after 7 d of cell culture was much higher than that after 1 day of cell culture. Very few dead cells were observed for all the culture time points. ALP staining was conducted after cell culture to study the cell differentiation (Figure 8 B). From the staining images, it could be seen that the 30PLLA scaffolds had the ability to promote MG‐63 cells differentiation, and the ALP activity increased with increasing culture time. Figure 8 The viability and differentiation of cells on the 30PLLA scaffolds after cell culture for different time intervals. A) Fluorescence morphologies of cells grown on the scaffolds. B) ALP staining images of cells on the scaffolds. Radius bone defect of New Zealand white rabbit was adopt as animal experimental model to evaluate the bone regeneration ability of the scaffolds. The experimental groups A and B were packed with 0PLLA scaffolds and 30PLLA scaffolds, respectively, while the blank group was an untreated defect as control ( Figure 9 A). No signs of inflammation or infection were observed for any animal after implantation. The rabbits were radiographed using X‐ray to evaluate the degree of bone regeneration after implantation for 4 and 8 weeks (Figure 9 B). In the experimental groups A and B, opaque calcified shadow was observed after implantation for 4 weeks, but the calcified density was lower than that of normal tissues. The area of shadow in experimental group A was larger than that in experimental group B, indicating that the 30PLLA scaffolds had good bone formation ability. After implantation for 8 weeks, the shadow in the bone defect region disappeared in experimental group B, and the bone defect region was completely connected to the host bone end. The shadow in the bone defect region remained in the experimental group A. The bone defect in the blank group remained and could not self‐repair. The bone regeneration ability was also studied by microscopic computed tomography (micro‐CT) (Figure 9 C). After 4 weeks, there was obvious new bone formation in the defect region implanted with the 30PLLA scaffolds. However, only minimal new bone formation for the 0PLLA scaffolds and the implanted scaffolds could be clearly observed. After 8 weeks, the defect was completely repaired for the 30PLLA scaffolds, while the defect was not repaired for the 0PLLA scaffolds. There was no evidence of bone defect repair for the blank group. Figure 9 Bone regeneration ability of the scaffolds. A) Surgical procedure for creating bone defect model in rabbits. B) X‐ray images of radius regeneration after 4 and 8 weeks of surgery. C) Micro‐CT images of new bone formation after 4 and 8 weeks of surgery. Histological analysis was performed using Masson trichrome and H&E staining to evaluate the bone defect regeneration quality in all groups. Masson trichrome staining images could give the information on new bone formation after 4 and 8 weeks of surgery ( Figure 10 ). After implantation for 4 weeks, the experimental group B had an obvious increased bone tissue formation as compared with the experimental group A. The new bone tissues grew from the margin toward the center of the scaffolds following with the degradation continued of the scaffold. After 8 weeks, only a small part of scaffold material remained in the bone defect region, and the residual scaffold material formed a coexistence structure with the new bone tissue. The formation ability of new bone tissue in the experimental group B was much higher than that in the experimental group A. While in the blank group, there was almost no new bone tissue formation. H&E staining images of the bone defect sections in the experimental groups A and B after 2, 4, and 8 weeks of surgery were also obtained ( Figure 11 A). The analysis further confirmed that the 30PLLA scaffolds were able to enhance bone defect repair relative to the 0PLLA scaffolds. The quantitative analysis of the new bone area at each implantation time was calculated from the corresponding H&E staining images (Figure 11 B). It could be noted that the new bone area progressively increased over time for all the scaffolds. The amount of new bone tissue in the 30PLLA scaffolds was higher than that in the 0PLLA scaffolds after the same implantation time ( P < 0. 01 or P < 0. 05). The results indicated that the 30PLLA scaffolds possessed excellent bone formation ability and high efficiency of bone regeneration. Figure 10 The masson trichrome staining images of bone defect sections in the experimental group A, experimental group B, and blank group after 4 and 8 weeks of surgery (SM: scaffold material; NB: new bone; MB: mature bone). Figure 11 The histological images and quantitative analysis of new bone formation. A) H&E staining images of the bone defect sections in the experimental group A and experimental group B after 2, 4, and 8 weeks of surgery (SM: scaffold material; NB: new bone; MB: mature bone). B) Quantitative analysis of new bone area (* P < 0. 05, ** P < 0. 01). The data from X‐ray and micro‐CT analysis indicated that the 30PLLA scaffolds induced more new bone formation than the 0PLLA scaffolds. And the histological evaluation further confirmed the results. The enhanced bone repair capacity of the 30PLLA scaffolds could be attributed to the degradation of PLLA. It could be observed from the degradation experiments that a lot of caverns formed on the 30PLLA scaffold due to the PLLA degradation. Previous studies had shown that the caverns on the scaffold surface were benefitted to cell migration and proliferation, thus promoting new bone tissue regeneration. 30 Besides, β‐TCP particles gradually exposed and degraded during the degradation of scaffolds, which provided an ideal microenvironment for osteoblast adhesion, proliferation, and differentiation. 31 From the cytocompatibility experiments, it had been demonstrated that the 30PLLA scaffolds had enhanced cell adhesion and accelerated proliferation compared with the 0PLLA scaffolds. As a result, the new bone formation ability was significantly improved. 3 Conclusion In the present study, PLLA powders were blended with the PEEK and β‐TCP powders, and the PEEK/β‐TCP/PLLA three‐phase scaffolds were fabricated via SLS for bone regeneration. The scaffolds had good biodegradability with the degradation of PLLA, and lots of caverns formed on the membrane. The wrapped β‐TCP exposed from the membrane and exchange ions with body fluid. As a result, a layer of apatite deposited on the scaffold surface, and the scaffolds displayed good bioactivity. Cell culture experiments demonstrated that the PEEK/β‐TCP/PLLA scaffolds exhibited improved osteoblast adhesion, proliferation, and differentiation, revealing good cytocomptibility. In vivo bone defect repair assessments confirmed that the scaffolds had excellent new bone formation ability and high efficiency of bone regeneration according to the micro‐CT, X‐ray, and histological evaluations. In summary, the PEEK/β‐TCP/PLLA scaffolds possessed good bioactivity, biodegradability, cytocompatibility, and new bone formation capacity, and therefore, have promising potential for bone repair applications. 4 Experimental Section Fabrication of the Scaffolds : Mixture powders were prepared by mechanical mixing of PEEK, β‐TCP, and PLLA powders in ethyl alcohol. The PEEK/β‐TCP/PLLA (5:2:3 wt/wt/wt) mixture powders served as an example, and its procedure was described as follows: 2 g of β‐TCP powders (Kunshan Chinese Technology New Materials Co. , Ltd, China) were added into 50 mL of ethyl alcohol. The mixture solution was magnetic stirring using a magnetic stirrer (Jintan Ronghua Instrument Manufacture Co. , Ltd, China) for 30 min followed by sonication using an ultrasonic cleaning device (Shanghai Kudos Ultrasonic Instrument Co. , Ltd, China) for 30 min. Then 5 g of PEEK powders (Dongguan Guanhui Plastic Materials Co. Ltd, China) and 3 g of PLLA powders (average particle size: 0. 2–5 µm, purchased from Jinan Daigang Biomaterial Co. , Ltd. ) were dispersed sufficiently in 150 mL of ethyl alcohol by constant magnetic stirring and sonication. The β‐TCP suspension and the PEEK/PLLA suspension were subsequently mixed and continuously sonicated for another 1 h with magnetic stirring. Thereafter, the mixture powders were collected through filtration, washed with deionized water, and dried at 60 °C in an electrothermal blowing dry box (Guangzhou Dayang Electronic Machinery Equipment Co. , Ltd, China). The mixture powders with different PEEK:β‐TCP:PLLA weight fractions of 7:2:1, 6:2:2, 5:2:3, 4:2:4, and 3:2:5 were prepared as described above. Scaffolds were fabricated using SLS according to the following procedure. Briefly, a cylindrical scaffold model (15 mm in diameter, 21 mm in height) was designed using Solidworks 2011 software (Solidworks Corporation, USA), and then converted into stereolithography (STL) format that could be recognized by the SLS system equipped with a CO 2 laser (Rofin‐Sinar Laser GmbH, Germany) and a dynamically focusing optical system (Sunny Technology Co. , Ltd, China). Laser sintering of the powders was produced using spot diameter of 500 µm, scanning speed of 120 mm s −1, scanning line interval of 950 µm, and layer thickness of 0. 1–0. 2 mm. After sintering was completed, the scaffolds were removed, and unsintered powders were removed by blowing compressed air. The PEEK/β‐TCP/PLLA scaffolds were labeled as 10PLLA, 20PLLA, 30PLLA, 40PLLA, and 50PLLA scaffolds for the 7:2:1, 6:2:2, 5:2:3, 4:2:4, and 3:2:5 ratio of PEEK:β‐TCP:PLLA, respectively. While the PEEK/β‐TCP scaffolds were labeled as 0PLLA scaffolds for the 8:2 ratio of PEEK:β‐TCP as control. Characterization of the Scaffolds : The morphologies were characterized using a SEM (FEI Co. , USA). Before SEM observation, the specimens were platinum sputter coated using an auto fine coater (JEOL, Ltd. , Japan) under argon atmosphere for 120 s. The elemental distribution of calcium (Ca), phosphorus (P), oxygen (O), and carbon (C) in the scaffolds was investigated using an EDS (Phenom World BV, Netherlands). The phase composition was investigated using a XRD diffractometer (German Bruker Co. , German) with Cu Ka radiation (λ = 0. 154056 nm, 40 kV, 40 mA). Before analysis, the scaffold specimens were fixed on a specimen holder by double side adhesive tape. The data were recorded in the 2θ range of 10–40° with scanning speed of 8° min −1. Compressive properties were evaluated using a universal testing machine (Shanghai Zhuoji Instruments Co. , Ltd, China) at a compression rate of 0. 5 mm min −1. All scaffolds were prepared in the form of cylinders (Φ15 mm × 15 mm) and vertically placed between two parallel plates. Modulus was computed with the initial slope of the stress–strain curve. Each experiment was repeated six times and the results were averaged. The thermal properties were determined by TGA and DSC using a synchronous thermal analyzer (Nanjing Dazhan Institute of Electromechanical Technology, China) under nitrogen atmosphere. Before measurement, the scaffolds were cut into short pieces, and the specimens with a weight of ≈8 mg were placed in an aluminum crucible. The measurements were carried out over a temperature range of 50–650 °C at a heating rate of 10°C min −1. The TGA and DSC data were collected simultaneously during the test. Biodegradability and Bioactivity of the Scaffolds : The biodegradation behaviors of the scaffolds were examined in PBS (Beijing Chemical Reagent Company, China) solution at 37 °C. Five scaffolds (Φ15 mm × 5 mm) for each group were weighed as W 0, followed by immersion in PBS at 37 °C for up to 28 d (7, 14, 21, and 28 d). After completion of each incubation period, the scaffolds were carefully extracted from the medium, rinsed with distilled water for removal of ions absorbed on scaffold surface, dried at 60 °C for 12 h to constant weight, and weighed as W 1. The weight loss (%) was determined using the formula: 32 Weight loss (%) = ( W 0 − W 1 )/ W 0 × 100%, where W 0 and W t are the dry weight of the scaffolds before and after immersion, respectively. The surface morphologies after the scaffolds soaked in PBS at different time intervals were observed using SEM, and the elemental composition and distribution were measured using EDS. The bioactivity was assessed by immersing the scaffolds in SBF solution at 37 °C for 7, 14, 21, and 28 d and testing the growth of apatite on their surface. The SBF with ion concentrations nearly equal to that of human blood plasma was prepared by dissolving NaCl, KCl, NaHCO 3, MgCl 2 ·6H 2 O, CaCl 2, K 2 HPO 4 ·3H 2 O, and Na 2 SO 4 (reagent grade, all from Beijing Chemical Reagent Company, China) into distilled water and buffered with trishydroxymethyl aminomethane and hydrochloric acid solution to pH 7. 4. 33 The scaffolds (Φ15 mm × 5 mm) were immersed in the SBF for different time intervals at a concentration of 0. 1 cm 2 mL −1 with a volume of 100 mL. After the predetermined time, the scaffolds were removed, rinsed three times with deionized water, and dried at 60 °C for 12 h. The formation of apatite on the scaffold surface was investigated using SEM and XRD. Cell Culture of the Scaffolds : MG‐63 cells (American Type Culture Collection, USA) of passage 3–4 were used to evaluate the cytocompatibility of scaffolds, and grown in Dulbecco's modified Eagle's medium, supplemented with 1 × 10 −3 m glutamine, 1% penicillin/streptomycin, and 10% fetal bovine serum (all from Cellgro Mediatech, Inc. , USA) in a 5% CO 2 incubator at 37 °C. Before cell seeding, scaffolds (Φ15 mm × 5 mm) were sterilized in 70% ethanol, washed with PBS, and placed under ultraviolet (UV) light for 30 min. Subsequently, the scaffolds were incubated in culture medium for 30 min, and then transferred into 12‐well culture plates. The cell suspension with a density of 1 × 10 5 cells mL −1 was seeded on the scaffolds. The cells/scaffolds constructs were cultured in 95% humidity atmosphere with 5% CO 2 at 37 °C for up to 7 d (1, 3, 5, and 7 d), and the medium was refreshed every two day. For cell adhesion, cells/scaffolds constructs were harvested and rinsed with PBS to remove nonadherent cells after culture for a given period. Afterward, they were fixed with 2. 5% paraformaldehyde (Sigma‐Aldrich Co. , USA), washed with PBS, dehydrated through a series of graded alcohol, and dried in the electrothermal blowing dry box at 40 °C for 24 h. After drying, the specimens were sputter‐coated with platinum and then examined using SEM. For cell viability, the cells/scaffolds constructs were separately exposed to 15 µg mL −1 calcein AM, 4. 5 µg mL −1 PI, and 0. 5 µg mL −1 DAPI (all from Sigma‐Aldrich Co. , USA) for 30 min. The stained specimens were analyzed using a fluorescence microscope (Olympus Corporation, Japan). Stained images were obtained, in which green, red, and blue represented live cells, dead cells, and cell nucleus, respectively. The images were analyzed using Photo Shop software (Adobe Systems Inc. , USA) to quantify the percentage of spread cells from six representative images. For cell proliferation, 20 µL Cell Counting kit‐8 (CCK‐8, Sigma‐Aldrich Co. , USA) solution was introduced to each well according to manufacturer's instructions, and then the cells/scaffolds constructs were incubated for 4 h at 37 °C in a CO 2 incubator after cell culturing. 100 µL of supernatant medium was transferred into a new 12‐well plate, and the optical density at 570 nm of the solution was evaluated using a microplate reader. The ALP activity was evaluated by a colorimetric assay using an ALP reagent with p‐nitrophenyl phosphate (p‐NPP, Sigma‐Aldrich Co. , USA) after cell culture. The staining was carried out using ALP staining kit according to manufacturer's recommendations, and the stained scaffolds were photographed using a microscope. In Vivo Bone Regeneration : Bone regeneration ability in vivo was investigated by measuring new bone formation using the model of rabbit radius bone defects. All animal experiments were performed at Xiangya Hospital of Central South University in accordance with protocols approved by the Institutional Animal Care and Use Committee. New Zealand white rabbits at 5 months of age and 2. 5–3 kg of weight were randomly divided into three groups corresponding to blank, 0PLLA scaffolds, and 30PLLA scaffolds. Twenty seven rabbits were used for each group and nine rabbits were assigned for different time intervals (2, 4, and 8 weeks). All the rabbits were anesthetized with pentobarbital, and a 20 mm longitudinal incision was made along the radius. After the skin and musculature were separated, a 10 mm bone defect was made using a reciprocating saw. The bone defect modes were established and divided into three groups. Experimental groups A and B were implanted with 0PLLA scaffolds and 30PLLA scaffolds, respectively, while the blank group was kept empty as control. The incision was closed using resorbable suture, and the rabbits were given three days of intramuscular injection of penicillin 10 000 units per day. The rabbits were sacrificed with an overdose of pentobarbital for tissue harvest and analysis after 2, 4, and 8 weeks of surgery. To evaluate new bone formation in the bone defect sites, the harvested specimens were radiographed using an IVIS Lumina XR instrument (Perkin‐Elmer, USA) at each time point. Radiographs were obtained at a suitable magnification, and the degree of new bone formation was determined by the grey scale from the X‐ray imaging system. For micro‐CT observation, the radius was scanned using a micro‐CT imaging system (SkyScan, Belgium) with 80 kV and 450 µA. After X‐ray and micro‐CT analyses, the harvested bone specimens were fixed in 10% formalin, dehydrated with a graded ethanol series, defatted with chloroform, demineralized with 10% disodium ethylenediaminetetraacetate dihydrate (all from Sigma‐Aldrich Co. , USA) solution, and embedded in paraffin blocks. Vertical sections with a 5 µm thickness were cut from the middle of defect using a microtome. They were stained with H&E and Masson's trichrome, and observed using a microscopically. New bone area was measured using the Photo Shop software (Adobe Systems Inc. , USA) and calculated by using the following equation: 34 New bone area (%) = A n / A o × 100%, where A n and A o are the new bone area and original defect area, respectively. For this analysis, six images were randomly obtained in the same section. Statistical Analysis : All quantitative data were given as mean ± standard deviation unless otherwise stated. Statistical analysis was performed using SPSS 19. 0 software (IBM Corporation, USA). Values of p < 0. 05 were considered significant, while p < 0. 01 were considered very significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201700931
| 2,018
|
Advanced Science
|
Blending Electronics with the Human Body: A Pathway toward a Cybernetic Future
|
Abstract At the crossroads of chemistry, electronics, mechanical engineering, polymer science, biology, tissue engineering, computer science, and materials science, electrical devices are currently being engineered that blend directly within organs and tissues. These sophisticated devices are mediators, recorders, and stimulators of electricity with the capacity to monitor important electrophysiological events, replace disabled body parts, or even stimulate tissues to overcome their current limitations. They are therefore capable of leading humanity forward into the age of cyborgs, a time in which human biology can be hacked at will to yield beings with abilities beyond their natural capabilities. The resulting advances have been made possible by the emergence of conformal and soft electronic materials that can readily integrate with the curvilinear, dynamic, delicate, and flexible human body. This article discusses the recent rapid pace of development in the field of cybernetics with special emphasis on the important role that flexible and electrically active materials have played therein.
|
A Pathway toward a Cybernetic Future Journal: Advanced Science Publish Date: 2018 Abstract: Abstract At the crossroads of chemistry, electronics, mechanical engineering, polymer science, biology, tissue engineering, computer science, and materials science, electrical devices are currently being engineered that blend directly within organs and tissues. These sophisticated devices are mediators, recorders, and stimulators of electricity with the capacity to monitor important electrophysiological events, replace disabled body parts, or even stimulate tissues to overcome their current limitations. They are therefore capable of leading humanity forward into the age of cyborgs, a time in which human biology can be hacked at will to yield beings with abilities beyond their natural capabilities. The resulting advances have been made possible by the emergence of conformal and soft electronic materials that can readily integrate with the curvilinear, dynamic, delicate, and flexible human body. This article discusses the recent rapid pace of development in the field of cybernetics with special emphasis on the important role that flexible and electrically active materials have played therein. Body: 1 Introduction The field of cybernetics emerged in the early 1960s to describe the possible merging of inanimate materials with living organisms. 1 The original definition of cybernetics has subsequently been expanded to also encompass technical healthcare monitors for the health‐conscious consumer and implants for the sick and disabled. Indeed, throughout the course of history, the survival of mankind has relied on its ability to develop materials such as textiles, alloys, metals, and various types of rubbers, gums, and clays that could address our biological limitations in relation to a sustainable livelihood in the unfriendly and changing habitats of the Earth. In a broad sense, cybernetics therefore represents an expansion of the material‐making industries of the past. However, in striking contrast to ordinary materials, cybernetic extensions are items that can integrate with the body to overcome the limitations of human biology in an even more daring manner compared to external clothing, tools, or machines for that matter. Therefore, cyborgs (short for “cybernetic organisms”) could ultimately be the logical evolution of humans into a more adaptable, smarter, and stronger organism ( Figure 1 ). Figure 1 Recent innovations in materials science are leading humanity on the road to a cybernetic future—a future wherein, the fine‐line between machine and human will slowly fade away and pave the way for cyborg‐like humans. These cybernetic extensions can monitor physiological signals, stimulate tissues, restore lost tissue functions, or even impose new superhuman abilities in their user. A great variety of cybernetic concepts are currently under development, some have already hit the healthcare market, while others are being carefully evaluated in laboratories all over the world. 2, 3, 4 Classified broadly, these devices can be organized into three major categories: a) wearable healthcare monitors that can provide the user with individualized health information; b) prosthetics that can replace disabled organs or body parts; and c) implants that possess the ability to transcend human biology beyond its current limitations ( Figure 2 ). Figure 2 The union of microelectronics, flexible materials, and living tissues has led to a great variety of cybernetic devices that can bring relief to the lives of those disabled from disease or aging, enhance the physical capabilities of human beings beyond normality, and expand human consciousness toward uncharted territories. The manufacturing of the abovementioned cybernetics relies on a number of challenges that have only recently begun to be addressed in depth. First, one of the key requirements of any device intended to interface with the human body is a proper integration between the device and the body to assure long‐term performance. To this end, a number of challenges related to the mechanical mismatch between conventional rigid electronic materials and the dynamic, soft, and curvilinear human body need to be solved as the elastic moduli of silicon and metal‐based electronics range from 1–170 GPa, 5 which are in a stark contrast to the moduli of soft tissues in the order of 1–1000 kPa. 6 Moreover, most of these electronic parts are made on 2D rigid silicon wafers with sharp edges that do not conform well to the curved human body. Finally, conventional rigid electronics are not flexible and typically break at ultralow strains (≈1%), 7 and therefore, they are unable to withstand the high physiological strains of organs and tissues. The mismatch between conventional electronics and the human body ultimately hinders a good physiological contact with biological tissues and is thus a limiting factor for proper device performance. To remedy the current situation, flexible bioelectronics have emerged, which enable an unusually facile and intimate integration with biological tissues, such as skin, heart, and brain. Notably, they have been applied as brain–machine interfaces 8, 9 to treat various neurological disorders, for real‐time monitoring of the beating heart, 10, 11 as skin‐based devices that can monitor important health‐markers 12, 13 and sophisticated prosthetics with the ability to reestablish a normal life for the hearing disabled, 14 the paralyzed, 15 and the blind. 16 These devices are unequivocally opening new possibilities for studying chronic diseases, improving surgical procedures, and empowering patients to self‐manage their own health. Building on these groundbreaking results, bioengineers from Harvard University have recently begun to interweave living tissues within intricate nanowire‐based materials to create hybrid constructs that are half man, half machine, 17, 18 or cyborg organic constructs (cyborganics) as the authors prefer to refer to them. Overall, the gateway into the field of flexible bioelectronics is currently heavily related to the development of new multifunctional materials, which are flexible, resilient, and electrically active at the same time. In this review, we will highlight the recent advances of such electronic and flexible materials with a special focus on how these formidable materials currently are being used by scientist all over the world to reshape the field of cybernetics into a readily implementable technology. To this end, we will exclusively focus on healthcare monitoring devices, cybernetic prosthetics, brain–machine interfaces, and the exciting emergence of cyborganics. The review will begin with a brief overview of flexible and conductive polymers, since these materials are instrumental in the engineering of flexible bioelectronics. We will then proceed to discuss the incorporation of liquid metals and 2D nanoelectronic materials into such polymers, as the research and development of such composite systems in our opinion will soon lead to a number of scientific breakthroughs in the field of cybernetics. We will initially focus on the application of these materials in wearable and implantable healthcare monitors before highlighting their use in cybernetic prosthetics. As a new standpoint, we have also reviewed the recent progress in the development of cyborganics and the possible new directions that this emerging field might spearhead in the future. Finally, some of the current material‐based limitations in the field of cybernetics and how they can be successfully addressed will be discussed. 2 Flexible and Insulating Polymers Flexible bioelectronics enable a great variety of biomedical applications that would otherwise be impossible to achieve using conventional rigid electronics, as these devices need to conform well to the human body in order to yield high‐quality recordings of physiological events. 19, 20 They also need to sustain their operational capacity in the dynamic in vivo environment without succumbing to the cyclic movements of tissues and organs. For these reasons, materials used in flexible bioelectronics must have exceptional flexibility, toughness, and good biocompatibility to blend with the human body. 7 Optical transparency is also instrumental in bioelectronics, as it makes wearable devices fit in visually with the human body, while at the same time providing the means for applications in which imaging is combined with multifunctional sensing. Polymers are promising material candidates in flexible bioelectronics, as they have many key advantages, including transparency, biocompatibility, and flexibility ( Figure 3 ). 21, 22, 23 In addition, polymers are inexpensive, scalable via roll‐to‐roll processing, and enable reasonable tradeoffs in optical transparency, chemical performance, and thermal stability compared to metals and glass‐based materials. 24 Flexible polymers are therefore currently considered as the material of choice for flexible bioelectronics, and thus they are the prime focus of this section. Figure 3 Special materials with special properties are required for proper interfacing between the electrical and biological components of cybernetic devices. 2. 1 Material Requirements Electronic devices are mainly fabricated at high temperatures; this may lead to expansion or shrinkage due to internal stresses in the underlying polymeric material and therefore dimensional instability of the device. 24 The relation between the thermal stability and device operation has made the coefficient of thermal expansion (CTE) and glass transition temperature important parameters for polymer materials in electronics. Obviously, the glass transition temperature of potential polymer candidates must be compatible with the process temperature to avoid dimensional changes during fabrication, while the CTE of the polymer should match with the low CTE of the conductive layers of the device to prevent strain accumulation and cracking during the fabrication process. Hence, polymers with a high glass transition point and low CTE are desired for flexible electronics. 24 Another key requirement is a high barrier property because moisture and gas permeation may lead to dimensional instability of the polymer material. Still, there are many applications for flexible bioelectronics where the materials need not to be fully impermeable—but semi‐permeable—to allow biosensors to operate, such as in smart bandages. 25 Polymers are also exposed to various chemicals and solvents during coating, patterning, etching, and other manufacturing processes and must have good chemical and solvent resistance. 26 Surface roughness is another important parameter in the fabrication of flexible electronics, since the current transport in such devices is limited by surface defects and cracks. 26 In a similar way, the surface defects can also catalyze crack formation when the device is bent and provide an easy pathway for unwanted diffusion of moisture and oxygen into the device. 24 Overall, a number of different organic or inorganic barrier coatings such as paralyne C, polyethylene terephthalate (PET), aluminum and indium tin oxides have been used to prevent solvent or moisture diffusion into flexible electronics, reduce surface roughness, and improve the adhesion between the electronic parts and polymer. 26 2. 2 Amorphous Polymers Polymers are essentially made from long chains of repeating monomer units held together by strong intermolecular interactions. 27 The intermolecular interaction type, which varies a lot from polymer to polymer, plays a prominent role in many polymer properties. Depending on whether the polymer chains are nicely ordered or disorganized, polymers are also classified as either crystalline, semi‐crystalline, or amorphous. 28 Notably, amorphous polymers typically exhibit a crystallinity <10% and crystalline polymers typically contain more than 80% crystalline structures, while semi‐crystalline polymers display a crystallinity ranging from 10% to 80%. 29 They are also categorized into these different groups in terms of their glass transition temperature ( T g ) and melting temperature ( T m ). In this respect, the amorphous polymers have no clear melting point and their glass transition temperatures are approximately between −125 and 350 °C. On the other hand, semi‐crystalline polymers have a distinct T m, while their T g in general is higher than amorphous polymers and typically in the range 75–260 °C. 24 Highly crystalline polymers, however, only exhibit a T m and do not have any T g associated to them. Crystalline polymers can especially have extraordinary properties such as high stability, strength, and good resistance toward the surrounding environment. 30, 31 These incredible properties of crystalline polymers are mainly associated with their well‐ordered molecular arrangements. For instance, such arrangements can severely restrict the molecular chain mobility, which in turn makes the polymer stronger mechanically and more rigid. However, in spite of this, crystalline polymers are in most cases not suited as materials for flexible electronics because of their rigidity, high surface roughness, and opaqueness. 24, 32 To this end, the transparencies of polymers are intimately linked to their crystallinity degree, due to the intensified light scattering taking place within crystalline regions. 33 Moreover, crystalline polymers have higher surface roughness due to the presence of growth facets on the crystallites. For these reasons, amorphous polymers are typically a preferred choice over crystalline polymers in flexible bioelectronics, as they are excellent transmitters of light, stretchable, and display smooth surface topographies. However, this comes at a high cost in terms of low barrier properties, dimensional instability, and high CTE coefficients. 24 Examples of amorphous polymers include polycarbonate and polyethersulfone, which have high optical transparency and good flexibility but show poor solvent resistance, high gas permeability, and dimensional instability at high temperatures. 24 In general, the CTE, optical transparency, and gas barrier properties of such amorphous polymers typically range between 54 and 75 ppm °C −1, 89–92%, and 50–80 g m −2 d −1, respectively. 24 These listed values are relatively higher than that of semi‐crystalline polymers. 24 Moreover, amorphous polymers exhibit modulus, flexural strength, and tensile strength in the range of 2. 1–2. 6 GPa, 93–115 MPa, and 50–84 MPa, respectively. 34, 35, 36 They tend to bend and deform, but are not highly stretchable, as they typically only can stretch up to 60–200%. 36 Polydimethylsiloxane (PDMS) is another interesting amorphous polymer that has found wide usage in bioelectronics due to its biocompatibility, nontoxicity, excellent optical transparency, and reasonable chemical resistance. 22, 37, 38 Due to its very low glass transition temperature ( T g ≈ −125 °C) PDMS is a highly flexible and stretchable (≈1000%) material with a relatively low shear modulus ranging between 100 kPa and 3 MPa and flexural modulus ≈54 MPa. 39, 40, 41, 42, 43 Importantly, the mechanical properties of PDMS are intimately linked with the way it is manufactured, 44 and it displays an unusually high CTE value (≈310 ppm °C −1 ) as compared to other amorphous polymers. 45 To date, PDMS has been used in many biological applications such as wound‐bandages, 46, 47 microfluidics, 48, 49, 50 and recently in various organs‐on‐a‐chip platforms. 51, 52 Polyimides (PIs) are an interesting class of amorphous polymers that offer most of the advantages of semi‐crystalline polymers along with a high glass transition temperature (350 °C) and low CTE (8–20 ppm °C −1 ). 26 They are also remarkably strong (2–4 GPa), flexible, and have an astonishing resistance toward chemicals and heat. 26, 53 PI has been used in the electronics industry for decades and has begun to gain momentum as a polymer for flexible bioelectronics in recent years. 9, 13, 54 Despite, the many advantages that PI‐based materials offer, they are not optically transparent (30–60%). 24 However, current cutting‐edge technology has enabled the incorporation of fluorine, sulfonic, or nonaromatic groups into PI to make it optically transparent. 55 PI‐based materials therefore hold great promise for flexible bioelectronics and have in recent years significantly advanced the field. 2. 3 Semi‐Crystalline Polymers Semi‐crystalline polymers are the most widely used group of polymers because they have unique properties that combine the best characteristics of amorphous and crystalline polymers. 56 Semi‐crystalline polymers are typically very flexible, whereas highly crystalline polymers are rigid, as briefly mentioned in the previous section. 31 Semi‐crystalline polymers also have a sharp melting point and do not soften gradually like amorphous polymers, which is a contributing factor to their superior dimensional stability at elevated temperatures as compared to amorphous polymers. 24 However, the sudden phase transition above the glass transition temperature significantly impairs the temperature operating range of semi‐crystalline polymers. Various semi‐crystalline polymers can be used for bioelectronics including PET, polyethylene naphthalate (PEN), and polyetheretherketone (PEEK). 24 These semi‐crystalline polymers show desirable mechanical flexibility, solvent resistance, high clarity, low CTE (≈20–45 ppm °C −1 ), and good moisture barrier properties (0. 1–0. 5%). 24 Notably, the mechanical and physical properties of semi‐crystalline polymers strongly depend on their morphology and the crystalline content within them. 57 Their elastic modulus, flexural strength, and tensile strength have been reported in the range of 2. 3–5 GPa, 70–170 MPa, and 55–200 MPa, respectively. 36, 58, 59, 60 Importantly, they are highly flexible and bendable, and can stretch up to 50–300%. 36 Although semi‐crystalline polymers have many of the aforementioned advantages, they do not possess the required upper operating temperature to match with the processing temperature of conventional electronic circuits (up to 350 °C), 24 since PET, PEN, and PEEK have glass transition temperatures in the range of 80–150 °C. 26 Another disadvantage of semi‐crystalline polymers is their high surface roughness, which significantly limits their performance in a range of applications. 24 However, flexible bioelectronics such as transparent electrodes, sensors, and actuators based on PET 61, 62, 63 and PEEK 64, 65 materials with acceptable performances have still been successfully fabricated. Parylene C is yet another promising semi‐crystalline polymer for flexible bioelectronics, as it is biocompatible 66 with a high T g point (80–100 °C), low CTE (35 ppm °C −1 ), 24, 26, 67 and displays excellent barrier properties toward water vapor, corrosive molecules, and various gases. 68, 69, 70 Moreover, parylene C is a low‐cost polymer that can be easily processed into a protective ultrathin layer on almost any material, making it an ideal choice for the fabrication of cheap and ultrathin conformal bioelectronics. 71, 72 Due to a growing technological demand, a significant amount of electronics are discarded and trashed every year. 73 There are serious environmental concerns about the hazardous and toxic materials present in such discarded electronics. 74 Therefore, the field is moving toward fabrication of flexible bioelectronics based on biodegradable and nontoxic materials in order to reduce the consequent accumulation of toxic waste. Flexible bioelectronics that can dissolve inside the body are also gaining momentum, as they can help to reduce the electronic waste materials and assure swift elimination of the bioelectronics once their mission in the body is accomplished. 75, 76 Extensive research has been done on flexible and degradable electronics made from cellulose (paper) and silk‐based biopolymers. The degradation pathway of cellulose is typically driven through hydrolysis mediated by the family of cellulose enzymes, whereas silk—in its low‐crystalline state—is a highly water‐soluble protein, which disperses in physiological relevant solutions. Out of these materials, paper is particularly recognized for being a good material candidate in flexible electronics due to its high availability, flexibility, light weight, and sustainability. 77 Various electronic items such as thin film transistors, 78, 79 organic solar cells, 80, 81 disposable radio frequency identification tags, 82 batteries, 83 and wearable diagnostic devices 2, 84 have been made from paper‐based materials. Nanocellulose 85, 86, 87 and silk 88, 89, 90, 91 are also interesting biopolymers, as they are simultaneously strong, flexible, biocompatible, thermally stable, and recyclable at the same time. For instance, the decomposition temperature of nanocellulose has, in some instances, been shown to exceed 300 °C, 92 while CTE values of around 8 ppm K −1 have been reported, 93, 94 which are in the same range as that of glass 95 and metals 96 and are significantly lower than most commercial plastics (>200 ppm K −1 ). 26 The amazing properties of nanocellulose have prompted its use as a material in electronics, with device performance comparable to its rigid counterparts. 97, 98, 99 Silk also displays similar remarkable thermal and mechanical properties as nanocelluose, and in addition to its biodegradability, it can be easily tuned, making it a prime candidate in implantable bioelectronics. 10, 75 Therefore, there is no doubt that various paper formats and biopolymers will open new avenues for fabrication of high‐performance flexible bioelectronics that are cost‐effective, nontoxic, biodegradable, and eco‐friendly. 3 Polymeric Conductors and Semiconductors Many important biological processes such as neuronal activity, 100 the synchronous beating of heart, 101, 102 and muscle contraction 103 are all tightly controlled by electricity. For this reason, most bioelectronic devices base their functionality on the conversion of biological signals into electricity 104, 105 —a signal format that can be easily recorded, modulated, and analyzed—and thus enable a great variety of bioelectronics. Electrodes play a major role in this direction, as they can connect the world of electronics with the world of living tissues. Among the many materials, gold (Au) and platinum (Pt) electrodes are still the standard choice in bioelectronics; however, their high cost and low flexibility represent critical issues for their application in flexible bioelectronics. Over the years, significant progress has been made by chemists to develop conductive, biocompatible, and flexible polymers that can address the aforementioned challenges. These multifunctional polymers are beyond doubt the primus motor behind the exponential speed at which the field of bioelectronics is currently evolving with. Therefore, the main focus of this section is directed toward conductive polymers. 3. 1 Conductive Polymers Electrically active polymers have emerged as a new class of mechanically robust and biocompatible materials for flexible bioelectronics. 20, 106, 107 One of the most remarkable properties of these conductive polymers are the many functionalization pathways that they offer to fine‐tune the electrical and mechanical properties. They also offer other interesting features because some of them have intrinsic semiconductor properties that are essential for the design of basic electronic components such as transistors and field‐effect transistors. 108 Due to their formidable multifunctional properties, the research and development of new polymeric conductors has dramatically expanded during the last decades; the most‐studied ones being polypyrrole (PPY), polyaniline (PANI), and poly(3, 4‐ethylenedioxythiophene) (PEDOT). 3. 1. 1 PPY PPY is considered to be one of the pioneering polymeric conductors in bioelectronics due to its ease of functionalization and unique electrochemical properties (high conductivity and stability in oxidized states). 105, 109 Depending on the conditions and reagents utilized in the oxidation, the electrical conductivity of PPy can range from 10 to 100 S cm −1. 110, 111 It has been widely used as the outer electrode material for neural implants 112 and for recording neural impulses. 105, 113 PPY has also been used as a material for controlling cell‐fate through electrical stimulation and is recognized for its ability to yield high‐resolution electrical recordings from electroactive tissues and cells. 105 Despite of their exciting electrical properties most PPy films are brittle and mechanically unstable due to their conjugated chain structure, which significantly limits their usage in many flexible electronics applications. 114 To remedy this drawback, efforts have been dedicated for enhancing the strength of the PPy matrix by optimizing its manufacturing process 115, 116, 117, 118 and reinforcing it with other polymers. 114, 119, 120 3. 1. 2 PANI PANI was discovered as early as 1862, 121 but it was not before the beginning of the 1980s that it started to garner the attention of scientists in the medical sciences. 122, 123 Due to its exceptional stability, 124 reasonable biocompatibility, 125, 126, 127 and high conductivity, 128 its range of applications encompasses a variety of fields including biomedical engineering, 129 flexible electronics, 130, 131 and electromechanical engineering. 132 Moreover, because of PANI's good biocompatibility and high conductivity, it also holds great promise for use in the engineering of electroactive tissues. 133, 134 To this end, the electrical conductivity of PANI has been reported within the range of 10–100 S cm −1 ; a parameter that can be fine‐tuned through molecular weight, 135 temperature, 136 oxidation level, 137 crystallinity degree, 138 degree of doping, 139 and film morphology. 137, 140 However, unfortunately, since PANI is synthesized from acidic solutions, it tends to degrade in physiological environments significantly impacting its electrical stability in such environments. 141 One possibility to address this problem is to functionalize PANI with specific dopants via either noncovalent or covalent approaches, or by utilizing PANI nanomaterials such as nanowires, nanofibers, and nanorods instead. 142 3. 1. 3 PEDOT PEDOT is perhaps the most‐investigated electroactive polymer to date, 129, 143 as it keeps its ability to conduct electricity over a broad pH range. 144 It is also electronically stable in physiological environments 105, 145 and can merge with in vivo tissues without inducing toxic and foreign body responses. 46, 146 The polymeric backbone of PEDOT can also be easily functionalized to increase its conductivity, biocompatibility, and stability through the incorporation of various dopants, counter ions, and biological moieties. 147, 148 The most frequently used PEDOT derivative is PEDOT doped with poly(styrene‐sulfonate) (PSS)—PEDOT:PSS—an optically transparent polymer with an electrical conductivity that can go as high as 4600 S cm −1. 149 Even though, PEDOT:PSS films can be stretched up to ≈60%, the electrical conductivity in such strain regimes is highly compromised, and ultimately can present a great hindrance for the utilization of PEDOT:PSS in flexible electronics. 150 Recently, this grand challenge has been tackled by incorporating ionic additives and various electrical conductivity enhancers to generate highly conductive and stretchable PEDOT:PSS films. 151 These materials display a conductivity that can reach 4100 S cm −1 under 100% strains, because of their enhanced crystallinity and more interconnected polymeric networks. 151 3. 1. 4 Poly(3‐hexylthiophene) (P3HT) P3HT is another interesting semi‐conducting polymer, which has been widely used in various types of electronics—in particular organic solar cells—since its discovery in 1980. 152, 153 P3HT is easy to modify, nontoxic, and conductive, and is typically generated from monomers of 2, 5‐polythiopene (2, 5‐PT) by using various metals to initiate the polymerization process. 153 Besides its usage in solar cells, P3HT has also recently found its ways into the field of flexible bioelectronics, as it has been used in several flexible electronic devices capable of sensing various biomolecules. 154 3. 2 Semiconductors Replacing inorganic semiconductors with their organic counterparts involves a number of trade‐offs. The advantages include decreased manufacturing expenses, higher flexibility, and light‐weight. 108, 155, 156 On the negative side, however, organic semiconductors can be electrically unstable in physiological environments, 157, 158 they are more fragile than their inorganic counterparts, 108, 159, 160 and they display a weak long‐term in vivo performance due to faster biodegradation in the body. Over the years, several modification strategies aimed at strengthening intermolecular polymer bonds have been used to increase the stability of organic semiconductors. Especially, PEDOT:PSS has garnered significant attention, as it can be easily functionalized and is susceptible to electrical dopants. 147 PEDOT:PSS is a p‐type organic semiconductor that is very sensitive to the surrounding electrolyte concentrations, 156 and it is able to electrically respond to electrolytes through its amazing cation uptake ability. 161 This property provides PEDOT:PSS devices with an unusually facile pathway to measure the many electrolyte‐sensitive events inside physiological environments. The unique electrical properties of PEDOT‐based polymers have sparked tremendous interest in the past few years with the primary focus being directed toward their applications as organic bipolar junction transistors and field‐effect transistors. 105, 156 In simple terms, a bipolar junction transistor consists of three semiconductors: a collector, base, and an emitter. 162 A field‐effect transistor (FET) is a further extension of a transistor, as its operating principle is basically the same as a transistor with one exception, namely, the inclusion of an electrode (gate) above the transistor channel (base). 162 By regulating the gate voltage, it is possible to control the current through the transistor channel and into the source. FETs therefore possess an extra feature compared to an ordinary transistor because of their ability to switch between two states, that is, a low current state (off) and a high current state (on) ( Figure 4 ). In most bioelectronic applications, the channel and the gate are separated by an electrolyte solution. By applying a positive potential to the gate, the “on” and “off” states, and ultimately the current across the channel, can be tightly regulated, as the positive gate potential drives cations from the electrolyte solution into the transistor channel (Figure 4 ), which in turn makes the FETs sensitive toward various forms of currents with a typical response time around 100 Hz. Another important electrical effect in some polymer‐based FETs is the formation of an electrical double layer, which can significantly improve the response time, as this double layer provides the FETs with high‐capacitance, and thus enables them to operate with much higher frequencies (≈10 kHz). 163 This current sensitivity enables various sensing and detection schemes depending on the application. For instance, in a landmark study led by Malliaras and co‐workers, a simple but yet elegant PEDOT:PSS transistor was developed for glucose monitoring. 164 The PEDOT:PSS transistor was capable of monitoring glucose concentrations through a mechanism that involved the enzymatic conversion of glucose into the byproducts gluconic acid and hydrogen peroxide (H 2 O 2 ) by glucose oxidase. The formation of hydrogen peroxide was quite essential, as it significantly alters the gate potential resulting in a huge current drop across the transistor channel. 165 In another study, the same principle was used to measure the lactate concentration in blood, which is a well‐known marker for monitoring the effect of exercise, wellness, and physical fitness. 166 Instead of glucose oxidase, another enzyme (i. e. , lactate oxidase) was used to convert lactate into pyruvate. The generation of pyruvate changed the gate voltage and thus the current across the transistor channel. Despite the many interesting applications of PEDOT:PSS‐based transistors, they are still behind inorganic transistors in terms of their conductivity and response time. Figure 4 Energy bands and semiconductor‐related devices. a) Energy‐band diagrams for insulators, conductors, and semiconductors. The current in semiconductors are either generated from electrons (−) or electron holes (+). b) A transistor is basically made from three different semiconductors. The current can only run in one direction in a transistor, and the current that passes through it is typically enhanced with a sustainable gain factor, making them suitable for various sensing applications. c) A qualitative illustration of the working principles behind a sensor that is based on a field‐effect transistor made from PEDOT:PSS, which is an organic semiconductor capable of absorbing electrolytes (anions) from a solution. The uptake of anions abolishes the mobile holes within PEDOT:PSS and thus changes the current that passes through it; it thereby enables it to sense biological processes that either diminish or increase the amount of electrolytes in the surrounding environment. In summary, organic semiconductors hold great promise in flexible bioelectronics due to their reasonable sensitivity, high flexibility, biodegradability, and low cost. They have already successfully been used in a wide range of biological and medical applications, and further applications are expected once their response time and conductivity are made to match their inorganic counterparts. 4 Conductive Polymer Composites One avenue to bridge the current gap between conductive polymers and their inorganic counterparts is to reinforce them with inorganic fillers. The prime components of most of today's inorganically reinforced polymers are 2D and 1D nanomaterials. 167 A wide selection of such materials exists; the most studied ones are 2D graphene sheets and 1D carbon nanotubes (CNT) and silicon nanowires, however, other emerging nanomaterials are also worthy of consideration, such as boron nitride, silicene, germanene, and phosphorene. 167 Owing to their remarkable electronic, thermal, piezoelectric, mechanical, and moisture‐sensing properties, these nanomaterials have found widespread importance in flexible electronics. 167, 168, 169 Indeed, it is anticipated that their inclusion in electronics could revolutionize the entire industry and facilitate the emergence of better, faster, and smarter electronics that can be readily implemented within existing electronic formats. 168 The unique portfolio of properties that nanomaterials bring to the table can also be readily utilized in flexible bioelectronics to yield even more flexible and electrosensitive devices. 169, 170, 171 Another area that avenues to improve the electrical properties of flexible polymers is the incorporation of liquid metals into them, as liquid metals are conductive, self‐healing, and reconfigurable. It has therefore been foreseen that their incorporation into polymers can lead to sophisticated electronic circuits that can spontaneously repair upon damage. Even though the research and development of the abovementioned inorganic fillers is still in its infancy, the achievements obtained to date are truly remarkable and have already surpassed those reported by conventional organic‐based polymers. We are therefore confident that these fillers will be a game‐changer in flexible bioelectronics. One concern, however, is their cytotoxicity and biocompatibility in vivo, which still has not been carefully evaluated. This is beyond doubt one of the major challenges that needs to be addressed before their potential can be translated into devices that can integrate seamlessly with the human body. 4. 1 Graphene In brief, graphene consists of a monolayer of carbon atoms that are packed densely into a 2D hexagonal honeycomb lattice with a carbon–carbon bond length of 1. 42 Å 172 and a thickness of only one atomic layer (≈0. 3 nm), which makes it the “thinnest” material ever discovered. 173 This ultrathin 2D nanomaterial displays really high electrical conductivity (2. 50 × 10 5 cm 2 V −1 s −1 ), 174 good thermal conductivity (≈3000 W m −1 K −1 ), 174 amazing mechanical flexibility (ultimate tensile strength (130 GPa), high Young's modulus (≈1 TPa), 174, 175 and low coefficient of thermal expansion (CTE) (≈−8 × 10 −6 K −1 ). 176, 177 Moreover, due to its high specific surface area (up to 2600 m 2 g −1 ), 178 large aspect ratio (up to 2000), 179 chemical reactivity, and tunable interface properties, graphene is much easier to functionalize compared to many other nanomaterials. Typically, graphene is synthesized by either top‐down or bottom‐up strategies. 180 In the top‐down approach, graphene sheets are exfoliated from the bulk graphite via chemical reactions or mechanical forces to get single or few‐layers of graphene ( Figure 5 ). In the bottom‐up strategy, graphene nanosheets are directly grown onto a substrate—starting from single atoms or molecules—by using chemical vapor deposition (CVD), organic synthesis, or solvothermal synthesis methods. 181, 182 Until now, different types of graphene have been successfully achieved by using these two strategies. These different varieties include graphene oxide (GO), graphene oxide quantum dots, graphene quantum dots, and reduced graphene oxide (rGO) (Figure 5 ). 182, 183 Among them, GO has received most attention owing to its highly oxidized nature due to its large numbers of surface residual epoxides, hydroxyl, and carboxylic acid group, which in turn provides many chemically reactive groups for various functionalization purposes. GO can also easily be reduced into highly conductive graphene (rGO) by eliminating oxygen‐containing functional groups through chemical, thermal, or irradiation treatment. 183 Figure 5 Various a) manufacturing methods of graphene. Reproduced with permission. 447 Copyright 2012, Nature Publishing Group, b) graphene oxide, reduce graphene oxide, and graphene oxide nanosheets are highlighted here. c) Graphene contains numerous functionalities, which can be used to firmly attach it to the backbone of polymers. Given these unique features of graphene, it has attracted tremendous attention as an electroactive and mechanical nanoreinforcer with the capacity to turn nonconductive polymer‐based materials into amazing conductors of electricity with impressive mechanical properties. 184, 185 However, it should be emphasized that the electrical conductivity and sensitivity of graphene are typically greatly affected by other factors such as the presence of adverse functional groups on the graphene sheets, various intrasheet and intersheet structures, intersheet junctions, the aspect ratio of the sheets, and the processing methods used to manufacture them. 184, 186, 187 Overall, they are therefore potential candidates for the development of even better conductive and polymer‐based electrode–tissue interfaces for flexible bioelectronics. As an example, graphene incorporation within PEDOT has resulted in significant improvement of mechanical and electrical properties, features that were used to enhance the performance of neural and microelectrode interfaces. 188 In this case, the positive charge of oxidized PEDOT chains was ionically bonded to the negatively charged group of GO to form a stable conductive polymer film. An advantage of this functionalization strategy is that it prevents GO from dispersing into the target tissue during electro‐physiological recordings, and can thus minimize any possible cytotoxicity caused by graphene inside the body. Aside from chemical incorporation of graphene into PEDOT, graphene can also be physically mixed with PEDOT polymer chains to form a free‐standing film. This approach relies on the polymeric structure of PEDOT, which consists of many conjugated π bonds, that enable strong π–π stackings between PEDOT molecules and graphene sheets. For instance in one study, it was shown that such free‐standing graphene–PEDOT composite films can lead to an almost sixfold increase in mechanical strength and more than twofold improvement in electrical conductivity as compared to a pristine PEDOT polymer film. 189 Graphene may also be used to effectively reinforce PEDOT:PSS polymer film for flexible bioelectronics. Toward this endeavor, studies have shown that the graphene–PEDOT:PSS composites can be formed by either in situ polymerization or various blending processes. 190, 191 In the in situ polymerization method, graphene is typically dispersed in the PSS solution after which the EDOT monomer is gently added. Then, the polymerization process is initiated in the presence of a Fe 3+. 190 Compared to PEDOT:PSS film, the electrical conductivity of the graphene–PEDOT:PSS composite film was increased with up to 41% (637 S cm −1 ) at only 3 wt% graphene loading. Although this approach is effective for homogeneous dispersion of graphene sheets in the polymer matrix, it is limited, as the rate of polymerization is decreased at high graphene content. By contrast, the solution blending method is the most straightforward approach to use for developing polymer films at high graphene concentrations. To this end, Seol et al. recently developed a stretchable and transparent conducting electrode based on mixing PEDOT:PSS with rGO. 191 Specifically, the authors managed to reduce the adverse agglomeration of rGO by functionalizing rGO with a surfactant—(phenyl isocyanate)—that could reduce possible π–π interactions between rGO nanosheets and PEDOT:PSS. This flexible composite system displayed significantly higher optical transmittance (≈86%) and a greater electrical conductivity (1010 S cm −1 ), as compared to rGO–PEDOT:PSS and pristine PEDOT:PSS. Graphene has also in recent years been used in wearable sensor systems due to its ability to improve the accuracy in position, acceleration, and velocity detection of its wearer at high strains and strain rates. 192, 193, 194 In an enlightening study, Boland et al. showed that by loading graphene into a natural rubber, it is possible to produce conducting composites with electrical conductivity as high as 0. 1 S m −1. 193 In this system, the graphene nanosheets could quickly respond to polymeric deformations caused by dynamic movements in a time dependent manner, due to mechanically induced changes in electrical conductivity within the polymeric matrix. Moreover, this flexible composite system could stretch up to 8 times its original length without losing electrical functionality and mechanical integrity. Other noteworthy application of graphene is its inclusion into piezoresistive/piezoelectric polymer‐based sensors for converting the kinetic energy of the moving body into harvestable energy and in electronic skin (e‐skin) devices. 169, 170, 171 Other properties of graphene such as electrical and mechanical stability can be further coupled with control over the specific interactions between graphene and polymers to generate a self‐healing and conductive composite system. 195 The impartation of self‐healing function empowers the electronics to be revamped not only mechanically but also electrically, which is of great interest for electrical functional restoration after damage during large mechanical strain regimes. To this end, graphene has been incorporated into a self‐healing polyvinyl alcohol (PVA) polymer matrix to generate highly stretchable and self‐healing strain sensors. 196 The conductive graphene‐hydrogel based strain sensors displayed fast electrical healing speed (within 3. 2 s), remarkable self‐healing performance (≈98%), and was able to sustain high elastic deformation (≈1000%) with gauge factor of 0. 92. 196 Although this system is unequivocally opening new possibilities for usage in robotics, healthcare monitoring, and various human motion detection systems, further efforts should be focused on enhancing their stability and sensitivity to bring this interesting technology into the mainstream market. 4. 2 Carbon Nanotubes (CNTs) CNTs belong to the fullerene family of nanomaterial's and consist of graphene sheets that are rolled‐up into high‐aspect ratio tubes (>1000). 197 They come both as single‐wall carbon nanotubes (SWCNTs) and multiwall carbon nanotubes (MWCNTs), and for the most part “the two” show similar properties, however, the tensile strength of SWCNTs is significantly lower than MWCNTs, which makes SWCNTs more flexible than their multiwalled counterpart. 198 Over the years CNTs, have found widespread application in diverse fields ranging from electronics, medicine, and drug‐delivery, as they pose unique properties such as formidable tensile strength (11–63 GPa), 199 high Young's modulus (1–1. 8 TPa), 200 excellent intrinsic conductivity (10 9 A cm −2 ), 201 high thermal conductivity (2000–6000 W m −1 K −1 at room temperature), 202 and are thermally stable up to 2800 °C in vacuum conditions. 203 They are also perfect reinforcers for flexible electronics, as they are readily bendable and squeezable, and display spring‐like properties under constant loading. Furthermore, CNTs can be synthesized via various methods with the most frequently applied ones being electrical arc‐discharge, laser ablation, and CVD ( Figure 6 ). 204, 205, 206 With each of these methods, the surface chemistry, surface area, surface charge, and CNT size distribution can be uniquely fine‐tuned to yield desired CNT batches for further downstream applications. 204 In laser ablation, laser pulses are applied to pure graphite blocks to vaporize them into ultrathin pieces that subsequently can generate CNTs on a water cooled collector, 204 while in the electrical discharge process a high current is applied between two graphene electrodes—anode and cathode—in the presence of metallic catalysts, which catalyze the growth of CNTs on the cathode (Figure 6 ). 204 The most standard method employed however is CVD, which is based on the pyrolysis of hydrocarbons in a tube furnace and the usage of metallic‐catalysts to polymerize them into CNTs. As the CVD method is the most practical, economical, and pure pathway for CNT synthesis, it is also the one most commonly used for commercial‐scale production of CNTs. 205 Figure 6 Various CNT a) production and b) functionalization strategies. c) A self‐healing and flexible PVA‐CNT based composite for human motion detection. Adapted with permission. 196 Copyright 2017, Wiley‐VCH. The outstanding properties of CNTs have turned them into widely used nanofillers in polymer matrixes for the construction of flexible, stretchable, and deformable electronics. 22, 207, 208, 209, 210 Until now, several methods have been proposed to incorporate CNTs into polymers such as solution mixing, melt processing, and in situ polymerization. 211 However, untreated CNTs are chemically inert and too hydrophobic to disperse with ease in organic or inorganic solvents, and are therefore unable to establish strong interactions with the polymer backbone. The key to overcome this obstacle is the functionalization (covalent or noncovalent) of CNTs with hydrophilic groups such as hydroxyl, carbonyl, carboxyl, and amines to improve the dispersion stability and chemical reactivity of CNTs, and ultimately yield reinforced polymers with good electronic, magnetic, optical, thermal, and mechanical properties. 212 Moreover, the performance of stretchable CNT‐based electronics could be enhanced by alignment of the nanomaterials. Indeed, studies have shown that horizontally or vertically aligned CNTs within polymer matrixes can significantly improve the conductivity and mechanical performance of flexible electronics. 213, 214, 215, 216, 217 Notably, CNTs typically form a percolation network within the matrix, which in turn significantly enhances the electrical conductivity within the polymer. For instance, several studies have shown that CNTs networks can be formed within polymers to yield highly conductive and stretchable strain sensors. 22, 207, 208, 216, 218, 219 In one paramount study, SWCNTs were embedded into a stacked nanohybrid structure within polyurethane (PU)–PEDOT:PSS to provide a transparent, stretchable, and patchable strain sensor. 208 This strain sensor displayed good optical transparency (≈63%), a high gauge factor (62. 3), and could stretch up to 100% before breakage. In a similar vein, SWCNTs were embedded into PDMS to generate a sensor that could sense mechanical deformations arising at the “bone–skin” interface to enable the detection of joint‐movements in the human body with a mechanical sensitivity ranging as low as 10 5 MPa −1 at 0. 9 MPa pressure. 22 Despite of its incredible sensitivity this system, however, could only perform at strains at 30% as the conductive CNT percolation network otherwise would breakdown. One avenue to improve the CNT percolation network is based on generating a PDMS foam consisting of 3D‐interconnected networks and then dip the entire polymeric foam into a CNT solution. 209, 220 In this procedure the CNTs will adhere to the PDMS network to generate a percolated network of conductive wires. For example, in a recent groundbreaking study, this concept was successfully used to develop a highly sophisticated wearable strain sensor with an excellent gauge factor of 134 at a 40% strain value. 209 Another strategy to achieve a more favorable CNT network within a polymer matrix is the use of graphene sheets to form a hybrid CNT/graphene structure, which efficiently inhibits the bending and bundling deformation of CNT networks during successive stretchings. 220 In this scenario, the graphene nanosheets are able to disperse and adsorb pristine CNTs, whereas the CNTs are capable to function as bridges and avert the restacking of graphene nanosheets through π–π interactions. As previously mentioned, the integration of self‐healing properties into polymers has speared a major paradigm shift toward the development of flexible and stretchable electronic devices with higher durability during successive loadings. Toward this endeavor, a self‐healing piezoresistive strain sensor device, which is able to detect the dynamic movement of the human body was recently developed by incorporating SWCNTs into a self‐healing PVA polymer matrix (Figure 6 ). 196 The self‐healing mechanism was built into the system via reversible hydrogen bonds in the PVA matrix mediated by borate ions through a simple one‐pot mixing of CNT, PVA, and borax. These weak hydrogen bonds could easily break and reform, and were thus one of the main driving mechanisms behind the self‐healing properties of the manufactured device. Specifically, the device was incorporated into a Scotch permanent clear mounting tape, which acted as an elastomeric substrate, and enabled the system as a whole to stretch 1000% with an amazing self‐healing efficiency of 98%, after only 3 s of healing time. In addition to this formidable stretching ability, which according to the authors was the highest value ever recorded for such device, the SWCNTs themselves also established spring‐like and conductive links between the individual polymer chains, and thus contributed to a significant improvement of the elasticity, conductivity, and flexibility of the composite system relative to pristine PVA. The sensor device was then mounted onto human joints, and the authors demonstrated its capacity to monitor various human motions in real time through changes in resistivity brought about by the mechanical strains, that the PVA‐CNT polymer experienced during human joint motions. 4. 3 Metallic Nanowires Existing bioelectronic devices primarily use metal electrodes such as Au, Pt, and Ag because of their suitable electrical properties, biocompatibility and corrosion resistance. 221, 222, 223, 224, 225, 226 Especially, Au has been extensively used for bioelectronics due to its high ductility, good corrosion resistance, good biocompatibility, and long‐term operational stability. Au‐based electrodes have found many interesting applications in bioelectronics; some of the most noteworthy are as electrical conductors for glucose biosensors, 227 cochlear implants, 228 and in electrodes that enable communication between the brain and various machine formats. 72 For instance, Au nanostructures in the form of nanowires have gained great interests in flexible and stretchable bioelectronics, owing to their remarkable aspect ratio (≈10 000), mechanical and electrical properties. 229, 230 To this end, Gong et al. fabricated a wearable and highly sensitive pressure sensor by sandwiching ultrathin gold nanowires (AuNWs) between PDMS sheets. 229 This system was able to detect pressures as low as 13 Pa with a response time of <17 ms, and sensitivity of 1. 14 kPa −1 in the pressure range of 0–5 kPa. Due to its excellent sensing, flexibility, and robustness, this device was used for real‐time monitoring of blood pressure and various acoustic vibrations. Although, AuNWs‐based devices are quite promising for the field of flexible electronics their conductivity could be improved even more. A possible approach for improving the conductivity of such devices is by combining them with conductive polymers. For example, PANI microparticles have been doped into AuNWs films to yield a tenfold improvement in conductivity and eightfold enhancement in electrical sensitivity in comparison to pristine AuNW‐based strain sensors. 231 Silver nanowires (AgNWs) are another interesting class of metallic nanowires, which has been widely used in various types of wearable and flexible bioelectronics. 226, 232, 233, 234 AgNWs can form highly conductive percolative networks to yield a good optical transparency and a high structural flexibility. For instance, Ho et al. developed a transparent stretchable strain sensor based on percolating networks of both AuNWs and AgNWs on an elastomeric PDMS substrate. 232 This strain sensor displayed good optical transparency (≈66. 7%), high gauge factor (≈236), and could stretch up to 70% of its original length. Due to their high conductivity, compatibility, and mechanical deformability, AgNWs‐based polymer‐based electronics have also been recognized as promising nanomaterials for electrophysiological recordings. 226, 235 For instance, in a recent study, AgNWs were patterned into styrene–butadiene–styrene (SBS) elastomers to form serpentine‐like meshes capable of mimicking the elastic and electrical properties of cardiac tissue. 226 In this scenario, the AgNWs formed a highly conductive percolation network, while the SBS rubber acted as a binder to maintain the mechanical elasticity. This stretchable cardiac mesh was readily integrated with the curvilinear and dynamic in vivo heart and could ultimately improve cardiac contractile function in a post‐myocardial‐infarction model. Although, AgNWs and AuNWs have demonstrated promising results in the field of flexible electronics, they are relatively expensive compared to their organic counterparts Hence, copper nanowires (CuNWs) have become an appealing alternative, since they are cheaper, but yet, display an electrical conductivity similar to that of silver (Ag) and Au. 236, 237 So far, different types of wearable and flexible devices have been developed by incorporating CuNWs into various elastic polymers, such as poly (acrylate), 238 polyurethane, 239 PVA, 240 Eco‐flex, 241 and SBS. 242 Although, CuNWs have many of the aforementioned advantages, CuNWs display some disadvantages related to their easiness to become oxidized, which significantly limits their electrical performance in a range of applications as oxidative layers are highly insulating. 237, 239 One strategy to overcome this obstacle is by coating CuNWs with corrosion‐resistant metals such as nickel, Pt or Ag. 237, 243, 244 For example, Song et al. coated CuNWs with nickel (Ni) to improve their oxidation‐resistant stability. 237 Subsequently, the CuNW–Ni composite was embedded into PDMS to provide a conductive elastomer composite with transparency of 80% and resistance of 62. 4 ohm sq −1. This composite system could endure up to 600 cycles of bending, stretching, and twisting tests without breaking. 4. 4 Silicon Nanowires 1D silicon nanowires (SiNWs) are also steadily gaining a foothold in bioelectronics—albeit to a lesser degree compared to graphene and CNTs—due to their unique electrical, mechanical, and optical properties. 245, 246, 247 Indeed, compared to the other nanomaterials discussed here, SiNWs have an advantage in terms of their better semiconductor properties, 248, 249 which in turn makes them amenable in the development of nanoscaled transistors and FETs. 250, 251 Notably, the nanoscale diameter and high‐aspect ratio of silicon nanowires significantly alter, and in some cases improve their electrical properties as compared to solid silicon, due to quantum effects arising from quantum confinement within the wires. The small size of silicon nanowires also makes it much easier to control their electrical properties. For instance, one can significantly widen the band‐gap of silicon nanowires by simply decreasing their diameter and the orientation of the wire axis also have an important say on the many interesting properties of silicon nanowires. 252, 253, 254 Various techniques based on both top‐down and bottom‐up manufacturing have over the years been developed to generate silicon nanowires. 245, 246 In the bottom‐up process individual Si atoms are lined up into silicon nanowires with diameters between a few nanometers to several hundred nanometers via CVD and vapor–liquid–solid based methods. In the top‐down approach different lithography methods such as electron beam lithography and reactive‐ion etching are employed to carve out ultrathin silicon nanowires from solid silicon wafers. Even though, top‐down approaches are the most attractive to employ due to the high precision and flexibility they offer, these benefits come at a huge cost, as conventional lithography methods are time‐consuming, costly, and difficult to upscale to meet an industrial scale production. Due to their amazing electronic properties, silicon nanowire devices display impressive sensitivity when it comes down to measuring bioelectrical signals in the body. 255, 256 For instance, silicon nanowires have recently been employed in implantable bioelectronics with the purpose of enabling even better electrophysiological recordings as well as controlled drug release in response to important biological events inside the body. 15, 17, 257 Despite of these interesting advances, which will be elaborated in a more detailed manner in Section 6. 4, silicon nanowires have not been as widely used in the field of bioelectronics as compared to their carbon‐based counterparts, and the focus have for their part so far mainly been directed toward photovoltatic, electronic, and energy storage devices. The authors therefore anticipate that the incorporation of silicon nanowires into biocompatible polymers represent an interesting area ripe for investigations, and could therefore enable significant breakthroughs in the field. 4. 5 Liquid Metals Another class of inorganic materials for flexible electronics is liquid metals, as these materials are liquid near room temperature and therefore flow readily in response to stress ( Figure 7 ). 258 This enables them to deform and stretch in response to stress in a reversible manner, which makes them ideal candidates for stretchable and self‐healable electronics. 259, 260, 261 Especially, in recent years, there have been tremendous advancements in the emerging technology of stretchable electronics based on liquid metals such as mercury and gallium oxide (Ga 2 O 3 ). 262 However, mercury is a well‐known environmental toxicant, 263 gallium oxide on the other hand has a relatively low toxicity making it ideal for flexible bioelectronics. 262 Moreover, gallium has a number of interesting properties that makes gallium an attractive component to include in flexible bioelectronics, 264, 265 as gallium has high electrical conductivity (2. 2 × 10 6 S cm −1 ), 266 good thermal conductivity (28W m −1 K −1 at ≈37 °C), 267 is highly stretchable, 258, 262, 265 and reconfigurable due its inherent oxide film that considerably decreases its surface tension without significantly impacting its other properties (Figure 7 ); 268, 269 gallium oxide thin film thus permits liquid metal droplets to wet polymer surfaces and also enables an inconsequential barrier to form for optimal electrical charge transport. Moreover, the wettable nature of gallium also allows it to adhere to the polymer surfaces to form almost any shape for soft and stretchable circuits. Other gallium‐based liquid metal alloys such as ternary alloys (galinstan; 68% gallium, 22% indium, and 10% tin) and eutectic gallium–indium alloys (EGaIn; 75% gallium and 25% indium) have also attracted much interest, as gallium alloys are easier to shape due to their below room temperature melting points, while displaying otherwise similar physical properties to conventional gallium oxide films. 270 Figure 7 Gallium oxide, its properties, and application in flexible bioelectronics. a) The chemical structure of gallium oxide. b) The many unique properties that gallium oxide has to offer. c) Incorporation into polymers to yield flexible and electrical circuits. d) A gallium embedded PDMS substrate with high‐fidelity and stretchable circuits. Reproduced with permission. 272 Copyright 2013, Wiley‐VCH. Due to their unique material properties and below room temperature melting point (gallium melts at 30°), gallium alloys have recently been used as conductive fillers in place of rigid filler particles to improve the electrical and mechanical properties of elastomeric polymers. 271, 272, 273 In brief, viscoelastic gallium droplets are shaped into the desired electrical circuits by injection of the liquid metal into premade hollow architectural geometries within elastomers. The gallium‐based circuit is subsequently hardened by freezing the composite system—as a whole—below the melting temperature of the liquid metal solution. Gallium‐based alloys can thus readily be used to generate highly complex stretchable circuits by simple injecting them into such elastomeric materials (Figure 7 ). 265, 274 For instance, injection of EGaln into hollow poly[styrene‐ b ‐(ethylene‐ co ‐butylene)‐ b ‐styrene] fibers have lead to stretchable and electrically conductive circuits that could stretch up to 800% before mechanical failure, without losing their electrical continuity. 265 These elastomeric circuits could find widespread importance in many exciting applications ranging from flexible electronics, electronic textiles, stretchable wires, and flexible bioelectronics. Liquid alloys composed of gallium are also ideal candidates for creating conductive circuits within elastomeric substrates, which spontaneously can self‐heal electrical and mechanical defects imposed on them during wear and tear. 260, 275 A recent study accomplished this daring task through a simple system composed of a hollow self‐healing polymer (Reverslink@) into which EGaln was injected. This marvelous system—in the advent of damage—could spontaneously heal its electrical properties and mechanical properties after 10 minutes. In another recent study flexible galinstan‐based electronics was generated through inkjet printing of the liquid metal onto a stretchable PDMS substrate. Notably, the galinstan‐based circuit could spontaneously heal itself after damage, and was engineered in a manner, that enabled it to remain electrically and mechanically stable even after 2000 stretching cycles at strain of 60%. 276 The incorporation of liquid metals within polymers and then the deposition of the entire system onto a conductive metal trace is another avenue for achieving a self‐healing circuit, as this ingenuity enables liquid metals to readily reconnect distal parts emerged during tear caused by either wear or mechanical failure. 261 In perspective, owing to its unique self‐healing properties, stretchability and conductivity, we anticipate that the integration of liquid metals into self‐healing elastomers will pave the way for many exciting opportunities for highly flexible bioelectronics in the near future. Indeed, the aforementioned multifunctional properties of gallium and its alloys have also made them attractive candidates for bioelectronics. For instance, it has been reported that the EGaIn electrodes can be used to both record and stimulate electrical activity in individual neurons. 277 Another particularly interesting application utilized the electrical and physical properties of Ga‐based liquid metals to achieve an injectable soft 3D electronic circuit that could be delivered to target tissues, such as heart and sciatic nerve. 278 In detail, this electrode was made from a Ga 67 In 20. 5 Sn 12. 5 alloy encased in a biodegradable gelatin hydrogel to achieve a minimally invasive integration between electrode and tissue. Although gelatin is widely recognized as a biocompatible and degradable polymer, it however is not conductive and stretchable. To this end, we anticipate that it is possible to use other types of polymers such as conductive PEDOT‐based polymers to improve electrical and mechanical properties of the system. Another interesting feature of the liquid metals is their ability to change shape via a variety of mechanisms such as mechanical and electrical stimuli, because this unique trait could be useful for a number of applications such as reconfigurable electronics and bioactuators. Furthermore, recent studies have also shown that the combination of fluidity, deformation reversibility, and conductivity in liquid metals can yield artificial microrobots that can move through blood vessels or intestines to fulfill numerous biomedical purposes. 279, 280, 281, 282 For instance, the ability to reshape liquid metals by near‐infrared irradiation has recently been utilized to develop an innovative toolbox for controlled drug delivery and optical manipulation of artificial blood vessels. 282 These studies have opened a new avenue for the construction of intelligent biorobots that could not be obtained through conventional rigid materials. 282 Nevertheless, the in vitro and in vivo application of flexible microrobots require rigorous testing before they can be applied to combat and monitor various diseases. 5 Healthcare Monitors for Empowering the Patient Currently, there are a growing number of breakthroughs in bioelectronics driven by the emergence of better and smarter materials that can readily integrate with the dynamic human body. These technological breakthroughs are aiming to empower the patient through technical health aids that enable real‐time monitoring of medical risk factors. 283 This is accomplished by providing the healthcare consumer with individualized health data for self‐diagnosis and self‐management of their personal health. Such wearable healthcare monitors will also allow doctors to check on patients remotely instead of costly and frequent in‐person visits at the clinic. As a result of their projected importance in the healthcare industry, the market for healthcare monitors is growing quickly, and this year alone it is anticipated that over 19 million of such devices will be sold over the counter and over 100 million devices by the end of 2018. 4 In spite of the great promise that they hold, the reliability and validity of the data obtained from wearable healthcare monitors are still under intensive investigation. At the moment, there is a lot of ongoing research in the design and development of flexible and biocompatible materials that ultimately can be integrated into the field of bioelectronics to enable more reliable healthcare monitors. Most of the current efforts are directed toward incorporating these materials into devices such as e‐skin, 284 smart wound bandages, 283 and tattoo‐based sensors ( Figure 8 ); 285, 286 however, the research and development of materials for invasive and flexible bioelectronics that enable the monitoring of the beating heart and neurological activity in the brain are also slowly gaining momentum. 287 Here, we will highlight the recent progress in these emerging areas and briefly outline the possible future directions that they may take. Figure 8 The field of patient empowerment is currently driven by wearable healthcare monitors (e. g. , smart bandages, electronic skin devices, tattoo‐based sensors) and implantable monitors (e. g. , flexible electrodes for electrocardiography [ECG] and smart stents for angioplasty). Made by Harder&Muller. 5. 1 Wearable 5. 1. 1 Electronic Skin (E‐Skin) The human skin can reveal important information about the overall health status of the patient as the mechanical metrics of skin are intimately linked to the circulatory function of the body 288 as well as various skin‐related diseases such as melanoma, psoriasis, and eczema. 4, 289 Therefore, conformal bioelectronics that enables real‐time monitoring of the mechanical properties of skin can be used to detect potentially life‐threatening and chronic diseases in the home rather than in the clinic. 290, 291 Over the years a wide‐range of these so‐called “e‐skin” devices have been developed for healthcare monitoring purposes. 12, 62, 284, 289, 292, 293 In simple terms, e‐skin devices are flexible sensing networks with accurate spatial mapping and detection capabilities that enable unmatched recordings of the mechanical metrics of skin. This utility stems in part from the incorporation of 2D nanoelectronics, organic light‐emitting diodes (OLED), and pressure sensors within biocompatible, conformable, and stretchable plastic‐like materials that can withstand high strain deformations while still maintaining their electrical performance. 290, 291 The nanoelectronics and pressure sensors work in coherence to transform the viscoelastic changes of the skin into electrical signals for OLED processing into pixelated signals. 13, 284, 293 In this direction it is essential to use a mode of synthesis that can yield uniform devices, wherein the individual components are matched perfectly within the e‐skin system. Some of the most noteworthy examples of e‐skin devices include an elastomeric pressure sensor based on the incorporation of CNTs within a PDMS substrate, 12 elastomeric dielectrics integrated within an organic field‐effect transistor for measuring the artery pulse from the wrist, 62 and a PI‐based system that converts viscoelastic changes of skin into a pixelated and user‐friendly format through an intricate interplay between CNT‐based transistors and OLEDs incorporated within the PI‐substrate. 13 Building on these results, an e‐skin system was recently developed with the capability of in‐depth characterization of various skin lesions related to invasive melanomas. 289 This concept could potentially be used for rapid characterization of pathological skin conditions and become a platform for at‐home management of skin‐related diseases ( Figure 9 ). Figure 9 An e‐skin device for monitoring melanoma and skin lesions. a) The device was fabricated through a layer‐by‐layer assembly of Pt/Au electrodes, a piezoelectric component, and a soft and biocompatible PI‐based polymer interfacing the device with the human skin. b) A bright‐field image of the generated e‐skin device. Mechanical mapping of various skin pathologies located c) below the breast, d) on the leg, e) around the nose, f) on the forehead, g) close to the eye, and h) on the neck. Adapted with permission. 289 Copyright 2015, Macmillan Publishers Ltd. Recently, more sophisticated circuits have been built to expand the sensing capabilities of e‐skin to encompass changes in temperature, 294, 295 humidity, 169 and chemical variables. 296 These additional features enable the patient to detect potentially dangerous foreign bodies from entering the body through the skin and to monitor the effect of various lotions targeted against pathological skin lesions. Moreover, the temperature of the skin is intimately linked to the blood‐flow and therefore presents an important operational parameter that ideally needs to be incorporated into e‐skin devices. 295 Such multimodal e‐skin devices have so far mostly been based on CNT and graphene‐based nanoelectronics 169, 296 due to the excellent sensitivity of CNTs and graphene toward temperature and humidity changes. To this end, an e‐skin device that is capable of detecting chemical, temperature, and pressure stimuli has recently been developed by sandwiching CNT‐based circuits between PDMS substrates to yield a piezocapacitive system with an ultralow pressure sensitive (0. 4 Pa) and fast response time (63 ms). 296 Notably, this e‐skin system enabled the detection of a range of chemical fluids and could therefore be used as a wearable electronic nose capable of detecting potentially dangerous fluids. Overall, the abovementioned advancements have led to the development of multifunctional and mechanically robust e‐skin platforms with formidable sensory capabilities for wireless diagnostics. 297 These advancements highlight the amazing potential of e‐skin technologies and foresee the introduction of e‐skin devices that have the capacity to perceive additional stimuli for various healthcare monitoring. 5. 1. 2 Smart Wound Bandages Chronic wounds represent a global healthcare challenge that is expected to grow at a tremendous speed in the coming years as the population ages. 298 With the current lack of methodology to properly treat the growing number of patients suffering from chronic wounds a cumbersome bottleneck is expected. 298 This hurdle is intimately linked to today's time‐consuming, costly, and passive wound management scheme, wherein the wound site, is neither monitored nor attended properly. To remedy the current situation, smarter solutions that offer a better insight into the healing process rather than passive wound management are needed. With the recent advancements in biosensors, a new generation of highly sophisticated wound dressings are rapidly emerging to revolutionize the classical wound care concept. 299 These platforms act as conformal point‐of‐care bandages that consist of sensors capable of detecting important biomarkers of relevance for the wound‐healing process. 300 They also provide the possibility of remote diagnostics through wireless communication technology and can therefore result in reduced nursing and hospitalization costs. 301, 302 Over the years, a comprehensive list of potential biomarkers for wound healing has been established, which include markers such as pH value, temperature, proteins, inflammatory mediators, cytokines, enzymes, hormones, and nutritional factors. 303 The most widely used markers are temperature and pH‐value, as they are intimately linked with the extent of inflammation and infection at the wound site. 299, 304 Therefore, a range of temperature and pH sensors have been incorporated into flexible, permeable and biocompatible materials to yield smart wound bandages. 305, 306, 307, 308, 309, 310, 311 A recent example, is a surgical suture made from silk—a water‐soluble protein—for wound closure and real‐time monitoring of wound‐healing processes. A silicon‐based temperature sensor was incorporated within this silk suture to enable high‐resolution sensing (≈0. 2 °C) of temperature changes caused by inflammation at the wound sites. 307 The system was tested in an animal model and showed promising results that highlighted its potential as a bioresorbable suture capable of monitoring the progress of inflammation within the target site. Building on these results, a multimodal wound bandage was developed with the ability to provide highly accurate temperature readings from the wound site. 308 The system was multimodal, since it provided readings on both temperature changes and temperature conductivity. As previously mentioned, temperature mapping of the wound site captures the inflammation progress; temperature conductivity, on the other hand, correlates with the moisture content of the wound, which is another important marker that is intimately linked to the state of the wound. In simple terms, this system consists of a PI substrate containing a sensor array that is connected with ultrathin and flexible copper wires. The conformability of the PI‐based device and its capacity to record temperature‐related readings from the wound site were confirmed on human subjects. The reported results were indeed promising and indicated that the developed wound bandage could be readily implemented in a clinical setting. In addition to monitoring the infection and inflammation status of a wound, other approaches can be explored to promote the healing process by administrating drugs and growth factors to the site. In a recent study by Bagherifard et al. , 312 thermoresponsive drug‐carriers were incorporated into a hydrogel‐based dressing with a flexible heater for controlled delivery of drugs and growth factors to the wound site. The platform enabled a controlled release of various compounds in response to temperature changes and could therefore potentially be used in smart wound bandages, wherein drugs and growth factors can be released in response to increases in temperature from inflammation and infection. Bandages with textile‐like materials have also been extensively used for covering wounds, since they are porous, biocompatible, and capable of delivering oxygen and removing exudates from the wound site. 313, 314 They can be fabricated in large scales using well‐known and simple techniques, such as weaving, knitting, and embroidering. Another key advantage of using textiles is that the mechanical properties of individual strands and the entire fabric can be tuned by changing the properties of the fibers or the architecture of the fabric during the manufacturing process. 314 To this end, a flexible and fiber‐based pH sensor was recently fabricated by loading pH‐sensitive microspheres into alginate‐based microfibers. 23 The level of acidity at the wound site was measured by taking images of the fibers with a smartphone camera and analyzing the images using an in‐house application. Such hydrogel‐based fibers can also be easily knitted into intricate native‐like architectures and therefore enable the generation of customized bandages for wound‐healing applications. In a another recent study, an advanced wound bandage was developed from conductive threads that were embroidered onto a textile to form interconnected electrodes. 315 A range of physical and chemical sensors was later incorporated into the bandage to enable high‐fidelity measurements of temperature, glucose, and pH from biological fluids. It was also shown that the developed platform could potentially be used for measuring the physiological conditions at the wound surface through an external device, such as a smartphone or personal computer. Despite the recent advancements in the development of new smart wound management systems, the available sensors for wearable diagnostic dressings are currently directed toward few physiological markers and thus cannot be specific about the intricate mechanisms occurring in the wound environment. Moreover, the suggested platforms are rarely commercialized and are still in early preclinical stages, choosing the right combination of incorporated sensors and the management of the huge amount of output data is another challenge that needs to be addressed in the future to provide even better wound management systems. 5. 1. 3 Tattoo‐Based Electrochemical Sensors Tattoo‐based electrochemical sensors hold significant potential for low‐cost healthcare monitoring applications, thanks to their ability to perform real‐time and noninvasive electrochemical analysis of important biomarkers present in body fluids. 285, 292, 316 These wearable sensors, which can be concealed in almost any artistic tattoo pattern, can offer substantial insight into the wearer's health. 317, 318 The majority of the developed tattoo‐based diagnostic devices have so far been based on body sweat due to the wide range of information that it can provide on patient health. This wealth of information include electrolyte imbalance, 319, 320, 321, 322 bone mineral loss, 323 presence of heavy metals in the body, 324 lactate monitoring, 325 muscular damage, 324, 326 and so forth. Recently, a range of landmark studies by Bandodkar and Wang has led to the development of a variety of noninvasive tattoo‐based sensors that can effectively monitor several electrolytes and metabolites including sodium, 319 ammonium, 286 lactate, 325 zinc, 324 glucose, 327 and 2, 4, 6‐trinitrotoluene 328 from human sweat. These advanced tattoo‐based sensors are generated through a conventional screen‐printing technique with smart tattoo inks—onto paper‐based materials made form cellulose—that are reinforced with CNT's to yield mechanically robust devices. Moreover, contrary to their textile and plastic‐based counterparts, these paper‐based sensors developed by Wang and co‐workers demonstrate high mechanical compatibility with skin and thus enable unprecedented skin integration for real‐time chemical and biosensing. 329 Tattoo‐based wearable sensors have also been developed to monitor electromyography (EMG) signals from the skin surface. The monitoring and subsequent control of such electrical signals have over the years been used for many clinical purposes, such as identifying neuromuscular disorders, 330 gait disorders, 331 studying muscle pain, 332 and also to serve as a control signal for various prosthetic devices. 333, 334 In practice, most EMG skin‐contact electrodes are made from Ag/AgCl gel electrode (pasted with adhesive tapes or straps on skin), however, they are limited in their functionality as they quickly dry out. Other disadvantages are patient discomfort, poor signal transmission at the electrode–skin interface, and the requirement for skin cleaning/preparation. 333, 335, 336 Therefore, to overcome these challenges, “tattoo electronics” or “epidermal electronics” was recently applied to create skin‐like EMG sensors with thickness and mechanical properties resembling that of the human skin. One of the first examples of this concept was proposed by Rogers and co‐workers, 284, 337 who packed electrodes, electronics, sensors, power supplies, and communication components into an ultrathin, stretchable, silicone‐based membrane that could attach to skin in a similar vein like a tattoo. This system was tested on almost every part of the human body (forearm, throat, face, forehead, back of the neck, and index finger) and was found to have sufficient quality for measuring EMG signals generated by the contraction of skeletal muscles. 284, 333 Although, such flexible EMG devices are well suited for electrophysiological recordings at various locations on the body, they still face significant challenges related to their stretchability and electrical response time. One noteworthy attempt to remedy this bottleneck was based on using graphene‐based electronic tattoos, wherein graphene was incorporated into a flexible thin (≈463 nm) poly (methyl methacrylate) substrate. 338 This ultrathin device could readily become attached on human skin via van der Waals interactions and exhibited stable electrical conductivity even when the device was stretched to 50% strain values. For EMG sensing, they placed the device directly on the human forearm without any skin preparation and compared it with commercial Ag/AgCl gel electrodes. The resulting comparison showed that the electronic tattoo–skin interface impedance was almost as low as that of commercial gel electrodes with similar susceptibility to human motion. Like‐wise, tattoo‐based e‐skin has been developed for controlling human prosthetics by stimulating the muscles in the vicinity of the implant electrically. 339 To this end, EMG signals generated by muscle contractions of the biceps or triceps have been harnessed by e‐skin tattoos to move artificial limbs. Indeed, we anticipate that such sophisticated systems could spearhead the development of patient‐friendly and less invasive human–machine interfacing with the purpose of controlling the movement of various prosthetic devices. Further advancements in 3D printing of electronic materials into highly complex circuits are expected to lead to miniaturized and electronically integrated tattoo‐based skin devices with higher mechanical and chemical potency that can cover a broader range of biomedical applications. Moreover, development of tattoo‐based energy harvesting devices 340 is anticipated to open up new attractive paths toward the fabrication of entirely self‐sufficient tattoo‐based sensors. 5. 1. 4 Electroencephalography (EEG) The human brain has always been one of the biggest mysteries in biology and is still in many ways an uncharted territory. Hence, different types of recording systems have been invented to collect information that can elucidate what really lies hidden beneath the skull and deep within the human brain. EEG is one of the gold standards for this endeavor because of its cost‐effectiveness, inexpensiveness, noninvasiveness, and high temporal resolution. In brief, EEG devices record the spontaneous electrical activity along the human scalp to detect neural oscillations from the brain—the so‐called “brain waves”—as abnormal brain wave activity typically coincides with most brain disorders. 341 Although EEG device has found use in several diagnostic applications including Parkinson's, 342 schizophrenia, 343 epilepsy, 344 and Alzheimer's, 345 some major challenges still remain unsolved. These challenges include its bulky design, inability to conform to the body over long time periods and the need of a conductive gel at the skin‐electrode interface for efficient electrical coupling. 346 Other disadvantages are patient discomfort, cumbersome procedures for mounting the electrodes and inability to perform chronic recordings due to drying of the conductive gels. Therefore, new strategies that can reduce the discomfort associated with mounting and wearing the EEG equipment are needed. One avenue for remedying the current situation is through the development of new and better materials that can yield conformable EEG electrodes capable of adhering onto the human skin without creating any discomfort for the patient. An achievement like this will undoubtedly initiate a more widespread use of EEG devices for many patient empowering applications, including at‐home management of brain‐related disorders, sleep monitoring, and cognitive control. Several noteworthy solutions have already been suggested to initiate this landmark change with the major focus being directed toward the inclusion of soft, conformable, and adhesive electronics into the existing EEG concept. 347, 348, 349 To this end, a conductive polymer‐based EEG electrode was recently developed that enables a much better conformal contact with the skin and higher quality recordings compared to gel‐based EEG systems. 348 The vastly improved electrical coupling with the skin was achieved by the deposition of PEDOT:PSS onto a flexible PI substrate. Interestingly, clinical recordings carried out on human subjects with this new EEG electrode significantly outperformed those retrieved from conventional Au electrodes. In another recent study, a conformable electrode system was developed and ear‐mounted for brain recordings ( Figure 10 ). 347 The mounting of an EEG device onto a topological complex organ like the ear was overcome by embedding a stretchable and serpentine‐like Au electrode into a flexible and adhesive PI substrate. The serpentine‐like electrode architecture resulted in excellent bendability (>180°) and stretchability (>50%) while maintaining the operational capacity of the EEG device. The intrinsic properties associated with the compliant electrode configuration resulted in high‐fidelity recordings and conformability even during intensive exercise, swimming, showering, and sleeping. Figure 10 A flexible ear electrode for brain–machine interfacing. a) Schematic illustration of the electrode and the principles underlying the EEG monitoring. b) Images showing the electrode and its mounting on the ear. c) Finitive element method (FEM) analysis of the strain on the device upon mechanical bending (180°). d) EEG alpha wave recordings were fairly stable for up to 14 days after mounting the ear electrode on the user. Adapted with permission. 347 Copyright 2015, National Academy of Sciences. Conformable, durable and user‐friendly EEG electrodes based on flexible, conductive and biocompatible materials are something that the health‐oriented consumer is currently craving, as EEG electrodes are already being used to improve the level of concentration 350 and cognitive abilities 350, 351 of the user. In the authors' opinion, the commercialization of the abovementioned EEG systems could therefore spark tremendous interest and open up new technology avenues with hitherto unseen consumer empowerment. 5. 2 Implantable 5. 2. 1 Electrocorticography (ECoG) Even though wearable EEG electrodes have provided neuroscientists with a wealth of information over the years, electrodes that are in direct contact with brain neurons enable much higher fidelity recordings in terms of spatial resolution (≈1 cm). 352 Among the different implantable strategies for monitoring brain activity, ECoG or intracranial electroencephalography (iEEG) has found the strongest foothold, as it is the least invasive option but yet the most accurate. The ECoG concept is based on an electrode that is implanted beneath the skull and in most cases on the surface of the brain. It is used to monitor brain functions in patients with epilepsy 353 and as a tool to assist surgeons in complicated brain surgeries. 354 However, the ECoG procedure can be invasive, as the electronic materials used at the moment are inflexible and thus do not conform well to the curvy brain architecture. ECoG devices based on materials that enable them to integrate with the curvilinear surface of the brain could offer a less invasive option for diagnosing, treating, and performing surgery on diseased brains. To address this issue, flexible and soft ECoG implants that can conform to the architectural intricacies of the brain have been developed. 8, 9, 355, 356, 357 One of the most noteworthy attempts to solve this problem was based on a flexible PI‐based electrode with significant spatial (≈500 µm spacing) and temporal resolution (>10 kS s −1 ). 9 This device was truly a piece of engineering art, as it was capable of incorporating thousands of interconnected silicon‐based sensors to yield exceptional fidelity without losing its flexibility. It was implemented in a feline model and used to record feline brain activity during sleep and electrographic seizures with hitherto superior recordings. In a similar vein, Malliaras and co‐workers developed a flexible electrode array, which was implemented with great success in a rat model. In simple terms, the ECoG device developed by Malliaras and co‐workers, 71 consisted of a PEDOT:PSS microarrays containing electrodes covering an area of 10 × 10 µm 2 and with a center‐to‐center distance of 60 µm. This high‐density configuration enabled unprecedented recordings of the electrical activity in rat brains along with good biocompatibility and flexibility. The developed ECoG electrode was also able to record signals that bear resemblance to epileptic spikes in a high‐fidelity manner. Building on these results, an even more dense and flexible electrode that could record action potentials in the brain was developed. 8 The electrode was coined the “NeuroGrid” and consisted of an ultradense electrode configuration with the area of the individual electrodes being 10 µm × 10 µm and a 30 µm spacing between them. The NeuroGrid was able to perform very high‐resolution recordings of action potentials from the surface of the brain cortex without even penetrating the delicate brain tissue, as was demonstrated both in rat and human subjects. Although, the neurogrid enabled stable brain activity recordings for up to 1 week within a host animal model, its long‐term electrical stability have not been tested yet, and accordingly the feasibility of this approach as a more permanent implantable solution for tracking and treating neurological disorders remains unclear. However, these types of implantable brain recording devices still represent a unique window into the brain and further advancements in this direction could result in breakthrough discoveries in neuroscience and ultimately enable us to better understand and cope with the most vital organ in our body; namely the brain. 5. 2. 2 Electrocardiography (ECG) Cardiac‐related diseases remain one of the major health issues in the world and are currently the number one cause of mortality in the developed countries. 358, 359 The ability to record electrical signals generated by the heart is vital for both understanding 360 and treating 361, 362 such diseases and also for detecting life‐threatening changes in the rhythm of the heart in the home rather than in the clinic. 363, 364 To this end, ECG has emerged as a simple and low‐cost procedure for recording the electrical activity of the heart, as the only requirements are a set of electrodes, which are either placed on the skin or in close proximity of the heart. 365, 366, 367 Albeit, electrodes placed on the skin are noninvasive and capable of real‐time monitoring of heart activity, the reliability and validity of the data obtained from such approaches are still under intensive investigation as these electrodes are not directly in contact with the heart tissue and thus unable to precisely record, measure, and stimulate the beating heart and control its electrical rhythm as compared to implantable ECG devices that are put in direct contact with heart tissue. 11, 368 Since most cardiac disorders cause irregular heart beating, the ECG method is a reliable tool for heart diagnosis, as it is pretty straightforward to distinguish between harmonic and nonharmonic electrical recordings from it. ECG is also a feasible method for electrically stimulating the diseased heart in order to mend it again. 369 These examples include restoring normal heart functions in patients with an abnormal heart rhythm or inefficient heart pumping arising from birth defects, age, or due to myocardical infarction. Conventional ECG electrodes are typically assembled on rigid substrates with sharp edges and therefore cannot integrate properly with the curvilinear and soft heart tissue to enable the needed electrical coupling between tissue and device for high‐fidelity recordings. 370 To address these challenges, 1D nanomaterials such as SiNWs, 18, 371 AgNWs, 226 and CNT 372 have been incorporated into biocompatible and soft elastomers to develop flexible and less invasive ECG devices. For instance, in a recent study, a silicon‐based nanocircuit was assembled onto a flexible PI substrate (50 µm thickness) to provide robust electrical recordings. 371 This device was able to conform to a beating embryonic chicken heart while enabling optical imaging and electrical recordings with excellent signal‐to‐noise ratios and high spatial resolution (in the µm range). In a similar vein, a flexible ECG device was recently engineered that consisted of an Au‐based electrode network (total of 288 measurement points) deposited onto an ultrathin (25 µm) flexible PI substrate. 54 The developed ECG device was implanted in a porcine animal model; thanks to its many electrical contact points (288 measurement outlets), it could record electrical signals with unusually high signal‐to‐noise ratio (≈34 decibels [dB]) and temporal resolutions (<2 ms). It was also readily integrated with the curvilinear and dynamic in vivo heart and displayed a formidable fatigue resistance that enabled the electrodes to maintain their electrical performance even after 10 000 bending cycles. Although the abovementioned flexible ECG devices are well suited for recording electricity from the curved heart surface, they are incapable of penetrating the many ripples and grooves of its surface, instead they are only able to establish conformal contact with localized regions underneath the heart surface. This current lack of methodology limits the recording and stimulation capabilities of ECG devices, and therefore even more cardiac‐friendly systems are called upon. One of the most obvious attempts to address this challenge is to replace the nonadhesive and biologically stable PI substrate used in most flexible ECG devices with a more conformable, adhesive, and bioresorbable material. 10 Recent work demonstrates that silk can be used as a highly adhesive carrier material for electronics and transistors, as it is soft, conformable, and displays a degradation profile, which can be fine‐tuned by either controlling the degree of crystalline β‐sheets within the silk film or through enzymatic linkages. 373, 374, 375 Hence, silk displays a unique set of properties of high interest for implantable cardiac devices, as it offers exceptional conformability and is capable of transferring and laminating electronic components onto the most rough tissues in the human body. 10, 357, 376 On this ground, an electrode network—capable of sensing bioelectricity—has been transferred onto a sacrificial, adhesive and highly conformable silk substrate to engineer an implantable ECG device that fully addresses the abovementioned challenges. 10 The silk‐based electrode could laminate ECG electrodes onto the most irregular and wrinkled regions of the heart surface, which thus far had been unattainable through conventional strategies. Notably, the laminated electrode array covered the curved surfaces of the heart in a 3D fashion while enabling a robust adhesion onto the surface with stable electrical recordings up to a ≈22% device strain. An alternative system to the silk‐based material strategies was recently developed to further improve recording/stimulation of the cardiac tissue by using ultrathin elastic membranes that matched the complex heart tissue architecture while simultaneously providing excellent electrical contact with important physiological areas on the heart tissue. 11 This—next‐generation—ECG device is capable of adhering and enveloping the heart without the need of sutures or tissue glue. Moreover, it is capable of high‐resolution readings of cardiac activity and is able to carry out localized electrical stimulation because of its conformal contact to all points on the heart. Furthermore, a uniform and high‐density distribution of sensing electrodes on the device enabled high‐fidelity electrical recordings from an ex vivo rabbit heart with excellent spatial control. Devices like this could spearhead new innovations in implantable cardiac monitors and stimulators, since they provide formidable flexibility and are capable of maintaining a strong contact with the heart tissue during contraction and relaxation. In summary, the abovementioned flexible and bioresorbable materials have enabled ECG systems for real‐time monitoring of cardiac activity with the grand goal of treating cardiac‐related disorders, as these materials have improved the devices so they can record electrical signals from previously unattainable electrical points within the ripples and grooves of the beating heart while, at the same time, enabling localized cardiac tissue stimulation to mend potentially life‐threatening chronic heart conditions. However, the stability of such electronic devices within biological fluids remains unclear, since biocorrosion at the electrode/tissue interface and the associated lowering of the device functionalities present a significant bottleneck within the field. 5. 2. 3 Electronic Stents (ES) An impaired blood circulatory system is among the major causes of mortality in the world, as it can disconnect vital tissues and organs in the body from receiving nutrients and oxygen. 377 In most cases, the formation of atherosclerotic plaques in blood vessels precedes these events by blocking the flow of nutrients and oxygen to the inflected body part. 377, 378 To this end, endovascular stent grafting has emerged as the gold standard treatment for patients suffering from diseases that arise from blocked or damaged blood vessels. 379, 380 This procedure works by implanting a small wire‐mesh tube into the inflicted area to enlarge the blocked arteries, support the weakened artery walls, and thus establish normal blood flow to the region that has been disconnected from the circulatory system. These traditional stent implants have, over the years, been made from stainless steel or other flexible and non/corrosive metals. 381 Although the conventional stents can be used to open blocked arteries, significant challenges remain to improve their biocompatibility and flexibility to enable them to fully meld in with damaged blood vessels. Moreover, an ideal stent device needs to be bioresorbable to avoid the many long‐term complications associated with nondegradable implants. To address these challenges, stents have been developed that are based on degradable polymers or various types of conductive magnesium alloys that altogether are capable of maintaining the structural integrity of blood vessels in a hitherto unprecedented manner and disperse in the body after normality has returned to the damaged tissue. 382, 383, 384 Some of these next‐generation stents have also used electronic/magnetic materials and can therefore be programmed to release drugs locally in a time‐dependent manner to prevent inflammation and to get wireless feedback from the progress of clearing blocked vessels. 385, 386 Most of these polymers typically degrade within the body—either enzymatically or through hydrolysis—while the degradation pathway of magnesium is based on corrosion, as magnesium quickly oxidize into corrosive magnesium hydroxide (Mg(OH) 2 ) in atmospheric air, which degrades into Mg 2+ and OH − in the body. 387 In a recent study, these concepts have been united into a sophisticated bioresorbable stent that is capable of giving real‐time feedback on postsurgery inflammation and blood flow through the targeted vessel. 385 This electronic stent was composed of a magnesium alloy coated with a degradable polylactic acid (PLA) film in which the conductive magnesium enabled wireless transmission of recorded data, and the PLA was used for controlled drug‐delivery. In this system, the recorded data was first stored in a memory module placed on the stent surface and then wirelessly transmitted to an external storage device with the aid of the antenna‐like magnesium alloy. Furthermore, the mechanical strength and structural integrity of the stent after implantation was maintained for a week. Multifunctional stents like this could spearhead new innovations in endovascular implants, since they provide formidable flexibility, degrade in the body after their job is completed and are able to reduce inflammatory reactions around the stent by providing anti‐inflammatory activity through localized drug delivery. In addition to the use of stents in treating circulatory related diseases they are also capable of recording and stimulating brain activity from within the narrow capillaries of the brain. 388, 389, 390 The keys for successful and noninvasive integration of such stents within the delicate capillaries of the brain are high flexibility and good biocompatibility. To this end, flexible and ultrathin electronic stents represent viable treatment options that can maintain the blood flow into the brain tissue without causing significant damage to the surrounding tissue while at the same time enabling neural recordings. 389, 391 For instance, Oxley and co‐workers 391 recently developed an electronic stent that could integrate with the ultranarrow capillaries of the brain for both brain activity recordings and neural stimulation ( Figure 11 ). In simple terms, this electronic stent was made from a self‐expanding and commercially available stent coated with biocompatible Pt electrodes for recording purposes and was coined the “stentrode”. The stentrode could readily be integrated within the curved blood vessels of the sheep brain for high‐fidelity recording of somatosensory evoked potentials (SSEPs) in close proximity of the superior sagittal sinus. 391 This type of electronic stents has also provided a new avenue for deep brain stimulation for up to 190 days with minimal trauma inflicted to the brain. 388, 391 Such implantable systems are indeed safe and can be delivered via conventional catheter angiography and therefore present a minimally invasive option compared with traditional open‐brain surgery. Figure 11 Engineering of an intracranial stent‐electrode (stentrode) array for recording brain activity. a) Preimplant Images showing the presence of intracranial lumens (blue arrow) and cortical veins (red arrow). b) Images showing the self‐expanding property of the stentrode. c) The integration of the stentrode within blood vessels of a sheep brain. The yellow arrows correspond to the electrodes, while the green arrows correspond to the delivery cathers. d, e) The stentrode was able to record high‐fidelity signals following the implantation. The recorded peak‐to‐peak amplitude was fairly stable for up to 28 days. f) Recordings of somatosensory evoked potential (SSEP). g) The position of the implanted stentrode in four different sheep models. h) A 3D representation of an implanted stentrode and its corresponding peak‐to‐peak amplitude recording. Adapted with permission. 391 Copyright 2016, Macmillan Publishers Ltd. In summary, the progress made in the field of electronic stents is truly remarkable and will undoubtedly address one of the biggest killers in our times: cardiac‐related diseases. Notably, endovascular stent‐electrode arrays for deep brain stimulation will undoubtedly facilitate breakthroughs in the field of brain–machine interfaces, as they in the authors opinion, will enable seizure prediction in patients with epilepsy and become viable treatment options for those suffering from Parkinson's disease. 6 Cybernetic Prosthetics Driven by the unification of prosthetics and flexible bioelectronics, the applications of cybernetic prosthetics are numerous and rapidly expanding. Examples include: spinal cord implants that enable the paralysed to move again; brain–machine interfaces that enable the immobilized to mobilize their thoughts to control man‐made limbs; retinal implants that bridge the gap between the brain and the blinded eye to restore vision in the patient; ear implants that make the hearing disabled hear again; and last but not least scientists are following a daring goal of uniting inanimate electronics with living tissues to create cyborganic transplants with unheard of properties ( Figure 12 ). It sounds too good to be true, but yet it is a reality that is currently bringing relief to the lives of disabled people. So far, the primary objective in this regard has been to engineer replacement parts for those disabled since birth through traumatic injuries or from neurological disorders. However, the cybernetic prosthetics could also be used for something entirely different, namely, enhancing human capabilities beyond normality—a topic that raises a host of unanswered ethical questions. Indeed, these are questions that need to be answered before the field of cybernetics can fulfil its true destiny, namely providing the means for mankind to enter the next step in the ladder of human evolution. Among the many key elements that have been used to create these fascinating devices are flexible materials and nanoelectronics. Especially, the union of the two into formidable machines that can integrate seamlessly with the human body has been the driving force behind the aforementioned quantum leaps in cybernetics. The following sections highlight the successful use of these tools by bioengineers to make a cybernetic future implementable in the healthcare system, society, and the private sphere. Figure 12 A depiction of some of the most noteworthy cybernetic prosthetics that we anticipate will spearhead the coming cybernetic revolution. Made by Harder&Muller. 6. 1 Flexible Spinal Cord Implants Traumatic injuries to the spinal cord significantly deteriorate the quality of life of those affected, and the affected ones typically suffer from chronic paralysis. 392 In brief, these injuries result in impaired signal conduction between brain and the body and can only be restored through approaches that enable the injured tissue to reconnect with the proper distal target. 393 One of the key strategies is based on sophisticated nerve implants that consist of microelectrodes that can convey signals from the brain cortex to neurons in a disconnected body part. 394, 395 However, the mechanical mismatch between today's conventional rigid electrodes and the compliant neural tissues significantly limits their long‐term performance and needs to be addressed to fully exploit their clinical potential. In fact, most electronic implants have a young modulus in the GPa range; a stark contrast to the very soft neural tissues of the human body (<1 kPa). 396 The research and development of soft and conformable materials for neural implants is still in its infancy and therefore presents an uncharted territory ripe for fruitful investigations. In a recent cutting‐edge study by a group of scientists from École Polytechnique Fédérale de Lausanne (EPFL), this daunting challenge was addressed by embedding flexible Pt electrodes and interconnects within a soft and biocompatible silicone‐based substrate, which they coined electronic dura matter or simply “e‐Dura” ( Figure 13 ). 15 The e‐dura implant is composed of a highly elastic silicone substrate measuring 120 µm in thickness, stretchable Au interconnects (35 nm in thickness) and soft electrodes coated with a Pt–silicone composite (300 mm in diameter), which altogether enable the device to mediate electrical stimulation and transfer electrophysiological signals from the spinal cord tissue to a microcomputed tomography (CT) scanner for assessment of whole‐body movements during daily activities. As a new feature, a microfluidic‐based delivery system (100 mm × 50 mm in cross‐section) was incorporated into e‐Dura for controlled delivery of various chemical substances to the injury site. Overall, this system was highly stretchable and exhibited an elastic modulus of ≈1. 2 MPa, which is quite similar to the mechanical properties of the native dura mater. 15 These unique mechanical properties of the e‐Dura allowed it to conform well to the curvilinear surface of brain and spinal cord tissues without damaging them. The e‐Dura implant could sustain numerous stretch cycles without breaking or losing functionality and was shown to restore the motor functions of rats suffering from spinal‐cord‐induced paralysation through a combination of chemical and electrical stimuli. The long‐term functionality and biointegration of the implant is evident, since the new neuroprostheses neither altered the cross‐section of the spinal cord nor triggered unwanted foreign body responses. On the other hand, conventional rigid electrodes lead to significant damage to the spinal cord and failed to completely restore lost motor functions in the operated rats. The e‐Dura implant is the first of its kind and enables the sought‐after long‐term application of spinal cord electrodes. It has huge potential, and it will be interesting to see how it will function in human trial studies. A successful outcome would undoubtedly encourage the millions of people currently suffering from spinal cord injuries and bring a long‐waited relief to their lives. The e‐Dura implant also holds great promise as a potential neuron‐to‐machine interface for real‐time disease monitoring and pain management for patients suffering from neurological disorders. Figure 13 Neural implants with native‐like elasticity. a) Schematics of the electronic dura (e‐Dura) mater implant and how it works in vivo. b) The elastic properties of e‐Dura and various native tissues. c) Postimplantation walking efficiency of rats, the circularity index of the operated spinal cord after 6 weeks, and the density of two cellular markers for foreign body response (astrocytes & microglia) in the spinal‐cord. d) The local longitudinal strain of the e‐Dura increased by much more as a function of applied strain as compared to a conventional rigid implant. This is indicative of a more compliant implant that can cope with the movement of the spinal‐cord region during daily activities. Adapted with permission. 15 Copyright 2015, AAAS. 6. 2 Noninvasive Brain–Machine Interfaces While spinal cord electrodes aim to restore mobility in paralysed patients by reconnecting the lost motor area with the brain cortex, brain–machine interfaces achieve this by rerouting motor‐related signals to manipulate prosthetic limps. 397 In general, brain–machine interfaces are sophisticated pieces of equipment that are capable of receiving and delivering feedback related to control and movement of various body parts. Their primary function is, therefore, to establish a communication link between neural activity in the brain and distal body parts. This concept can also be used to control inanimate objects with the mind through a communication link between the brain–machine interface and electronics inserted within the host. 398 Examples include opening doors, controlling the lighting in a room and moving an item from one place to the other. Brain–machine interfaces have also demonstrated the potential to treat a wide variety of neurological and psychological disorders, such as Parkinson's disease, 399 Alzheimer's disease, 400 psychiatric disorders, 401 and multiple sclerosis. 402 This is achieved through deep brain stimulation with electrodes that regulate and redirect abnormal neural impulses in the brain. 9, 403, 404 In an interesting example from 2015, 404 a self‐powered brain–machine interface for deep brain stimulation was introduced ( Figure 14 ). The interface was completely self‐sufficient, and it used a flexible piezoelectric component to harvest energy from mechanical motions in the body. It also showed promising results in terms of harvesting sufficient energy to efficiently stimulate the motor cortex inside the brain of mice. Self‐powered deep brain stimulators can address many issues related to conventional battery‐driven implants, because they are 100% self‐sufficient and can therefore circumvent the many invasive interventions that battery‐driven brain–machine interfaces require in terms of repeated battery changes. Figure 14 Self‐powered brain–machine interfaces. a) Schematics of the piezoelectric energy harvester and photographs showing the device in its original, bending, and release state. b) The electrical signal measured from the device during bending and unbending in the forward connection and c) reverse connection. d) Depiction of the animal experiment, and data related to the stimulation of paw movement of mice through bending and unbending of the flexible energy harvester. Adapted with permission. 404 Copyright 2015, the Royal Society of Chemistry. However, unfortunately, many of the above‐referenced brain–machine interfaces do not meld well with the human brain due to a mechanical and biological mismatch. 405 This is possibly a result of the inability of rigid electrodes to conform to the swelling and contraction of the human brain during day‐to‐day activities. 9 Over the years, several studies have addressed this challenge through flexible and soft electrodes that can integrate easily with compliant brain tissue. 406, 407 Although flexible brain–machine interfaces hold great promise as a less invasive treatment than their rigid counterparts, they still rely on invasive and costly surgical procedures. To address this challenge, a method is required that is completely noninvasive to enable a seamless merger between flexible electronics and the brain. To address this grand challenge, a methodology was recently introduced by Lieber and co‐workers, 408, 409 which was based on the delivery of flexible silicon‐based circuits into the brain for monitoring and recording neuronal activity. The process involved a thin syringe needle loaded with freestanding and flexible electronic components that were injected into the target tissue ( Figure 15 ). 409 Once inside the body, the flexible electronics could reshape into the desired layout and yield high‐resolution recordings of brain activity. The unique structural and mechanical property of the injected mesh electronics also resulted in an unusually facile integration with the brain tissue, which displayed an unmatched portfolio of formidable properties once inside the brain, as it was more conformable and smaller than any other electrodes that have been implanted into the brain. This cutting‐edge technology will unequivocally lead to a momentous advance—not only as a platform for more sophisticated brain–machine interfaces—but also as a new technology with the capacity to reshape the entire field of biomedical engineering. Figure 15 Syringe‐injectable electronics for neural recording. a) Depictions of syringe‐injectable electronics. b) Schematics showing the injection of the electronics into the brain of mice. c) Photographs depicting the injection process into a three‐month‐old mouse brain. d, e) Schematics showing the areas of the mice brain, wherein the mesh electronics was injected. f) Bright‐field microscopy imaging of the brain region into which the mesh electronics was injected five weeks after injection. g) Bright‐field and epi‐fluorescence images corresponding to the region indicated by a white box in (f). h) Fluorescence image corresponding to the region indicated by a blue box in (f). i, j) Electrical recordings from the mouse brain using the injected mesh electronics. Adapted with permission. 409 Copyright 2015, Macmillan Publishers Ltd. 6. 3 Retinal Implants Age‐related and disease‐related degeneration of the human eye affects millions of people worldwide and is expected to grow rapidly in the coming years due to the ageing global population. 410, 411 Retinal prosthesis provides an opportunity to restore the vision in those who have lost sight through such degenerative diseases. 412, 413 In brief, a retinal prosthesis replaces the function of the damaged eye by exciting nerve impulses from healthy neurons in the eye; this is accomplished with the aid of microelectrode arrays on the prosthesis. The excited nerve impulses are subsequently transmitted to a chip in the virtual cortex of the brain to create artificial visions of the surrounding world. 414 In order to prevent physical damage to the retina at the implant–tissue interface, it is desirable to have implants made from materials with properties such as bendability, foldability, and biocompatibility. 415, 416 Currently, most retinal electrodes are made from silicon‐based microelectronics 416, 417 with relatively good biocompatibility. However, the silicon‐based electrodes can be mechanically rigid and bulky, 416 and they are difficult to mold into minimally invasive and ultrathin configurations. Therefore, these substrate types could in the long run lead to severe tissue damage due to a mechanical mismatch between the rigid implant and the surrounding retinal tissue. PI‐based electronics have emerged as promising flexible materials for retinal prosthesis, as they both can accommodate for the movement of the eye during day‐to‐day routines and are compatible with native tissues. 418, 419 PI's can accomplish this daring task, since they are biocompatible and moldable into the ultrathin and delicate electronic devices required of retinal prosthetics. They also recover to their original shape even after rolling or folding in stark contrast to silicone‐based electronics. 420, 421 Furthermore, other type of polymers such as parylene C and PET have been proposed for retinal prosthetic, however, they have higher CTE and lower T g compared to PI, as described in Section 2. 2 (PI) and Section 2. 3 (PET and parylene C). 422, 423 Therefore, they have much more difficulties with withstanding the high processing temperatures during the manufacture of metal‐based electrode arrays. Importantly, a high CTE can also facilitate strain accumulation within the polymeric film and thereby increase the risk of crack formation between the electrode layer and substrate at elevated temperatures. One noteworthy example of PI‐based retinal devices is the Argus II, which consists of 60 Pt microelectrodes embedded on a flexible PI‐substrate. 16, 424, 425 The Argus II was the first prosthetic retinal device to receive regulatory approval in both USA and Europe 16 and is currently an available and commercial option for treating people with impaired eye vision. 16, 426 At the time of this writing, more than 100 retinal devices have been implanted worldwide; 16 nonetheless, some challenges are still being investigated or remain unsolved at present, and thus further research is required in this direction. For instance, with the Argus II system, it is difficult for patients to perform complex visual tasks, such as face recognition, orientation in unknown environments or text reading. 425 Therefore, we anticipate that future research will focus on the design and fabrication of material platforms that can incorporate even more electrode units and thus increase the number of excited retinal neurons to obtain the needed complex signaling for complex visualization. 6. 4 Cyborg Organic Constructs (Cyborganics) Whereas the aforementioned applications of flexible electronics were primarily directed toward establishing conformal electrical contacts with tissues in vivo to restore and/or enhance the functions of the human body, even more impressive cybernetic systems are currently under development to create cyborg organic constructs (cyborganics). In its essence, “cyborganics” represents an emerging concept at the crossroads of tissue engineering, materials science, electronics, and chemistry with the daring aim to create a new class of hybrid organs and hydrogel carriers made from a combination of inanimate and biological matter. 17, 18, 427 Cyborganics can be divided into two different classes: 427 i) permanent hybrid organs consisting of stable matter and living tissue and ii) hydrogel carriers embedded with biological matter and degradable nanomaterials with the grand goal of regenerating dysfunctional tissues. The concept was recently exploited by Liebers and co‐workers at MIT in a landmark study, wherein stem cells were matured into functional tissues within a web of silicon nanowires. 17 These remarkable hybrid constructs were capable of real‐time monitoring of markers related to tissue growth and functions. To probe the physiological environment of the living tissue, silicon nanowire FETs, which are capable of recording physiological signals, were embedded within the hybrid constructs. The sensor network was a truly outstanding detector of important physiological events once the encapsulated stem cells had turned into cardiac tissue and premature blood vessels, as it enabled electrical recordings from the beating cardiac tissue with incredible resolution (≈ms) and PH‐recordings from within the vessels. The concept of “cyborganics” was also recently applied to create cochlea‐shaped electrodes consisting of cartilage tissue interweaved with inductive coil antennas ( Figure 16 ). 428 This hybrid implant had similar properties to a native ear and was capable of auditory sensing in the radio frequency range. In brief, an ear‐like construct was generated via additive manufacturing of cartilage cells and Ag nanoparticles embedded within an alginate hydrogel matrix. The three components were merged into the syringe‐extrusion of a 3D printer, which deposited the mixture into a bioelectronic hybrid ear. Over time, the alginate matrix was degraded and the cartilage cells retained the ear‐shaped morphology that was originally 3D printed. Indeed, this concept of additive manufacturing of living cells with electronic materials is leading a highly innovative direction in the field of cybernetics. Still, the generation of such hybrid constructs is still a big challenge in the field. In the authors' opinion, this concept deserves much future attention, as it could enable the generation of a wide range of replacement parts with performances beyond what human organs typically are capable of delivering. Figure 16 A cyborg ear for enhanced auditory sensing. a, b) Schematics showing the 3D printing of the artificial ear. c) Photographs of the printed ear before and after culturing. d) Chrondogenic cell viability and secretion of important chrondogenic markers. e) Audio signals were transmitted through the right ear (R) and received through the left ear (L) with good fidelity. Adapted with permission 428 Copyright 2013, American Chemical Society. 7 Outlook 7. 1 Flexible Materials and Electronics Flexible and conductive polymers will unequivocally reshape the field of bioelectronics and herald a new age in modern electronics. However, there are many challenges that need to be addressed, since the thermal stability, solvent resistance, CTE coefficients, and conductivity of polymer materials still differ a lot from their inorganic counterparts. 26 Perhaps the most promising approach to remedy the current situation is through the incorporation of multifunctional nanomaterials into polymers. Polymer materials that are stronger than steel, optically transparent, flexible, and highly conductive have been developed through the unification of nanomaterials and polymers. 225, 427, 429, 430, 431 In one paramount study, carbon nanotubes functionalized with Ag nanoparticles were incorporated into a matrix of polyvinylidenefluoride and solvent cast into an ultrathin sheet (<140 µm). The hybrid film displayed unusual electromechanical properties with a conductivity at 5710 and 20 S cm −1 at 0% and 140% strains, respectively. 225 Building on these results, a microscale paper‐cutting approach, based on a top‐down plasma etching technique, was developed to engineer even higher plasticity into carbon‐based nanocomposite polymers. 432 Through this approach, the ultimate strain of the nanocomposites was increased from 4% to 370% while preserving their high electrical conductance. This unique property stems from stress delocalization mediated by the notches introduced into the film through the paper‐cutting process. In other studies mineral‐based platelets such as laponite, sumecton, and montmorillonite have been successfully incorporated into various polymer matrixes to yield films with incredible barrier properties and CTE coefficients similar to that of glass. 430 Moreover, properties such as high‐field‐effect carrier mobility and long‐term stability under mechanical, electrical, and environmental stress have made inorganic semiconductors and nanomaterials suitable for various flexible electronic applications. 273, 433 In recent years, they have also been integrated into a great variety of organic polymer matrices to yield multifunctional flexible devices with improved electrophysiological sensing capabilities. 284, 339, 434 Notably, these properties and functionalities of inorganic components are greatly affected by their crystal structure, phase, size, shape, and their chemical attributes. 217 We envision that the integration of inorganic components into biointegrated systems represent an interesting area ripe for investigations, and could therefore enable significant breakthroughs in the field. Despite these encouraging results, multifunctional polymers with all of the abovementioned properties are yet to hit the consumer market for flexible electronics. However, the era of nanoreinforced polymers is still in its infancy; we therefore anticipate that further research in this direction will yield polymer composites with even more extraordinary properties than those previously reported. 7. 2 Healthcare Monitors for Empowering the Patient The unsystematic healthcare monitoring of the populace is one of the primary driving forces behind many chronic diseases. Examples include the stealthy plaque buildup in blood vessels caused by unhealthy eating habits, the well‐documented correlation between diabetics 2 and high blood glucose levels and the intimate link between a number of debilitating neurodegenerative diseases and unnoticed cases of neuroinflammation. A number of smart technical health‐aids have already been developed with the ability to gather such vital health data from the patients from home without the need of costly and time‐consuming in‐clinic visits. Indeed, Rogers and co‐workers are currently working hard to spearhead some of these technologies into the consumer market through a newly funded company, namely MC10. MC10 specializes in wearable physiological monitors that are made from materials that enable them to effortlessly adhere and integrate onto the skin; they recently launched the “BioStamp, ” which is a conformable e‐skin device that can link physiological data recorded from the body to portable devices for efficient readout and analysis of those data. In a similar vein, Google is currently working on a new wrist‐worn device that will enable ECG, skin temperature, and pulse monitoring. In addition to recordings associated with circulatory health and skin wellness it would be interesting to expand the operational capacity of the commercially available healthcare monitors to also include EEG measurements from the brain, similar to the study from Rogers and co‐workers that we reviewed in Section 5. 1. 4, 347 wherein a conformable electrode was plugged into the ear for high‐fidelity EEG recordings. We note that such devices, if completed, would constitute complete real‐time healthcare monitors of individualized health data, since they enable heart activity mapping, brain activity recording, and real‐time monitoring of skin‐related diseases. Stretchable and elastomeric bioelectronics has also been gradually realized in other more specialized sensorial and therapeutic applications. For instance in recent study by Mannoor et al. a flexible material interfaces that can be used to monitor bacterial contaminants on tooth enamel was developed. 435 The core of this flexible biosensor was an intricate Au‐graphene conduit that was deposited onto a silk‐based substrate through shadow mask‐assisted electron beam evaporation, and subsequently functionalized with antimicrobial peptides. While the printed graphene‐based conduits form the primary “circuit” of the nanosensor, the peptides are aimed to detect appropriate pathogens. In the event of a bacterial attachment to these peptides, the resistance of the graphene coil changes—thanks to the relatively charged cellular membrane of the bacteria—and elicits an inductive effect, which can be detected via a wireless device that utilizes an impedance analyzer. This platform opens up a new paradigm for monitoring pathogens and could find applications in food industry, hospitals, implantable personal healthcare devices, and many more. In another recent study, a multifunctional and transparent endoscope capable of temperature/pH sensing, drug delivery, and performing real‐time optical imaging was developed. 436 This transparent electronic system consisted of 1) a graphene and Au‐based sensor that can detect pH and temperature fluctuations in tissues, 2) light sensitive dyes that can be used for imaging purposes, and 3) drug‐loaded nanoparticles with therapeutic potential. The system was implemented in a mouse model and displayed remarkable potential for diagnosing and treating colon cancer in patients. In summary, there is currently a push in the field of bioelectronics driven by the emergence of a new class of flexible electronic materials, and in the authors' opinion, it seems like its worldwide applicability is closer to becoming a reality than ever before. As this trend continues, we envision a truly remarkable healthcare revolution in the form of personalized health monitoring devices with impeccable accuracy. 7. 3 Cybernetic Prosthetics Extensive work on animals has paved the way for implementing a wide range of the aforementioned cybernetic extensions in human subjects. However, a large amount of work is still needed in this direction to develop even better material interfaces to reach the grand goal of realizing human cyborgs with supernatural abilities and powers. Especially, the engineering of electrode materials that can bridge between the brain and human consciousness as well as more sophisticated cyborganics are called upon to bring cyborgs out from the realm of science fiction and into real‐world applications. Among the most noteworthy applications of brain–machine interfaces is to treat and possibly cure neurological and physiological disorders. Malfunctions in the neural circuitry of the brain and neurotransmitter imbalance are key factors behind these kinds of brain disorders. One avenue for restoring order in the disordered brain is, as briefly mentioned in Section 5. 2, to use brain–machine interfaces for electrically stimulating the brain into a balanced state again. Building on this concept, the field of optogenetics has emerged to enable even better control and stimulation of the brain through light‐activated release of neuroactive reagents from neurons that have been genetically altered to become sensitive to light. 437, 438, 439 Despite being a relatively young field of research, optogenetics have already facilitated many groundbreaking discoveries. Most noteworthy is its application in systems that can be used to control the motion behavior of transgenic insects. 440, 441 In brief, these systems consist of optically transparent insects with modified brain neurons that can release important behavioral molecules from the brain in response to light‐activation. However, this form of brain stimulation is less straightforward in larger animals, and therefore other approaches are called upon to open the light‐activated gateway into the brain of more complicated organisms. To address this challenge, optogenetic systems based on flexible, light sensitive, biocompatible, soft, and mechanically compliant electrode materials have been developed. 442, 443, 444 In a recent example, flexible and wirelessly controllable LED devices were implanted beneath the skull of mice to stimulate dopaminergic neurons to release dopamine. 442 This study demonstrated that the movement of mice could be easily controlled and re‐directed with the aid of light and subsequent release of dopamine. Overall, these results clearly show that it is feasible to use such microelectrodes for controlling human behavior, curing various physiological disorders and facilitating cognition enhancement in a hitherto unprecedented manner. However, in the authors' opinion, we need to approach these technologies with caution and carefully evaluate the risk and opportunities associated with their use in human subjects, as they can be misused in an unethical manner. So far, the prime‐usage of cyborganics has been in various diagnostic and proof‐of‐concept applications. However, the union of electronic materials and living tissues can pave the path for impressive devices that can replace dysfunctional organs and enhance bodily functions beyond normality. We anticipate that 3D printing of living cells along with 2D nanomaterials such as graphene or CNTs can yield even stronger tissue constructs that are part animate and part inanimate. Moreover, the deposition of electronics and cells into architectures similar to those found in native tissues through various microprinting techniques is a feature that definitely will spark unprecedented progress in the field of cybernetics. 427, 445, 446 Along these thoughts, Shin and co‐workers recently managed to 3D print a mixture of cardiac cells and CNT‐based circuits into flexible electronics with cyborganic‐like features. 445 Specifically, it was shown that cardiac cells embedded within these hybrid constructs were viable and able to express important cardiac markers. This technology can in the future be expanded to yield cyborg cardiac‐like patches or maybe even spearhead the next‐generation of pacemakers. The prospect of 3D printed hybrid constructs consisting of neurons and flexible microelectrodes could also result in less invasive and more advanced brain–machine interfaces, wherein the fine line between machine and brain is abolished due to the presence of both native‐like brain matter and electronics. Given the fact that muscle tissue consists of highly aligned and organized muscle bundles, the combination of 3D printed carbon‐made fibers and muscle cells into anisotropic tissue‐like architectures can lead to alternative treatments for those suffering from scleroses and muscular dystrophy. These cyborg muscles are also anticipated to give the wearer formidable strength, similar to the supernatural abilities of the cyborgs, that have been presented to us through science fiction movies. Ultimately, the engineering of flexible electronics that can blend right in with the human body has driven the field of cybernetics to reach the capacity to deliver truly outstanding human–machine interfaces, that in the authors opinion will yield more advanced and versatile beings in the foreseeable future. 8 Conclusion The widespread usage of sophisticated prosthetics such as cochlea, retina, brain–machine, and neural implants along with the projected importance of technical health aids in our daily‐life‐activities clearly indicates that we are slowly entering a cybernetic era. These advances are a direct consequence of flexible and electronic materials that can meld seamlessly with the soft and curved tissues of the body. The development of brain–machine interfaces that enable thought‐controlled movement of objects has especially stirred a lot of frenzy in the field and we anticipate that these cybernetic extensions could potentially expand human cognition to a new level. Another emerging topic (i. e. , cyborganics) can change the way synthetic material parts and tissue transplants are viewed and ultimately herald a new era in tissue engineering, wherein more durable, better, and stronger tissues are custom‐made in the laboratory to hack the biology of humans beyond our current imagination. In the authors' opinion, further advances in this direction are intimately linked with 3D printing of hybrid neural circuits and nanoreinforced designer materials that can impose exceptional abilities on the user. The advantages that the 3D printing technology bring to the field of cybernetics are first localized printing of cells, materials, and nanoelectronics onto flexible substrates, and second the ability to custom‐make cyborg organs with native‐like tissue architectures. Other areas of research ripe for investigations are cybernetic devices that can harvest energy from the moving body without the need of any interventions or revision surgeries to replace old batteries. In summary, the ongoing progression of research in the field of cybernetics has been made possible from the beautiful trinity between biology, materials science, flexible electronics, and tissue engineering. However, we should be wary of the fact that the implementation of some of the suggested cybernetic inventions need to be carefully and critically evaluated to make sure they follow existing ethical guidelines. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201800006
| 2,018
|
Advanced Science
|
Histone Deacetylase 7‐Derived Peptides Play a Vital Role in Vascular Repair and Regeneration
|
Abstract Cardiovascular disease is a leading cause of morbidity and mortality globally. Accumulating evidence indicates that local resident stem/progenitor cells play an important role in vascular regeneration. Recently, it is demonstrated that a histone deacetylase 7‐derived 7‐amino acid peptide (7A, MHSPGAD) is critical in modulating the mobilization and orientated differentiation of these stem/progenitor cells. Here, its therapeutic efficacy in vascular repair and regeneration is evaluated. In vitro functional analyses reveal that the 7A peptide, in particular phosphorylated 7A (7Ap, MH[pSer]PGAD), could increase stem cell antigen‐1 positive (Sca1 + ) vascular progenitor cell (VPC) migration and differentiation toward an endothelial cell lineage. Furthermore, local delivery of 7A as well as 7Ap could enhance angiogenesis and ameliorate vascular injury in ischaemic tissues; these findings are confirmed in a femoral artery injury model and a hindlimb ischaemia model, respectively. Importantly, sustained delivery of 7A, especially 7Ap, from tissue‐engineered vascular grafts could attract Sca1 + ‐VPC cells into the grafts, contributing to endothelialization and intima/media formation in the vascular graft. These results suggest that this novel type of peptides has great translational potential in vascular regenerative medicine.
|
1 Introduction Cardiovascular disease (CVD) represents a major cause of morbidity and mortality globally. Intimal injury is a common pathological basis of the process of CVD, and the recovery of injured endothelium is essential for protection against CVD. 1, 2 Dysfunction of endothelial cells (ECs) is thought to play a pivotal role in injured endothelium, and EC health has been suggested to be maintained by a consistent supply of supporting growth factors and activation of local stem/progenitor cells that are capable of differentiation toward ECs. 3, 4 It is well documented that vascular progenitor cells (VPCs), which are mainly located in the adventitia, contribute to the repair of the injured endothelium via migration toward the endothelium and differentiation into an EC lineage. 5, 6 These VPCs are also an important source for the regeneration of tissue‐engineered vascular grafts. 7, 8 Vascular bypass surgery has been successfully employed for the therapy of vascular diseases. Unfortunately, in the case of small‐diameter vascular grafts (diameter less than 6 mm), a high rate of restenosis limits the clinical use of commercially available products. The main reason for this is the poor regeneration of vascular tissue including slow and poor endothelialization. 9, 10 In addition, there is a compelling need for novel strategies to revascularize ischaemic tissues, and critical limb ischaemia continues to be a severe health problem globally. In many cases, surgical or graft‐based procedures are not possible, necessitating the development of molecular or cell‐based therapies to promote angiogenesis. 11, 12 Histone deacetylases (HDACs) are a family of enzymes that remove acetyl groups from N‐acetylated lysine residues on histones and are involved in gene transcriptional regulation through modulating chromatin structures. HDAC7, a member of the class II HDACs, is specifically expressed in the vascular endothelium and contributes to the maintenance of vascular integrity during early embryogenesis. 13, 14, 15 Recently, we demonstrated that a 7‐amino acid (aa)‐peptide (7A, MHSPGAD) could be alternatively translated from a short open reading frame within the 5′‐terminal untranslated region of mouse Hdac7 mRNA in stem cell antigen‐1 positive (Sca1 + ) VPCs in vitro and in vivo. The serine residue within the 7A peptide could be phosphorylated by the activated kinase MEKK1 S393, which in turn could transfer the phosphate group to the Thr145 site of 14‐3‐3γ, forming a novel MEKK1‐7A‐14‐3‐3γ signaling pathway downstream of vascular endothelial growth factor (VEGF). This novel signaling pathway contributed to the activation of the Sca1 + ‐VPCs. 16 In the present study, we focused on the functional analysis of this 7A peptide and its translational potential in vascular repair and regeneration by using different disease and transplantation models. 2 Results 2. 1 The 7A Peptide Increased VPC Migration and Differentiation toward the EC Lineage Recently, we found that endogenous 7A peptide could only be detected in in vitro cultured Sca1 + ‐VPCs or Sca1 + cells in injured femoral arteries and not in Sca1 + cells in normal vessels. Therefore, it is very important to investigate the relationship between 7A expression and the activation of Sca1 + VPCs. Considering 7A can function as a phosphate carrier, a synthetic phosphorylated 7A (7Ap) (MH[pSer]PGAD) was included in the functional analyses. As controls, scramble 7S (MPHASGD) and 7Aa (MHAPGAD), in which the serine of the 7A is substituted by alanine to totally abolish phosphorylation, were also included in this study. As shown in Figure 1 A, 7A especially 7Ap accelerated Sca1 + ‐VPCs migration. As expected, 7Aa retarded Sca1 + ‐VPCs migration. Wound healing can be a combined effect of cell migration and proliferation. To assess whether 7A/7Ap‐increased migration was derived from the combined effect, a Br‐dU incorporation assay was performed. Under the serum‐free condition, the 7A peptide by itself could not support Sca1 + ‐VPC proliferation (Figure S1A, Supporting Information). Further experiments revealed that 7A had no effect on cell survival under oxidative stress (Figure S1B, Supporting Information). Figure 1 7A and 7Ap peptides increased migration and differentiation toward the EC lineage. A) 7Ap significantly increased VPC migration in a wound healing model. Wound was introduced into confluent VPCs by tip scratching, and incubated with DMEM medium containing 2% FBS and 1 ng mL −1 of 7S, 7Aa, 7A, or 7Ap peptide. Images were taken at 0, 12, 24, and 36 h postscratching (left, Scale bar: 100 µm). The migrated cells in scratched area were counted from three views per scratching, three scratchings per well, and three wells per peptide (right). Data presented are representative images or mean of three independent experiments. *: p < 0. 05 (7Ap vs 7S) ( n = 6, one‐way‐ANOVA followed by Tukey's post hoc analysis). B, C) 7A/7Ap increased VPC differentiation toward the EC lineage. The 3 d spontaneously differentiated VPCs were incubated with differentiation medium containing 1 ng mL −1 peptides and 10 ng mL −1 VEGF for 4 d, followed by quantitative RT‐PCR analysis with GAPDH as house‐keeping gene (B) or tube formation assay (C, Scale bar: 200 µm). 1% BSA was included as vehicle control. The data presented are representative images or mean of three independent experiments ( n = 6, two‐way ANOVA followed by Dunnett's multiple comparison tests). D) MEKK1‐7A‐14‐3‐3γ mediated 7aa‐peptide‐induced VPC differentiation toward EC lineage. The VPCs were transfected with siRNA and cultured in differentiation medium for 3 d, followed by further differentiation with same protocol described above for 4 d, followed by quantitative RT‐PCR analysis of Pecam1, Cdh5, and Tagln mRNA levels with Gapdh as house‐keeping gene. 1% BSA was included as vehicle control. The data presented are representative images or mean of three independent experiments. *: p < 0. 05. **: p < 0. 01 ( n = 6, two‐way ANOVA followed by Dunnett's multiple comparison tests). The vessel wall resident Sca1 + ‐VPCs can differentiate toward both EC and smooth muscle cell (SMC) lineages. 17, 34 When embryonic stem cells differentiate into SMCs, the Hdac7 transcript variant 2 undergoes further splicing, leading to the incorporation of 7A into the far N‐terminal end of the HDAC7 protein. 18 Thus, we assumed that 7A might be involved in the cell differentiation process. To test this, Sca1 + ‐VPCs were cultured in differentiation medium in the presence of 7‐aa peptides and/or VEGF, followed by quantitative RT‐PCR analysis of EC and SMC marker expression. As shown in Figure 1 C, 7A increased Pecam1 (CD31) and Cdh5 (CD144) mRNA levels, which were significantly enhanced by VEGF, although VEGF alone had only a slight effect. The effect of 7Ap alone on Pecam1 and Cdh5 expression was comparable to the combined effect of 7A and VEGF (Figure 1 B left and middle). 7A alone had no effect on Tagln (SM22) expression but significantly decreased Tagln expression in the presence of VEGF, although VEGF had a slight increasing effect. 7Ap alone significantly decreased Tagln expression. These results suggest that 7A, especially 7Ap, favors Sca1 + ‐VPC differentiation toward EC but not SMC lineages. The EC differentiation was further confirmed by tube formation assays (Figure 1 C). Interestingly, although 7Ap alone was very effective in EC differentiation, the effect was somewhat decreased in the presence of VEGF (Figure 1 B, C). The effect of 7Ap on mature EC/SMC migration and proliferation was assessed with transwell migration and Br‐dU incorporation assays on human umbilical vein endothelial cells (HUVECs) and human vascular smooth muscle cells, respectively. The scramble 7Sp (MPHA[pS]GD) was included as a control. As shown in Figure S2A (Supporting Information), 7Ap increased Sca1 + ‐VPCs migration but had no effect on the migration of either ECs or SMCs. Furthermore, Br‐dU incorporation assays showed that 7Ap had no effect on cell proliferation of all three cell lines at a concentration of 1 ng mL −1 under serum‐free conditions (Figure S2B, Supporting Information). To test whether 7A/7Ap‐mediated Sca1 + ‐VPCs differentiation toward EC was also through the novel MEKK1‐7A‐14‐3‐3γ signal pathway, Mekk1 and 14‐3‐3γ knockdown via siRNA transfection was introduced into Sca1 + ‐VPCs, followed by analyzing EC/SMC differentiation markers via quantitative RT‐PCR. As shown in Figure 1 D, Mekk1 knockdown abolished 7A/VEGF‐induced EC differentiation but had no effect on 7Ap‐induced EC differentiation, while 14‐3‐3γ knockdown ablated both 7A/VEGF and 7Ap‐induced EC differentiation. These results suggest that the MEKK1‐7A‐14‐3‐3γ signal pathway is indeed involved in Sca1 + ‐VPCs differentiation toward EC lineage. Strikingly, knockdown of Mekk1 especially 14‐3‐3γ significantly increased Tagln expression (Figure 1 D). 2. 2 The 7aa‐Peptide Enhanced Vascular Injury Repair and Angiogenesis in Ischaemic Tissues In Vivo As described above 7A, especially 7Ap, could facilitate VPC mobilization and differentiation toward the EC lineage, suggesting that the 7A peptide may have therapeutic potential in vascular injury repair and angiogenesis in ischaemic tissues. To test this, the wire‐guided femoral artery injury model was created in ApoE −/− mice. 19 Pluronic‐127 gel containing 10 ng mL −1 of peptides or PBS was applied surrounding the adventitia of the injured vessels, which were harvested four weeks postsurgery. The hematoxylin and eosin (H&E) staining of cryo‐sections revealed that 7A or 7Ap administration significantly reduced neointima formation ( Figure 2 A). Figure 2 The peptide 7Ap increased vascular injury repair and angiogenesis in ischemic tissues in vivo. A) The peptides 7A and 7Ap attenuated neointima formation in a mouse femoral artery wire‐guided injury model. The left panel shows the H&E staining images of the injured vessel sections four weeks postsurgery. Scale bar: 100 µm. The right panel shows the average intima plus media area, with the value for the PBS group set as 1. 0 ( n = 6, one‐way‐ANOVA followed by Tukey's post hoc analysis). B) The peptide 7Ap increased foot blood perfusion in mice with a hindlimb ischemia model, which was introduced into six‐month‐old C57bl/6 mice, with 200 µL Pluronic‐127 gel containing 1 ng mL −1 peptides applied around the injured vessels. The left panel shows the Doppler Scanner images. The right panel shows the average ratio of blood flow in the right side (injured) to the left side (uninjured) ( n = 6, two‐way ANOVA followed by Dunnett's multiple comparison tests). C) The peptide 7Ap increased Sca1 + (green) cell migration into the ischemic tissue and differentiation into CD31 + cells (red). DAPI was included to counterstain the nuclei. The left panel depicts representative immunofluorescence staining images on a skeletal muscle section from the injured leg. Arrow indicates the Sca1 + cell niche. The right panel shows the mean ± SEM CD31 + or Sca1 + cells from six 20× views. Scale bar: 20 µm. *: p < 0. 05. **: p < 0. 01 ( n = 6, two‐way ANOVA followed by Dunnett's multiple comparison tests). Furthermore, a hindlimb ischaemia model was introduced in eight‐month (Figure 2 B) and ten‐week (Figure S3, Supporting Information) old C57BL/6J mice, in which Pluronic‐127 gel containing 10 ng mL −1 peptides was applied surrounding the injured vessels. 20 Foot blood perfusion was measured by a Doppler Scanner on day 7 and day 14 postsurgery. Young mice showed better foot blood perfusion recovery compared to aged mice (PBS group in Figure S3, Supporting Information, vs Figure 2 B). Importantly, 7Ap significantly promoted recovery in both groups of mice, which might be due to the increased formation of a new Sca1 + cell niche in the ischaemic tissues (Figure 2 C). 2. 3 Fabrication of Peptide‐Loaded Tissue Engineered Vascular Grafts (TEVGs) To evaluate whether the 7A peptide has therapeutic potential, TEVGs were fabricated by a co‐electrospinning technique. 9 The TEVGs were sized at 2. 0 mm diameter and 600 µm wall thickness ( Figure 3 A). They showed a micro/nanohybrid fibrous structure, consisting of PCL microfibers (≈6 µm) to provide mechanical support and structural maintenance and collagen nanofibers (≈600 nm) to act as a carrier for delivery of peptides (Figure 3 B). The loading amount is about 417. 5 ng for the graft of 1 cm length that has been calculated from the feeding ratio of electrospinning. Both fibers were distributed homogeneously across the graft wall as revealed by scanning electronic microscopy (SEM) (Figure 3 B middle). The distribution of the two types of fibers could also be clearly identified by double labeling showing PCL in red and collagen in green (Figure 3 B right). The peptides (7S, 7A, 7Ap) released from the grafts as a result of the degradation of the collagen fibers exhibited a sustained profile over a period of 30 d. The cumulative release reached ≈43 ± 7% (s. d. ) of the theoretical value (Figure 3 C) in 7A group, and a similar releasing profile has been observed in 7S and 7Ap groups (Figure S4, Supporting Information). The TEVGs (1. 0 cm in length) were further evaluated by in vivo implantation in a rat abdominal artery replacement model (Figure 3 D left). The patency of the implanted grafts was first examined by ultrasound at different timepoints (Figure 3 D, right). Most of the grafts had good patency without aneurysms or bleeding and patency rates were higher in the 7A/7Ap groups than in the PBS and 7S groups (Figure 3 E). 21, 38 Further analysis by stereomicroscopy revealed that the luminal surfaces of all groups were smooth, clean and free of thrombi at 12 weeks (Figure S5, Supporting Information). Figure 3 The fabrication of the peptide‐loaded tissue engineered vascular grafts. A) The images show the morphology of the tubular vascular grafts, the microstructure as revealed by scanning electronic microscope, and B) (upper panel) the distribution of the two types of fibers of PCL revealed as red by DiI and collagen revealed as green by DiO. B) (lower panel) The graft parameters are summarized. C) In vitro release of 7A peptide from the TEVG ( n = 3). D) Implantation of the graft in a rat to replace the abdominal artery (left), and the ultrasound scanning image of the implanted TEVGs (right). E) The patency rate of implanted TEVGs at different timepoints. 2. 4 7Ap Promoted Vascular Regeneration and Remodeling in TEVGs The endothelialization on the lumen side of the implanted TEVGs was first observed by scanning electronic microscopy at three locations selected continuously from the anastomotic site to the midportion. As shown in Figure 4 A, the lumen side of the 7A and 7Ap‐loaded TEVGs was already covered by a layer of cells at two weeks postimplantation. In contrast, the PBS and 7S‐loaded TEVGs were only partially covered, and bare fibers could still be identified in the midportion. After four weeks postimplantation, all of the grafts were almost fully covered by a confluent layer of neo‐tissue. In the 7A group, especially the 7Ap group, the cells oriented along the blood flow, resembling the endothelium of the native artery. 4′, 6‐diamidino‐2‐phenylindole (DAPI) staining revealed that a number of cells had incorporated into the wall of all of the TEVGs (Figure 4 B). Immunofluorescence staining demonstrated that there was a layer of ECs on the lumen side (Figure 4 B). The statistical analysis of the CD144 + cell‐covering area via immunofluorescence staining on the longitudinal sections revealed that 7Ap‐loaded TEVGs had much higher endothelialization rate compared to other groups (Figure 4 C and Figures S6 and S7, Supporting Information). After 12 weeks postimplantation, all of the grafts were almost fully endothelialized without detectable differences (Figure S8, Supporting Information). These results suggest that 7Ap can facilitate endothelialization of TEVGs. Figure 4 7Ap effectively increased endothelialization of the TEVGs. The TEVGs were implanted into rats to replace the abdominal artery and collected at two and four weeks postimplantation, followed by observation of the endothelialization on the lumen side of the TEVGs using A) scanning electronic microscope or B) CD144 staining of the sections. C) The endothelialized coverage was calculated by the CD144 positive cells covered area as a percentage of the entire lumen area. DAPI was used to counterstain the nuclei. Data are represented as the mean ± SEM for each group. *: p < 0. 05. **: p < 0. 01 ( n = 6, one‐way‐ANOVA followed by Tukey's post hoc analysis). In the native arterial vessels, ECs form a thin layer of intima while one to several layers of SMCs lie in the media. To assess whether the endothelialization of the TEVGs was accompanied by SMC layer formation, immunofluorescence staining was first performed on the cross sections with anti‐α‐SMA antibody. As shown in Figure 5 A, a layer of SMA + cells could be observed after four weeks, and the SMCs were well organized. The evolution of α‐SMA + layers from 4 to 12 weeks showed a divergent tendency among the four groups, that is, their thickness decreased in the 7A and 7Ap groups and increased in the PBS and 7S groups (Figure 5 B). This indicates that 7A, especially 7Ap, could reduce the overproliferation of SMCs that lead to adverse intimal hyperplasia. Figure 5 7Ap increased EC‐SMC interactions in the implanted TEVGs. Immunofluorescence staining was performed on TEVGs sections to observe the recruitment of SMCs by using A) anti‐α‐SMA and B) anti‐MHC, and quantification of the average layer thickness of B) α‐SMA and D) SM‐MHC positive cells. E) The interaction between ECs and SMCs was observed by double immunofluorescence staining with anti‐α‐SMA (green) and anti‐vWF (red). DAPI was included to counterstain the nuclei. Data are represented as the mean ± SEM for each group ( n = 6 for 4 weeks, and n = 4 for 12 weeks). **: p < 0. 01 (one‐way‐ANOVA followed by Tukey's post hoc analysis). The regeneration of mature smooth muscle (SM‐MHC + ) did not show any significant differences at four weeks postimplantation (Figure 5 C). After 12 weeks, the thickness of the SM‐MHC + layer was remarkably higher in the 7Ap group than in the PBS one (Figure 5 D), suggesting that 7Ap could modulate the infiltrated SMCs into a more mature contractile phenotype. Furthermore, double immunofluorescence staining with anti‐vWF (red) and anti‐α‐SMA (green) antibodies exhibited better tissue regeneration in the 7Ap group compared to the control group, showing a layer of neo‐endothelium and several layers of aligned SMCs beneath (Figure 5 E). At the same time, we noticed that some of the vWF‐positive cells were SMA‐positive as well. To address whether there is endothelial‐to‐mesenchymal transition, further detailed investigation is required. Successful vascular grafts need capillary vessels to support cells in the graft wall. Immunofluorescence staining with anti‐CD31 antibody on the cross sections of the TEVGs revealed that the addition of 7A especially 7Ap effectively increased capillary vessel formation even at two weeks postimplantation (Figure S9A, Supporting Information). After four weeks, the density of capillary vessels in 7Ap group was significantly higher than that in other groups (Figure S9B, Supporting Information). These results suggest that 7A especially 7Ap can increase capillary vessel formation. The 7Ap‐loaded TEVGs resembled native vasculature more closely. 2. 5 7Ap Stimulated Local Vascular Stem/Progenitor Cells Migration and Differentiation into the EC Lineage As our in vitro studies have shown that 7A, especially 7Ap, could increase Sca1 + ‐VPC migration and differentiation toward the EC lineage, we wondered whether the local VPCs contributed to the neo‐vasculature formation in the TEVGs. To test this, immunofluorescence staining was performed with anti‐Sca1 antibody on sections from two‐ and four‐week implanted TEVGs. As shown in Figure 6 A, the infiltration of Sca1 + ‐VPCs within the graft walls could be detected at two weeks postimplantation. As time went on, the Sca1 + cell number increased. Very importantly, 7A, especially 7Ap, enhanced this process, since the 7Ap group had the highest number of Sca1 + cells (Figure 6 B). Further experiments with double immunofluorescence staining using CD144 (red) and Sca1 (green) antibodies (Figure 6 C) revealed that the majority of the CD144 + cells were also Sca1‐positive (Figure 6 C, magnified image), indicating that these new ECs are derived from the Sca1 + ‐VPC differentiation. Figure 6 7Ap increased Sca1 + ‐VPC differentiation toward ECs contributing to endothelialization of TEVGs. A) Immunofluorescence staining was performed to detect the recruitment of Sca1 + ‐VPCs using anti‐Sca1 antibody on the implanted TEVGs sections at time indicated, and B) quantification of the infiltrated Sca1 + cells within 0–100 µm depth from the lumen by randomly selecting six views (40×) from each cross section. C) The contribution of the Sca1 + progenitor cells‐derived ECs to the endothelialization was detected by double immunofluorescence staining with anti‐Sca1 (green) and anti‐CD144 (red) antibodies on the four weeks implanted TEVGs sections. DAPI was used to counterstain the nuclei. Data are represented as the mean ± SEM for each group ( n = 6 at 2 and 4 weeks, and n = 4 at 12 weeks). *: p < 0. 05; **: p < 0. 01 (one‐way‐ANOVA followed by Tukey's post hoc analysis). It has been established that both local resident and circulating stem/progenitor cells contribute to vascular remodeling following vascular injury. 22, 23, 24 To explore the origin of the infiltrated Sca1 + cells and their contribution to the vascular remodeling in TEVGs, a bi‐layered vascular graft was employed in this study ( Figure 7 A, upper panel). In addition to the regular vascular grafts that act as an inner layer, an external layer composed of PCL nanofibers was introduced as an outside barrier (Figure S10, Supporting Information). 25 Due to its low porosity, this external layer could effectively restrict the infiltration of cells from the surrounding tissues after implantation for two weeks (Figure 7 Aa, b). 26 Quantitative analysis based on DAPI staining showed that the average cell number was significantly decreased within the graft wall, especially in the area near the lumen (Figure 7 Ac). The infiltration of Sca1 + ‐VPCs was further compared between the two groups, and statistical data showed that the density of the Sca1 + cells was markedly reduced due to the outside barrier at both two and four weeks (Figure 7 B). All of these results confirm that surrounding tissues are an important source for the Sca1 + ‐VPCs infiltration within the vascular grafts. Figure 7 Bi‐layered PCL TEVGs revealed that the Sca1 + ‐VPCs were mainly derived from the surrounding tissues. A) An illustration of the bi‐layered PCL fabricated TEVGs. B, C) The outer layer with small pores reduced Sca1 + ‐VPCs recruitment into the TEVGs wall and the lumen endothelialization of the TEVGs. B) The single PCL and bi‐layered PCL fabricated TEVGs were implanted into rats to replace the abdominal arteries, and harvested at two and four weeks, respectively, postimplantation, followed by immunofluorescence staining with anti‐Sca1 antibody to show the recruitment of Sca1 + ‐VPCs in the TEVGs wall or C) with anti‐CD31 (green, upper) and anti‐CD144 (red, lower) antibodies to show the lumen endothelialization. DAPI was used to counterstain the nuclei. Data are represented as the mean ± SEM. *: p < 0. 05; **: p < 0. 01; ***: p < 0. 001 ( n = 6, two‐way ANOVA followed by Dunnett's multiple comparison tests). Endothelialization of the two types of vascular grafts was compared by SEM observation (Figure S11, Supporting Information) and immunofluorescence staining (Figure 7 C, Figures S12 and S13, Supporting Information). The results indicated that the outside barrier slowed down the endothelialization process, and the average coverage ratio of neo‐endothelium was significantly lower in the outside barrier (bi‐graft) graft than in the control one (Figure 7 C). These results suggest that both the adjacent ECs 21, 27 and the circulating and local resident stem/progenitor cells migrate into and contribute to vascular remodeling in the TEVGs. To further verify the effect of 7Ap peptides on the migration of Sca1 + ‐VPCs from surrounding tissues into the grafts, Matrigel containing GFP‐Sca1 + ‐VPCs was applied surrounding the adventitia of vascular grafts ( Figure 8 A). Immunofluorescence images of the cross sections showed that after 3 d postimplantation the number of cells migrated to the graft wall was significantly ( p < 0. 05) lower in the control group compared to the 7Ap group (Figure 8 B, C). The fate of infiltrated Sca1 + ‐VPCs from surrounding tissues has also been investigated by seeding the GFP‐Sca1 + ‐VPCs into the graft wall directly (Figure 8 D). Double immunofluorescence staining using CD144 and Sca1 antibodies demonstrated after 7 d postimplantation some Sca1 + ‐VPCs differentiated into ECs, and 7Ap significantly ( p < 0. 05) enhanced the number of double‐positive cells on the lumen side compared to the control group (Figure 8 E, F). All of these results confirmed that Sca1 + ‐VPCs from the surrounding tissues can migrate transmurally and differentiate into the EC lineage in the TEVGs. Figure 8 7Ap increased in vivo migration and differentiation of Sca1 + ‐VPCs preseeded in the TEVGs. A) Matrigel containing GFP‐Sca1 + ‐VPCs was applied surrounding the adventitia of the vascular grafts. B) Immunofluorescence staining was performed to detect the migration of seeded VPCs within 0–200 µm depth from the lumen (dotted area) after 3 d postimplantation, and C) the corresponding quantification on GFP‐Sca1 + ‐VPC. D) GFP‐Sca1 + ‐VPCs were seeded in the graft wall directly. E) Double immunofluorescence staining was performed to detect the differentiation of seeded VPCs after 7 d postimplantation using anti‐GFP (green) and anti‐CD31 (red) antibodies, and the corresponding quantification on GFP + CD31 + VPC on the lumen side from five randomly selected views of each cross section. F) DAPI was included to counterstain the nuclei. Data are represented as mean ± SEM, *: p < 0. 05 ( n = 5, two‐tailed unpaired Student's t ‐test). 3 Discussion Tissue‐engineered vascular grafts prepared by traditional strategies are time‐consuming and costly, which often restricts their clinical application. Instead, a new in situ (cell‐free) concept has received increasing attention, that is, a neo‐artery is completely generated in situ by taking advantage of the regenerative potential of the body. 28 To stimulate or maximize this in situ regeneration capability, vascular grafts have been engineered in terms of physical structure 25 and surface chemistry, 29, 30 as well as the delivery of vasoactive molecules. 31 In this study, we demonstrated that an Hdac7‐ derived 7‐aa peptide can serve as a vasoactive molecule to mobilize local resident progenitor cells, contributing to in situ neo‐artery formation. The origin of cells participating in the regeneration of vascular tissues remains a major obstacle restricting the success of in situ tissue engineering strategies. We tend to believe that vascular cells (ECs and SMCs) migrate from the anastomotic native tissues (transanastomotic ingrowth). Breuer and co‐workers have reported that the adjacent vessel wall is the principal source of these endothelial and smooth muscle cells when implanted a polymer‐based graft as an inferior vena cava interposition in mice. 27 However, the results are often limited by many factors, including graft structure, implant site, and animal species. For example, cell migration from other directions (such as surrounding tissues) has been restricted due to the dense structure of the graft walls. It has been accepted that endothelial cell migration from the anastomotic sites is generally limited to 1 cm, 32 which is far less than the clinical requirement of re‐endothelialization of relatively longer vascular grafts. 33 Therefore, there is a need for recruitment of cells (both mature ECs and stem/progenitor cells) from sites beyond the anastomotic border via circulation or through migration from the surrounding tissue. Vascular tissue‐resident stem cells have been discovered to display the capacity to differentiate into vascular cell lineages, which may contribute to the regenerative process. 34 More specifically, the adventitial resident VPCs play a major role in the repair of the injured endothelium via migration toward the endothelium and differentiation into an EC lineage. 5, 6 HDAC7 plays an important role in the maintenance of endothelium integrity. 35, 36, 37 In a parallel study, we have demonstrated that mouse Hdac7 mRNA can undergo alternative translation to produce a biological active 7‐aa peptide, which functions as a phosphate transfer carrier in cellular signal transduction. In this study, we have evaluated its application potential in peptide‐loaded tissue engineered vascular grafts by incorporating the 7A/7Ap peptides into the micro/nanofibrous vascular grafts. In vitro studies demonstrated the 7Ap peptide is very powerful in Sca1 + ‐VPCs‐derived EC differentiation. However, we also noticed that 7Ap seemed less effective during EC differentiation in the presence of VEGF compared to 7Ap alone. This phenomenon may be due to the diverse functions of VEGF. In in vitro differentiation process, the increased EC cells can be derived from both the newly differentiated ECs and the proliferation of the differentiated mature ECs. When mature ECs undergo proliferation, the EC marker especially those involved in cell‐to‐cell contact such as CD31 ( pecam1 ) and CD144 ( Cdh5 ) will be downregulated to relieve the connection among adjacent cells. This will result in an observable decrease in EC marker expression at mRNA levels and tube formation. Although 7Ap had no effect on EC proliferation, VEGF is a mitogenic factor for EC. Thus, the combination of 7Ap and VEGF may show a decrease in EC marker expression and tube formation compared to 7Ap alone. An intact endothelial cell monolayer has proven to be an antithrombogenic interface to maintain the patency of the implanted vascular grafts. 38 In the meantime, the recruitment of SMCs is vital to stabilize the newly formed endothelium. 32, 39, 40, 41, 42 In this study, we have demonstrated that the inclusion of 7A, especially the phosphorylated version 7Ap, significantly increased endothelialization. The majority of the ECs, especially those in the midportion of the TEVGs, may be derived from the local resident Sca1 + progenitor cells of the surrounding tissues, because 7Ap can serve as a chemoattractant for Sca1 + ‐VPCs and direct their differentiation toward the EC lineage. The ECs on the lumen of TEVGs will secrete relevant cytokines (such as PDGF or TGF‐β1) to recruit SMCs to stabilize the newly formed endothelium. 40 The interactions between ECs and SMCs are very important for the function and homeostasis of the regenerated neo‐artery. Considering that 7Ap suppresses Sca1 + ‐VPCs differentiation toward SMCs and has no effect on SMC migration and proliferation, the recruitment of SMCs may be mainly mediated by ECs on the lumen of the TEVGs. After all, precise lineage tracing may be required to verify the contribution of Sca1 + ‐VPCs. Therefore, 7Ap may help with the organization and maturation of SMCs within the TEVGs, which could also explain the discrepancy between in vivo SMC regeneration and in vitro SMC differentiation. Importantly, the Sca1 + progenitor cells also contribute to capillary vessel formation in the TEVGs wall, which is critical for the functional maintenance of the neo‐artery following scaffold degradation. The dose‐dependency of the 7A/7Ap peptide is an important factor affecting therapeutic efficacy. 7A/7Ap was effective in EC differentiation at a concentration of 0. 1 ng mL −1. For cell migration, it was more effective at a relatively higher concentration, and therefore 1 ng mL −1 was used. In in vivo studies, we assumed that there would be a concentration gradient extending from the carrier (F‐127 gel or collagen fiber), and the final concentration within the ischaemic tissue reached ≈0. 1 ng mL −1. For future studies, it will be necessary to detect whether there is dose dependency. In summary, the clinical potential of histone deacetylase 7 (HDAC7)‐derived 7‐amino acid peptides in vascular repair and regeneration has been systematically evaluated. The effect of the 7A peptide, especially the phosphorylated 7A peptide (7Ap), on VPC migration and EC differentiation has been investigated by in vitro functional analyses. Local delivery of 7A/7Ap increased re‐endothelialization and suppressed neointima formation in the femoral artery injury model, and promoted foot blood perfusion recovery in the hindlimb ischaemia model. Degradable tissue‐engineered vascular grafts loaded with peptides have been successfully prepared and evaluated in an abdominal aorta replacement model of rats. Sustained delivery of 7A, especially 7Ap, from tissue‐engineered vascular grafts enhanced endothelialization and intima/media formation in the vascular graft via inducing Sca1 + ‐VPC migration and differentiation. All of these results suggest that 7A/7Ap provides a promising therapeutic strategy for ischaemia diseases and vascular graft bypass surgery. Because the human orthologue of Sca1 has not been identified yet, it will be necessary to investigate whether 7A/7Ap has similar effect on human adult stem/progenitor cells, which will provide direct evidence for its possible application in translational medicine. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800252
| 2,018
|
Advanced Science
|
DIY 3D Microparticle Generation from Next Generation Optofluidic Fabrication
|
Abstract Complex‐shaped microparticles can enhance applications in drug delivery, tissue engineering, and structural materials, although techniques to fabricate these particles remain limited. A microfluidics‐based process called optofluidic fabrication that utilizes inertial flows and ultraviolet polymerization has shown great potential for creating highly 3D‐shaped particles in a high‐throughput manner, but the particle dimensions are mainly at the millimeter scale. Here, a next generation optofluidic fabrication process is presented that utilizes on‐the‐fly fabricated multiscale fluidic channels producing customized sub‐100 µm 3D‐shaped microparticles. This flexible design scheme offers a user‐friendly platform for rapid prototyping of new 3D particle shapes, providing greater potential for creating impactful engineered microparticles.
|
Engineered microparticles with nonspherical shapes have attracted increased attention recently due to different shape‐dependent functionalities that emerge from specific particle geometries. 1 Examples of these shape‐dependent functionalities are seen in biology where the shape and surface area of drug carrier particles can alter cellular uptake, 2 and the shape of implanted hydroxyapatite particles can change the body's inflammatory response. 3 Additionally, complex double‐cone particles have allowed for programmable release of multiple active ingredients for therapeutic applications. 4 Besides for bioinspired applications, nonspherical particles can self‐align in a confined flow 5 or change the rheological properties of a solution. 6 Also, particles with interlocking shapes can create low‐density loadbearing structures from particle jamming. 7 Although, it remains challenging to create complex 3D‐shaped particles to meet new emerging demands. Microfluidic methods such as stop flow lithography (SFL) 8 and optofluidic maskless lithography (OFML) 9 have shown great potential for producing complex‐shaped particles, however, the approach mostly resulted in 2D extrusion shapes with uniform side profiles. By modifying the SFL process, we previously demonstrated the creation of 3D‐shaped particles with a fabrication scheme called optofluidic fabrication. 10 Optofluidic fabrication creates an infinite set of complex 3D‐shaped particles simply by varying inertial flow 11 and light conditions. However, our 3D particles generated from the use of this method have been mainly on the mm scale because of large mm scale channels used to lower pressure drops, limiting the method's potential usage. Similar particle fabrication methods termed optical transient liquid molding (TLM) 12 and Dean flow‐based optofluidic fabrication 13 were reported generating 3D‐shaped particles with overall sizes of hundreds of micrometers. These methods created sub‐millimeter particles by using smaller channel molds, even though their dimensions are still large for potential applicability in numerous biotechnological applications. 14 Here, we present a novel method termed next generation optofluidic fabrication (NG‐OF) for creating 3D‐shaped microparticles an order of magnitude smaller than our previously demonstrated particles, while still utilizing primarily mm scale channels for ease of fabrication and low pressure drops. As shown in Figure 1 a, NG‐OF uses multiscale channels where the inertial flow shaping is achieved in a mm scale channel upstream, 10 and passes a tapered reduction section while keeping its flow shape. The flow is then rapidly stopped, ultraviolet (UV) polymerization is initiated in the reduction section (Figure 1 b; and Figure S1, Supporting Information), and a 3D microparticle(s) is generated (Figure 1 c–e). Furthermore, we present a new rapid channel fabrication scheme to effectively create fluidic channels with custom pillar configurations. When a new particle shape is needed, a new channel with a different pillar configuration is required. Thus, instead of preparing new channels either by photolithography or 3D printing, 10, 15 NG‐OF uses an on‐the‐fly approach that creates pillars at a desired channel location by local UV polymerization (see Figure 4 a and more details below) that is used for creating 3D‐shaped microparticles. Figure 1 Operating principles of next generation optofluidic fabrication: a) The flow cross‐section of inert (i and iii) and UV‐reactive streams (ii) are shaped by fluid inertia as they flow past pillars at Re ≈ 5. The flow cross‐sections at slices S0 and S6 are shown at the inlet and after passing six pillars. After inertial flow shaping, fluid streams enter a tapered reduction zone that reduces the channel area 100‐fold, scaling down the flow cross‐section to the µm scale. b) After inertial flow shaping, the flow is quickly stopped, and patterned UV light is illuminated on the channel reduction section creating a 3D microparticle. c) A rendering of a microparticle resulting from a T‐shaped UV light pattern with an E‐shaped flow cross‐section. d) Experimental setup showing the microparticle polymerization process. From left to right: Far‐field view of UV illumination on the channel reduction; From top to bottom: Before, during, and after 20 ms UV illumination. e) 3D view of an experimentally fabricated microparticle. Scale bar represents 75 µm. As mentioned, NG‐OF is based on inertial flow shaping from our previous optofluidic fabrication. 10 Figure 1 a shows inertial flow shaping of three streams in a fluidic channel containing six pillars. The initial and final flow shapes are shown at corresponding slices labeled S0 and S6. The channel design is comprised of mm and µm scale channels connected through a tapering section. This design is adopted to avoid excessively high input pressures when small microchannels are used for inertial flow (Note S1, Supporting Information), which is often not compatible with conventional Polydimethylsiloxane (PDMS)‐glass fluidic devices. With our design, in upstream mm scale channels, inertial flow shaping can occur under reasonable pressure drops while maintaining the flow cross‐section through the tapering zone. The flow is then rapidly stopped, and a 3D particle is generated by illuminating patterned UV light at the reduction section (labeled “R”), as shown in Figure 1 b–e. To generate a well‐defined particle, the flow needs to be stopped fully to allow sufficient UV exposure and minimize diffusion between UV‐curable and inert fluids. The stopping time relies on two timescales: 1) hydraulic capacitance and 2) viscous dissipation. 12 First, the hydraulic capacitance timescale is dominated by the flexible channel walls which inhibits fast changes in pressures due to expansion and relaxation of the PDMS channel walls under pressure. 8, 16 When the input pressure is lowered to stop the flow, the channel relaxes and induces unwanted squeeze flow. 8, 16 We address this issue by using stiffer thermally cured PDMS channels, however, more curing agent or other rigid oxygen permeable materials 17 can further reduce this hydraulic capacitance timescale. Second, the viscous dissipation timescale is determined by viscous diffusion of momentum, proportional to D h 2 /ν, where ν is the kinematic viscosity. This implies that smaller channels will result in faster viscous dissipation leading to shorter flow stopping time. However, the higher pressures required by smaller channels often become too high for conventional PDMS‐glass channels to withstand. As a solution, NG‐OF utilizes the tapered reduction design where the channel cross‐sectional area decreases by a factor of 100. This allows a moderate input pressure of 1200 mbar to still generate sufficient inertial secondary flows at Re = 5 (where Re is the Reynolds number and see Note S1 (Supporting Information) for detailed definitions of all associated parameters) with stop times less than 400 ms (see Figure S2, Supporting Information). For NG‐OF microparticle generation, we first prepared a channel mold using a 3D printer (photocurable inkjet printer) since photolithography is not practical for creating multiscale channels with a vertical tapering design. To understand and characterize the inertial flow in our channel, entire channels were simulated using COMSOL Multiphysics. Note that the 3D printed mold displayed a rounded cross‐sectional channel shape in the reduction section due to resolution limits of the 3D printer (Figure S3, Supporting Information), and this feature was taken into consideration for simulations. Simulation details can be found elsewhere, 15 but briefly, the steady‐state incompressible Navier–Stokes and convection‐diffusion equations were solved simultaneously with each model containing 8. 9 × 10 6 degrees of freedom. Normalized concentration slices c* of photoinitiator concentration are presented in Figure 2 a–c where three channels with different pillar configurations are modeled: A channel with six pairs of half‐cylindrical pillars located along the side‐walls (Figure 2 a), six pillars located along the center (Figure 2 b), and pillars that create an “E”‐shape flow cross‐section (Figure 2 c). Channel top views are also presented for an x – y slice 5 µm above the channel bottom, and more slices can be found in Figure S4 (Supporting Information). The z – y slices S0‐S6 show the flow shape evolution, and the final reduction slice labeled “R” is shown with the area scaled up 100 times for easier visual inspection. Figure 2 Numerical versus experimental analysis of inertial flow shaping with reduction channels. a) Normalized concentration plots of inertial flow shaping from a channel with half‐cylindrical pillars located along the side walls. The x – y view shows the concentration 5 µm above the channel bottom. z – y slices at S0‐S6 show the flow shaping achieved upstream. After the reduction section, slice R shows the miniaturized flow cross‐section with the area scaled up 100 times for easier visual inspection. b) Similar concentration plots are shown for a channel with pillars located along the central axis and c) for a complex pillar configuration that results in an “E” shape flow profile. d) Thin particles are polymerized and flipped on their sides at locations S6 and R to reveal flow shape changes. Scale bars for row S6 represent 1 mm, while scale bars for row R represent 100 µm. Simulation cross sections at similar locations as d) are shown using an upper 50% concentration threshold. e) To compare cross‐sectional shapes pre‐ and postreduction, the reduction cross‐sections were scaled up. Error bars represent standard deviation from 15 particles created in 3 different channels. These simulation concentration slices were compared with experiments. The simulation concentration slices were thresholded similar to our previous work 15 using the upper local 50% concentration range to compare with the experimental particle shapes. For experiments, a thin slit of UV light was illuminated at locations S6 and R across the entire width of the channel, generating a particle whose shape represents the flow cross‐section. Simulation and experimental cross‐sections for the three channel designs are shown in Figure 2 d, and more particle images can be seen in Figure S5 (Supporting Information). Channel heights were normalized between S6 and R (Figure S6, Supporting Information) to allow direct comparison between cross‐sectional area measurements, and a detailed comparison is plotted in Figure 2 e. Although general cross‐sectional area trends matched, slight cross‐sectional area differences were seen. These cross‐sectional area differences were smaller for experimental cases compared to simulations cases. Since the steady state simulation results do not take into account the flow stopping step, diffusion occurring during flow stoppage creates a larger experimental particle shape, especially in the channel reduction where particle sizes approach relevant diffusion scales. One interesting feature was seen in the center‐pillar case (Figure 2 b) where the experimental area was increased while the simulation predicted a decrease. This may be because the numerical prediction does not include particle polymerization or oxygen inhibition. For the center‐pillar channel, the flow shape is isolated from the side walls, preventing loss of polymerization from oxygen‐inhibition near the side walls, 18 allowing the full uninhibited flow shape to polymerize. Also, it was assumed that the main channel was perfectly rectangular. Under internal pressure, PDMS channels bulge, leading to shape discrepancies which could be corrected with more accurate geometry modeling. With the channel design validated as a method to miniaturize the inertially shaped flow, the same three channels were used to create 3D‐shaped microparticles as shown in Figure 3. The full 3D particle shape could be adjusted by simply changing the light pattern or changing the location the light pattern was illuminated. In Figure 3 a, UV light was illuminated off‐center, and 3D microparticles with overall dimensions of width × depth × height approximately equal to 70 × 70 × 104 µm 3 are generated. By using the center‐pillar channel in Figure 3 b, the UV‐reactive stream is directed away from the channel walls, allowing for a smaller 3D particle. Although the flow cross section is not as intricate from this pillar configuration, overall particle dimensions are as small as 30 × 50 × 35 µm 3. To demonstrate increased particle shape complexity, the E‐channel was used (Figure 3 c) that created particles with dimensions of ≈70 × 70 × 104 µm 3. Note that since the photomasks were aligned manually, some shape discrepancies can be seen from the particle side‐views. However, these mask alignment discrepancies could be fixed completely with a digital micromirror device (DMD) UV system to accurately control the location of UV illumination. [[qv: 9b]] These 3D microparticles can potentially be made smaller and finer by overcoming challenges with diffusion between coflowing fluid streams. As particle sizes become smaller, diffusion length scales can approach particle sizes, either requiring faster flow stopping to limit diffusion or particles with less intricate shapes. With the experimentally tested conditions of Re = 5, the average transit time through the main channel section is 2. 2 s, leading to a total diffusion time of ≈2. 6 s, including flow stopping. The theoretical particle resolution can be inferred by calculating the diffusion length L D from Equation (1) below (1) L D = D PI t D where D PI is the diffusivity of photoinitiator and t D is the total diffusion time. With the DMPA diffusivity of 4 × 10 −11 m 2 s −1 15 in poly(ethylene glycol) diacrylate (PEG‐DA) and a diffusion time of 2. 6 s, an isometric diffusion length of ≈10 µm is expected, representing the particle resolution limit. By incorporating a rigid channel, 19 higher allowable pressures and flow rates could reduce transit and stop times, leading to finer particle features. For example, a Reynolds number of 20 reduces the transit time fourfold to 0. 55 s. Assuming a faster stop time of <100 ms from a rigid channel, the diffusion length could be reduced to L D ≈ 5 µm, allowing for particle features half the size. Figure 3 Examples of 3D‐shaped microparticles created using NG‐OF. a) The top‐view, side‐view, and 3D‐view of experimentally fabricated 3D microparticles are shown by combining three different UV light patterns with part of the “I”‐shaped flow cross‐section. b) The center‐pillar design creates a smaller UV‐reactive flow shape. c) The E‐pillar channel demonstrates more complex 3D‐shaped microparticles. Renderings of the 3D particles are shown below. The position of UV illumination is also illustrated with the UV light path intersecting the channel cross section. Particles are false‐colored to enhance contrast with background. Scale bars represent 75 µm. It should be mentioned that to fabricate a new particle shape, a specific pillar configuration should be determined. For this task, two open‐source programs, Flowsculpt 20 and uFlow, 21 can be used to easily design a new channel. Flowsculpt calculates the necessary pillar configuration to achieve a desired flow shape, and uFlow quickly displays the flow shape from user inputted pillars. Although these tools are not designed specifically for the NG‐OF process due to the interpillar distance difference, they still provide quick and very close approximations to flow shapes and channel designs that greatly aid in the design process. Then, a specific channel can be prepared using a 3D printer; however, the 3D printing preparation process can be costly and time‐consuming whenever new channels are required. Therefore, instead of printing new channels, a simple method is hereby presented by creating custom pillars in blank channels via UV polymerization. As shown in Figure 4 a–c, a blank channel is first filled with UV‐reactive fluid. Then a circular pattern of UV light is illuminated to create a pillar (Note S2, Supporting Information) at a desired location. By using a programmable motorized microscope stage (Movie S1 and Figure S7, Supporting Information), multiple pillars were created sequentially with high precision in a fully automated manner. We refer to this custom UV‐polymerization pillar scheme as “Do‐it‐yourself” (DIY) pillars. Although a blank channel mold originally needs to be printed, any pillar configuration (size, location, and spacing) can be prepared from this single blank channel mold. Figure 4 Overview of on‐the‐fly pillar fabrication for microparticle generation. a) A schematic representation of the on‐the‐fly pillar fabrication process used to create custom DIY pillars from UV‐polymerization. 1) An empty channel is filled with UV‐reactive fluid. 2) A circular photomask is illuminated to polymerize a custom pillar. 3) A motorized stage moves to the next pillar location, and this process is repeated until all pillars are created. 4) Then streams of UV reactive precursor and inert fluids are injected for to initiate flow shaping. b) Far field image showing a DIY pillar being polymerized. c) Experimental images showing the DIY pillar polymerization process to replicate d) the E‐channel with PDMS pillars from the 3D printed channel mold. e) The lateral pillar locations and f) the diameter of DIY and original PDMS pillars are compared to the intended channel design for the E‐channel. Error bars represent the standard deviation from 3 separate channels. g) Thin particles are polymerized and flipped on their sides at locations S6 and R to reveal flow shape changes. Scale bars for column S6 represent 1 mm, while scale bars for column R represent 100 µm. The reduction cross‐sections were scaled up to set the channel heights equal. h) Cross‐sectional areas are plotted for the original channel, simulation thresholded results, and DIY E‐particles. Error bars represent standard deviation from 15 particles created in 3 different channels. i) 3D‐shaped microparticles created using a DIY Channel. Scale bar represents 75 µm. Note that under normal UV polymerization conditions, the DIY pillars will not remain fixed during flow due to the oxygen permeable layer. Thus, we overexpose UV light and use a slight channel modification made during the channel preparation step. An oxygen diffusion barrier is added on top of the channels by simply inserting a conventional glass slide into the liquid PDMS during the curing process. This glass slide becomes embedded within the PDMS channel structure and controls oxygen diffusion into the channel. Using the DIY‐pillar generation scheme, we created the same channel layout as the E‐channel used in Figures 2 and 3 (Movie S1, Supporting Information). The lateral location and diameter of the six sequential pillars were analyzed (Figure S8, Supporting Information) and plotted in Figure 4 e, f. Both the original 3D printed channel molds and the DIY channels were compared. For both the original channels and the DIY channels, the lateral positions of the pillars and errors in diameter were compared (Figure 4 e, f), and resulted in less than 3% error, showing a high precision pillar fabrication scheme. Simulation and experimental areas were analyzed (Figure 4 g) at S6 and R for the DIY channel and compared with the E‐channel data from Figure 2 d. The E‐shape is stretching relatively more during inertial flow shaping and also shows slight asymmetry in the middle “tip. ” This stretched inertial flow shaping may be caused from the more rigid DIY pillar channels since the embedded glass diffusion barrier does not allow as much channel expansion during flow. The different E‐particle shapes might also be caused from slight concave shape (Figure S9, Supporting Information) of DIY pillars due to the numerical aperture (NA) of the 2. 5× objective (NA = 0. 085). The experimental cross‐sectional areas were plotted for the DIY case and shown in bar plots in Figure 4 h next to the original channel and simulation area results. Unlike the PDMS and simulation area changes, the DIY‐E particle area increases 13. 0% after the taper. However, the DIY E‐shapes are actually matching more closely to the simulations compared to the original E particles, possibly since the DIY pillar diameters are closer to the intended computer‐aided design (CAD) than the original PDMS pillars (Figure 4 f). 3D microparticles were also created using the DIY E‐channels with four different photomasks, shown in Figure 4 i. Similar to the original channels, these 3D particles had dimensions of ≈70 × 70 × 104 µm 3. In summary, the presented next generation optofluidic fabrication provides a simple platform to create 3D‐shaped microparticles by overcoming many design challenges faced when scaling down channel sizes with inertial flows. By utilizing primarily mm scale fluidic channels, low operating pressures allow for fast flow response times while still using standard PDMS channels. NG‐OF can rapidly create DIY channels with any pillar configuration from a single blank channel mold. The versatility of NG‐OF allows for biological components such as cells or drugs to be added to precursor streams and become encapsulated within microparticles for studies of cell‐particle interactions with complex‐shaped particles. 3D particles from NG‐OF can also benefit from the increased surface area per unit volume to be used for highly sensitive RNA/DNA sensing over 2D microparticles. 22 Moreover, other additives such as magnetic nanoparticles can be added for particle control using an external magnetic field. 12 The presented work uses 3D printed molds to create multiscale channels. By using 3D printers with finer resolution, or by using a process such as micromachining, smaller reduction channels could be created that easily allow smaller 3D microparticles. With the ability for NG‐OF to quickly create a variety of 3D‐shaped microparticles, this technique can be easily adopted to meet the demands of new emerging fields. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201800261
| 2,018
|
Advanced Science
|
A Modular Assembly of Spinal Cord–Like Tissue Allows Targeted Tissue Repair in the Transected Spinal Cord
|
Abstract Tissue engineering–based neural construction holds promise in providing organoids with defined differentiation and therapeutic potentials. Here, a bioengineered transplantable spinal cord–like tissue (SCLT) is assembled in vitro by simulating the white matter and gray matter composition of the spinal cord using neural stem cell–based tissue engineering technique. Whether the organoid would execute targeted repair in injured spinal cord is evaluated. The integrated SCLT, assembled by white matter–like tissue (WMLT) module and gray matter–like tissue (GMLT) module, shares architectural, phenotypic, and functional similarities to the adult rat spinal cord. Organotypic coculturing with the dorsal root ganglion or muscle cells shows that the SCLT embraces spinal cord organogenesis potentials to establish connections with the targets, respectively. Transplantation of the SCLT into the transected spinal cord results in a significant motor function recovery of the paralyzed hind limbs in rats. Additionally, targeted spinal cord tissue repair is achieved by the modular design of SCLT, as evidenced by an increased remyelination in the WMLT area and an enlarged innervation in the GMLT area. More importantly, the pro‐regeneration milieu facilitates the formation of a neuronal relay by the donor neurons, allowing the conduction of descending and ascending neural inputs.
|
1 Introduction Spinal cord injury (SCI) causes massive death of neurons and glia, leaving a hostile microenvironment that impedes nerve fiber regeneration and tissue repair. 1 Although endogenous neural stem cells (NSCs) have the potential for spinal repair, their bias toward astrocytic differentiation hinders the restoration of lost neurons and oligodentrocytes. Moreover, astrocyte differentiation of NSCs also contributes to glial scar after spinal cord injury. 2, 3 Yet, thus far, there is no effective treatment to this condition. The recent development of stem cell–based tissue engineering technology has provided a promising strategy for treating SCI. 4 With the insights gained from developmental biology and the advancement in 3D culturing, tissue engineering has entered a new era of constructing of multiple tissue types or organoids. 5 It is believed that tissue engineering would help realize constructing a functional tissue or organoid through the combination of stem cells, functional biomaterials, and neurotrophic factors to treat diseases of the central nervous system, including SCI. 6 It has been reported that significant motor sensory recovery may be achieved when donor NSCs were chemically induced into neurons in vivo to reconnect the neural pathways of the injured spinal cord. 7 It is believed that the exogenous stem cell–derived neurons can contribute to a newly established “neuronal relay” that would reform the synaptic connection of the interrupted neural pathway and reconstruct the neural transmission circuit. Thus, the use of “neuronal relay” strategy to repair the injured spinal cord has received increasing recognition 2, 8 in recent years. When stem cells were combined with biomaterials supporting sustained release of neurotrophic factors, it is believed that the postinjury microenvironment could be improved, allowing longer axonal growth and more synaptic connections with the host neurons. 9 However, safety issue remains a major concern for direct transplantation of stem cells into the spinal cord because of the possibility of migration and ectopic proliferation, and the uncertainty of the differentiation of the grafted cells. 10 In our previous study, we have adopted the “tissue engineering neuronal relay” strategy to repair the rat SCI by tissue engineering approach. To strengthen the safety, the stem cells were induced into young neurons in vitro, prior to transplantation. 11 The preconstructed functional neuronal network had provided a favorable microenvironment that increased the survival of donor cells, thereby fostering the regeneration of the host nervous tissue. 11, 12 More importantly, this in vitro tissue construction strategy is believed to minimize the uncertainty of stem cell differentiation in vivo. [[qv: 6b]] Using the same tissue engineering strategy, we have also constructed NSC‐derived oligodendrocytes that formed distinct multilayered lamellae both in vitro and in vivo. 13 These two tissue construction strategies have allowed the donor cells to differentiate with specific function for repair of injured spinal cord. Buoyed by these advancements in spinal cord tissue engineering, it is presumable that a spinal cord organoid that recapitulates the major morphological and functional properties of both white and gray matters would provide reinforced therapeutic efficacy in SCI repair, but such configuration has not yet been achieved. In the present study, we have combined the neuronal and oligodendroglial induction techniques and developed a novel spinal cord–like tissue (SCLT) in vitro by modular assembly of the white matter–like tissue (WMLT) and the gray matter–like tissue (GMLT) from NSCs with structural and functional simulation to the rat spinal cord. The neural cells in the SCLT exhibited targeted repair in the injured white matter and gray matter, respectively; moreover, both the repaired white matter and gray matter functioned synergistically to rebuild the neural pathway and to improve paralysis hind limb motor function of rats after SCI. It is also envisaged that the SCLT would serve as a useful in vitro platform for future study of neuropharmacology and neurodevelopment of the spinal cord. 2 Results 2. 1 Modular Assembly of SCLT Through a combination of gene modification, 3D culture technique, and a modular assembly protocol ( Figure 1 a), NSCs were induced in the SCLT into the major cell types (neuron, oligodendrocyte, and astrocyte) in the gray and white matters of the spinal cord. The GMLT module was constructed by promoting neuronal differentiation of neurotrophin‐3 (NT‐3)/TrkC gene–modified NSCs in the collagen sponge column, while the WMLT module was constructed by inducing the maturation of the ciliary neurotrophic factor (CNTF) gene–modified oligodendrocyte precursor cells (OPCs) derived from NSCs in the collagen sponge ring (Figure 1 a). After separated culturing for 7 days, the two modules were assembled into one entity, which was allowed to continue culturing for another 7 days for further maturation and mutual interaction (Figure 1 a). Immunocytochemistry was used to validate the expression of anticipated markers during stepwise construction of the modular assembly and maturation process by cells in the GMLT, WMLT, and SCLT after 14 days of culture. Nestin positive NSCs from the neurospheres (Figure S1a, Supporting Information) were transfected with a vector to either express NT‐3 or its receptor TrkC, and then seeded in a 3D collagen sponge column for 7 days of culture for the construction of GMLT (Figure S1b, Supporting Information). Immunocytochemical analysis revealed that ≈70% of the GMLT cells expressed the neuronal marker Map2, and 20% of the cells expressed the astrocyte marker glial fibrillary acidic protein (GFAP) at this time point (Figure S1c, Supporting Information). Meanwhile, NSCs were induced into OPCs with exogenous triiodothyronine (T3) and platelet‐derived growth factor (PDGF), and the induced cells showed a high purity of neuroglycan 2 (NG2, a marker for OPCs, Figure S1d, Supporting Information). The OPCs were induced for differentiation to construct the WMLT. After 7 days of culture in the collagen sponge ring, the cells were able to differentiate into myelin basic protein (MBP) positive cells with ubiquitous expression of CNTF (Figure S1e, Supporting Information). It is estimated that 80% of the differentiating cells expressed oligodendrocyte marker MBP, and 20% of them expressed GFAP at this time point (Figure S1f, Supporting Information). Figure 1 Assembly of the SCLT and its phenotypic similarity to normal adult spinal cord. a) A schematic diagram of the assembling process of a SCLT. b) An overview of the SCLT showing good integration of the WMLT (arrowhead) and the GMLT (arrow) with no obvious gap. c) Distinct partition of the GMLT module (GFP positive) and the WMLT module (GFP negative, MPB positive) after being assembled into a SCLT for 7 days. d, d1, d2) H&E staining of a SCLT showing good cell distribution in both WMLT and GMLT region. e, f) Surface view of a SCLT as seen by scanning electron microscopy. g) After 14 days of culturing, protein expression of individually cultured GMLT or WMLT, SCLT, and normal spinal cord (SC). h) Cells (arrows) in the GMLT were triple‐stained for Map2, PSD95, and SYP. GFP and Map2 positive cells in the GMLT expressed i) ChAT, j) glutamate (Glu), or k) GAD67. l) Bar chart showing the percentage of ChAT, Glu, and GAD67 positive cells among all GFP positive cells. Asterisks indicate p < 0. 05. m) Relative expression levels of different proteins in the GMLT, WMLT, SCLT, and SC groups, and asterisks, #, and & symbols indicate p < 0. 05 when the SC group was compared with the GMLT, WMLT, or SCLT groups, respectively. n) Q‐PCR was used to detect the difference in expression of specific mRNA between the SCLT and normal spinal cord. Asterisks indicate p < 0. 05 when the SC group was compared with the SCLT group. Scale bars = 500 µm in panels (b)–(d) and (e), 40 µm in panels (d1) and (d2), 5 µm in panel (f), 10 µm in panel (h), 50 µm in panels (i)–(k). Thus, following 14 days culturing in vitro, the assembled SCLT roughly simulated the anatomical partitions of a spinal cord, i. e. , the white matter and the gray matter (Figure 1 b–d). The assembly exhibited a good integration with no discernible gap between the two modules (Figure 1 b, phase‐contrast imaging of a SCLT in the culture dish). Additionally, cells seeded in each module did not seem to migrate significantly into the other module at this time point, as shown in Figure 1 c, where most of the cells in the GMLT (from green fluorescent protein (GFP) transgenic Sprague‐Dawlay (SD) rats) or the counterparts in the WMLT (from non‐GFP donor, immunostained by MBP, a myelin marker) remained within the same module. Hematoxylin and eosin (H&E)‐stained section sampled from the center of a serial of SCLT transverse sections showed the dense population of cells in both WMLT and GMLT modules (Figure 1 d). A closer look from this section revealed good cell viability and enriched cell contacts (Figure 1 d1, d2). Scanning electron microscopy (SEM) presented the surface view of a SCLT (Figure 1 e). Higher magnification image showed the abundant cell contacts amid the dense cell population in the GMLT (Figure 1 f). 2. 2 Phenotypic Similarities between the SCLT and Adult Spinal Cord Tissue Following 14 days single culture of GMLT or WMLT, or 7 days culture after they were assembled to SCLT, protein expression profile was measured by Western blotting. The results showed that the GMLT predominantly presented neuronal phenotypes, as demonstrated by the expression of a battery of molecular markers that include neurofilament (NF) as a pan mature neuronal marker, postsynaptic density protein 95 (PSD95), and synaptophysin (SYP) as post‐ and presynaptic markers, respectively, and neurotransmitter synthesizing proteins such as choline acetyltransferase (ChAT), glutamate decarboxylase 67 (GAD67). In contrast, the WMLT were enriched with cells expressing MBP, a myelin marker (Figure 1 g, m). When the two modules were assembled into SCLT, the protein expression profile showed a high similarity to that of a mature spinal cord (a positive control) in markers of NF, PSD95, ChAT, GAD67, GFAP, SYP, and MBP, except for the exogenous genes like TrkC, NT‐3, and CNTF (Figure 1 g, m). After the assembly of SCLT, Map2 positive neurons in the GMLT region exhibited intense expression of PSD95 and SYP, suggesting the establishment of synaptic connections between neurons within SCLT (Figure 1 h). Moreover, the detection of neurotransmitter glutamate (Glu) or neurotransmitter synthesizing enzymes ChAT and GAD67 in Map2 positive neurons indicated the neurochemical divergence following the neuronal induction (Figure 1 i–k). Using quantitative real‐time polymerase chain reaction (Q‐PCR), the messenger RNA (mRNA) expression level for neuron and oligodendrocyte in SCLT after 14 days of culture resembled most of the phenotypes for terminal differentiation, despite a few immature ones, when compared with the normal adult rat spinal cord (SC, Figure 1 n). 2. 3 Myelination, Vesicle Releasing, and Neuronal Electrophysiological Activities inside the SCLT After 14 days single culture of GMLT or WMLT, or 7 days culture after they were assembled to SCLT, the samples were prepared for morphological and functional assessments as follows. By electron microscopy, thin myelin laminae were observed in the WMLT even without being assembled with GMLT, after 14 days single culturing ( Figure 2 a). It is noteworthy that myelination within the WMLT may be independent from the axon since some myelin sheaths were devoid of axonal profile (Figure 2 a). The neurons inside the 14 days singly cultured GMLT appeared to make many contacts with the neighboring cells (Figure 2 b). Some of these intercellular contacts resembled an immature synapse feature, as indicated by the existence of small and dense vesicles detected inside the presynaptic component; however, postsynaptic density (PSD) in the postsynaptic component was not evident (Figure 2 b1, b2). When the WMLT and the GMLT were assembled into the SCLT, multilayered myelin sheaths enwrapping an axonal profile were observed in the WMLT (Figure 2 c). Moreover, more mature synaptic features, such as an increased number of synaptic vesicles at the axon terminals, and the formation of PSD at the postsynaptic component, were detected in neurons within the GMLT region (Figure 2 d, d1, d2). Figure 2 Myelin and synaptic formation potential of the SCLT. a, b, b1, b2) The singly cultured WMLT or GMLT module presents features of immature myelin sheath (a), or synaptic connections (arrows indicate small and dense vesicles in (b), (b1), (b2)) under the electron microscope, respectively. c) The SCLT shows multilamellar myelin sheaths (arrows in (c)) enwrapping an axonal profile (asterisk in (c)) with (d), (d1), (d2) seemingly more mature features of synapses, like the presynaptic vesicles (arrows in (d1), (d2)) and the postsynaptic membrane density (arrowheads in (d1), (d2)). Scale bars = 200 nm in panels (a), (b1), (b2), (c), (d1), and (d2), 1 µm in panels (b) and (d). To further assess the functionality of neurons in the SCLT, we performed FM1‐43 dye releasing assay, calcium imaging, and patch clamp recordings. We found that SCLT neurons could internalize the FM1‐43 fluorescent dye following an initial dose of high [K + ] and release it after a second dose ( Figure 3 a–c). This is consistent with the fashion of mature neurons in FM dye loading and releasing following high [K + ] stimulation. 14 Ca 2+ oscillations activity could be detected when neurons are loaded with calcium dye. 15 Neurons in the SCLT presented spontaneous calcium surges in individual cells (Figure 3 d1–d6). Transient increase of calcium surges was observed following an excitatory glutamate (Figure 3 e1–e3) or high [K + ] stimulation (Figure 3 f1–f3). The spontaneous calcium surges could be blocked by tetrodotoxin (TTX, Figure 3 g1–g3), suggesting that TTX‐sensitive voltage‐gated sodium channels, which contribute to the forming of action potentials in mature neurons, 16 might exist in the neurons in the SCLT as well. On the other hand, neurons in the singly cultured GMLT did not present vesicle releasing capability following high [K + ] stimulation nor Ca 2+ oscillations activity, suggesting that they may not resemble mature neuronal properties (Figure S2, Supporting Information). In light of these findings, patch clamp experiment was carried out to study the electrophysiological properties of the neurons in the SCLT. We found that the capability of firing action potential strings by the neuron was increased when the culturing time of the SCLT was extended from 7 days (Figure 3 i) to 14 days after the assembly (Figure 3 j). Meanwhile, miniature excitatory postsynaptic currents (mEPSCs) and miniature inhibitory postsynaptic currents (mIPSCs) were also detected in the neurons inside the SCLT, given 14 days of culturing time (Figure 3 k, l). Figure 3 Functional assessment of the SCLT. a, b) Neurons in the GMLT were observed to unload prelabeled FM1‐43 dye (green) following membrane depolarization triggered by high [K + ] stimulation, as shown by (c) the steep drops of fluorescence intensity after the stimulation. d1–d6) Neurons in the SCLT show spontaneous calcium surges (arrows) during Fluo‐4 calcium live cell imaging. Single‐cell tracing of calcium surges reveals that calcium activities in the neurons can be excited by (e1)–(e3) glutamate or (f1)–(f3) high [K + ], and suppressed by (g1)–(g3) tetrodotoxin (TTX). h) Whole‐cell patch clamp recording exhibited the increased firing of action potentials of neurons from the culturing time (i) 7 days to (j) 14 days after the SCLT were assembled. k, l) High‐frequency miniature excitatory postsynaptic current (mEPSC in (k)) or miniature inhibitory postsynaptic current (mIPSC in (l)) were detected in neurons in the SCLT. Scale bars = 10 µm in panels (a), (b), and (d1)–(d6), 100 s in panels (e1)–(g3). 2. 4 Structural Integration between the SCLT and Organotypic Dorsal Root Ganglion (DRG) or Muscle Cells We then investigated the potential of SCLT to establish synaptic connections with peripheral nerve cells or muscular junctions by organotypic cocultures with DRG or muscle cells, respectively ( Figure 4 a). To better illustrate the relationship between the SCLT and the organotypic DRG, we have designed the experiment as follows: 1) DRGs (derived from GFP transgenic SD rats) + SCLTs (from wild‐type SD rats). This part was to study the innervations of DRGs into the SCLTs (Figure 4 b–d). 2) SCLTs (GFP) + DRGs (wild type). This part was to study the interactions between nerve fibers from DRGs and SCLTs (Figure 4 e). 3) SCLTs (with the WMLTs from GFP and the GMLTs from wild type) + DRGs (wild type). This part was to study the interactions between nerve fibers and myelin (Figure 4 f, g). Figure 4 The SCLT establishes connections with DRG or muscle cells. a) A schematic diagram showing a SCLT establishes connections with the organotypic cultured DRG or muscle cells. b) GFP donor–derived DRG extended NF positive neurites into the c, d) SCLT. e) The neurons in GFP donor–derived GMLT region form close contacts with DRG (from GFP negative donor), with PSD95 expression at the contact sites. f, g) GFP donor derived WMLT of a SCLT‐expressed MBP (arrows) wrapping NF‐positive DRG neurites (GFP negative in (f)) or neurites (GFP negative in (g)) extending from the neurons in the GMLT region. h) The outgrowing NF positive neurites from GFP positive SCLT form close contacts with α‐BT‐expressing muscle cells with bouton‐like enlargement (arrows) at the terminal. Scale bars = 500 µm in panel (b), 50 µm in panels (c) and (d), 20 µm in panels (e)–(h). Following 7 days of coculturing, DRG neurites (GFP positive, Figure 4 b) were observed in the SCLT block (Figure 4 c, d). The neurites from the DRG neurons (GFP negative) were in contacts with nerve fibers emanated from neurons of the SCLT (GFP positive). The expression of postsynaptic marker PSD95 in the latter may indicate the potential of synapse forming between the two components (Figure 4 e). Additionally, NF positive nerve fibers extending from the DRG (GFP negative) or GMLT (GFP negative) were well ensheathed by MBP positive myelin‐forming oligodendrocytes (GFP positive, arrows) inside the WMLT region (Figure 4 f, g). On the other hand, muscle cells differentiated from mouse myoblasts (C2Cl2) appeared to attract the growth of nerve fibers from neurons in the SCLT to grow and make contacts with them. Confocal microscopy suggested that NF positive nerve fibers extending from the SCLT (from GFP transgenic SD rats) established axonal bouton–like contacts (arrows) with α‐bungarotoxin (α‐BT) positive motor endplate–like structures in the desmin positive muscle cells (Figure 4 h). 2. 5 Transplantation of SCLT Improved Hind Limb Motor Function Immediately following the transected SCI with 2 mm spinal cord tissue removed in the adult SD rats, the SCLT or the gelatin sponge scaffold (SF) was implanted into the injury gap. Two months after the transplantation, the rats in the SCLT group showed significant hind limb motor function and electrophysiological presentation improvement compared with those in the control groups (the SF group and the SCI group). The body weight support capability of the hind limbs during immobile posture for rats in the SCLT group was noticeably better than those in the other two groups ( Figure 5 a1–a3), suggesting a better preservation of muscular strength. Additionally, compared with those from the other two groups, rats from the SCLT group exhibited more frequent weight‐bearing stepping during mobile courses, such as grid climbing or locomotion in the open field (Figure 5 b1–b3, c1–c3). Basso, Beattie, and Bresnahan (BBB) open‐field locomotor test was used to quantify the hind limb motor function for rats in the SCLT, SF, and SCI groups ( n = 8 in each group). The results showed a distinctively improved motor function for rats in the SCLT group than those in the other two groups beginning at the 3rd week after SCLT transplantation. The BBB score for rats of the SCLT group was above 9 points at 8 weeks, indicating there was weight‐bearing locomotion (Figure 5 d). As a widely used clinical evaluation, the peak and the latency of the cortical motor evoked potentials (CMEPs) are interpreted conventionally as major parameters reflecting the number of excited axons and the conduction velocity of the nerve, respectively. In the SCLT group, increased amplitude and shortened latency of cortical motor evoked potential (EP) were observed, suggesting a better repair for the motor pathway (Figure 5 e–g). Figure 5 Behavior and electrophysiological improvement following the SCLT implantation and the integration of SCLT with the host spinal cord. a–d) Rats in the SCLT presented an overall improvement of hind limb motor function in the glass cube observation (a1), the inclined grid climbing test (b1), and the open field observation (c1), when compared with the Scaffold (SF) group (a2), (b2), (c2) or the transection spinal cord injury (SCI) group (a3), (b3), (c3). d) BBB scoring showed a significant increase in the hind limb motor function following the SCLT implantation, asterisks indicate statistical significance when compared with the SCI group ( *p < 0. 05 in (d)) and hash symbols indicate significant difference when compared with the SF group ( # p < 0. 05 in (d)). e–g) Rats in the SCLT group have increased amplitudes and shortened latency of the CMEPs, resembling more similarities to those in the Normal (Nor) group (* p < 0. 05, when compared with the Nor group; # p < 0. 05 when compared with the SCLT group; & p < 0. 05 when compared with the SF group). Asterisks indicate statistical significance when compared with the Nor group. h) GFP donor derived WMLT of a SCLT integrated with the host spinal cord 8 weeks after implantation. i, j) GFP positive donor cells expressed MBP and enwraped NF positive regenerating nerve fibers in the injury/graft site of spinal cord. k) Immunoelectron microscopy (IEM) showed that myelin sheath (stained by DAB, arrows) formed by GFP donor cell enwrapped an axon (asterisk). Host myelin sheath did not show DAB staining (arrowheads). l) GFP donor derived GMLT of a SCLT integrated with the host spinal cord 8 weeks after implantation. m) The GFP and NF double‐positive nerve fibers were surrounded by MBP positive cells. n) IEM showed that GFP positive axons (DAB labeled, asterisks) were enwrapped by myelin sheathes. Scale bars = 500 µm in panels (h) and (l), 10 µm in panels (i) and (j), 20 µm in panel (m), 200 nm in panels (k)–(n). Within 8 weeks of observation time, rats did not show any sign of distress, such as significant body weight loss, aggression, or vocalization when touched, or porphyrin staining following the implantation of SCLT. 2. 6 SCLT Implant Conferred Targeted Structural Repair in Injured Spinal Cord Eight weeks after transplantation, the collagen scaffold degraded, leaving with no identifiable debris in the injury/graft site as detected by H&E staining (Figure S3, Supporting Information). The SCLT implant was endowed good histocompatibility and tissue repair as evident by significantly reduced cavity formation in the injured spinal cord when compared with the SF and SCI groups (Figure S3, Supporting Information). GFP donor cells were used to separately construct the WMLT module or the GMLT module of a SCLT, so as to evaluate the structural integration between the host tissue and the two components of the implant two months after transplantation. In the present study, we did not observe any significant donor cell migration outside of the implant. The fluorescence signals emitted from the GFP donor–derived WMLT were mainly detected at the peripheral region of the injury/graft of spinal cord (Figure 5 h). The majority of the donor cells in this region maintained MBP expression (Figure 5 I, j and Figure S4a, Supporting Information) and formed sheaths enwrapping the NF positive axons (Figure 5 i, j). Immunoelectron microscopy (IEM) showed that the GFP immunopositive sheaths (labeled by 3, 3′‐diaminobenzidine (DAB), arrows in Figure 5 k) presented a myelin sheath feature enwrapping the axon (asterisk in Figure 5 k). Taken together, the WMLT component of the SCLT showed robust capability of myelinating the nerve fibers at the injury/graft site of spinal cord, representing an expected white matter repair after SCI. The fluorescence signals for the GFP donor–derived GMLT concentrated at the central region of the injury/graft (Figure 5 l). The majority of the donor cells in the GMLT region of a SCLT were Map2 positive neurons or GFAP positive astrocytes (Figure S4b, Supporting Information). Similar to the neurotransmitter profile in vitro, the donor cells in the GMLT region exhibited the expression of GAD67, Glu, and ChAT when transplanted into the injured spinal cord (Figure S4c–f, Supporting Information). The nerve fibers outgrowing from the donor neurons (GFP positive), together with the host regenerating ones (GFP negative), helped increase the innervation of the injury/graft site of spinal cord. Samples derived from the implanted SCLT showed that most of the nerve fibers of donor neurons made close contacts with MBP positive sheaths (Figure 5 m). IEM confirmed that GFP immunopositive axons (DAB labeled, asterisks in Figure 5 n) were enwrapped by myelin sheaths. The results suggested that the GMLT component of the SCLT helped partially restore the loss of neuronal population in the injured area after SCI. Thus, it can be concluded that transplantation of the SCLT has helped repair the white matter and the gray matter damage simultaneously two months after the SCI. 2. 7 The Donor Neurons Structurally Integrated with the Host Neural Circuits The 5‐HT positive nerve fibers descending from the brainstem severed after the transection injury were observed to regenerate and penetrate through the injury/graft site of spinal cord to areas several millimeters caudal to the injury/graft site in the SCLT group ( Figure 6 a–c). On closer examination, the 5‐HT positive nerve fibers longitudinally penetrated the WMLT region (GFP negative zone, Figure 6 d), suggesting the re‐establishment of passageway to the long descending supraspinal fibers. Some of the nerve fibers made contacts with the cells in the GMLT region, in a way similar to their segmental innervation in the physiological condition (GFP positive cells, Figure 6 d). IEM confirmed the structural contacts between a 5‐HT positive nerve fiber (nanogold labeled, superimposed with light purple in Figure 6 e, e1) and the adjacent donor cell (DAB positive, asterisk in Figure 6 e, e1). In addition, the regeneration of sensory nerve fibers was assessed by calcitonin gene‐related peptide (CGRP) immunoreactivity ( Figure 7 a). CGRP positive nerve fibers were widely detected in areas rostral (Figure 7 b) and caudal (Figure 7 c) to/in the injury/graft site in the SCLT group (Figure 7 d). A large number of these nerve fibers made contacts with GFP/PSD95 double‐positive donor cells, indicating the potential of synapse forming. IEM showed many cell contacts established between the donor cells (DAB labeled, asterisks in Figure 7 e) and the surrounding host CGRP positive cells (nanogold labeled, superimposed in light purple in Figure 7 e). Some of the contacts displayed the features of a typical synapse such as the presence of round agranular vesicles (arrows in Figure 7 f, g) in the presynpatic component (i. e. , the CGRP positive nerve fiber terminal) and the focal membrane thickening in the postsynaptic component (GFP immunopositive donor cell labeled by DAB, Figure 7 f, g). Figure 6 The donor neurons in the SCLT integrate with descending 5‐HT nerve fibers. a) A longitudinal section of spinal cord containing GFP donor derived GMLT of a SCLT. b, c) 5‐HT positive descending nerve fibers passed through the injury/graft site of spinal cord and extended several millimeters into the caudal area to the injury/graft site 8 weeks after SCLT implantation (c). d, d1) Most 5‐HT nerve fibers traveled longitudinally through the WMLT region, while some of them formed close contacts with the donor cells in the GMLT (arrows). e, e1) IEM showed that a 5‐HT nerve fiber (superimposed in light purple and labeled by nanogold particles, arrowheads in (e1)) formed close connection with GFP positive donor cells (DAB labeled, asterisks) and the host cell (superimposed in yellow). Scale bars = 500 µm in panel (a), 40 µm in panels (b)–(d), 20 µm in panel (d1), 1 µm in panel (e), 200 nm in panel (e1). Figure 7 The donor neurons in the SCLT integrate with the ascending CGRP positive nerve fibers. a) A longitudinal section of spinal cord showed massive CGRP positive nerve fiber regeneration 8 weeks after the SCLT implantation. A large portion of CGRP nerve fibers traversed the WMLT region (arrowheads). b, c) CGRP signals can be observed in areas rostral (b) and caudal (c) to the injury/graft site of spinal cord. d) The CGRP positive nerve fibers (arrows) formed close contacts with PSD95 expressing GFP donor cells in the GMLT region of a SCLT. e–g) IEM showed that nanogold particle labeled CGRP nerve fibers (superimposed in light purple, arrowheads in (g) and (f)) closely contacted the transplanted cells (DAB labeled, asterisks). Scale bars = 500 µm in panel (a), 40 µm in panels (b) and (c), 20 µm in panel (d), 0. 5 µm in panel (e), 200 nm in panels (g) and (f). Quantitative analysis showed that the number of NF, 5‐HT, and CGRP positive nerve fibers in the rostral and caudal areas to/in the injury site of spinal cord was significantly higher in the SCLT group, when compared with that in the SF or SCI groups ( p < 0. 01, n = 5 in each group; Figure S5, Supporting Information). 3 Discussion Tissue engineering–based regenerative medicine has now entered a new era of applying implantable prebuilt tissue or organoid to repair tissue loss. 17 Buoyed by the success in constructing functional neural networks from adult stem cells in vitro to repair SCI in our previous studies, 11, 18 the present study has constructed a SCLT simulating the major cellular composition of the gray matter as well as the white matter of the spinal cord. When transplanted into the injured spinal cord, the SCLT integrated with the host tissue and exhibited targeted repair to the gray and white matter damage after injury. Thus, the modular assembly of SCLT may serve as a potential construct to help the structural and functional restoration after SCI. Hydrogels are considered to be the most widely used bioscaffold to construct the organoids. [[qv: 5b, 19]] In view of the weak mechanical property of the hydrogel, the volume of the organoid is often limited, and the maintenance of the scaffolding requires complex treatments. 20 This property has made the hydrogel‐based organoid more suitable for an in vitro model, rather than an implant for in vivo treatment. [[qv: 6a, 14]] On the other hand, the collagen sponge is mechanically stronger, and hence is easily tailored into column or ring shape. Additionally, the porous structure provides sufficient space for cell growth and differentiation; meanwhile it facilitates readily exchange of the nutrients and the wastes. As a result, cells residing in the SCLT presented a good viability and were able to survive in vitro up to one month. More importantly, the vast surface of the sponge offers the scaffold with ample space conducive for neuronal differentiation, nerve fiber extension, and establishment of intercellular contacts, including the synapses. The nonimmunogenic property of the collagen sponge makes it a safe biomaterial to use in the central nervous system as approved by the US Food and Drug Administration (FDA). The good mechanical property, along with its cytocompatibility and histocompatibility of the collagen sponge has allowed the construction of an optimal scaffold for SCLT. CNTF plays an essential role in the differentiation and maturation of oligodendrocyte from its precursor OPCs. 21 However, it is critical to predifferentiate NSCs into OPCs before applying CNTF so as to avoid astrocytic differentiation of NSCs as triggered by CNTF effect. 22 Our previous studies focused on using NT‐3 and its receptor TrkC gene modification technique to promote neuronal differentiation of NSCs with synaptic transmission capability. 11, 23 Here, we have adopted the same technique to construct the GMLT. By using this method, a mixture of neurons with the potentials to produce various neurotransmitters was generated, as evidenced by immunocytochemistry detection of GAD67, Glu, and ChAT. Electron microscopy showed synapse profiles forming between two neurons in the GMLT region. Along with the finding from FM1‐43 labeled vesicle releasing assay, our findings suggest that there may be synaptic transmission of impulses via neurotransmitters between two neurons. As further support, electrophysiology assessment of neurons in the GMLT region presented a mature neuron feature in bursting action potentials. Robust membrane potential fluctuations as recorded as EPSCs or IPSCs reflected the dynamic postsynaptic responses to the excitatory or the inhibitory neurotransmitter, respectively. These results indicated that the neurons in the GMLT region of a SCLT were synaptically connected to each other and formed a neuronal network capable of synaptic transmission. The interactions between neurons and glial cells are vital for mutual maturation during development and maintenance of homeostasis. 24 In this study, we also observed a dynamic interaction between the GMLT and the WMLT component when they were assembled into the SCLT. For example, although the singly cultured WMLT module exhibited a potential of forming myelin as evidenced by the detection of MBP immunoreactivity, there was a paucity of multilayered myelin sheath as observed under the electron microscopy, suggesting the structural immaturity of the oligodendrocytes. This might highlight the importance of axonal inputs in guiding the myelination process although the initiation of myelination could be independent from axon. 25 Similarly, neurons in the singly cultured GMLT presented several immature features, as manifested by the synapse ultrastructure. When the SCLT was assembled, remarkably, multilayered myelin sheath was observed with axon being enwrapped in the WMLT region. Indeed, the delicate balance between the inhibitory and promoting molecules derived from axon is believed to be essential to guide the thickness of the myelin sheath. 26 Meanwhile, the oligodendrocyte offers a variety of trophic factors that help the development and survival of neuron. 27 Moreover, myelin aids axon metabolism 28 and integrity. 29 As we observed that focal membrane thickening, namely, the postsynaptic density, was evident at the synaptic contacts between two neurons in the GMLT region after being assembled into the SCLT. Moreover, aggregation of axons, dendrites, synapses containing at least two different types of vesicles, i. e. , the small clear vesicles and the small dense‐core vesicles, and the surrounding glial cell processes, were common in the border area between the GMLT and the WMLT; such a configuration and ultrastructural features resemble the neuropil in the central nervous system (CNS). All this suggests that the assembly of the SCLT represents not only a structural combination of the WMLT module and the GMLT module, but also allows a dynamic integration of function between the two components that enhances the further maturation of each other. Therefore, the SCLT may serve as a platform for study of the interactions between the neurons and the glia cells in vitro. After 14 days of culture, the SCLT shared many similarities to the normal adult spinal cord in terms of gene and protein expression as evaluated by Q‐PCR and Western blotting, respectively. The differences between these two entities lied in the expression level of some mature neuron markers, for example, the expression of SYP, encoding a compositional presynaptic protein, was observed highly expressed in the adult spinal cord at mRNA and protein levels. On the contrary, some immature/developmental markers, such as the tubb3, encoding class III β‐tubulin, whose expression levels were higher in the SCLT. This suggests that, despite the mature oligodendrocyte or neuron phenotypes and the functions observed, the SCTL might retain the potential for further development and maturation after 14 days of culture. This notion is supported by the findings from the coculturing of the SCLT with the organotypic DRG or muscular cells. The structural integration between the SCLT and DRG or muscle cells suggests that, like the developing spinal cord, the SCLT is able to establish new connections with the DRG, which represents the afferent inputs from the peripheral, and with the muscle cells, which may be regarded as the targets of the efferent nerve fibers. The intermediate stage of the SCLT between the immature and the mature stage would render it a promising application value when implanted into the injured spinal cord. This is because it is generally believed the immature cells adapt to the post SCI microenvironment better than the mature ones. [[qv: 8a, 30]] Two months after SCLT transplantation, donor cells, both in WMLT and GMLT, survived well and maintained oligodendroglial or neuronal phenotype similar to that observed in vitro. The strategy of transplantation of in vitro prebuilt SCLT consisting of differentiated cells is believed to increase the safety of donor cells 31 and dodge the uncertainty of cell differentiation amid postinjury microenvironment when compared to direct transplantation of stem or progenitor cells. [[qv: 10a, 32]] Following SCLT transplantation, the overall hind limb motor function was significantly improved, compared with the control groups; in fact, it was superior to that of what we reported before using NSC‐derived neural network transplantation (equivalent to the GMLT transplantation). 12 Consistent with previous observation, the GMLT module of the SCLT was capable of making contacts with the host descending or ascending neural circuits that would potentially serve as “tissue engineering neuronal relay” to conduct signals passing through the injury/graft site of spinal cord. Furthermore, the WMLT module may provide an additional benefit in motor pathway repair. Indeed, the oligodendrocytes in the WMLT region were able to myelinate the nerve fibers in the injury/graft site of spinal cord. This could help repair the white matter loss and presumably contribute to the shortened latency as demonstrated in the present CMEP recording. Another feature worthy of note is that the WMLT region may offer inductive microenvironment for the regeneration of descending supraspinal nerve fibers, for example, exuberant 5‐HT nerve fibers were observed traversing the WMLT region and reinnervating the spinal cord segment caudal to the injury/graft site of spinal cord. 4 Conclusions All in all, the assembly of the SCLT has realized a dynamic interaction and functional integration of the WMLT module and the GMLT module. When transplanted into the injured spinal cord, the WMLT and the GMLT regions of the SCLT exhibited targeted repairing to tissue loss of the white and gray matter after SCI. As a result, improved paralysis hind limb motor function was observed following the SCLT transplantation. Transplantation of the SCLT is expected to provide a novel therapeutic approach for the structural and functional repair in the spinal cord transected completely or missing a segment spinal tissue. In addition, it may be used as an in vitro platform for study of neural development and neuropharmacology. 5 Experimental Section NSCs and OPCs Culture and Identification : NSCs were isolated from SD rats or GFP transgenic SD rats (Osaka University, Osaka, Japan), as described previously. 32 Briefly, rats (1–3 days old) were anesthetized and the whole hippocampus was dissected and dissociated into single cell suspension. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM)/F12 (1:1) supplemented with 1× B27 (Life Technologies, USA) and 20 ng mL −1 basic fibroblast growth factor (bFGF, Life Technologies, USA). Typically, cells were grown as neurospheres in suspension, which were passaged by mechanical dissociation approximately once each week. Nestin immunoreactivity was assessed and confirmed for all the neurospheres. Oligodendrocyte precursor cells (OPCs) were obtained according to the published protocols with slight modifications. 25, 33 Briefly, NSCs at passage 2 were plated on polylysine‐coated culture dishes and cultured with DMEM/F12 containing 10 ng mL −1 platelet‐derived growth factor AA (PDGF‐AA, Life Technologies, USA), 10 ng mL −1 bFGF (Life Technologies, USA), 30 ng mL −1 triiodothyronine (T3, Sigma‐Aldrich), and 1% fetal bovine serum (FBS, Gibco) for 3 days. The cells were passaged when they reached 90% confluency, and were purified by differential digestion/adhesion technique. The cells were tested for the expression of NG2 (EMD Millipore). The purity of the OPCs used in all subsequent experiments was ≈80% NG2 positive. Modular Assembly of SCLT : NSCs were transfected with lentivirus carrying a puromycin resistance gene and a neurotrophin‐3 (NT‐3) coding sequence (pLent‐EF1a‐NT‐3‐Flag‐CMV‐GFP‐P2A‐Puro) or its receptor TrkC (pLent‐EF1a‐TrkC‐Flag‐CMV‐GFP‐P2A‐Puro) gene (Vigene Biosciences). For OPC transfection, CNTF gene was delivered via lentivirus (pLent‐EF1a‐CNTF‐Flag‐CMV‐GFP‐P2A‐Puro, Vigene Biosciences). After 48 h of incubation, fresh medium containing 2 µg mL −1 puromycin was added for another 48 h to select the transfected cells. For GMLT construction, a collagen sponge (BIOT Biology) scaffold was tailored into a short column (2 mm in diameter and 2 mm in length). A mixed cell suspension in 20 µL containing equal amount of NT‐3‐NSCs and TrkC‐NSCs (2 × 10 5 for each scaffold) was dripped into the prewet scaffold. The GMLT module was maintained in DMEM/F12 (1:1) with 1× B27 supplement (Life Technologies, USA) and 1% FBS. For WMLT construction, a collagen sponge ring (3 mm in outer diameter, 2. 0 mm in inner diameter, and 2 mm in length) was made by a biopsy punch. A total of 2 × 10 5 CNTF–OPCs in 20 µL culture medium were seeded to the scaffold. The WMLT module was maintained in DMEM/F12 (1:1) with 1× B27 supplement and 1% FBS. The WMLT and GMLT modules were cultured separately for 7 days before they were assembled into the SCLT. The SCLT was maintained in DMEM/F12 (1:1) with 1× B27 supplement and 5% FBS for another 7 days. Organotypic Coculturing of SCLT with DRG or Muscular Cells : DRGs for organotypic culture were established from 1 day old newborn SD rats. Dissection of DRGs from the spine column was done in ice‐cold DMEM/F12. The surrounding connective tissue and the adherent dura mater were removed. Three DRG blocks were placed on the top of each SCLT and were cultured for 7 days in DMEM/F12 medium containing 1× B27 supplement and 5% FBS with medium change every 2 days. Mouse myoblasts (C2Cl2, a gift from Prof. H. Liao, Department of Human Anatomy, Southern Medical University, Guangzhou, China) were cultured in DMEM/F12 with 15% FBS at a density of 20 cells mm −2. When cells reached 80% confluency, they were induced to differentiate into muscle cells with DMEM/F12 plus 2% horse serum (Gibico). On day 7, SCLTs were placed above the muscular cells. The co‐culture system was maintained in SCLT culture medium. Vibration of the culture dish was minimized to allow extension of SCLT neurites to the muscle cells. Western Blotting and Quantitative Real‐Time PCR Analysis : After 14 days of culturing, scaffolds ( n = 3) from SCLT, GMLT, WMLT, or normal adult rat SC (positive control) were used for intracellular and extracellular protein extraction. Equal amount of proteins were loaded onto a 10% polyacrylamide gel. Proteins were separated by electrophoresis, followed by transferring them onto a polyvinylidene fluoride (PVDF) membrane. The membrane was incubated with primary antibodies at 4 °C overnight, followed by incubation with horseradish peroxidase (HRP)‐conjugated secondary antibodies. The bands were detected with an enhanced chemiluminescence (ECL) Western blotting substrate kit. The amount of glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) protein was used as loading control. Q‐PCR was used to compare the gene expression between SCLT (the SCLT group) and normal spinal cord (the SC group). Total cellular RNA in the SCLT or SC groups was extracted with TRIZOL reagent (Takara Bio Inc. , Otsu, Japan). After reverse transcriptional reaction into complementary DNA (cDNA) (DRR037S, Takara Bio Inc. , Otsu, Japan), Q‐PCR was carried out using Takara SYBR Premix Ex TaqTM mixture (DRR041S, Takara Bio Inc. , Otsu, Japan), in a Bio‐Rad iCycler iQ5 PCR machine (BioRad). GAPDH was used as internal reference. Reaction protocol was as follows; 95 °C for 30 s, and 40 repeated cycles of 95 °C for 5 s, followed by 60 °C for 31 s. Each experiment was repeated for 3 times with a duplicate in each time. Primers for Q‐PCR are listed in Table S1 (Supporting Information). Whole‐Cell Patch Clamp : The whole‐cell configuration was used to record the electrical activities of neurons in SCLT with a HEKA EPC amplifier 10 (HEKA Inc. ). The results were analyzed by Patchmaster software (HEKA Inc. ). Signals were filtered at 1 kHz and sampled at 5 kHz. The external solution contains 140 × 10 −3 m NaCl, 5 × 10 −3 m KCl, 2 × 10 −3 m CaCl 2, 1 × 10 −3 m MgCl 2, 10 × 10 −3 m 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES), and 10 × 10 −3 m glucose (320 mOsm, pH set to 7. 3 with Tris‐base). The patch electrodes had a resistance of 3–5 MW, when filled with pipette solution, containing 140 × 10 −3 m CsCl, 2 × 10 −3 m MgCl 2, 4 × 10 −3 m ethylene glycol‐bis(β‐aminoethyl ether)‐ N, N, N ′, N ′‐tetraacetic acid (EGTA), 0. 4 × 10 −3 m CaCl 2, 10 × 10 −3 m HEPES, 2 × 10 −3 m magnesium adenosine triphosphate (Mg‐ATP), and 0. 1 × 10 −3 m guanosine triphosphate (GTP). The pH was adjusted to 7. 2 with Tris‐base, and the osmolarity was adjusted to 280–300 mOsm with sucrose. Electrophysiological recordings were performed at room temperature (22–24 °C). Mini postsynaptic currents (mPSCs) were counted and analyzed using Fitmaster (HEKA Inc. ). Live‐Cell FM1‐43 and Calcium Imaging : FM‐143 [( N ‐3‐triethylammonmpropyl)‐4‐(4‐(dibutylamino) styryl)] (Life Technologies, USA) was used to evaluate synaptic vesicle releasing SCLT after high [K + ] (50 × 10 −3 m ) stimulation. 34 The first dose of high [K + ] solution stimulated the recycling of endocytic synaptic vesicles that contained the FM1‐43. After 3 times of rinsing (15–20 min for each) with culture medium in the absence of FM1‐43, the endocytic/exocytotic activities were brought down to the basal level and the nonspecific labeling of cytoplasmic membrane was also eliminated while the synaptic vesicles kept the labeling of FM1‐43. After this, the second dose of high [K + ] solution unloaded FM‐143 from the cells by inducing the depolarization process. Release of FM1‐43 labeled synaptic vesicles was captured by LSM780 confocal laser scanning system (Zeiss). A control experiment to assess nonspecific bleaching of fluorescence was performed simultaneously. Live‐cell calcium imaging was performed using a LSM780 confocal laser scanning system (Zeiss), equipped with temperature and CO 2 control module. For calcium imaging, Fluo‐4 (Life Technologies, USA) was prepared according to the manufacturer's instruction and was applied in the SCLT for 60 min in darkness. The supernatant was then removed and the SCLT was washed with a bath solution for 3 times. Imaging was performed at 494 nm excitation. Photos were taken every 20 s for 100 frames. [[qv: 5b, 15]] Data analysis of calcium imaging was performed using HCimage Live 4. 2. 0. Region of interest (ROI) was manually selected and the mean fluorescence for each ROI was calculated at each time frame. Changes in fluorescence was enumerated as follows: Δ F / F = ( F − F basal ))/ F background, where F basal was the lowest mean fluorescence value of one frame and F background was the average mean fluorescence of all frames. Neuropharmacological drugs (100 × 10 −3 m glutamate, 50 × 10 −3 m KCl, and 1 × 10 −3 m TTX) were delivered by perfusion and kept for a 10 min incubation time before washing out. Surgery and SCLT Transplantation : Three days before surgery, all animals were given cyclosporine A, intraperitoneally. Adult female SD rats (220–250 g, supplied by the Experimental Animal Center of Sun Yat‐sen University) were used in this study. Following laminectomy, a 2 mm cord segment including the associated spinal roots was completely removed at the T10 spinal cord level. After hemostasis was achieved, SCLT or collagen sponge scaffolds (the SCLT group or the SF group) were used to fill up the gap. The transection group (the SCI group) had 2 mm cord segment removed without filling in any biomaterials. Cyclosporine A (Novartis) was administrated once every day for two months. All experimental protocols and animal handling procedures were approved by the Animal Care and Use Committee of Sun Yat‐sen University, and were in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Assessment of Locomotor Performance : Hind limb function of the rats was assessed weekly after surgery, using the BBB open‐field locomotor test, 35 the glass cube locomotor function observation (the rat was put in a glass cube with 30 cm in height), and the inclined‐grid climbing test. 36 BBB test was used to quantitatively evaluate the voluntary movements and the body weight support capability ( n = 8 in the SCLT, SF, and SCI groups, respectively). Glass cube locomotor observation was used to generally observe the hind limb standing. Inclined‐grid climbing test was used to assess the accuracy of foot placement and coordination during locomotion. Two independent investigators who were blind to treatments determined the BBB scores. Electrophysiology : At the end of the experiment, EP were recorded as described previously to assess the functional status of motor signal conduction ( n = 8 in each group). Basically, following general anesthesia and exposure of the sciatic nerves and sensorimotor cortex (SMC), the electrodes (BL‐420E Data Acquisition Analysis System for Life Science, Taimeng, Chengdou, China) were connected to the sciatic nerve and SMC, respectively. The CMEPs were calibrated first, and then recorded as per the standardized protocols. 37 Perfusion and Tissue Preparation : All rats were deeply anesthetized with 1% pentobarbital sodium (50 mg kg −1, intraperitoneally (i. p. )) and intracardically perfused with physiological saline containing 0. 002% NaNO 2 and 0. 002% heparin, followed by 4% paraformaldehyde. After perfusion, the spinal cord was dissected, postfixed overnight in the same fixative, and dehydrated in 30% sucrose/phosphate buffered saline (PBS). Longitudinal sections of the selected spinal cord segments were cut at 25 µm thickness using a cryostat. All sections were stored at −30 °C until further processing. Immunofluorescence Staining : Specific proteins were detected using immunofluorescence staining as described in the previous publications. 32, 37 Briefly, the sections were incubated with primary antibodies mixed in 0. 3% Triton X‐100 at 4 °C overnight, followed by the incubation with secondary antibodies. The slides were then examined under a fluorescence microscope. A summary of antibodies used is provided in Table S2 (Supporting Information). Ultrastructure Observations : For SEM, scaffolds with cells were first washed 3 times with PBS, fixed in 2. 5% glutaraldehyde for 90 min, dehydrated with a series of graded ethanol, and then freeze‐dried for 2 days. The dried samples were coated with gold and examined under a scanning electron microscope (Philips XL30 FEG). For transmission electron microscopy (TEM), scaffolds were fixed with 2. 5% glutaraldehyde at 4 °C for 1 h and postfixed with 1% osmic acid for 1 h. They were dehydrated through graded ethanol and embedded in Epon812 overnight, followed by polymerization at 60 °C for 48 h. Semithin sections (2 µm thickness) were cut on a Leica RM2065 microtome, mounted on glass slides, stained with toluidine blue (5%, in a borax solution) and mounted using neutral balsam before observation. Ultrathin sections (100 nm thickness) were cut, double stained with lead citrate and uranyl acetate, and examined under an electron microscope (Philips CM 10). For IEM, rats were transcardially perfused with 0. 1 mol L −1 of sodium phosphate buffer containing 187. 5 units per 100 mL of heparin, followed by perfusion with 4% paraformaldehyde, 0. 1% glutaraldehyde, and 15% saturated picric acid. The dissected spinal cord was postfixed overnight at 4 °C in fresh fixative and subsequently cut into 50 µm sagittal sections on a vibratome. To improve the penetration of antibodies, vibratome sections were transferred into cryprotectant solution containing 25% sucrose and 10% glycerol in 0. 1 m PBS overnight at 4 °C, followed by a quick freeze–thaw in liquid nitrogen 3 times. After washing with PBS, the sections were treated for 1 h with 20% goat serum (Tris buffer, pH 7. 4) to block nonspecific binding of the antibody. Sections were first incubated with primary antibodies (anti‐GFP combined with anti‐5‐HT or anti‐CGRP, n = 3) in 2% normal goat serum solution at 4 °C for 24 h, then incubated with secondary antibodies overnight at 4 °C, and postfixed in 1% glutaraldehyde for 10 min. The sections were detected by SABC‐DAB Kit and gold enhanced with Gold EnhanceTM EM Plus Kit (NanoProbe 2114, USA), osmicated, dehydrated, and embedded in Epon812. The Epon blocks were sectioned and examined under the electron microscope (Philips CM 10). Morphological Quantification : For in vitro quantification of immunopositive cells, one in every ten of the whole series of sections from each scaffold was selected ( n = 5 in each group). After immunostained with the respective markers, five areas (0. 7 mm × 0. 5 mm including four corners and one center) for each of the sections were chosen. The percentage of immunopositive cells were calculated by counting the total number of immunopositive cells. The numerical value obtained was then divided by the total number of GFP positive cells. For in vivo quantification, areas that were 300 µm rostral or caudal to the injury/graft site, or that in the injury/graft site of each of the horizontal sections were scrutinized. One in every ten sections from each animal was processed; a total of five sections per rat were analyzed ( n = 5 in each group). Three 0. 7 mm × 0. 5 mm areas for each of the sections cut through the rostral or caudal to the injury/graft site along with those for each of the sections cut through the injury/graft site were chosen. The percentage of immunopositive cells was calculated by counting the total number of immunopositive cells. The value obtained was then divided by the total number of GFP positive cells. For quantification of NF, 5‐HT, and CGRP positive nerve fibers, the immunopositive fibers with a length greater than 20 µm in the selected fields were counted. Statistical Analysis : All statistical analyses were performed using the statistical software SPSS13. 0. Data were presented as means ± standard deviation (S. D. ). When three sets of data were compared, one‐way analysis of variance (ANOVA) with a least significant difference (LSD)‐ t (equal variance assumed) or Dunnett's T3 (equal variance not assumed) was performed. A statistically significant difference was accepted at p < 0. 05. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800361
| 2,018
|
Advanced Science
|
Nanoscale Coatings for Ultralow Dose BMP‐2‐Driven Regeneration of Critical‐Sized Bone Defects
|
Abstract While new biomaterials for regenerative therapies are being reported in the literature, clinical translation is slow. Some existing regenerative approaches rely on high doses of growth factors, such as bone morphogenetic protein‐2 (BMP‐2) in bone regeneration, which can cause serious side effects. An ultralow‐dose growth factor technology is described yielding high bioactivity based on a simple polymer, poly(ethyl acrylate) (PEA), and mechanisms to drive stem cell differentiation and bone regeneration in a critical‐sized murine defect model with translation to a clinical veterinary setting are reported. This material‐based technology triggers spontaneous fibronectin organization and stimulates growth factor signalling, enabling synergistic integrin and BMP‐2 receptor activation in mesenchymal stem cells. To translate this technology, plasma‐polymerized PEA is used on 2D and 3D substrates to enhance cell signalling in vitro, showing the complete healing of a critical‐sized bone injury in mice in vivo. Efficacy is demonstrated in a Münsterländer dog with a nonhealing humerus fracture, establishing the clinical translation of advanced ultralow‐dose growth factor treatment.
|
1 Introduction Active lifestyles in the young, obesity, diabetes, and osteoporosis in the elderly are driving an increase in the occurrence of traumatic injury. In 2010, >250 000 people in the USA over the age of 65 were hospitalized for hip fractures, 1 with an increasing number of patients experiencing nonhealing (known as nonunion) fractures. 2 Rates of nonunions from 2 to 5% have been recently suggested, 3, 4 with important variations with type of fracture, age, gender, and risk factors. Nonunion fractures account for the most inpatient hospital days of all musculoskeletal injuries at a cost of $9. 8 billion annually in the USA. 5 With an ageing population, >1 million fractures are expected to occur annually by 2050. 5 Thus, there is an urgent need to develop new bone repair therapies that are safe, cost‐effective, and efficacious. Currently, therapeutic approaches for the treatment of nonunion fractures include growth factor (GF)‐based treatments, 6, 7 stem cell therapies, 7 and magnetic field treatments. 7, 8 GFs, in particular bone morphogenetic proteins (BMPs), are commonly used in fracture treatments. However, they are not without limitations, including partial efficacy 9 ; uncontrolled and nonlocalized delivery, which can produce potentially harmful, nonspecific, systemic side effects 10, 11, 12 ; and high cost due to the large doses of GFs used. 9, 13 BMP‐2 has been used for over a decade in bone regenerative therapies, loaded into collagen sponges at high concentrations (1. 5 mg mL −1 ). 14 Despite the US Food and Drug Administration (FDA) releasing a notification of the life‐threatening complications associated with the therapeutic use of high‐dose BMP‐2 for bone repair, including ectopic bone formation, neurological problems, and high risk of cancer, the use of GF therapies continues. 15 Nevertheless, new advanced technologies are being developed to replace existing GF‐based treatments by exploiting the bioactive properties of materials. 16, 17, 18, 19, 20, 21, 22 Still, the translation of materials‐based platforms from in vitro and in vivo lab testing through to clinical applications remains limited due to overengineering, use of novel chemistry unlikely to gain regulatory approval, and/or complex grafting of biologicals. Structural scaffolds based on FDA‐approved materials can potentially be used in conjunction with bioactive polymer coatings. One such bioactive polymer is poly(ethyl acrylate) (PEA). 17 We have previously shown that while fibronectin (FN) typically adsorbs onto polymers in globular conformation, PEA promotes the spontaneous organization of FN into physiological‐like networks. When assembled, these networks present both integrin‐binding (III 9–10 ) and GF‐binding (III 12–14 ) regions to cells. 17, 23 Importantly, the GF‐binding region can stably present GFs, such as BMP‐2, at ultralow doses. 17 However, the application of bioactive polymer coatings, such as PEA, to 3D scaffolds and their clinical translation have been hindered by several limitations. First, the current predominant technique for coating a surface with PEA is spin coating, 17, 24 which is difficult to achieve on a 3D scaffold. Second, PEA is nonbiodegradable. One solution to this problem would be to coat a biodegradable polymeric scaffold with a layer of PEA that is thin enough (<10s of nm) for the body to metabolize after the scaffold has degraded. 25, 26 Standard polymer coating techniques such as spin coating typically yield a PEA coating with a thickness of ≈1 µm on a flat surface. This is too thick for clinical purposes, even if spin coating was to be successfully implemented on a 3D scaffold. Spin coating also requires PEA to be dissolved in an organic solvent and thus traces of harmful solvents can remain. We therefore sought to develop a solvent‐free coating method 24 with nanoscale coating depth that could be applied to a variety of 3D applications, such as scaffolds and grafts. To achieve this, we present a simple, robust, and translational approach to promoting bone regeneration in nonunion bone defects. We report the development of a plasma–polymerization strategy for coating 3D scaffolds with thin layers of PEA. We show that this highly facile method can be used to coat 2D surfaces and complex 3D scaffolds, including 3D‐printed scaffolds and microparticles, and that this coating can assemble FN in the form of biomimetic networks. We investigate whether these FN networks support synergistic interactions between integrins and BMP‐2 receptors that promote osteogenesis in vitro. Importantly, this nanoscale coating can be applied to 3D biodegradable materials and we investigate their use as an implant to advance bone healing in a mouse model of a critical‐sized, nonhealing bone defect. We also report the first veterinary application of this technology to treat a 2‐year‐old Münsterländer dog with a nonhealing and infected fracture of the humerus. We treat this nonunion using PEA‐coated decellularized bone chips seeking to promote rapid and robust healing of this bone defect and prevent limb amputation. Together, we introduce a next‐generation GF‐based technology for effective bone repair in vivo that is likely safer in a clinical setting given the ultralow dose of BMP‐2 used. 2 Results 2. 1 Plasma‐Polymerized PEA Coatings Promote FN Assembly and Effective BMP‐2 Presentation The plasma‐based polymerization of ethyl acrylate into plasma‐polymerized PEA (pPEA) coatings is illustrated in Figure 1 a. We used spin‐coated PEA (a nontranslatable approach, as explained above), denoted as SC‐PEA, as a positive control. When FN is coated on SC‐PEA (open FN chain, illustrated in Figure 1 b), BMP‐2 binding, crosstalk with integrins, and osteogenesis are promoted, as previously demonstrated. 17 The chemistry of pPEA was confirmed using X‐ray photoelectron spectroscopy (XPS). The spectrum of pPEA shows a prominent disappearance of the C—O and C—O—C moieties at 286. 6 and 533. 5 eV, respectively, which differs from that of SC‐PEA (Figure 1 c, d). 27 It is known that plasma polymerization leads to partial loss of functional groups and crosslinking. 24 By fine‐tuning the parameters of the plasma system, such as chamber pressure, power, and polymerization time, we were able to coat pPEA onto glass at a rate of as low as 10 nm min −1 (Figure S1a, Supporting Information). The thickness of the deposited pPEA coatings (≈300 nm) guarantees that the XPS spectra are due to pPEA and not to the underlying substrate, as the sensitivity of XPS is ≈10 nm. 24 The stability of the films was confirmed by monitoring the water contact angle (WCA) after deposition and up to 14 d. Figure S1b in the Supporting Information shows that WCAs remain unchanged at different time points after deposition (Figure S1b, Supporting Information). We measured the strength of interaction between pPEA coatings and the underlying substrates using a normalized pull‐off test that resulted in an adhesion strength of 7. 14 ± 0. 35 MPa (Figure S1c, Supporting Information). Changes in plasma settings resulted in coatings of different thicknesses but with similar XPS spectra, indicating that the properties of the pPEA deposited at different conditions were comparable (Figure S2, Supporting Information). Figure 1 Physicochemical characterisation of plasma PEA coatings, FN, and BMP‐2 adsorption on pPEA. a) Schematic representation of a custom‐made plasma polymerization chamber. b) FN structure, showing its three domain types (I, II, III) and functions. Domain III region (III 9–10 ) contains the RGD (Arg‐Gly‐Asp) sequence that facilitates cell adhesion via integrin binding, and region III 12–14 binds various GFs, including BMP‐2. c) XPS characterization of SC‐PEA and pPEA. High‐resolution C1s and O1s spectra are shown with fitted components in colored dotted lines. d) Chemical structure of PEA, with labelled carbon and oxygen atoms corresponding to components in panel (c). AFM phase images of FN adsorbed for 10 min on e) SC‐PEA and f) pPEA. Thin fibrillar networks were observed on SC‐PEA, whereas thick, dense networks were observed on pPEA. g) Surface density of FN adsorbed at different concentrations onto SC‐PEA and pPEA for 1 h. h) Relative exposure of integrin‐binding and GF‐binding domains on FN adsorbed on different surfaces, measured using ELISA. i) Relative adsorption of BMP‐2 on FN‐coated surfaces, measured using ELISA. j) Cumulative BMP‐2 release from surfaces coated with pPEA, FN, and BMP‐2 during 2 weeks. k) AFM height images of BMP‐2 labelled with gold nanoparticles on SC‐PEA and pPEA coated with FN. White arrows indicate gold nanoparticles showing BMP‐2 distribution. l) Schematic representation of immunogold assay for GF detection. All data are presented as mean ± SD, n = 3, one‐way ANOVA with Tukey's test for multiple comparisons. * p < 0. 05, ns = not statistically significant. Atomic force microscopy (AFM) was performed to visualize the organization of FN on both SC‐PEA and pPEA (Figure 1 e, f). Poly(methyl acrylate) (PMA), which has one fewer carbon in its side chain, is a good control for PEA as it results in FN adsorption in globular conformation. 28, 29, 30 (Data on interactions with PMA are presented in Figure S3a, b in the Supporting Information. ) FN assembly into nanonetworks on SC‐PEA (Figure 1 e) has been well characterized previously. 28, 29 It occurs when FN is adsorbed onto SC‐PEA at a sufficiently high concentration (>10 µg mL −1 ) so that when FN unfolds, it can contact neighbouring FN molecules. This allows nanonetworks to assemble via interactions among FN matrix‐binding regions. AFM showed that pPEA coatings do indeed induce this FN assembly but with denser, thicker morphology compared with those formed on SC‐PEA (Figure 1 f and Figure S3c, Supporting Information), although the measured surface densities of FN adsorbed on pPEA and SC‐PEA were similar (Figure 1 g and Figure S3d, Supporting Information). Note that spin‐coated PMA (Figure S3e, Supporting Information) did not induce fibrillogenesis (Figure S3a, b, Supporting Information). To assess the activity of FN on pPEA, we quantified the availability of the integrin‐binding and GF‐binding regions using enzyme‐linked immunosorbent assay (ELISA) with monoclonal antibodies against the respective regions (Figure 1 h). We looked at the FN(III 9–10 ) domain, which includes the RGD (Arg‐Gly‐Asp) sequence for integrin binding, 30, 31, 32 and the FN(III 12–14 ) domain, 17 which binds a variety of GF families including BMP‐2. 16, 23 The availability of both domains was significantly higher on pPEA coatings than on SC‐PEA surfaces ( p < 0. 05). Given that we normalized the graphs to total surface FN (Figure 1 g and Figure S3f, Supporting Information), we propose that the reported differences are not due to the total amount of FN adsorbed on the surface (as the densities are similar between SC‐PEA and pPEA), but instead to the enhanced availability of specific binding sites on FN with pPEA. Next, we used ELISA to quantify the amount of BMP‐2 adsorbed on FN‐coated surfaces (on SC‐PEA and pPEA coatings) from a solution at a concentration of 50 ng mL −1 BMP‐2 (Figure 1 i); this is the concentration used in subsequent biological experiments. We also assessed the stability of the adsorbed BMP‐2 on FN‐coated pPEA and quantify the release from the surface as a function of time. More than 90% of the adsorbed BMP‐2 remains in the surface after 14 d. (Figure 1 j). In line with the above data on GF binding‐site availability, more BMP‐2 was adsorbed on FN coated on pPEA than on SC‐PEA (Figure 1 i and Figure S4, Supporting Information). To assess the spatial distribution of BMP‐2 molecules on the FN nanonetworks, immunogold staining was performed on SC‐PEA‐ and pPEA‐treated surfaces coated with different concentrations of FN and exposed to 25 ng mL −1 BMP‐2 (Figure 1 k). The localization of BMP‐2 is indicated by bright dots on the AFM images, corresponding to gold nanoparticles with a diameter of ≈15 nm conjugated to primary antibodies specific for BMP‐2 (Figure 1 l). Gold nanoparticles indicative of BMP‐2 (Figure 1 k) interact specifically with individual nanofibers of FN on both SC‐PEA and pPEA and confirm the specific localization of BMP‐2 on FN. (Note that lower concentrations of FN and BMP‐2 than those used in cell experiments were used here to clearly depict this localization. ) 2. 2 Plasma‐Polymerized PEA Coatings Drive Synergistic Signaling and hMSC Osteogenesis To investigate whether pPEA‐induced BMP‐2 presentation drives cell adhesion, enhances synergistic integrin/GF signalling, and is osteoinductive at low GF concentration, we cultured human mesenchymal stem cells (hMSCs) on pPEA surfaces first coated with FN and then 50 ng mL −1 BMP‐2. First, cell attachment and spreading after 24 h were significantly enhanced on pPEA‐coated surfaces with FN compared to cells on SC‐PEA‐coated surfaces with FN (Figure S5, Supporting Information), as evidenced by the higher number of attached cells, greater degree of spreading, and enhanced focal adhesion (FA) formation. This correlated well with the better availability of FN(III 9–10 ) domains on pPEA (Figure 1 h). Next, synergistic integrin and GF signalling were evaluated, initially by assessing whether BMP receptor 1A (BMPR1A) colocalizes with focal adhesions. Vinculin (which stains focal adhesions) immunofluorescence colocalized with BMPR1A immunofluorescence in single cells ( Figure 2 a), indicating that BMPR1A and integrins are in close proximity, which would enable crosstalk to occur between the adhesion and GF pathways (Figure 2 b). (The staining of specific integrins is shown in Figure S6 in the Supporting Information. ) Next, the phosphorylation of small mothers against decapentaplegics (SMAD) and focal adhesion kinase (FAK) was examined in order to investigate BMP‐2‐related signalling and focal adhesion‐related signalling pathways, respectively. 33, 34, 35 The expression of phosphorylated SMADs (pSMAD) and phosphorylated FAK (pFAK) in hMSCs after 1 h of culture on the test versus control surfaces is shown in Figure 2 c. Figure 2 hMSC signalling and differentiation. a) Colocalization assay of BMP receptor 1A (BMPR1A, green) and FAs (red). White arrows on the merged image show areas of colocalization in yellow. Scale bar = 20 µm. b) A schematic representation showing that synergistic signalling between integrin and GF receptors can occur when the integrin‐binding (III 9–10 ) and GF‐binding (III 12–14 ) domains of FN are in close proximity. c) Western blotting of pSMAD 1/5/9 and pFAK, expressed by hMSCs after 1 h in culture on SC‐PEA and pPEA, with and without FN and BMP‐2. Quantified blots to show relative expressions of d) pSMAD and e) pFAK, both normalized using total protein amount, from hMSCs after 1 h in culture on SC‐PEA and pPEA, with and without FN and BMP‐2. hMSCs cultured with soluble BMP‐2 and on glass alone were used as controls. Data are presented as mean ± SD, n = 3, one‐way ANOVA with Tukey's test for multiple comparisons. * p < 0. 05. f) Normalized ALP expression in hMSCs after 12 d in culture on SC‐PEA and pPEA surfaces, with and without FN and BMP‐2, from a fluorescent ALP assay. Data are presented as mean ± SD, n = 3, one‐way ANOVA with Tukey's test for multiple comparisons. * p < 0. 05. Immunofluorescent labelling of g) OPN and h) OCN in hMSCs cultured on SC‐PEA and pPEA, with and without FN and BMP‐2, for 21 d. Phalloidin stains actin cytoskeleton in green and DAPI stains nuclei in blue. Scale bar = 50 µm. SMADs 1, 5, and 9 can be phosphorylated by BMPR1A, leading to their nuclear translocation and the activation of RUNX2 (runt‐related transcription factor 2, the osteogenic master transcription factor). 33, 36 Quantification of pSMAD expression by western blotting, normalized by total protein amount (Figure 2 d and Figure S7, Supporting Information), showed a significant upregulation of pSMAD on pPEA‐coated surfaces with FN and BMP‐2 (pPEA + FN + BMP‐2) compared with pPEA without BMP‐2 (pPEA + FN); a similar trend was observed on SC‐coated surfaces (Figure 2 d). In both cases, the presentation of BMP‐2 on PEA resulted in enhanced SMAD signalling compared to the soluble administration of the GF alone (soluble BMP‐2, Figure 2 d). We also examined the expression of pFAK in hMSCs, which was normalized using total protein amount (Figure 2 e and Figure S7, Supporting Information). pFAK is involved in integrin‐related signalling 35, 37 and was significantly upregulated ( p < 0. 05) in hMSCs cultured on surfaces coated with pPEA + FN + BMP‐2 relative to those cultured without BMP‐2 (pPEA + FN) or with soluble BMP‐2 alone (again, a similar trend was observed for the control material, SC‐PEA). We note that in hMSCs cultured with soluble BMP‐2, pFAK is induced to the same level as it is in cells cultured on pPEA without BMP‐2, which suggests that increased pFAK expression is focal adhesion related and occurs independently of BMP‐2. From these findings, we propose that enhanced synergistic adhesion and BMP‐2 signalling on pPEA + FN + BMP‐2 occurs as a consequence of the simultaneous occupancy of integrins and BMP‐2 receptors. 35, 37 We next investigated the potential of this synergistic signalling cascade, driven by PEA‐induced assembly of FN, for osteogenesis. Alkaline phosphatase (ALP) expression was measured after 12 d of hMSC culture (Figure 2 f). ALP, which plays a role in bone mineralisation and is an early marker of osteogenic differentiation, 38 was expressed by hMSCs cultured on pPEA + FN + BMP‐2 at a significantly higher level than it was by cells cultured on pPEA + FN without BMP‐2 coating. Long‐term culture was also performed for 21 d and two osteogenesis‐related proteins, osteopontin (OPN) and osteocalcin (OCN), were visualized by immunofluorescence (Figure 2 g, h and Figure S8, Supporting Information). Both OPN and OCN (shown in red) were more highly expressed in hMSCs cultured on pPEA + FN + BMP‐2 than they were by cells cultured on pPEA‐FN surfaces without BMP‐2. Moreover, comparing SC‐PEA and pPEA, it is clear that when both FN and BMP‐2 were present, OPN was expressed more distinctly on the pPEA‐coated surfaces. Von Kossa staining further revealed the presence of mineralized deposits on the surfaces coated with BMP‐2 (Figure S9, Supporting Information). As proof of concept that pPEA can be employed as a coating for 3D biomaterials, further in vitro analysis was performed on 3D polycaprolactone (PCL) scaffolds (cylinders of 5 mm diameter, 800 µm high). The scaffolds were fabricated by 3D printing (to generate filaments of 200 µm in diameter, with 500 µm separation, layered at 90° to the last layer, see Figure S10a in the Supporting Information). Once fabricated, the scaffolds were coated with pPEA for 30 min before adsorbing FN to generate a fibrillar network (inset, Figure S10a, Supporting Information). FN adsorption was confirmed by ELISA (Figure S10b, Supporting Information). The expression of BMPR‐2, RUNX2, and osterix (OSX) was then quantified by quantitative polymerase chain reaction (qPCR) (Table 1, Supporting Information). qPCR results showed that by day 7 in culture, OSX expression (but not that of BMPR‐2 and RUNX2 at this time point) was significantly increased in hMSCs cultured on PCL + pPEA + FN + BMP‐2, compared to cells cultured on PCL + pPEA or on PCL + pPEA + FN (Figure S10c, Supporting Information). It is likely that BMPR‐2 and RUNX2 upregulation had occurred at an earlier time point (previous reports have shown that BMPR‐2 levels are highest at day 3, and RUNX2 levels highest at day 5 of culture, while OSX expression occurs later at days 7–11 39 ). With respect to RUNX2, a significant increase in its phosphorylation was seen in PCL + pPEA + FN + BMP‐2 as early as day 5 in culture, relative to other controls (PCL, PLC + pPEA, PCL + pPEA + FN) (Figure S10d, Supporting Information). After longer‐term culture (21 d), mature bone nodules could be observed by Alizarin red staining on the scaffolds coated with PCL + pPEA + FN + BMP‐2 (Figure S10e, Supporting Information). 2. 3 Plasma‐Polymerized PEA Coatings Drive Regeneration in a Murine Nonhealing Radial Bone Defect Model To investigate the translational potential of the osteogenic pPEA coatings, we used an adult mouse model of a critical‐sized, nonhealing radial bone defect. 17 This bone repair model has significant advantages: i) the 2. 5 mm defect does not spontaneously heal, providing a rigorous critical‐sized model, ii) it allows for simple in vivo imaging approaches (e. g. , microcomputed tomography, µCT), and iii) the ulna provides sufficient stabilization of the defect and no fixation plates/hardware are required. This simplifies the surgical procedure and reduces the risk of infection, a major advantage over the rat calvaria and segmental femur defect models. 21 Polyimide tube implants of 4 mm were used in this model as they form a biocompatible but nonbioactive scaffold that fits well over the defect (Figure S11, Supporting Information). These were coated with degradable PCL by solvent casting, creating a layer of an FDA‐approved biomaterial within the tube. This was followed by polymerization of a very thin layer of pPEA (estimated at <100 nm) to cover the underlying PCL polymer. Then, FN or FN + BMP‐2 was adsorbed on the cylindrical polymer surface. BMP‐2 was used at low concentration, i. e. , ≈15 ng of BMP‐2 on the wall of the coated tubes (surface density of 100 ± 8 ng cm −2, as measured using ELISA). 17 This BMP‐2 concentration is at least 100% lower than that used in advanced materials systems previously tested in murine models that are based on integrin‐specific polyethylene glycol hydrogels loaded with BMP‐2. 21 Note that even if humans and rodents do not metabolize biologics at the same rate, the amount of BMP‐2 used was ≈300 fold lower than that of the clinical gold standard. 14, 17 Implant tubes without BMP‐2 coated only with the PCL layer or with PCL + pPEA were used as negative controls. FN and BMP‐2 adsorbed on solvent‐casted PEA were used as positive controls based on previous data. 17 Bone formation was evaluated by X‐ray ( Figure 3 a) and by 3D µCT reconstructions, which displayed the total length of the radius scanned (Figure 3 b). The 3D µCT images showed increased bone growth along the defect in the presence of BMP‐2, to the extent where the bone gap is fully bridged (Figure 3 b). These results show that higher levels of bone regeneration occurred in the presence of BMP‐2, even when it was only applied to the walls of the implant tube, as performed here. Most importantly, similar levels of bone regeneration are seen on samples coated with solvent‐casted PEA (positive control) and on the test material pPEA, which indicates enhanced in vivo effects of pPEA, even with the use of low doses of BMP‐2 (Figure 3 b). Further, limb function was restored in all treated mice. The quantification of bone formation from µCT images shows enhanced bone surface density in the presence of low‐dose BMP‐2 (both on PEA and pPEA) compared to conditions where no BMP‐2 was present in the implant (Figure 3 c), albeit the same bone volume was found in all other conditions tested (Figure 3 c). The discrepancy between bone volume and bone surface density between groups is a reflection of how bone is organized within the defect, allowing or not full bridging. A way to measure the architecture of new bone is by analyzing the 3D information provided by µCT and quantify the total surface of bone formed. This is particularly relevant for areas of the implant functionalized with BMP2, as a new layer of bone grows in contact with them. These results suggest the different organizations of bone, leading to the full bridging of defects in the pPEA + FN + BMP‐2 condition (Figure 3 b). Figure 3 Bone regeneration in a murine model of a critical‐sized radial bone defect with low doses of BMP‐2. a) X‐ray images at 0, 4, and 8 weeks after surgery. b) 3D reconstructions from the µCT images showing the radius in the area of the defect, 8 weeks after introduction of the PCL‐pPEA implant (with or without FN and BMP‐2). c) Quantification of the volume and specific surface of new bone. Data are presented as mean ± SD, minimum n = 3. Two‐tailed t‐test was used to analyze data. * p < 0. 1. d–h) Hematoxylin‐Safranin O‐fast green staining of histological sections in the area of the defect. The tissue is organized in structures resembling bone marrow (rounded white structures in panels (d) and (e)) versus fibroblast‐like morphology (extended and aligned) in the center of the defect in panels (f)–(h). Arrow points to red staining that indicates cartilage formation in panel (h). The histological sections in the area of the defect support the findings obtained from the µCT images. Implant tubes (solvent‐casted PEA and pPEA) coated with FN and BMP‐2 induced higher levels of new bone formation, relative to controls (Figure 3 d, e). Both proximal and distal bone extended into the implant, and structures that resemble the bone marrow were observed along the entire defect (Figure 3 d, e). However, a larger gap between proximal and distal bone was observed in the absence of BMP‐2 (Figure 3 f, g), with less bone healing observed for PCL‐only control implants (Figure 3 h). Moreover, fibroblast‐like structures were observed filling the gap of the defect in PCL‐only control implants, with cartilage tissue present (red staining in Figure 3 h), indicating the differentiation of cells toward the chondrogenic lineage. Taken together, the in vitro and in vivo data indicate that the synergistic integrin/BMP‐2 receptor osteoinductive response is generated by pPEA‐coated scaffolds. From these findings, we propose this polymerization approach as a potential method for coating material implants in 2D and 3D scaffolds. 2. 4 Veterinary Case Study: pPEA Supports Healing of a Nonunion Humeral Fracture in a Dog To demonstrate the translatable potential of the pPEA system, we performed a first trial of this approach in a veterinary case study. A 2‐year‐old female Münsterländer dog ( Figure 4 a and Figure S12 and Video S1, Supporting Information) was presented to the Small Animal Hospital, University of Glasgow, in July 2016 for the management of a comminuted fracture of the diaphysis of the right humerus, sustained when she was hit by a car (Figure 4 b). The fracture was stabilized surgically using a standard open reduction and internal fixation technique (Figure 4 c). The dog made a slow early recovery, retaining a significant degree of lameness and developed a discharging sinus on the medial aspect of the distal humerus around 2 months postoperatively, indicative of an infection at the fracture site. This was confirmed when Staphylococcus spp. was cultured from this sinus. The infection was treated with antibiotics (potentiated amoxicillin, 500 mg by mouth three times daily) for 6 weeks based on the sensitivity profile. Despite this, the dog remained persistently lame. Five months after surgery, there was no convincing evidence of fracture healing. Radiographically, areas of osteolysis and implant loosening indicated the presence of osteomyelitis and delayed union (Figure 4 d). Figure 4 Humeral fracture healing in a dog treated with bone chips coated with pPEA, FN, and BMP‐2. a) A schematic representation of a 2‐year‐old female Münsterländer dog, showing a comminuted fracture of the diaphysis of her right humerus, sustained when she was hit by a car. b) Radiograph of the comminuted fracture of the diaphysis of the right humerus. c) Radiograph showing the surgical stabilisation of the fracture using standard open reduction and internal fixation technique. d) Five months after c), osteolysis at the fracture site was evident, consistent with osteomyelitis and delayed union. e) Restabilization of the fracture using an external skeletal fixator. f) Six weeks after (e), there was no evidence of fracture healing. g) Radiograph showing fracture nonunion, 8 months after the injury. h) Prior to surgery, decellularized bone chips were coated with pPEA to form pPEA‐chips, which were subsequently coated with FN and BMP‐2. i) Bone marrow was harvested from the humeral head on the left side and mixed with 5 cc of coated pPEA‐chips. j) Bone plates and screws were used to stabilize the fracture, and the combined graft materials were placed within the fracture gap. k) Postoperative radiograph shows the fracture gap filled with graft. l) Evidence of fracture union 7 weeks after surgery performed in panels (g) to (j). We removed the implants on the basis that a bacterial biofilm was likely covering them and perpetuating the infection. The fracture was restabilized with an external skeletal fixator (Figure 4 e) and antibiotic treatment continued (potentiated amoxicillin, 500 mg by mouth twice daily) for a further 2 weeks, as Staphylococcus spp. cultured directly from the metal implants showed the same sensitivity profile as previously shown. Radiographs taken 6 weeks after revision surgery showed no evidence of fracture healing and significant osteolysis (Figure 4 f). We were concerned that this fracture was developing an atrophic nonunion, and radiographs taken 10 weeks postrevision (8 months following the original injury) were consistent with this (Figure 4 g). Because of the poor prognosis, we considered limb amputation but instead elected to revise the fracture fixation once more using allograft microparticles coated with pPEA on which FN and BMP‐2 were adsorbed at 50 µg mL −1. This concentration is 30‐fold lower than that used in clinical standards (BMP‐2 loaded in a collagen sponge at a concentration of 1. 5 mg mL −1 ). 14 We note that while this is a higher concentration than that used in our in vivo mouse bone defect model, it is still only 10% of the dose (0. 5 mg mL −1 ) typically used in complex fractures in dogs in veterinary applications. 40 Decellularized bone chips (2–4 mm) were coated with pPEA using plasma polymerization to form pPEA‐chips and subsequently coated with FN and BMP‐2 (Figure 4 h). A standard medial approach was made to the humerus. Approximately 1 cm of the nonhealing ends of the fracture was excised, along with all soft issue within the fracture gap. Bone marrow was harvested from the humeral head on the left side and mixed with 5 cc of pPEA‐chips (Figure 4 i). The fracture site was thoroughly debrided and then stabilized using two bone plates and screws and the combined graft materials were placed within the fracture gap (an ≈2 cm defect) (Figure 4 j); soft tissues were closed as per routine protocol. Postoperative radiographs showed the fracture gap filled with graft and the appropriate placement of the implant (Figure 4 k). The dog made a good recovery from the procedure. There was a moderate degree of postoperative swelling of the affected limb, which resolved within a week, but no other adverse effects were observed. The tissue removed from the fracture site was submitted for bacterial culture, which showed the persistence of a Staphylococcus infection. Antibiotics were prescribed (cephalexin 300 mg by mouth twice daily) and continued for 4 months postoperatively. The dog's use of the affected limb steadily improved, and radiographs taken 7 weeks after surgery were consistent with fracture union (Figure 4 l). By 6 months postoperatively, the dog had resumed normal exercise, and there was no clinical evidence of recurring infection. 3 Discussion Notwithstanding significant off‐target effects, BMPs have been used in patients for the management of nonunion fractures where traditional approaches have failed. 14 Here, we present a simple and facile technology that allows materials with complex geometries to be coated with ultralow and thus safer levels of BMP‐2 for use in bone regeneration‐promoting therapies to treat critical‐sized, nonunion bone defects. Plasma polymerization presents a solvent‐free coating method that can be used for a variety of 3D applications, such as scaffolds and grafts. This new approach produces thin PEA coatings (layers of 30–200 nm are possible) which are strongly attached to the underlying substrate (≈7. 10 MPa – Figure S1, Supporting Information) that can assemble FN in the form of biomimetic nanonetworks (Figure 1 f), rather than as a globular structure. This unfolding of the FN molecule via nanonetwork formation exposes sites responsible for FN function, such as integrin‐binding (III 9–10 ) and GF‐binding (III 12–14 ) regions. 17, 23, 30, 31 In in vitro assays, synergistic signalling was observed in hMSCs cultured on surfaces coated with pPEA + FN + BMP‐2, together with the upregulation of pSMADs (Figure 2 c, d), indicating the onset of canonical BMP‐2 signalling mediated by BMP receptors, as shown previously. 17 Increased pFAK with BMP‐2 stimulation from the surface but not in soluble form indicates integrin ligation. 35 Previous research has shown that intact integrin function is crucial for BMP‐2 activity and that inhibition of FAK activation blocks SMAD signalling activation by BMP‐2. 37 The upregulation of pFAK on BMP‐2‐coated surfaces is shown in Figure 2 c, e, signifying that the combined effects of pSMAD and pFAK signalling could be a consequence of the potency of BMP‐2 in enhancing downstream osteogenic activities. Previous evidence has demonstrated that crosstalk occurs between GF receptors and integrins, which plays a crucial role in regulating mechanotransduction in the extracellular matrix. 17, 37, 41, 42, 43 The colocalization of BMP‐2 receptors with various types of integrins has been demonstrated previously, 17, 37, 39 and this colocalization is confirmed in our study (Figure 2 a). We show that pPEA + FN + BMP‐2 promotes bone regeneration in a critical‐sized defect in the mouse radius. The effect is similar (full bridging of the defect and histological evidence of new bone growth) to the use of bulk PEA deposited using solvent‐casting. 17 We found that pPEA was not effective in promoting regeneration if only coated with FN, which reveals the importance of the minimal amount of BMP‐2 used (≈15 ng) (Figure 3 b). It is noteworthy that the use of pPEA and pPEA + FN in the mouse model came closer to bridging the bone defect than did the use of PCL alone (Figure 3 b), indicating that passive adsorption of adhesion proteins and GFs from the mouse host might also occur onto the scaffold. Nevertheless, our data indicate that engineering adhesion in this way can only enhance osteogenesis to a certain extent. To regenerate bone and repair large defects in a reasonable time, synergistic signalling between integrins and GF receptors is required in vivo. Further, we translate the technology to demonstrate the application and efficacy of this approach in promoting bone regeneration in a veterinary case of an infected nonunion fracture. In a previously reported recent case study, dogs suffering from nonunion long‐bone fractures were treated with a compression‐resistant matrix soaked in a solution of 0. 5 mg mL −1 BMP‐2. 40 Here, we report the successful repair of a nonunion fracture in a dog using decellularized bone chips treated with a novel pPEA + FN + BMP‐2 coating using only 10% of the BMP dose previously reported. 40 The bone chips themselves facilitated healing of the defect only through osteoconduction, as this kind of graft material has no inherent osteoinductive capacity. 44 Fracture nonunion in our hospital is uncommon, and these cases are highly variable, so it was not possible to treat a control animal with similar injury, infection, and compromised bone healing. Our conclusions should therefore not be overstated; however, based on clinical experience, we feel that the rapid and robust bone healing observed would not have occurred had the surgery been supplemented using bone chips alone. We note that the treatment of infected critical‐sized defects, to which we demonstrate a regenerative approach, is a major clinical challenge in orthopedic surgery. Other strategies to present GFs from a material surface, including protein engineering techniques, the use of peptides that bind heparin and GFs, 45 and the use of layer‐by‐layer technologies, 41 are reported to be more effective than the soluble administration of GFs. 46 However, these approaches do not exploit the synergy between integrin and GF receptors to enhance accelerated regeneration. 47 4 Conclusion To conclude, we show here that the delivery of BMP‐2 at ultralow doses in synergy with integrin‐binding regions of FN can overcome the current obstacles facing orthopedic treatments, such as the regeneration of infected bone. This system has the potential to be developed into a safe, efficient, and cost‐effective therapeutic approach for delivering BMP‐2 at low doses to stimulate bone regeneration and demonstrates the clinical potential of new biomaterials. The study represents a first translation of a polymer/biological interface that targets reproducible molecular control of cell phenotype through multiple cell receptor targeting. This strategy can be potentially used in other applications where the use of growth factors is important to achieve cellular effects, such as cardiovascular tissue engineering, regeneration of osteochondral defects, and nerve repair. 5 Experimental Section Plasma Polymerization : Circular 12 mm diameter microscopy cover glasses were sonicated for 25 min in ethanol and treated in air plasma for 5 min before being exposed to monomer plasma. Plasma polymerization of the ethyl acrylate monomer (E9706, Sigma‐Aldrich) onto the substrates was carried out in a custom‐made capacitively coupled plasma installation for low‐pressure plasma in a 15‐L T‐shaped reactor made of borosilicate and stainless steel end plates sealed with Viton O‐rings. Vacuum was produced by a rotary pump or a scroll pump (both BOC Edwards), with operating experiment pressures for the monomer plasma from 0. 09 to 0. 45 mbar. The plasma was initiated via two capacitively coupled copper band ring electrodes situated outside of the reactor chamber and connected to a radio frequency power supply (Coaxial Power System Ltd. ) that works at 13. 56 MHz up to 300 W. The monomer pressure was controlled via speedivalves (BOC Edwards) and monitored with a pirani gauge (Kurt J. Lesker). Control samples were prepared by spin‐coating PEA dissolved in toluene at 4% wt/wt. PEA sheets were prepared by radical polymerization of ethyl acrylate solution using 1% benzoin as photoinitiator. A 100 µL droplet of the polymer solution was placed on the glass coverslip, and spin coating was operated at a speed of 3000 rpm and an acceleration of 3000 rpm s −1 for 30 s. The PEA‐coated coverslips were then subjected to solvent extraction under vacuum at 60 °C for at least 2 h to ensure that no traces of solvent remained on the surface. XPS : X‐ray photoelectron spectra were obtained at the National EPSRC XPS Users' Service (NEXUS) at Newcastle University, an EPSRC Mid‐Range Facility. XPS was performed using a K‐Alpha apparatus (Thermo Scientific), with a microfocused monochromatic Al Kα source (X‐ray energy = 1486. 6 eV) at a voltage of 12 kV, current of 3 mA, power of 36 W, and spot size of 400 µm × 800 µm. Spectra analysis and curve fitting were performed using CasaXPS software version 2. 3. 16. AFM : Human FN (1918‐FN, R&D Systems) was prepared at 20 µg mL −1 in Dulbecco's phosphate‐buffered saline (DPBS) and a 200 µL droplet was placed on the surface of glass coverslips treated with either spin‐coated PEA (SC‐PEA) or pPEA. The protein was allowed to adsorb for 10 min and the remaining liquid was thereafter removed from the surface. The surface was then washed twice with DPBS and once with Milli‐Q water and dried under a stream of nitrogen before AFM imaging. A JPK Nanowizard 4 (JPK Instruments) apparatus was used for imaging in tapping mode using antimony‐doped Si cantilevers with a nominal resonant frequency of 75 kHz (MPP‐21 220, Bruker). The phase signal was set to 0 at a frequency 5–10% lower than the resonant frequency. Height and phase images were acquired from each scan. The JPK Data Processing software version 5 was used for image analysis. Protein Adsorption Assays : ELISA was performed to assess the exposure of specific domains on the FN molecule. After substrates had been coated with 20 µg mL −1 FN in DPBS for 1 h, they were blocked for 30 min with 1% bovine serum albumin (BSA, A7979, Sigma‐Aldrich) in DPBS. Next, antibodies for the FN(III 9–10 ) domain (HFN7. 1, mouse monoclonal, 1:330, Developmental Studies Hybridoma) or FN(III 12–14 ) domain (P5F3, sc‐18 827, mouse monoclonal, 1:30, Santa Cruz Biotechnology) were added onto the surfaces and incubated for 1 h. The surfaces were thereafter washed 3 × 5 min with 0. 05% Tween 20 in DPBS (PBST). Then, a horseradish peroxidase (HRP)‐conjugated antimouse antibody (626 520, 1:200, ThermoFisher) was added onto the surface and incubated for 1 h in the dark, following by washing for 3 × 5 min with PBST. A substrate solution (DY999, R&D Systems) was then added onto the surfaces and the samples were incubated in the dark for 20 min, followed by the addition of a stop solution (DY994, R&D Systems). The absorbance of the colored solution was read at 450 and 540 nm and the data were used to determine the relative exposure of the FN domains. All procedures were performed at room temperature. ELISA was also performed to quantify BMP‐2 adsorption. After FN was coated at 20 µg mL −1 for 1 h, the surfaces were blocked with 1% BSA in DPBS for 30 min, followed by adsorption of BMP‐2 (355‐BM, R&D Systems) in DPBS for 1 h. The surfaces were washed and blocked again with 1% BSA for 30 min. Next, primary antibodies against BMP‐2 (ab14933, rabbit polyclonal, 1:2000, Abcam) were added onto the surfaces and incubated for 1 h. After washing for 3 × 5 min with PBST, a biotinylated antirabbit antibody (BA‐1100, 1:10 000, Vector Laboratories) was added onto the surfaces and incubated for 1 h. The samples were then washed again for 3 × 5 min with PBST, and a streptavidin‐HRP solution (DY998, R&D Systems) was added and incubated for 20 min. After a final 3 × 5 min wash with PBST, a substrate solution was added onto the surfaces and the samples were incubated in the dark for 20 min, followed by the addition of a stop solution. The absorbance of the colored solution was read at 450 and 540 nm and the data were used to determine the relative adsorption of BMP‐2. All procedures were performed at room temperature. Quantification of BMP‐2 Release from pPEA‐Coated Surfaces : To determine the degree of BMP‐2 release from pPEA‐FN‐coated surfaces, BMP‐2‐loaded samples were incubated at 37 °C for 1 h. At 14 different time points (2 h and 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, 13, and 14 d), supernatant was collected and samples were replenished with fresh buffer. The amount of BMP‐2 in the supernatant was measured using sandwich ELISA (DY355, R&D Systems) following the manufacturer's instructions. Briefly, ELISA plates were coated with the capture antibody and then blocked with BSA for 1 h. After appropriately diluted supernatants were added, bound BMP‐2 was detected with biotinylated antihuman BMP‐2. Streptavidin‐conjugated HRP was then added to the plates. Enzyme substrate (tetramethylbenzidine and peroxide) was treated for 20 min, and the reaction was stopped by adding an acidic solution. Absorbance was measured at 450 nm with wavelength correction at 570 nm. The standard curve was calculated using a four‐parameter logistic (4‐PL) curve fit. The amount of BMP‐2 was calculated from a standard curve based on known concentrations of BMP‐2. Experiments were performed with three replicates of each time point. Immunogold Characterization of FN‐GF Interactions : For immunogold characterization using AFM, immunogold staining was performed on SC‐PEA‐ and pPEA‐modified glass surfaces coated with 3 µg mL −1 FN and subsequently with 25 ng mL −1 BMP‐2. The samples were washed three times with DPBS and fixed with 4% formaldehyde. They were then incubated with primary antibodies against human BMP‐2 (MAB3551, mouse monoclonal, 1:50, R&D Systems) for 1 h at room temperature. After washing the samples three times with 0. 5% Tween‐20 in DPBS (wash buffer), an antimouse immunogold reagent conjugated to 15 nm gold nanoparticles (815. 022, 1:50, Aurion) was added to the samples for 1 h at room temperature. The samples were then rinsed with wash buffer and fixed again with 4% formaldehyde and imaged using AFM. Cell Culture : hMSCs (PromoCell) were used for the experiments. Cells were cultured in Dulbecco's Modified Eagle's Medium (D5671, Sigma‐Aldrich) with 10% fetal bovine serum (10 500‐064, ThermoFisher), 0. 4% penicillin/streptavidin (P0781, Sigma‐Aldrich), 1 × nonessential amino acids (11 140‐035, ThermoFisher), and an antibiotic mix consisting of 1 × 10 −3 m sodium pyruvate (S8638, Sigma‐Aldrich), 1 × 10 −3 m L‐glutamate (G7513, Sigma‐Aldrich), and 0. 5% Fungizone (15 290‐018, ThermoFisher). Cells were incubated in a 5% humidified CO 2 atmosphere at 37°C. Cells were used at passages 2–3 and experiments were performed in triplicates. For all experiments, cells were cultured initially without serum for 3 h. Colocalization Studies : Cells were washed with DPBS and fixed with 4% formaldehyde solution at 4 °C for 15 min. Cells were then permeabilized with a solution of 0. 1% Triton X‐100 in DPBS at 4 °C for 10 min. A 1% BSA solution was added and the cells were incubated at room temperature for 30 min to block nonspecific binding. After blocking, primary antibodies (antivinculin, V9131, mouse monoclonal, 1:400, Sigma‐Aldrich; anti‐BMPR1A, PA5‐11 856, rabbit polyclonal, 1:50, ThermoFisher) were added to the cells and incubated at 37 °C for 1 h. Cells were then washed with 0. 5% Tween‐20 in PBS (PBST) for 3 × 5 min. Thereafter, AlexaFluor 488 donkey α‐mouse (A21202, 1:200, ThermoFisher) and Cy3‐conjugated goat α‐rabbit (711‐165‐152, 1:100, Jackson ImmunoResearch) were added to the cells and incubated at 37 °C for 1 h, followed by PBST washing for 3 × 5 min. The nuclei of the cells were stained using VectaShield‐4′, 6‐diamidino‐2‐phenylindole (DAPI) (H‐1200, Vector Laboratories), while samples were mounted on glass slides for fluorescence microscopy. Western Blot : Cell lysates were harvested after culturing on various surfaces for 1 h, using radioimmunoprecipitation (RIPA) buffer supplemented with protease inhibitors (S8830, Sigma‐Aldrich) and phosphatase inhibitors (78 427, ThermoFisher). Western blotting was performed using the same amount of protein in each sample in denaturing conditions for pSMAD 1/5/9 (12656T, rabbit monoclonal, 1:1000, Cell Signaling) and pFAK (05‐1140, mouse monoclonal, 1:500, EMD Millipore). After membranes were washed in Tris‐buffered saline with Tween 20 for 6 × 5 min, antirabbit (7074, 1:2000, Cell Signaling) and antimouse (NA931VS, 1:10 000, Abcam) secondary antibodies were added, and a chemiluminescent HRP substrate for immunodetection (Millipore, WBKLS0500) was used before X‐ray detection. Protein expression was quantified using ImageJ and normalized with total protein amount measured by BCA (23 225, ThermoFisher). ALP Assay : Cells grown on coverslips were rinsed twice with DPBS and harvested prior to ALP assay. Each coverslip with attached cells was broken into pieces and transferred to a 1. 5 mL Eppendorf tube. Then, 500 µL of ice‐cold Tris‐HCl was added to each tube. The sample was sonicated twice for 10 s each at 5 W and centrifuged at 10 000 × g for 5 min. The supernatant was transferred to a new tube and the protein concentration of each sample was measured by BCA assay (23 225, ThermoFisher). Then, a sample volume corresponding to 25 µg of protein (adjusted to 25 µL with DPBS) was added to each well of a black 96 well plate, and 100 µL of a fluorescent 4‐methylumbellifeyl phosphate disodium salt (MUP, M8168, Sigma‐Aldrich) substrate solution, made by mixing 500 µL of 1 m NaHCO 3, 2 mL of 50 × 10 −3 m diethanolamine, 7. 5 mL of deionized water, and 34 µL of 25 mg mL −1 MUP stock, was added to each well. After incubating at 37 °C for 1 h in the dark, the fluorescence of the plate was read at an excitation wavelength of 360 nm and an emission wavelength of 465 nm. ALP standards were prepared using serial dilutions of a 10 mU µL −1 ALP solution (P6774, Sigma‐Aldrich) to obtain a calibration curve. Immunofluorescence of Osteogenesis‐Related Markers : Cells were washed with DPBS and fixed with 4% formaldehyde solution at 4 °C for 15 min. Cells were then permeabilized with a solution of 0. 1% Triton X‐100 in DPBS at 4 °C for 10 min. Blocking was performed using 1% BSA at room temperature for 30 min. Primary antibodies (anti‐OPN, sc‐21 742, mouse monoclonal, 1:100, Santa‐Cruz Biotechnology; anti‐OCN, sc‐73 464, mouse monoclonal, 1:100, Santa‐Cruz Biotechnology) were added to the cells and incubated at 37 °C for 1 h. Cells were then washed with PBST for 3 × 5 min. Thereafter, AlexaFluor 488 donkey α‐mouse (A21202, 1:200, ThermoFisher) and Cy3‐conjugated goat α‐rabbit (711‐165‐152, 1:100, Jackson ImmunoResearch) were added to the cells and incubated at 37 °C for 1 h in the dark, followed by PBST washing for 3 × 5 min. The nuclei of the cells were stained using VectaShield‐DAPI, while samples were mounted on glass slides for fluorescence microscopy. Implant Preparation for In Vivo Study of Nonunion Critical‐Sized Bone Defect : Thin, porous polyimide sleeves (Microlumen) were coated by solvent casting of a PCL solution, creating a polymer layer on the walls of the tube, followed by the deposition of a thin layer (<1 µm) of pPEA to cover the underlying PCL polymer. Then, FN or FN + BMP‐2 was adsorbed on the cylindrical polymer surface. Implant tubes without BMP‐2 and tubes coated with the sole polymer layer of PCL or PCL + pPEA were used as controls. Implant tubes covered with solvent‐casted PEA coated with FN and BMP‐2 were used as a positive control, this condition being the one that showed a higher potential in promoting bone regeneration in the preliminary experiment. 17 At least three replicates per group were used in the experiment. It was important to note that implant tubes coated was used with the polymer of interest instead of 3D scaffolds, due to the size limitations of the small animal model. Nonhealing Bone Defect Model in Mice : All in vivo experiments were conducted under the Animals (Scientific Procedures) ACT 1986 (ASPel project license n° 70/8638). Male B6‐129 mice (8–10 weeks old, Charles River) were used for these studies. Mice were fully anesthetized using isoflurane gas as an anesthetic agent. To create bone defects, the midsection of the radius and ulna in the right front paw of the mouse was exposed. A custom‐built double‐blade bone cutter was used to precisely generate a 2. 5 mm segmental defect on the radius without disturbing the ulna. The implant tube (4. 0‐mm long) was fitted over the ends of the defect. After repositioning the muscle and skin, the wound was closed and the mouse was completely ambulatory following recovery. Figure S11 in the Supporting Information shows a scheme of the segmental bone model used. Analysis of New Bone Formation : Bone growth in the area of the defect was evaluated 8 weeks after implantation. Bone samples were explanted and analyzed by X‐ray (SARRP, Perkin Elmer 0820 detector panel) and µCT scanning (Bruker Skyscan Micro X‐ray CT). Then, tissue samples were decalcified and embedded in paraffin and histological sections were stained with hematoxylin‐Safranin O‐fast green. 3D reconstructions of 4 mm length of the bone radio in the area of the defect were obtained by contouring 2D slices from the µCT scans. Quantification of the volume and specific surface of new bone within the defect was performed using the free CTAn software from Bruker. In order to ensure that only new bone formation was measured, the volume of interest (VOI) was selected to evaluate a central 2. 0 mm length in the area of the defect. The 3D volume measurement was based on the marching cubes volume model of the binarized objects within the VOI. The specific surface was evaluated as the ratio of solid surface to volume measured in 3D within the VOI. This was a basic parameter to assess the thickness and complexity of structures, and thus was useful to characterize the complex porous structure of the bone. Preparation of Bone Chips for Implantation : Commercially available dog bone chips (Veterinary Instrumentation Ltd. ) were coated with pPEA. A total of 5 cc of cancellous canine chips, with a chip size of 2–4 mm, were spread on a glass petri dish and placed in the aforementioned custom‐made plasma chamber. The chips were coated with a plasma power of 100 W for 30 min. Monomer pressure during plasma was 1. 8 to 2. 4 × 10 −1 mbar. After plasma, the chips were sterilized under UV light for 15 min and then coated with adsorbed FN and BMP‐2. First, 10 mL of FN in DPBS at 20 µg mL −1 was used to coat the chips, using vacuum to ensure that the liquid reached the entire surface of the chips. After 1 h of adsorption, the remaining liquid was removed using a pipette. Then, 6 mL of BMP‐2 at 50 µg mL −1 was added to the chips. Vacuum was once again used to ensure full surface coating. The chips were moved into the operating theater and after 2 h of GF adsorption, the chips were spread on a surgery gauze to soak all the remaining liquid and thereafter directly used by the veterinary surgeon. Clinical Veterinary Case— Details of Surgery and Postoperative Care : The local ethics committee was consulted regarding this novel procedure, and the informed consent of the dog's owner was obtained prior to the procedure. The Münsterländer dog (2‐year‐old female, neutered, body weight 21 kg) was anesthetized using a standard protocol: premedication with methadone (6 mg IM) and medetomidine (100 µg IM), induction with propofol (30 mg IV), and maintenance with isoflurane in oxygen. The entire right humerus and proximal third of the left humerus were clipped and aseptically prepared for surgery. For analgesia, a brachial plexus nerve block was performed preoperatively using 20 mg of levobupivacaine, and methadone (6 mg IV) was given once during surgery. A standard surgical approach was made to the fractured humerus, on its medial aspect. Approximately 1 cm of the nonhealing ends of the bone were excised, along with all soft tissue within the fracture gap. The fracture was stabilized using a 3. 5 mm locking compression plate (Synthes) placed medially and a 2. 7‐mm locking compression plate (Synthes) placed cranio‐medially. A standard surgical approach was made over the greater trochanter of the left humerus and a hole was drilled in the lateral cortex using a 4. 5 mm drill bit. A curette was used to harvest cancellous bone from the humeral head, which was then mixed with 5 cc of pPEA‐chips (decellularized bone chips coated with pPEA using plasma polymerization and subsequently with FN and BMP‐2, as previously described). The combined graft materials were placed within the fracture gap using Debakey forceps. The muscle layer was closed using polydioxanone in a simple continuous pattern and skin was closed using poliglecaprone 25 in an intradermal pattern. Surgical time was 4 h and 5 min and recovery from anesthesia was rapid and smooth. For postoperative analgesia, methadone (6 mg IM) was given every 4 h for the first day postoperatively and meloxicam (2 mg PO SID) was prescribed for 2 weeks. Seven weeks after surgery, the dog was sedated with medetomidine (200 µg IV) and butorphanol (2 mg IV) for radiography of the humerus, and recovered well from this procedure. The dog attended twice weekly physiotherapy sessions for 5 months postoperatively, and daily physiotherapy exercises were encouraged at home. For the first 7 weeks, physiotherapy included the application of pulsed electromagnetic field therapy to the fracture site (50 Hz, constant pulse, for 30 min twice daily, Biomag 2 Therapy Unit, Westville Therapy). The dog's exercise was restricted to short controlled lead walks for around 3 months after surgery; it was then progressively increased as limb function improved. Statistical Analysis : Experiments were performed in triplicates ( n = 3) and data were expressed as mean ± SD. Statistical analysis was performed by one‐way analysis of variance (ANOVA) with Tukey's test for multiple comparisons using OriginPro 8. Statistical significance was defined as p < 0. 05. For in vivo experiments, data were presented as mean ± SD, minimum n = 3. Two‐tailed t‐test was used to analyze data, p < 0. 1. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201800448
| 2,018
|
Advanced Science
|
Recent Advances in Engineering the Stem Cell Microniche in 3D
|
Abstract Conventional 2D cell culture techniques have provided fundamental insights into key biochemical and biophysical mechanisms responsible for various cellular behaviors, such as cell adhesion, spreading, division, proliferation, and differentiation. However, 2D culture in vitro does not fully capture the physical and chemical properties of the native microenvironment. There is a growing body of research that suggests that cells cultured on 2D substrates differ greatly from those grown in vivo. This article focuses on recent progress in using bioinspired 3D matrices that recapitulate as many aspects of the natural extracellular matrix as possible. A range of techniques for the engineering of 3D microenvironment with precisely controlled biophysical and chemical properties, and the impact of these environments on cellular behavior, is reviewed. Finally, an outlook on future challenges for engineering the 3D microenvironment and how such approaches would further our understanding of the influence of the microenvironment on cell function is provided.
|
1 Introduction In vivo, stem cells reside in a complex, specialized, and dynamic microenvironment, or “microniche. ” 1 Although these microenvironments are extremely diverse, they share a number of characteristic features of function and composition. 2 The microenvironment serves as a structural support for cells, but also offers various biochemical (e. g. , cell–cell contact, cell adhesion sites, and insoluble factors) and biophysical (e. g. , topography, porosity, and rigidity) cues that together regulate cell behavior, including cell spreading, migration, differentiation, and self‐renewal. The extracellular matrix (ECM), a key constitutive part of the microniche, plays an essential role in regulating cell behavior, 3 and supports cell or organ development, function, and repair. The physical properties of the ECM (topography, porosity, rigidity) all impact on biological functions that are related to cell spreading, division, migration, or tissue polarity. In addition, the ECM provides biochemical signaling cues that regulate cell phenotype ( Figure 1 ). Figure 1 Niche interactions known to modulate stem cell phenotype. The biochemical composition, mechanical properties, and microstructure of the ECM are all known to modulate stem cell behavior, with optimal properties dependent on both the stem cell type of interest and the desired phenotypic output. Stem cells, including pluripotent stem cells, embryonic stem cells (ESCs), mesenchymal stem cells (MSCs), hematopoietic stem cells, and neural stem cells, have been widely used for investigating fundamental interactions between cells and the ECM, and have potential applications in translational regenerative medicine or stem cell therapy. Thus, controlling stem cell fate (the ability to maintain the stemness, or to differentiate into different cell types) through engineered microniches is becoming particularly important in cell biology and tissue engineering field. Recently, numerous studies have shown that engineered microniches that mimic different aspects of the native stem cell niche can promote maintenance of stem cell quiescence (which is necessary for long‐term culture of stem cells to generate disease models), 4 facilitate stem cell expansion (which is needed for stem cell delivery and stem cell therapy), 5 and regulate stem cell differentiation (which can be used for tissue engineered constructs). 6 In this review, we will discuss the role of the microniche in controlling cell function, with a specific emphasis on the importance on the role of the ECM. We will start with a short overview on different properties of the ECM that regulate cell fate, and then examine the differences between 2D and 3D cell culture. We will also provide an overview of the techniques used for investigating the interactions between ECM and stem cells in 3D, and discuss current advances toward designing 3D engineered niches. 2 The Stem Cell Microniche The stem cell niche consists of a myriad of interacting components (Figure 1 ), which may include the ECM, other cells, growth factors, and heterologous cell types (e. g. , endothelial cells). These components provide biophysical and biochemical inputs that regulate cell behavior such as adhesion, spreading, migration, division, self‐renewal, quiescence, and differentiation. This section reviews recent progress in studying the effect of different ECM properties on regulating cell fate determination and engineering approaches to control the stem cell microenvironment. 2. 1 Extracellular Matrix Mechanics The native ECM is a network of fibrillar proteins and polysaccharides that anchors cells within their specific microenvironment. Cells are mechanically coupled to the ECM through transmembrane proteins known as integrins. 7 These integrins bind specific cell‐adhesive ligands presented by ECM proteins, connecting the ECM to the intracellular actin cytoskeleton. During cell spreading and growth, the ECM can be mechanically deformed and remodeled by cells, 8 the mechanical properties of the ECM alter the ability of cells to generate tension, modulating cell spreading, nuclear shape, and intercellular signaling pathways. Different types of mechanics can influence cell behavior in different ways, including bulk stiffness, local stiffness, strain‐stiffening, and stress‐relaxation. 2. 1. 1 Bulk Stiffness Substrate stiffness, typically characterized by the elastic or Young's modulus, has emerged as one of the most important mechanical features in controlling cell fate. This means that cells can sense the resistance of the substrate (typically a hydrogel) toward deformation. Modifications to the bulk stiffness of ECM‐coated hydrogels give rise to a range of responses in stem cells. On 2D substrates, mesenchymal stem cells typically show differentiation toward osteoblasts on stiff substrates while lineage selection on soft substrates favors adipocytes[[qv: 6a]] ( Figure 2 a). Figure 2 Bulk stiffness regulates stem cell fate. a) The differentiation of MSCs toward particular lineages is regulated by substrates with stiffness that is similar to native tissues. b) Mechanotransduction pathways inside cells regulate cell fates. c) Actin cytoskeleton organization depends on substrate stiffness. The different colors indicate different orientations of actin filaments. Reproduced with permission. 9 Copyright 2015, Nature Publishing Group. d) A micropatterned platform that limits cells to a stiff region stimulates durotaxis. Reproduced with permission. 10 Copyright 2014, National Academy of Sciences (United States). e) Spatially patterned matrix elasticity directs stem cell differentiation. Reproduced with permission. 11 Copyright 2016, National Academy of Sciences (United States). f) Stiffness triggers nuclear YAP localization by regulating transport across nuclear pores. Reproduced with permission. 12 Copyright 2017, Cell Press. g) Stiffness gradient affects cell migration, cells can migrate from soft to stiff. Reproduced with permission. 13 Copyright 2017, National Academy of Sciences (United States). h) Stiffness determines embryonic stem cell differentiation. Reproduced with permission. 14 Copyright 2016, Cell Press. During mechanotransduction, mechanical stimuli, such as stretching, shear stress, or substrate rigidity, are converted into chemical signals that control cell fate. 15 Key in this process are focal adhesions (FAs) 16 and cell–cell interactions (involving, among others, β1‐integrin 7 and E‐cadherin 17 ), mechanosensors (such as talin 15 ) and nuclear signaling elements (for example, yes‐associated protein (YAP)/transcriptional coactivator (TAZ) 18 and lamin A/C 19 ), which together act to modify protein and gene expression profiles (Figure 2 b). Until now, substrates with stiffnesses ranging from a few hundred Pa to MPa have been prepared in a range of model substrates, including natural material such as chitosan, hyaluronic acid, gelatin, alginate, and agarose, or synthetic hydrogels such polyethylene glycol (PEG), poly(vinyl alcohol) (PVA), or polyacrylamide. Cells cultured on these hydrogels are responsive to the degree of stiffness by altering their adhesion, spreading, morphology, and migration characteristics. For instance, fibroblasts or endothelial cells cultured on a relatively stiff substrate (>2–3 kPa) display significant spreading and generate greater actin stress fibers compared with those on a relatively soft substrate (<2–3 kPa). 20 The orientations of actin filaments strongly depend on substrate stiffness, with stiffer substrates can leading to more aligned actin filaments (Figure 2 c). 9 Cell spreading is also affected by stiffness, and by preparing a rigid domain of one large adhesive island, adjacent to a soft area of small adhesive islands grafted in an otherwise nonadhesive soft hydrogel, researchers have shown that cells spread and probe substrate stiffness by using filopodia extensions (Figure 2 d). 10 Matrix stiffness often shows local heterogeneities at different length scales within the natural niche. 11, 21 Yang et al. fabricated a hydrogel with regions of spatially varied and distinct mechanics, and they found that hMSCs cultured on hydrogels with higher concentrations of stiff regions showed more spread, elongated cell morphologies, higher nuclear YAP localization, and higher osteoblast differentiation, indicating that local variations in the underlying substrate mechanical properties might regulate cell adhesion, spreading, and nuclear transcription effectors (Figure 2 e). 11 The effect of stiffness of cell function can often be related to the activity of certain nuclear transcription factors (YAP/TAZ), 18 and it was shown recently that stiffer substrates give rise to nuclear flattening, thereby stretching nuclear pores, and reducing their mechanical resistance to molecular transport, and finally increasing YAP nuclear import and localization (Figure 2 f). 12 In addition, cell migration is also affected by stiffness. 13, 22 When subjected to a stiffness gradient, cells display directed migration toward stiffer regions. The anisotropic mechanical properties lead to directional epithelial growth and trigger cells to migrate at the direction where the stiffness is larger, a behavior termed durotaxis, which is considered to contribute to the repair of tissue (Figure 2 g). [[qv: 22a]] It has been shown that matrix stiffness also guides the spreading and differentiation of ESCs (Figure 2 h), 14 where softer substrates enhance mesoderm differentiation of human ESCs. However, much of our knowledge about stiffness‐induced stem cell differentiation on 2D cell cultures cannot be directly translated to a 3D environment. For example, it was recently reported that hMSCs, encapsulated in a stiff cross‐linked hyaluronic acid hydrogel, showed reduced cell spreading and nuclear localization of YAP/TAZ. 23 These results indicate that mechanotransduction signaling in a 3D environment is not merely regulated by bulk stiffness, but is sensitive to other parameters such as dimensionality and degradability of the gel. 2. 1. 2 Local Microenvironment Since the local microenvironment is quite different from the bulk ECM, researchers have started to realize the importance of the local microenvironment of cells. Unlike bulk stiffness, where increased stiffness always promotes cell spreading, materials with soft local stiffness have greater flexibility in changing their conformations to optimize cell contact, and thereby inducing the formation of FAs and relevant cellular signals to trigger cell spreading ( Figure 3 a). If the fiber stiffness is higher, the transfer of cellular traction forces to nearby fibers will be limited. Consequently, cells are not able to build up sufficient tension, which may suppress cell spreading and migration (Figure 3 b). The fibrous nature of the ECM creates a unique microenvironment that enables long range mechanical cell–cell communication via cell‐induced remodeling of the network (Figure 3 c). 24, 26 Recently, Baker et al. fabricated a synthetic fibrous material with tunable fiber mechanics by using electrospinning. They found a critical role for fiber recruitment in the cellular response to fibrous materials, where lower fiber stiffness promoted cellular tension to deform and recruit surrounding fibers, greatly increasing the ligand density around the surface of cells, facilitating the formation of FAs and subsequent signaling events (Figure 3 d). 25 Figure 3 Cell response to local fiber stiffness. a) Schematic image shows how cells response to soft fibers. b) Soft local stiffness increases ability of cells to deform and recruit fibers, thus enhancing concentration of local ligand and activating FAs and related cellular signals to induce cell spreading. c) Cells deform collagen fibers to form bundled structure. Reproduced with permission. 24 Copyright 2017, Nature Publishing Group. d) Cells failed to spread on substrate with low bulk stiffness, in contrast, increasing fiber stiffness suppressed cell spreading. Reproduced with permission. 25 Copyright 2015, Nature Publishing Group. Cells are capable of sensing and responding to local mechanical properties in a 3D microenvironment. Recent efforts have focused on producing collagen materials with tunable properties. By controlling the collagen gelation temperature, collagen hydrogels of different fiber stiffnesses can be prepared. 27 Collagen fiber bundling and diameter can be increased by decreasing the gelation temperature, which results in increasing local fiber stiffness. It was shown that increased local fiber stiffness can withstand the repetitive contractile pulling at cell adhesion sites, which reinforces the stability of cellular adhesion and maturation of human foreskin fibroblasts. [[qv: 27c]] By adding gold nanorods into collagen hydrogels, the nanoscale stiffness of the collagen hydrogel can be tuned without changing the bulk mechanical properties, and increased local collagen stiffness was shown to upregulate β1‐integrin‐mediated signaling pathways. 28 These emerging insights into how cells respond to local stiffness rather than bulk stiffness have critical implications for the development of new biomaterials for engineering the cell microenvironment in 3D. 2. 1. 3 Strain‐Stiffening Many filamentous biopolymers (fibrin, F‐actin, microtubules, or vimentin) display nonlinear elasticity, typically strain stiffening (when the applied strain to the matrices is increased beyond the critical strain, the materials become stiffer with increasing strain) ( Figure 4 a). However, the effects of nonlinear elasticity mechanical properties on cell behavior have barely been studied. Recently, it was shown that hydrogels with nonlinear elasticity facilitate long‐distance communication between cells, 31 regulate the ways of 3D cell migration, 32 and control stem cells differentiation. 29 For example, Janmey and co‐workers demonstrated that fibroblasts and hMSCs displayed an elongated morphology when cultured on soft fibrin gels, indicating that the gels can be deformed by cell traction force, allowing access to the high strain moduli in the regimes of strain stiffening. 31, 33 Shear rheology measurements showed that the cells increased the stiffness of the fibrin gels (Figure 4 b). 30 In a recent study, hydrogels based on the polyisocyanopeptide (PIC) were produced with precisely controlled strain‐stiffening behavior. The critical strain of the PIC hydrogels was increased by increasing the PIC polymer chain length, while the adhesion‐ligand density and the stiffness of PIC bulk hydrogel were kept constant. When cells were cultured in 3D PIC hydrogels, hMSCs preferred to differentiate into osteoblasts when the critical strain was increased, a process apparently mediated by microtubule‐associated protein DCAMKL. 29 Taken together, these results highlight the strain‐stiffening property as an important element in fabricating 3D microenvironments. Figure 4 Nonlinear mechanical properties determine cell fate. a) An overview of strain‐stiffening properties for different materials. Reproduced with permission. 29 Copyright 2016, Nature Publishing Group. b) Cells can generate cell traction forces to actively stiffen fibrin gel. Reproduced with permission. 30 Copyright 2013, Cell Press. The traction strain was quantified by measuring the displacements of embedded fluorescent beads inside the fibrin hydrogel (black lines). The elastic modulus was measured by real time rheology (red line). The blue curve shows elastic modulus for a pure fibrin gel without cells. c) An overview of stress‐relaxation properties for different materials (including hydrogels and tissues). Reproduced with permission. [[qv: 6b]] Copyright 2016, Nature Publishing Group. d) Hydrogels with faster stress relaxation property can promote cell spreading and proliferation. Reproduced with permission. [[qv: 6b]] Copyright 2016, Nature Publishing Group. 2. 1. 4 Stress Relaxation The natural ECM is not an ideal elastic solid. Most hydrogels and soft tissues that are based on biopolymers display viscoelastic (or dissipative) properties. 34 These hydrogels show stress relaxation (the stress decreases in response to the constant applied strain with increasing time) or creep (the strain increases in response to the constant applied stress with increasing time). 35 Figure 4 c shows stress‐relaxation tests for different materials, including hydrogels and native tissues. Living tissues all exhibit stress relaxation behavior. However, the effects of stress relaxation properties on cell behavior have often been overlooked. In recent years, a number of groups have designed hydrogels with tunable stress relaxation properties by changing the hydrogel composition or concentration, [[qv: 35b]] molecular weight, 36 cross‐link type or density, 37 and degradation. 38 Recent studies demonstrated that hydrogel stress relaxation properties could have significant effects on cell fate decisions. For example, Cooper‐White and co‐workers found that hMSCs morphology, proliferation, and differentiation were influenced by modifications to substrate creep. 39 Chaudhuri and co‐workers prepared alginate matrices with controllable viscoelastic or elastic features through covalent or ionic cross‐linking, and it was found that when hMSCs encapsulated in the 3D alginate hydrogels with faster relaxation properties, showed enhanced spreading, proliferation, and osteogenic differentiation (Figure 4 d). It is thought that integrin signaling, ECM ligand bundling, cell contractility, and nuclear YAP localization all play a role in these processes. [[qv: 6b, 40]] Since most biopolymers show both stress‐relaxation and strain‐stiffening properties, it should be noted that changes in viscoelasticity and nonlinear elasticity are often coupled, which makes it difficult to decouple the two. Chaudhuri and co‐workers found that collagen and fibrin hydrogels exhibited both stiffening and faster stress relaxation upon increasing the strain, an effect attributed to the dissolution of weak cross‐links that are dependent on the force. 41 Thus, future studies are needed to engineer the 3D cell microenvironment with purely nonlinear elasticity or viscoelasticity behavior, and explore potential applications of these hydrogels in tissue engineering and regenerative medicine. 2. 1. 5 Surface Receptors Several recent papers have argued that mechanical feedback of the linkage between ECM substrate and cell surface receptors could influence cell adhesion, spreading, and differentiation. 42 For example, Trappmann et al. found that cell spreading and differentiation were unaffected by the stiffness of polydimethylsiloxane (PDMS) substrates, but were strongly dependent on the modulus of polyacrylamide (PAAm). The authors proposed that soft PAAm hydrogels were more porous than stiff gels and this will lead to differences in anchoring densities, thereby altering the mechanical feedback of the collagen. [[qv: 42b]] Recently, Navajas and coworkers developed a hydrogel with precisely controlled rigidity and nanometer‐scale distribution of ECM ligands. [[qv: 42e]] They found that when cells were cultured on low‐rigidity substrates, FAs formation could be upregulated by increasing the spacing between ligands, while on high‐rigidity substrates, adhesion collapsed. Moreover, disordered ligand distribution on the substrates significantly increased the stability of adhesion formation, but reduced the rigidity threshold for adhesion collapse. [[qv: 42e]] On the one hand, these results show that the precise nature of the mechanical properties of the link between cells and the substrate must be taken into account when designing substrates for regulating cell fate. On the other hand, cells are very complex systems, and how exactly insoluble physical cues from the cellular environment affect cell behavior still poses a considerable challenge. 2. 1. 6 Degradability Many natural materials, such as collagen or fibrin hydrogels, are enzymatically degradable, enabling cells to degrade and remodel their microenvironment. The effect of degradation has a significant effect on cell behavior, especially in 3D microenvironments. Lutolf et al. 43 highlighted the importance of matrix degradability in studies of cellular invasion into degradable and adhesive synthetic hydrogels. Khetan and Burdick 44 have shown that cell spreading was limited in hydrogels with a high density of nondegradable cross‐links. They further demonstrated that in 3D covalently cross‐linked hyaluronic acid hydrogels, the differentiation of hMSCs was regulated by the generation of degradation‐mediated cellular traction force, independent of matrix mechanics or cell morphology. 45 Burdick and co‐workers recently investigated the effect of degradation and stiffness on neural progenitor cell stemness in a 3D hydrogel. The hydrogel was made from elastin like protein and functioned with cell‐adhesive peptide. By changing the protein concentration and cross‐linking density, the stiffness and degradability of hydrogels could be independently tuned. They found that neural progenitor cell stemness did not depend on gel stiffness, but strongly related with degradability. Degradability could increase cell‐mediated matrix remodeling and then enhance neural progenitor cell self‐renewal and potency. This study provided an evidence for the important role of degradability in maintaining neural progenitor cell in 3D microenvironments. [[qv: 4b]] Overall, these results highlight the important role of degradability in regulating cell fate. It should be noted though that controlling the degradation kinetics and the formation of degradation byproducts remains challenging, especially since degradation leads to softening of the ECM, and thus making it harder to present cells with ECM of the right stiffness. 2. 1. 7 Confinement Cells in the body are confined by other cells or by components of the ECM. Therefore, studying the cellular response to confinement is very important for fundamentally understanding the interactions between cells and the ECM. Recently, a lot of in vitro models, including microchambers, grooved substrates, microfluidic channels, microcontact printed substrates, and 3D hydrogels, have been engineered to study the effect of confined environment on cell spreading, migration, and signaling[[qv: 22a, b, 46–50]] ( Figure 5 a). Figure 5 The effect of confinement on cell behavior. a) Schematics of engineered models of confining microenvironments. b) Cells migrate fast in confined environments because of low adhesion. Reproduced with permission. 46 Copyright 2015, Cell Press. c) Mechanical confinement regulates cartilage matrix formation by chondrocytes. Reproduced with permission. 47 Copyright 2017, Nature Publishing Group. d) Confinement affects cell migration. Reproduced with permission. [[qv: 22a]] Copyright 2012, National Academy of Sciences (United States). e) Confinement environment is sufficient to induce patterned differentiation of embryonic stem cells. Reproduced with permission. 48 Copyright 2014, Nature Publishing Group. f) Geometric confinement induced self‐organizing human cardiac microchambers. Reproduced with permission. 49 Copyright 2015, Nature Publishing Group. Cell confinement has been used in a number of different studies, including, for example, studies into the relationship between cell cytoskeleton and cell polarity[[qv: 22b, 51]] and cell migration under confinement. [[qv: 50c, 51a, 52]] On 2D substrates, cells can form distinct FAs and stress fibers to spread and migrate. Conversely, cells in confined environments typically show fewer FAs and suppressed stress fiber formation. [[qv: 51a]] Furthermore, cytoskeletal structures and nuclear elongation are aligned with the confining axis. For example, actin accumulation and stress fibers formation were suppressed under confinement environment, regardless of substrate stiffness. [[qv: 22b, 51, 52d]] Confinement can also alter the type and morphology of cell adhesions. The homogeneous expression of phosphorylated focal adhesion kinase (pFAK) and p‐paxillin will be inhibited under increased confinement. [[qv: 51a]] Similarly, when cells are limited to 1D fibronectin lines that are generated by microcontact printing, FAs will be distributed along the cell body. 53 Vinculin will be also homogeneously dispersed over the cell body in cells that are vertically confined 46 (Figure 5 b). By culturing cells in 3D hydrogels, Lee et al. found that when chondrocytes were cultured in hydrogels with slower stress‐relaxation, cell volume expansion was limited by the spatial confinement, resulting in lower cell proliferation rate (Figure 5 c). 47 The influence of confinement on cell migration behavior has also been extensively studied. Cells migration in confinement is typically straight (Figure 5 d), [[qv: 22a]] and migration speed is significantly higher in microchannels than on 2D substrates. [[qv: 51a, 54]] Fully confined cells display a sliding migration, [[qv: 46, 51a, 55]] but it remains unclear whether vertical and lateral (confinement affects cell migration equally). Geometric confinement can also influence stem cell differentiation. For example, when human embryonic stem cells colonies were geometrically confined on circular Matrigel micropatterns, they reproducibly differentiated into an outer trophectoderm‐like ring, an inner ectodermal circle, and a ring of mesendoderm that expresses primitive‐streak markers (Figure 5 e). 48 Ma et al. exploited confinement conditions to link spatial cell‐fate specification and the formation of a beating 3D cardiac microchamber, which can be used to mimic certain aspects of early stage heart development (Figure 5 f). 49 Taken together, these studies clearly show that confinement gives rise to marked changes in the cellular cytoskeleton structure, cellular adhesion distributions, cell migration behavior, and stem cell differentiation, indicating that cells are responsive to physical confinement. 2. 1. 8 Geometrical Cues In native tissue, different cell types vary greatly in their size and shape, and these geometrical cues are important factors in cell fate regulation. The influence of these cues can be studied by culturing cells on micropatterned ECM (for example, collagen, fibronectin, lamin, Matrigel) islands of defined geometries, which can be fabricated with various techniques, for example, microcontact printing/stamping, microwells with different geometries and sizes, and cell printing. When culturing cells on these 2D ECM islands, the cells generate tension forces, and spread until they arrive at the island perimeter. 62 Cells prefer to generate larger tension at curvature, partially because of the confinement, 63 and this will lead to upregulation of FAs and actin formation ( Figure 6 a). The molecular mechanism of cell‐geometry‐dependent regulation of differentiation has been elucidated in some cases. 64 A recent study suggested that cell geometry regulates cell signaling via modulation of plasma membrane order. Changes in plasma membrane order due to geometric cues affect stem cell fate through a newly identified signaling mechanism involving the serine/threonine kinase Akt/protein kinase B. 65 Studies on cell geometry have shown that cell fate can be guided between apoptosis, growth, and differentiation by altering the extent to which the cell can physically expand and flatten (Figure 6 b). 56, 66 Recent studies demonstrated that the differentiation of MSCs could be switched between osteoblast and adipocytes in a shape‐dependent manner (Figure 6 c), 57 which is partially dependent on the localization of YAP/TAZ (Figure 6 d). 18, 67 Cell geometry also plays a very important role in nuclear events. It has been shown that confining cells on patterned surfaces could significantly alter the structural organization of the nuclear lamina compared with cells on flat surfaces (Figure 6 e). 19 Substrate topography (e. g. , grooves, steps, pits, etc. ) also strongly controls MSC shape and lineage selection. For example, Desai et al. fabricated a substrate with spatially organized multiple adhesive ligands patterns, and found that cells can sense surface geometry by segregating single integrins on the surface of cells to regulate ECM‐specific binding. 58 (Figure 6 f) Cell geometry can also regulate nuclear geometry, which may generate a new way to control stem cell lineage commitment on the subcellular level 59, 68 (Figure 6 g). Apart from single cells, tissue growth is also strongly affected by the geometrical features of the matrix. Human epidermal stem cells seeded on 100 µm diameter circular collagen‐coated disks, self‐assembled into a stratified microepidermis. Like the small islands that accommodate single cells, larger islands with a nonadhesive center still supported microepidermis assembly. 69 Figure 6 The effect of geometry and topography on cell fate decisions. a) Schematic image shows how cells sense sharp curvature. b) Cell spreading area determines stem cell differentiation. Reproduced with permission. 56 Copyright 2004, Cell Press. c) Differentiation of hMSCs is determined by cell contractility triggered by different geometries. Reproduced with permission. 57 Copyright 2010, National Academy of Sciences (United States). d) Cell spreading area directs YAP/TAZ localization. Reproduced with permission. 18 Copyright 2011, Nature Publishing Group. e) Cell spreading area determines nuclear lamin localization. Reproduced with permission. 19 Copyright 2015, Nature Publishing Group. f) Substrates with spatially organized multiple adhesive ligands patterns can be used for investigating the effect of various integrin bindings on cell adhesion and migration. Reproduced with permission. 58 Copyright 2011, Royal Society of Chemistry (United Kingdom). g) With the increase of pillar height, nucleus was deformed, FAs and actin cytoskeletons were densely distributed around the micropillars and became obscure. Reproduced with permission. 59 Copyright 2016, Elsevier. h) Geometry determines tissue growth rate. Reproduced with permission. 60 Copyright 2008, The Royal Society (United Kingdom). i) Geometric cues affect cell proliferation rate. Reproduced with permission. 61 Copyright 2005, National Academy of Sciences (United States). Cells in microtissues detect and respond to radii of curvature and when grown in polygonal channels, new tissue started in the corners (Figure 6 h). 60 The tissue in sharp corners (for example, triangular channel) was thicker than those in square and hexagonal channels, following the decrease of local curvature and indicating that increasing local curvature can increase the rate of proliferation (Figure 6 I). 61 Although the idea of 3D micropatterned systems is not novel, technical limitations of these endeavors have limited the feasibility of studying single cell behavior in 3D microenvironments. Recently, we demonstrate the first method to constrain stem cell size and geometry in a systematic and quantitative manner, by encapsulating cells in 3D hyaluronic acid hydrogel microniches. 70 This method differs from previous studies on 2D micropatterned substrates and microwells, as it can provide cells with a completely nonpolarized microenvironment of precisely defined volume, and it also allows for rapid acquisition of confocal microscopy images on large numbers of individual cells in identical microenvironments. By using this method, we found that cytoskeletal organization in cells in 3D microniches has a preferred size and geometry. Furthermore, we found that key proteins and mRNA concentrations were diluted in larger cells. Separate studies show that geometrical cues also affect the orientation of cell motility, initiated by the formation of actin filaments, lamellipodia and filopodia ( Figure 7 a). 71 Polarity axes as defined by the internal and cortical cell asymmetry were controlled by the adhesive geometry (Figure 7 b). 72 When cells were cultured on ECM islands with square or rectangle geometry, FAs and actin stress fibers would be inclined to situate along the cell's diagonal axes (Figure 7 c, d). 73, 74 The alignment of stress fibers and FAs is partially a result of actomyosin contractility (Figure 7 e). 75 Moreover, it was found that all fibers were connected to each other instead of being isolated and cell relaxation was induced by means of local ablation of one fiber (Figure 7 f). 76 The cell shape within tissue can reflect the past physical and chemical signals that the cells have run into, and the cellular phenotype can also be controlled by the cell shape information. Ron et al. used microfabricated 3D biomimetic chips to demonstrate that 3D cell shape can control cell phenotype via cell tension (Figure 7 g). 77 In addition, it appears that the interplay between actin and microtubuli arrangement plays an important role in cell polarization. In cells spreading on either soft, ECM‐coated gels, or stiff cadherin‐modified substrates, the rearward actomyosin (partially) prevents microtubuli penetration at the leading edge on both soft and stiff substrates. 79 By contrast, when cells were allowed to spread unconstrained on stiff ECM‐coated substrate, microtubule (MT) aligned in parallel with actin stress fibers, and reached all the way to the leading edge of the cell (Figure 7 h). 78 Figure 7 The effect of geometrical cues on actin organization. a) Schematic image showing how geometry directs cytoskeleton organization. Reproduced with permission. 71 Copyright 2012, Cell Press. b) Organization of polarity is governed by cell adhesive microenvironment. Reproduced with permission. 72 Copyright 2006, National Academy of Sciences (United States). c) Organization of the stress fibers, FAs, and ECM within cells on a patterned square ECM island. Reproduced with permission. 73 Copyright 2002, FASEB. d) Cell aspect ratio changes affect organization of actin stress fibers and FAs. Reproduced with permission. 74 Copyright 2012, Nature Publishing Group. e) Actin, myosin II, and α‐actinin staining for different cell types. Reproduced with permission. 75 Copyright 2015, Nature Publishing Group. f) Dissipation of elastic energy in severed stress fibers depends on fiber length. Reproduced with permission. 76 Copyright 2017, National Academy of Sciences (United States). g) (Left) Scanning electron microscopy (SEM) images show in vivo podocytes with branched structure; (Right) F‐actin staining for cells cultured on glass, box, and microchannels. Reproduced with permission. 77 Copyright 2017, Nature Publishing Group. h) Microtubule growth trajectories are correlated with F‐actin bundles controlled by cell geometry. Reproduced with permission. 78 Copyright 2012, The Company of Biologists (United Kingdom). A range of techniques have been used to control geometric cues on substrates and study their influence on stem cells cultured on such substrates. However, challenges remain. First, it is necessary to assess the influence of geometrical control, after long‐term culture when the cells produce their own ECM and loose direct links with microscale or nanoscale geometrical cues. Second, it remains unclear whether findings on 2D substrates can be applied to 3D. Finally, the underlying molecular mechanisms by which cells sense and respond to the geometric cues, and how the mechanical properties of cells result in the cytoskeleton tension and contractility of cells, are not fully understood. 3 Taking Dimensionality into Consideration: From 2D to 3D As mentioned above, different properties of ECM can be designed to regulate cell fate determination. However, most of these studies involved 2D platforms, which present, by necessity, grossly oversimplified environments compared to the in vivo 3D scenario. In 3D, cells form adhesive connections on all sides, providing an unpolarized environment for cells to grow. The polarized environment and extremely asymmetric distribution of adhesions on 2D substrates may lead to unnatural apical−basal cell polarity and corresponding alterations in cell functions. Besides, cell spreading and adhesion on 2D substrate are unlimited, which allows free spreading and migration of cells without any physical limits. Those fully embedded cells are sterically hindered when they spread and migrate, as they are confined by the surrounding matrix. Cells must penetrate the matrix pores, or degrade the matrix around them, before spreading and migration becomes possible. On 2D substrates, the speed of migration is determined by the actin polymerization, integrin‐mediated adhesion, and myosin‐mediated cellular contraction. However, in a 3D matrix, the contribution effectors to cell migration are very complex, involving, for example, the activation of the nuclear piston, 80 local ECM stiffness, [[qv: 27c]] membrane tethered protease degradation, [[qv: 27a, 81]] the ability to squeeze the nucleus through matrix pores, 82 and microtubule dynamics. 83 As a result, the speed of cell migration and its response to stiffness are quite different in 2D compared to 3D. Furthermore, on 2D substrates, cell culture medium, soluble factor, and cell‐secreted factors can undergo free diffusion, whereas in 3D matrices, diffusion of oxygen, proteins, and small molecules can be limited, resulting in gradients. It is likely that cells cultured in 3D display behavior more relevant to in vivo conditions. Khetan et al. demonstrated that when hMSCs were cultured in covalently cross‐linked HA hydrogels, hMSCs differentiation was controlled by the generation of cellular traction forces mediated by hydrogel degradation, regardless of cell morphology and hydrogel stiffness. These outcomes emphasize the critical role of degradability in 3D as a parameter separate from the influence of cell morphology or substrate. 45 Recent efforts 84 on 3D tumor spheroids, aimed at recapitulating the natural tumor microenvironment, showed that 3D tumor spheroids better mimic tumor cell development than traditional 2D monolayer models. Zernicka‐Goetz and co‐workers have shown that by culturing embryonic and extraembryonic stem cells inside a 3D Matrigel, the cells self‐organized into a synthetic embryo, whose development and structure were very similar to those of the natural embryo. 85 4 Technologies to Engineer 3D Stem Cell Niches As discussed above, cells can sense and respond to myriad signals from their 3D microenvironment. Over the past decades, a wide range of sophisticated in vitro cell culture platforms have been developed that control the presentation of biochemical and mechanical cues in 3D. One of the key points to consider in the fabrication of a 3D environment for cells is to allow oxygen and nutrients reach to the compartmentalized cells, while excreted waste products are released. A broad range of fabrication approaches have been employed to control cell–matrix and cell–cell interactions in 3D ( Figure 8 ). In this section, we discuss recent work on bioengineering approaches for controlling interaction between cells and the microenvironment in 3D. Figure 8 Schematic overview of the major methods used to achieve 3D cell culture. 4. 1 Hydrogel‐Based Technology Hydrogels, which are water‐swollen cross‐linked polymeric systems, can be prepared from a variety of natural biomaterials and synthetic polymers ( Table 1 ), presenting a wide range of mechanical and chemical features. Many methods can be used to regulate the physical and chemical properties of hydrogels. 2, 86 Table 1 Representative materials that can be used for 3D cell culture studies Materials Gelation method Featured properties Ref. Natural‐derived materials Collagen Raising the temperature and the pH can initiate collagen fibril self‐assembly Fibrous structure Exhibits structural and mechanical properties (strain‐stiffening) reminiscent of native tissues Displays native cell adhesion ligands [[qv: 24, 26b, 41, 51b]] Fibrin Thrombin can initiate self‐assembly of insoluble polypeptide chains of fibrinogen into a fibrillar network Fibrous structure Enzymatically degradable Strain‐stiffening property 30, 93 Gelatin Gelatin gel can be formed by lowing the temperature or photo‐cross‐linking (for methacrylated gelatin, GelMA) Stiffness can be controlled Enzymatically degradable 94 Alginate Alginate hydrogels can be formed by cooperative binding with divalent cations such as Ca 2+ or Ba 2+ Should be functioned with adhesive proteins for cell adhesion and spreading Stress‐relaxation property [[qv: 6b, 47, 95]] Hyaluronic acid Modified HA can form gels by photo‐cross‐linking or enzymatically cross‐linking It contains a high degree of chemical modification that enables considerable tunability 45, 96 Chitosan Gels can be formed by adjusting the pH Excellent biocompatibility and immunostimulatory activities 97 Dextran Dextran gels can be formed by chemically cross‐linking Commercially available Cross‐linked dextran can act as a microcarrier 98 Agarose Cooling initiates the aggregation of double helices by the entanglement of anhydro bridges Tunable elastic moduli Viscoelastic properties 99 Matrigel Gels can be formed irreversibly and rapidly between 24 and 37 °C Gelling speed depends on the concentration and gelation temperature A heterogeneous composition 100 Synthetic materials Polyethylene glycol (PEG) PEG gels can be formed under both physiological pH and temperature Can be engineered to present different adhesive ligands and to degrade via passive, proteolytic, or user‐directed modes 87 Poly(vinyl alcohol) (PVA) Modified PVA can form gels under photo‐cross‐linking Satisfactory biocompatibility and sufficient mechanical properties 101 John Wiley & Sons, Ltd. Naturally derived hydrogels for cell culture are mainly made of proteins and ECM elements, for example, collagen, fibrin, hyaluronic acid, or Matrigel, as well as materials derived from other biological sources such as chitosan, alginate, gelatin, agarose, or dextran. [[qv: 86b, 87]] Most of these hydrogels are inherently biocompatible and bioactive, since they are naturally derived. [[qv: 86a]] Some of them (for example, collagen, fibrin, and Matrigel) have binding sites for cells to interact with, and such interactions have some benefits for the viability, proliferation, cell migration, differentiation, and remodeling of the gel matrix. 88 However, hydrogels made from those natural materials have some disadvantages in isolating certain cell responses and determining exactly which signals are promoting cellular function. For example, Matrigel is comprised of entactin, laminin, and collagen, but also contains a variable and uncharacterized fraction of growth factors. 89 Furthermore, it is difficult to independently tune the physical and chemical properties for these natural hydrogels. [[qv: 89b]] For example, there is no way to regulate the stiffness of collagen or fibrin gels without changing the adhesive ligand density, pore size, and porosity of the hydrogel. Finally, the shape and size of individual cells cannot be controlled inside hydrogels, and we cannot use hydrogels to make direct comparisons with the outcomes on 2D substrates. Alternatively, hydrogels composed of synthetic polymers, for example, PEG, can be used for long term cell culture, and allow for ECM deposition as they degrade, suggesting that synthetic gels can be used as 3D cell culture platforms, even when there is no integrin‐binding ligands. Hydrogels made from those synthetic materials are highly reproducible, the mechanical properties can be easily adjusted, and can be conveniently processed. However, they lack the endogenous factors that facilitate cell behavior. These synthetic scaffolds offer a minimalist approach with which the mammalian cells can be cultured in vitro for the purpose of clinical applications and the basic researches of cell physiology. The ECM is a very dynamic system. To properly mimic the native ECM, some of its complexity (for example, dynamics) must be taken into consideration when designing these hydrogels. Recently, instead of mimicking the static aspects of the cellular microenvironment, researchers started to adopt more dynamic hydrogels. External stimuli can be used to change the chemical and physical properties of hydrogels to better mimic the dynamic native cellular microenvironment. For instance, mechanically dynamic hydrogels that can be stiffened, 90 softened, 91 or reversibly stiffened and softened, 92 have been developed to investigate the effect of stiffness changes on cellular responses. These mechanically dynamic substrates enable us to study the effect of mechanical dosing on cell fate decisions, which is of particular interest for the mechanobiology community. 4. 2 Microwell‐Based Technology Microwells are a widely used and simple platform to structurally engineer the 3D cell microenvironment. Microwell arrays can be produced by means of direct etching into silicon, or by photolithography, or through molding of hydrogel materials using soft‐lithography. Many different cell types (such as human hepatoblastoma cells, fibroblasts, adipose‐derived stem cells, embryonic stem cells)[[qv: 50b, 106]] can be cultured in microwells to form cell spheroids in a high‐throughput manner. For example, embryonic stem cell aggregates can be formed inside microwells of different sizes ( Figure 9 a). [[qv: 50b]] People found that cardiogenesis was enhanced in larger embryoid bodies (for example, 450 µm in diameter), while the differentiation of endothelial cells was increased in smaller embryoid bodies (for example, 150 µm in diameter). These cell spheroids can be taken as components for bottom‐up tissue engineering applications or serve as efficient 3D in vitro models for research on drug toxicity or cancer invasion. Lutolf et al. 102 modified and functionalized inside surfaces of microwells with different biomolecules to examine in vitro self‐renewal of hematopoietic stem cells as well as the regulation of this process by recombinant protein signals (Figure 9 b). Furthermore, cell density, porosity, and mechanics of the hydrogel as well as the concentration of coated ECM components can be combinatorially regulated in these microwells, which enables a study on the effect of cell–cell interactions as well as hydrogel stiffness on the fate of MSCs. 107 Figure 9 Microwells in cell biology studies. a) ESCs cultured in PEG microwells with different diameters for 7 d. Reproduced with permission. [[qv: 50b]] Copyright 2009, National Academy of Sciences (United States). b) High‐throughput platform based PEG microwells for investigating single cell fate. Reproduced with permission. 102 Copyright 2009, Royal Society of Chemistry (United Kingdom). c) Confocal images show cells cultured in PDMS microwells with different shapes. Reproduced with permission. 103 Copyright 2007, Royal Society of Chemistry (United Kingdom). d) Controlling spatial organization of multiple cell types in microwells with certain 3D geometries. Reproduced with permission. 104 Copyright 2012, Wiley. e) Microwells can be used for creating microparticle arrays with complex building blocks, green particles are assembled before red particles. Reproduced with permission. 105 Copyright 2017, Nature Publishing Group. Changing the sizes and geometric features of microwells can provide tunable confined spaces for controlling cell differentiation. Moreover, by culturing cells in microwells, the influence of cell shape, substrate stiffness, and dimensionality can be decoupled (Figure 9 c). For example, Tsurkan et al. 108 fabricated microwells and microchannels with defined architectures using microlens array photopatterning technology, and they identified that neural precursor cell differentiation is dependent on the degree of spatial confinement. However, most reported microwell cultural systems are immobile, limiting their possibilities to actively operate encapsulated individual cells. Recently, microwells with varied dynamically adjustable geometries have been designed by using biocompatible polymers that are responsive to temperature, such as polycaprolactone (PCL). 109 The dynamic alterations in microwell geometries resulted in dramatic changes in the cytoskeletal architecture and differentiation patterns of stem cells. Tekin et al. prepared dynamic microwells with tunable shape transformation properties under different temperatures by using poly( N ‐isopropylacrylamide) (PNIPAAm), a thermoresponsive polymer. This feature was exploited to pattern multiple cell types at different temperatures in dynamic circular and square microwells 104 (Figure 9 d). Cellular microwell arrays can provide high throughput platforms for deconstructing the multicomponent cues that regulate cell function, and can be used to create large‐scale microparticle arrays with complex motifs 105 (Figure 9 e). However, the limitation of using microwells for cell culture is that they are pseudo‐3D models that cannot really mimic the in vivo 3D environment; therefore, more advanced and integrative technologies should be developed to engineer the biophysical microenvironment of cells. 4. 3 Microgel‐Based Technology Inspired by observing different organs or tissues that consist of repetitive building blocks (think of hepatic lobules or nephron architecture), microgels have been fabricated for 3D cell encapsulation. To date, microgels have been fabricated with different shapes and sizes by using different methods. For example, a patterned photomask could be used to fabricate microgels with an array of shapes. By expanding this method, Fan et al. 110 presented a two‐step method based on photolithography technology to encapsulate single neuron cells in gelatin microgels, and found that axonal circles formed in these hydrogel rings mimicking self‐synapse diseases. Another common approach to create microgels involves the use of a micropatterned mold. For example, by using a patterned PDMS stamp, HA microgels containing the cells could be molded under UV cross‐linking. 111 By using the same method, more complex 3D cell microenvironments over multiple size scales can be fabricated. 112 Recently, Ma et al. engineered a hyaluronic acid microgel that contains fibrinogen by using droplet‐based microfluidics. 113 The microgels serve as a 3D microenvironment for culturing of single hMSCs, and could be cultured up to 4 weeks with different stiffness (0. 9–9. 2 kPa). 114 One recent study from Weitz and co‐workers 115 shows that by using microfluidic technology, single cells could be encapsulated in 3D alginate microgels, and cells remained viable in microgels over three days. It was found that the osteogenic differentiation of encapsulated cells was determined by the cell density or gel stiffness, and the work also demonstrated that by injecting the singly encapsulated marrow stromal cells intravenously into the mice, the clearance kinetics were postponed and the donor‐derived soluble factors in vivo were maintained. Therefore, encapsulation of individual cells in microgels might be useful in the field of regenerative medicine applications and tissue design. 5 Conclusions and Outlook The use of engineered 3D cellular microenvironments enables us to look into the way in which cells interact with and react to the external environment. However, much work remains to be done. Cells are rarely in equilibrium, and understanding how cells accumulate information about their environment over time, how external stimuli are translated molecularly into cell face decisions, and how these decisions manifest themselves in changes in cell phenotype remain core questions for cell biology. Controlling the environment as much as possible can help answer these questions. To provide a very small glimpse, we recently found that cell spreading dynamics could provide a strong indication of future cellular behavior. 116 Future work should focus on developing new ways to track and observe single cell dynamics over extended periods of time, while building up a molecular picture of the changes occurring in the cell. It should be noted that changes do not only occur inside cells, cells also modify their surroundings. A very promising development there is the engineering of biomimetic materials with time‐regulated properties that react to external stimuli. 90, 92, 117 However, most of these dynamic materials only result in mechanical or topographical changes, which is oversimplified when compared to the in vivo cell microenvironment dynamics. Therefore, future work should focus on developing new materials that allow the real‐time control of cell microenvironments and fully capture cell dynamics. Ultimately, we need all of this information to understand how we can engineer synthetic microenvironments for developing and maintaining living tissues inside synthetic compartments. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201800450
| 2,018
|
Advanced Science
|
Stretchable Multiresponsive Hydrogel with Actuatable, Shape Memory, and Self‐Healing Properties
|
Abstract Smart hydrogels with responsive behaviors have attracted tremendous attention. However, it is still a challenge to synthesize stretchable hydrogels capable of changing their original properties in response to multiple external stimuli. Here, integration of actuation function, shape memory, and self‐healing capability in a highly stretchable hydrogel under triple external triggers is achieved by rationally engineering multiple functional moieties. The hydrogel exhibits high stretchability (average relative strain (mm/mm) is >15) and excellent fatigue resistance during 100 loading cycles of 100% strain. Incorporating a moisture‐insensitive polymer film with the hydrogel, hydroactuated functionality is demonstrated. Moreover, shape memory and self‐healing abilities of the hydrogel are realized by the formation of ionic crosslinking or dynamic borate ester in conditions of multivalent cations and pH, respectively. Deformable plastic flowers are displayed in this work as a proof‐of‐concept, and it is believed that this smart hydrogel could be used in plenty of frontier fields, such as designing electronic devices, soft robotics, and actuators.
|
Functional materials, like actuation hydrogel, 1 shape memory, 2 and self‐healing polymers, 3 have drawn increasing attentions and perform extensive applications in fields of tissue engineering, 4 drug delivery, 5 smart coating, 6 and soft robotics. 7 Up to now, efforts devoted to the development of functional materials mainly focus on combing different moieties to maximize certain type of function. Taking self‐healing hydrogels for example, combinations of dynamic covalent bonding, supramolecular interactions, and polymer chemistry have been investigated to improve healing efficiency. 8 Even though these researches are beneficial to broaden the range of functional hydrogels, diversity of functionalities is also a key factor that deserves further investigation to improve the versatility of hydrogels. In recent years, stimuli‐responsive hydrogels, 9 also called smart hydrogels, have been explored to meet the demand of multifunctionality, these materials are capable of adapting structure and property changes in a controlled manner to external stimuli, including moisture, pH, temperature, light, and so on. The advent of smart hydrogels significantly boosts the development of functional materials endowing hydrogels with postresponse properties upon external triggers. 10 Many functional materials whose functionality stems from responsiveness have been reported. For example, surface modeling of polyacrylamide (PAM) hydrogel into microarrays, microstructure hydrogel showed microactuation functionality in responding to moisture changes. 11 Bio‐inspired hydrogel generated by boronate–catechol complexation displayed excellent self‐healing performance in weak alkaline environment. 12 And multiresponsive supramolecular hydrogel which consists of polyacrylic acid, phenylboronic acid, and agar presents shape memory functionality under triple stimuli. 13 However, capability of current responsive hydrogels is commonly too monotonous to meet various requirements, either showing one functionality upon individual trigger, 14 or multiresponsive behaviors all resulting in same property. 15 Meanwhile, stretchability of hydrogel, which could be enhanced by interpenetrating network structure, 16 acts an essential role in researches referring to soft robotics, electronic skin, and artificial muscle. 17 Although multifunctional materials have been reported, like Harada and co‐workers 18 engineered a supramolecular hydrogel with redox‐responsive shape memory and self‐healing abilities by employing two host–guest interactions, 19 poor stretchability restricted the usage of such hydrogels. However, the fabrication of stretchable hydrogel with diverse properties which can respond to multiple external stimuli still remains a challenge. Here, we described a novel multiresponsive hydrogel with actuatable, shape memory, and self‐healing properties corresponding to three types of common triggers: moisture, multivalent cations, and pH. This smart hydrogel is formed by two interpenetrating polymer chains: polyacrylamide and phenylboronic acid grafted alginate (alginate‐PBA), which is abbreviated to i ‐PAP hydrogel. Among the gel, PAM is chosen because of its extensively studied properties and excellent compatibility with other brittle polymer networks to reinforce the final mechanical performance. 20 Alginate, a kind of natural polysaccharides, can be rapidly cross‐linked by various multivalent cations (Ca 2+ was used in this work) thus possessing ionic responsiveness. 21 In particular, numerous hydrophilic groups in PAM and alginate chains greatly improve water absorption of hydrogel, then the huge volume changes caused by water content variation could act as the driving force of actuation functionality. We selected phenylboronic acid to modify alginate on account that boronic acid group is able to dynamic covalently interact with diol groups of alginate in mildly alkali conditions (pH is about 9–10) 22 showing self‐healing behavior without destroying the original nature of alginate thoroughly. 23 With these compatible multiple functionalities and high stretchability, this gel demonstrates promising prospects in aspects of soft robotics, electronic skin, and actuators. Figure 1 illustrates synthetic process of i ‐PAP hydrogel. The detailed procedure is provided in the Supporting Information. Pristine i ‐PAP gel is considered as control sample in the context. The resultant gel was completely transparent with naked eyes (Figure 1 b) and held a typical porous structure (Figure 1 c), which differed from the lamellar structure of pure alginate‐Ca 2+ hydrogel. 24 Hydrogels immersed with CaCl 2 or NaOH solution were regarded as conditioned hydrogels (Ca‐conditioned for CaCl 2 and alkaline‐conditioned for NaOH). The formation process of conditioned gels is schematically demonstrated in Figure S1 (Supporting Information). Comparing with control gel, two kinds of conditioned samples exhibited more compact microscopic structures by scanning electron microscopy (SEM) images (Figure S2, Supporting Information). To verify differences of physiochemical properties between control and conditioned hydrogels, we carried out Fourier transform infrared spectra and rotational rheometer characterizations (Figures S3 and S4, Supporting Information). Figure 1 Schematic illustration of i ‐PAP hydrogel synthetic process and microscopic structures of hydrogel. a) Schematic to detail synthetic process of i ‐PAP hydrogel. b) Photography of hydrogel sample twisted on glass tube. c) Cross section and surface scanning electron microscopy (SEM) images of lyophilized i ‐PAP hydrogel. Inset demonstrates the surface structure of same sample. Scale bar: 50 µm. Owing to PAM scaffold and two interentangling polymer chains, our hybrid hydrogel demonstrated excellent stretchability ( Figure 2 a), which is more remarkable than any single of its constituents, the polyacrylamide and alginate‐PBA gels (Figure S5, Supporting Information). Tensile ability of hydrogels was tested by a home‐made device (Figure S6, Supporting Information) and evaluated through comparing critical relative strain (defined as maximum stretching displacement without rupture to its original length (mm mm −1 )). The average critical relative strain of gels is >15 in statistically (Figure 2 b, c), with 16 for control hydrogel, 24 for Ca‐conditioned hydrogel (Figure 2 b) and 20 for alkaline‐conditioned hydrogel (Figure 2 c), respectively. The enhanced stretchability of hydrogels treated with Ca 2+ or OH − was probably derived from the ionic crosslinking or dynamic boronate ester bonds, thus dissipating the applied energy and enhancing the stretchability of conditioned hydrogels. To best of authors' knowledge, this is the largest stretching degree compared with the state‐of‐the‐art multistimuli‐responsive hydrogels. 13, 25 Nevertheless, neither the concentration of CaCl 2 (Figure 2 b) nor immersion time of NaOH (Figure 2 c) had obvious effects on the stretching performance. This result implied that bonding sites for Ca 2+ in control gel were finite and easy to be occupied in short time. It is worth to be mentioned that immersion time of alkaline solution is limited to 1 h because polyacrylamide tends to hydrolyze under alkaline conditions. Figure 2 Stretching performance of original and conditioned i ‐PAP hydrogels. a) Tensile demonstration of original i ‐PAP hydrogel. b, c) Maximum tensile ability of i ‐PAP hydrogel varies with concentration of CaCl 2 and immersion time of 1 mol L ‐1 NaOH. The results are illustrated as the mean ± s. d. of four independent experiments. d) Tensile stress–relative strain curves of three hydrogels until relative strain achieves 12. e, f) Tensile stress–relative strain curves of Ca‐conditioned hydrogel and alkaline‐conditioned hydrogel samples subjected to loading–unloading tensile cycles under 200, 400, and 600% strains, successively. g–i) Tensile stress changes with time of i ‐PAP hydrogel, Ca‐conditioned hydrogel, and alkaline‐conditioned hydrogel samples loading with 100% strain for 100 cycles. The insets show details of corresponding curves. To quantify tensile stresses of hydrogels, a tensile tester with a 100 N load cell was employed to monitor stretching process. As exhibited in tensile curves (Figure 2 d), the larger tensile stress of Ca‐conditioned hydrogel indicated the strong ionic crosslinking interaction. Then, we loaded a sample of conditioned hydrogels up to certain tensile strains (200, 400, and 600%), unloaded the gel to zero force, and followed with a second and third loading to study recovery of hydrogels after tensile. The remarkable hysteresis loop of Ca‐conditioned hydrogel (Figure 2 e) reflected the effective energy dissipation. In contrast, negligible hysteresis was detected in alkaline‐conditioned gel sample (Figure 2 f). On the other hand, loading–unloading hydrogel samples in 100% strain for 100 cycles were operated. Control (Figure 2 g) and alkaline‐conditioned (Figure 2 i) hydrogels exhibited superior fatigue resistance 26 as no significant tensile stress changes were observed. The reformation of hydrogen bonds between polymer chains and dynamic boronate ester bonds could account for the reinforced fatigue resistance. Maximum tensile stress of alkaline‐conditioned hydrogel (1. 5 kPa) was smaller than control sample (2. 0 kPa), indicating these gels were more susceptible to subtle external force (Figure 2 i). Under the same conditions, tensile stress of Ca‐conditioned gel was decreased by almost 50% undergoing 100 tensile cycles (Figure 2 h). Therefore, Ca‐conditioned hydrogels were unable to recover to its original state with distinct deformation after large stretching, while alkaline‐conditioned hydrogels made it at the macroscale. This could be explained that ionic crosslinking of alginate‐PBA was different from dynamic interactions and hard to reform to result in crack propagation. With predominant stretchability and fatigue resistance advantages, i ‐PAP hydrogels had a promising potential in substrate material of wearable electronics with high sensitivity and durability. Apart from excellent stretchability, i ‐PAP hydrogels manifested multiple functionalities related to external stimuli. First, actuatable functionality was achieved via responsiveness to moisture. In hydration and dehydration cycle, the i ‐PAP hydrogel showed reversible response to moisture with great volume changes. The volume of original hydrogel fiber expanded fivefolds after complete swelling in water, while returned to the original state after dehydration in air ( Figure 3 a). To demonstrate actuation function, classical asymmetric bilayer structures 27 were used by assembling tailored rectangular i ‐PAP hydrogel samples with polydimethylsiloxane (PDMS) sub‐millimeter (0. 5 mm) films. The PDMS films were hydrophilic modified by oxygen plasma. The Janus assembly dehydrated rapidly in 60 °C oven and bent spirally contributing to different physical properties of i ‐PAP hydrogel and PDMS. 28 During dehydration, contraction force generated by volume shrinkage resisted to the gravity of the whole assembly and deformed it. Deformation patterns changed with degree of dehydration until i ‐PAP hydrogel divorced from PDMS sheet (Figure 3 b). Once rehydrated, deformed assembly induced by hydrogel dehydration was able to return to its original shape, just as a simple actuator. Nevertheless, this actuation function of pure PAM gel or alginate‐PBA gel has hardly been reported to our knowledge. Figure 3 Actuatable functionality of i ‐PAP hydrogel in response to moisture stimuli. a) Digital photograph of reversible swelling–shrinking transformation of i ‐PAP hydrogel. b) Deformation process of Janus assembly containing hydrogel and hydrophilic PDMS during dehydration and rehydration. c) Diverse patterns of “plastic flowers. ” Not only spirally bending deformation achieved without intervention, sophisticated patterns would realize by elaborately preorganizing. As shown in Figure 3 c, a series of attractive “plastic flowers” were designed. PDMS strips served as petal and hydrogel as the bud of flowers. Before dehydration, original petals were horizontal with center of flower poking to the tip of syringe needle. The factors which change the driving forces, like contact areas, geometrical shape, and relative arrangement, all played essential roles in determining the eventual morphologies of flowers. The bigger size of bud which meant greater driving forces, the higher bending curvature of petal, thus sample c′ presented the smallest bending curvature in vertical. And, different geometrical shapes of hydrogel could result in more than one bending orientations. For example, comparing sample a′ and b′ in Figure 3 c, there also appeared horizontal spiral 29 in sample a′ except for bending vertically, and sample b′ with isosceles triangle bud displayed axial bending at the tip of petal. Additionally, to get flower with uniform bending curvatures, programing mini‐gel strips on PDMS petal regularly was useful, like Figure 3 c′, d′, bending curvatures in four arms of petal were more uniform than other patterns. The other flowers we had fabricated are exhibited in Figure S7 (Supporting Information). It could be reasonably deduced that infinite patterns should be precisely designed beyond the configurations demonstrated above. Taking advantage of responsiveness to Ca 2+ which could interact with vacant guluronic acid (G unit) of alginate moieties, shape memory of hydrogel was explored. We regulated the shape of i ‐PAP hydrogel through introducing or eliminating Ca 2+ inside the gel. To conduct shape writing procedure, two samples of i ‐PAP hydrogel with different thickness (thickness: 1. 5/0. 75 mm, width: 1 cm) were tailored and curled in spiral (type A, 1. 5 mm thick) or annular cylindrical (type B, 0. 75 mm thick) on glass rods. The thinner (0. 75 mm) film sample was enwound in annular cylinder layer‐by‐layer to better support its own weight since it was much softer than thicker gels. With ionic bonding between Ca 2+ and alginate‐PBA as the reversible crosslinks, this temporary shape could be recorded by immersing hydrogel in 0. 1 mol L ‐1 CaCl 2 solution and easily eliminated by being transferred into EDTA solution ( Figure 4 a). Shape programing and immersing procedures made the straight samples shape morphed, and helical shape or annular cylinder was memorized even assisting glass rod was released. Preprogrammed i ‐PAP hydrogel maintained a stable helical configuration when put into CaCl 2 solution (see Video S1, Supporting Information). Moreover, when resoaked in the solution of CaCl 2, decalcified hydrogel was capable of morphing into spiral shape (Figure S8, Supporting Information). Interestingly, Na 2 CO 3 basic solution was beneficial to weaken the temporary shape of hydrogel as well (Figure S9, Supporting Information). The constructed helical hydrogel recovered to its original state because of the releasement of calcium ions. Figure 4 Shape memory and self‐healing properties of i ‐PAP hydrogels respond to external Ca 2+ and alkaline conditions, respectively. a) Reversible shape memory effect occurs in the presence of Ca 2+ and could be erased by immersing in EDTA·2Na solution. b) Hydrogel images of before and after healing. c) Loading the healing hydrogel with biaxial tensile force for several cycles, the gel still maintains an intact. In spite that shape memory and self‐healing are two kinds of contradictory properties, 30 automatic self‐healing ability of i ‐PAP hydrogel in alkaline conditions was achieved. As have been extensively studied, boronic acid groups have high binding affinity with diols when pH of surroundings is above the p K a of boronic acid, forming dynamic boronate esters. 31 Considering that phenylboronic acid and diol units coexisted in the gel, self‐healing between cut surfaces of i ‐PAP hydrogel was expected. Before cut into two halves, hydrogel sample was immersed in NaOH for a few seconds to offer a suitable alkaline condition for generating diol–boronate anion complex. The incisions of the cut halves were put together and contacted with each other. Not surprisingly, automatic self‐healing at 4 °C was observed after a week (Figure 4 b). In spite of being loaded to biaxial tensile force, the healing gel resulting from dynamic boronate ester bonds was strong enough to withstand several tensile cycles without rupture (Figure 4 c and Video S2, Supporting Information). The stress–strain curve of healed gel coincided partially with result of virgin alkaline‐conditioned gel and the joint reformed between the cut surfaces could resist a fracture stress up to 4. 0 kPa (Figure S10, Supporting Information). The i ‐PAP hydrogel treated with NaOH possesses better recovery and self‐healing property and is more preferable substitute material demanding for antivulnerable and repeatable. In conclusion, this work successfully provided a smart hydrogel with multiple functionalities and stretchability. This multiresponsive hydrogel showed responsiveness to three categories of ordinary stimuli, namely, moisture, multivalent cations, and pH, exhibiting amusing actuation function, excellent shape memory, and automatic self‐healing properties in low temperature. Based on the mismatch of volume change ability after dehydration, we fabricated a series of “plastic flowers” with various patterns by assembling hydrogel strips with hydrophilic PDMS films to validate actuation function of hydrogel. In theory, appearance of flower could be elaborately regulated by tuning size, geometry, and distribution of hydrogel sample. With compatible multifunctionality and stretchability, this smart hydrogel broadens the scope of functional materials and improves the flexibility and versatility of hydrogel applications, which have great prospects to be used in fields of actuators, soft robotics, electronic skin, and other applications. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201800471
| 2,018
|
Advanced Science
|
Elucidating Self‐Assembling Peptide Aggregation via Morphoscanner: A New Tool for Protein‐Peptide Structural Characterization
|
Abstract Self‐assembling and molecular folding are ubiquitous in Nature: they drive the organization of systems ranging from living creatures to DNA molecules. Elucidating the complex dynamics underlying these phenomena is of crucial importance. However, a tool for the analysis of the various phenomena involved in protein/peptide aggregation is still missing. Here, an innovative software is developed and validated for the identification and visualization of b ‐structuring and b ‐sheet formation in both simulated systems and crystal structures of proteins and peptides. The novel software suite, dubbed Morphoscanner, is designed to identify and intuitively represent b ‐structuring and b ‐sheet formation during molecular dynamics trajectories, paying attention to temporary strand‐strand alignment, suboligomer formation and evolution of local order. Self‐assembling peptides (SAPs) constitute a promising class of biomaterials and an interesting model to study the spontaneous assembly of molecular systems in vitro. With the help of coarse‐grained molecular dynamics the self‐assembling of diverse SAPs is simulated into molten aggregates. When applied to these systems, Morphoscanner highlights different b ‐structuring schemes and kinetics related to SAP sequences. It is demonstrated that Morphoscanner is a novel versatile tool designed to probe the aggregation dynamics of self‐assembling systems, adaptable to the analysis of differently coarsened simulations of a variety of biomolecules.
|
1 Introduction The spontaneous organization of initially chaotic biological systems has helped scientists to explore the origins of Life on Earth. 1, 2, 3, 4, 5 Indeed, it is widely accepted that the self‐assembling propensity of DNA, RNA, proteins, and peptides is one of the main molecular mechanisms that may have sparked Life as we know it. The so called “bottom‐up” approach to design self‐assembling materials, is directly inspired by this fascinating phenomenon. 6, 7, 8 Self‐assembling materials are also used as models to study molecular mechanisms that cause abiogenesis and Alzheimer's disease. 1, 2, 3, 4, 5 Moreover, self‐assembling peptides, inspired by the properties of biomolecules, have been developed for applications in diverse nanoscience sectors such as electronics, material science, and regenerative medicine. 9, 10 The identification of stable domains that act as novel structuring motifs is critical for the development of self‐assembling biomaterials. 7, 8, 10 Furthermore self‐assembling peptides (SAPs) are promising building blocks for tissue engineering due to their favorable biocompatibility, tailorability and biomimetic properties. 11, 12, 13, 14 In the last decade our efforts have been focused on the development of SAP hydrogels for nervous regeneration and we designed different classes of SAPs such as functionalized SAPs, 15 complementary coassembling peptides (CAPs) and BMHP1‐derived SAPs. 16, 17 These peptides, featuring a promising pro‐regenerative potential in neural tissue engineering applications, form differently β‐structured filaments depending on their sequences and, in the case of BMHP1‐derived SAPs, depending on the presence of biotin at the N‐terminus. 17, 18 Others demonstrated that, before assembling into nanoscaled filaments, SAPs initially self‐aggregate into oligomeric molten globules that are shaped by hydrophobic interactions. 19, 20, 21 Within such molten globules peptides can adopt conformations of packing similar to paracrystalline and ordered nuclei. According to the protein nucleation mechanism this organization is due to the interplay between the forced spatial confinement of peptides and optimized (in terms of potential energy) nonbonded interactions. 22 Smith and colleagues demonstrated that amyloid cross‐β structures of the peptide Aβ(16–22) can assemble through a dynamic conformational phylogeny. In their work isotope‐edited infrared spectra allowed to quantify the relative distribution of paracrystalline intermediates formed from intermolecular molten globules in which nucleation occurred previously. 19 Furthermore, molecular dynamics simulations are a powerful tool to study peptide self‐assembly, and showed that molten peptide oligomers could act as incubators for β‐structuring. 23, 24, 25, 26, 27, 28, 29, 30, 31, 32 Nonetheless, at the molecular level it is still unclear how oligomer‐to‐fibril transition emerges. Indeed, currently available tools to analyze molecular dynamics do not allow to track key events of the self‐assembling process such as the evolution of secondary structure patterns over time. Coarse‐grained molecular dynamics (CG‐MD), enabling the simulation of larger systems on longer simulation times, showed great potential for high throughput screenings of the self‐assembling propensity of biomolecules. In addition, CG‐MD simulations allowed to estimate self‐assembling propensity of different peptide sequences for a wide latitude of potential applications. 33, 34, 35, 36 Nonetheless, analytical tools for the quantitative tracking of secondary structure patterns (such as β‐structures) over MD trajectories are still lacking. This is an important limitation, given that knowledge of the time‐dependent formation of secondary structures is crucial for a deeper understanding of the self‐assembling phenomenon. In order to recognize β‐structuring domains, we developed a topological pattern recognition software based on the multilayer graph theory, named Morphoscanner, and we validated it on diverse protein structures. Here, we applied Morphoscanner to MARTINI CG‐MD simulations of peptide systems featuring spatial dimensions that are typical of molten particles. 1, 19, 20 Simulations have been designed to mimic the experimental conditions that trigger peptides to self‐assemble into nanostructured hydrogels. Thanks to the high adaptability of our software‐suite to different sequences and system sizes, we could demonstrate that SAPs exhibit sequence‐dependent intraoligomer organization in agreement with previously described self‐assembly models. 15, 16, 17, 18, 19, 20 2 Results and Discussion 2. 1 Morphoscanner Network analysis and graph theory have found many applications in the study of protein structures and dynamics. Many studies focused on the development of algorithms for recognition or prediction of secondary structures such as α‐helix and β‐sheet arrangements. Due to of the large number of information to be taken into account (e. g. , solvent accessibility, contact potentials, residue types), 37, 38, 39, 40 however, the main application of the currently available algorithms consists in the analysis of single Protein Data Bank (PDB) structures. Indeed, many software packages cannot be applied to analyze molecular dynamics (MD) trajectories. Furthermore, currently available MD analysis tools for secondary structure analysis do not include a software adaptable to different coarsening levels of the simulated systems. 41, 42, 43 To identify β‐sheet arrangements and to study their relative alignments in MD simulations we developed the Morphoscanner tool. The classical flat β‐sheet arrangement was first described by Pauling and Corey as a rectangular flat shape. However, crystallographic studies showed that β‐sheets tend to fold into saddle‐shaped surfaces as result of the interplay of individual peptide twisting and interchain hydrogen bonding. While the recognition of the hydrogen bond pattern is important for the identification of β‐structures, hydrogen bonds (H‐bonds) are sometimes not explicitly modeled in MD models. To identify both flat and twisted arrangements compatible with β‐sheets, Morphoscanner represents the peptide system as a 2D‐lattice graph defined on two axes: 42, 44 one runs parallel to the backbone direction, the other one goes parallel to the H‐bonds direction. The edges along backbone direction represent covalent bonds. Instead, each edge along H‐bonds direction represents a β‐contact if the center‐of‐masses of two backbone group‐of‐atoms belonging to different peptides are closer than 4. 7–5. 3 Å, which is the typical inter β‐strand distance in β‐sheet structures (see Supporting Information). 45 The β‐contacts network has been numerically represented trough a matrix (dubbed BB matrix) which is used to dynamically rationalize the global and local amount of the “in & out‐of‐register” mutual disposition of strands in the system. 46 For this purpose, the BB matrix has been coarsened from residue‐to‐residue to strand‐to‐strand interaction level, thereby yielding a strand potential β‐interaction matrix named P matrix. Such “potential β‐interaction” is used to underline that the topological organizations identified by Morphoscanner are compatible with β‐structures. For sake of brevity “β‐interactions” will be used along the text instead of “potential β‐interactions. ” The calculation of β‐interactions was achieved through a pattern‐matching algorithm comparing the BB matrix with a set of shift matrices describing all possible mutual alignment between two strands. 47 Indeed, each shift matrix describes a single mutual alignment of two adjacent strands in function of the k parameter called shift value (see the Supporting Information for details) indicating their respective degree of sliding. For sake of clarity the possible parallel and antiparallel alignments (with positive or negative shifts) of the tested peptides are represented in Tables S6 and S7 of the Supporting Information. The dynamic reconstruction of topologies compatible with β‐structuring has been faced with a dynamic multilayer network approach. 48 The interaction network formed by all residues in the system is investigated using the β‐contacts (BB matrix) and β‐interactions ( P matrix), then the potential β‐structures are identified through the following heuristic: 1) a triplet of consecutively adjacent strands is identified in the P matrix, 2) and that same triplet has to satisfy in the BB matrix the minimum conditions for the number and distribution of inter‐strands H‐bonds. In brief, in a triplet of strands, the same portion of one strand must form a minimum of three β‐contacts with each of two neighboring strands. The iteration of such continuity criteria allows to identify both stable and evolving β‐sheets. 2. 2 Morphoscanner Validation The primary input for Morphoscanner is a series of contact map derived from PDB structures or molecular dynamics trajectories. Morphoscanner requires the following information: the number of strands (S) and the number of amino‐acid residues per strands (strand length, SL) in which the protein sequence should be divided. Morphoscanner returns different outputs and calculates the β‐strand percentage (%Ms) as follows (1) % Ms = Number of β –strands Number of strands ∗ 100 Also, we introduced an intuitive graphical representation called “shift profile” in order to highlight the preferential arrangement of strands. To validate Morphoscanner we analyzed some protein structures as reported in Figure 1. As our main focus was to characterize coarse‐grained systems of SAPs, protein structures from PDB were CG‐mapped according to the MARTINI model and subsequently analyzed using Morphoscanner (given the intrinsic versatility of Morphoscanner, this procedure could have been performed on other levels of structural coarsening). Thanks to web server STRIDE, the secondary structures assignment for each PDB structure could be readily computed. The STRIDE output files were evaluated through an in‐house developed “R script” which works similarly to Morphoscanner. The above cited script returns the β‐strand percentage (%S*), similarly to the Morphoscanner output (2) % S ∗ = Number of β –strands Number of strands ∗ 100 Figure 1 Morphoscanner validation on different protein structures. A series of PDB structures were analyzed with STRIDE‐based R script and Morphoscanner. In the first column, the reference PDB structures are represented as cartoon using VMD. In the second column, CG structures are visualized highlighting β‐sheets identified through Morphoscanner. β‐strand percentages calculated via R script (%S*) and Morphoscanner (%MS) are shown in the third column. In addition, shift profiles were used to quantify strand displacement in each structure. In the last column we depicted just the predominant shift profile. P = parallel alignment; A+ = antiparallel alignment with positive shift; A‐ = antiparallel alignment with negative shift. The analyses of 2mxu (SL = 32, S = 12) were in agreement and showed that strands were parallel aligned. In 2fkg (SL = 9, S = 35) strands were preferentially antiparallel aligned. The same conclusions were reached for 1d2s (SL = 10, S = 34) and 3bep (SL = 6, S = 122) analysis. The comparison between %S* and %Ms was used to determine if Morphoscanner successfully identified β‐sheet structures. Amyloid plaques are a defining characteristic of Alzheimer's disease. The Aβ(1‐42) fibrils is the initial and predominant constituent of amyloid plaques. We investigated the Aβ(1‐42) amyloid fibrils (2mxu) structure. To analyze this structure, SL and S were set to values of 32 and 12, respectively. The ssNMR analyses of Aβ(1‐42) amyloid fibrils revealed parallel β‐strands arrangement. 49 The Morphoscanner analyses were fully in agreement with the abovementioned structural investigations, and they were confirmed by a STRIDE analysis of PDB structures (%S* = 100, %Ms = 100), as shown in Figure 1. Indeed, all the potential β‐sheet structures showed a preferential parallel in‐register alignment. We further tested Morphoscanner on an engineered Boriella OspA structure (2fkg), a β‐sheet rich self‐assembly mimicry. Its structure consists of β‐hairpin repeats connected by turn motifs. 50 To analyze the 2fkg structures, SL and S were set to 9 and 35, respectively. The analyses through STRIDE and Morphoscanner provided similar results (%S* = 85, %Ms = 77). In agreement with structural analysis performed by Makabe et al. , Morphoscanner identified an antiparallel out‐of‐register β‐strands organization (see Figure 1 ; Figure S1, Supporting Information). In addition, β‐sheet profiles revealed a different progressive strand displacement accountable by twisting and bending between different strands. Morphoscanner was also tested on different biological assemblies that have pivotal roles in diverse metabolic pathways. Laminin are high‐molecular weight proteins of the extracellular matrix and constitute the biologically active part of the basal lamina, influencing cell differentiation, migration, and adhesion. These proteins consist of different subunits comprising the lamin‐g‐like module (see 1d2s in Figure 1 ) which mediates the binding to different molecules such as heparin and the cell surface receptor alpha‐dystroglycan (alpha‐DG). 51 To perform the analysis of the 1d2s structure, SL and S were set to 10 and 34, respectively. The Morphoscanner and STRIDE analyses reported the same results (%S* = 88, %MS = 91). In addition, as demonstrated by crystallographic analyses, Morphoscanner revealed a preferential antiparallel alignment among potential β‐strands (see Figure S1, Supporting Information). Escherichia coli beta clamp is a subunit of the DNA polymerase III holoenzime which consists of two identical subunits, made of 366 residues each. 52 To obtain a periodic division of 3bep structure, SL and S were set to 6 and 122, respectively. Morphoscanner identified the antiparallel β‐sheet structures, as shown in Figure 1 ; Figure S1 (Supporting Information), which could be perfectly superimposed to β‐sheet representation obtained by VMD. This was also demonstrated by comparison between Morphoscanner and STRIDE statistics (%S* = 56, %Ms = 56), as shown in Figure 1. 2. 3 Modeling of Assembling Systems: BMHP1‐Derived SAPs, CAPs, and (LDLK) 3 Looking for a broad SAP analysis, CG‐MD simulations were used to study the self‐assembling propensity of seven punctually mutated BMHP1‐derived SAPs, 18 the almost neutral (LDLK) 3 SAP, 53 and the two complementary charged (LDLD) 3 + (LKLK) 3 CAPs (see Table 1 ). 16 Systems comprised a total of identical 100 peptides for BMHP1‐derived SAPs and (LDLK)3, and 50 plus 50 oppositely charged peptides in case of mixed CAPs. Table 1 All BMHP1‐derived SAPs (2, B3, 4, B24, B26, 30, 31) systems were simulated at 3% (w/v) to mimic the standard empirical conditions enabling nanostructured hydrogel formation. (LDLK) 3 and CAP concentration was 1% (w/v). SS parameters of all residues were set to extend. Three CGMD simulations have been carried out for each system up to 500 ns: one simulation per each set was prolonged to 2000 ns. Lastly, simulations were further prolonged to 4500 ns for (LDLK) 3, (LDLD) 3 + (LKLK) 3, B24 and 30 Sequence ID Sequence Box size [nm] CG ions beads (NA+/CL−) CG Water beads N° of peptides N° sim x time [ns] 2 GGGPFSSTKT 17. 56 50/150 43770 100 1 × 2000 2 × 500 B3 Btn‐GGGPFSSTKT 18. 6 60/160 52213 100 1 × 2000 2 × 500 4 WGGGPFSSTKT 18. 61 60/160 52480 100 1 × 2000 2 × 500 B24 Btn‐GGGAFASTKT 18. 35 58/158 50314 100 1 × 4500 2 × 500 B26 Btn‐GGGPFASTKT 18. 52 59/159 51038 100 1 × 2000 2 × 500 30 WGGGAFASTKT 18. 4 58/158 50311 100 1 × 4500 2 × 500 31 WGGGAFSSTKT 18. 42 58/158 50612 100 1 × 2000 2 × 500 (LDLK) 3 LDLKLDLKLDLK 28. 8 0/0 200058 100 3 × 4500 (LDLD) 3 + (LKLK) 3 LDLDLDLDLDLD+LKLKLKLKLKLK 28. 8 184/184 199649 50 + 50 3 × 4500 John Wiley & Sons, Ltd. In this work molecular interactions in CG‐MD simulations were modeled by the MARTINI force field that has recently showed promising potential for the high‐throughput screenings of SAPs. 54, 55, 56 In MARTINI, four heavy atoms are usually represented by one CG bead, while a lower ratio is used for atoms involved in rings. Bonded interactions are described with bond, angle and dihedral energy functions, while nonbonded interactions are described through Lennard‐Jones and Coulomb functions. 55, 56, 57, 58 Given that some of the BMHP1‐derived SAPs include N‐terminal Biotin‐tag, which was not yet available for the MARTINI force‐field, we parametrized the biotin tag as follows: structural and interaction parameters were extrapolated from previous UA simulations and validated through octanol/water partition coefficient (logP) calculations (see the Supporting Information for details). 18, 54, 55 Experimental and calculated logP values did not show significant differences. Notably, in MARTINI the secondary structure of molecules is fixed throughout the simulation, therefore the choice of the secondary structure (SS) parameters is crucial for the reliability of the modeled system. A fully extended secondary structure was adopted for both (LDLK) 3 and CAPs because of 1) the presence of equally spaced identical or opposite charges along the same short peptides and 2) their typical β‐sheet signature in circular dichroism spectra. 16, 53 In case of BMHP1‐derived SAPs, in line with previously published works, 15, 17, 18, 59 the chosen secondary structure assignment was initially derived by comparing united atom (UA) and CG simulations of octameric systems. In CG simulations of octamers, three different secondary structure sets (all extended, all coil or sampled conformational distribution of monomers in UA simulations) were combined with two starting structures distributions: all extended and sampled configurations of monomers in UA. After comparing gyration radius, aggregation order, and alignment degree (see the Experimental Section for details and Figures S2–S4, Supporting Information) of UA and CG simulations of octamers, we chose fully extended secondary structures and sampled structural configurations for the subsequent CG simulations of BMHP1‐derived peptides 100‐mer systems. 55 2. 4 Using Morphoscanner for the Analysis of Self‐Assembled Peptidic Aggregates In MARTINI CG‐MD simulations, the fixed SS parameters allow to discriminate between various secondary structures, however, they do not allow to detect any secondary structure transitions. Notwithstanding this limitation, it is possible to evaluate the movement of secondary structure elements in the simulated systems. 54, 55 Morphoscanner was used for the analyses of the CG‐MD simulations of SAPs in Table 1 with extended secondary structure parameters. S and SL parameters were set equal to the number of peptides and of backbone grains per peptide, respectively. The organization of the simulated systems over time was schematized into a count of both total the β‐interactions in the system and the percentage of peptides taking part in potential β‐sheets formation ( Figures 2 and 3 ). We used a shift profile representation over time (Figures 2 and 3 ) to track peptides preferential arrangement during self‐assembling. Lastly, the shift profile approach was adopted to monitor peptides arrangement within β‐sheets structures (Figure S5, Supporting Information). Figure 2 β‐interactions and β‐structuring of SAPs in CG‐MD simulations with extended SS parameters. The onset of β‐interactions does not warrant the formation of β‐sheet structures. This is clearly evident from the comparison among peptides 2, 4 and B26. The above‐mentioned SAPs reached the same number of β‐interactions, but B26 had the lowest degree of β‐structuring propensity, followed by 2 and 4. Such features are attributable to their sequences and, in particular, to N ‐terminal functionalization. Figure 3 Analysis of mutual alignment of peptides featuring diverse self‐assembling propensities. Peptides mutual alignment shift profiles of SAP 2, B24, and 30 which were simulated with extended secondary structure parameters (see Table 1 ). P refers to parallel alignment, A+ to antiparallel alignment with positive shift, A‐ to antiparallel alignment with negative shift. BMHP1‐derived SAPs preferentially shifted by one residue in P alignment, but (LDLK) 3 and CAPs showed much stronger alignment in both P and A‐ alignments at one residue shift. This feature was likely due to the electrostatic interactions among their complementary charged side‐chains. On the other hand, the mutation of Pro and Ser with Ala increased the number of β‐interactions in B24 and 30 assemblies if compared to SAP 2 (see Table 1 ). Biotinylation also slightly improved β‐sheet structuration propensity in B24 in respect to 30. Notably, CAPs and (LDLK) 3 showed less β‐interactions than BMHP1‐derived SAPs. This was due to the different shapes of supramolecular aggregates; (LDLK) 3 and CAPS formed bilayered β‐sheet‐rich aggregates. BMHP1‐derived SAPs formed ovoid aggregates where peptide strands could simultaneously interact with multiple surrounding peptides. The total number of β‐interactions in BMHP1‐derived SAPs, CAPs, and (LDLK) 3 was 150 to 240. B24 showed the highest number β‐interactions, while the lowest numbers were found in CAPs and (LDLK) 3 (Figure 3 ; Figure S5, Supporting Information), caused by different peptides arrangement within oligomers. CAPs and (LDLK) 3 assembled into bilayered structures made of peptides packed side‐by‐side. Instead peptide B24, similarly to other BMHP1‐derived SAPs, assembled in ovoid oligomers where interactions among neighboring peptides were favored. BMHP1‐derived SAPs preferentially aligned in parallel out‐of‐register of one residue (≈10–15% of total β‐interactions). CAPs and (LDLK) 3 were preferentially shifted by one residue in P (≈10%) and in A‐ alignments (≈25%). As shown in Figure 3 and Figure S5 (Supporting Information), all the potential β‐interactions in aggregates formed by CAPs took part in β‐sheet structures. Indeed, all peptides contributed to β‐sheet formation in all simulations within 50 ns (data not shown). BMHP1‐derived peptides generally showed a variable β‐structuring propensity related to the punctual mutations in their sequences (see Table 1 ). SAP 2, made of the BMHP1 motif and a triplet of Gly, did not show a good β‐structuring propensity. Only 10% of the total simulated SAP 2 peptides took part in β‐sheets formation (Figure S1, Supporting Information) and they preferentially aligned in parallel out‐of‐register by 1 residue (Figure 4 ; Figures S6 and S7, the Supporting Information ). The introduction of Trp at the N‐terminus improved the β‐structuring propensity of SAP 4, with 25% of peptides involved in β‐structuring (Figure S5, Supporting Information). Biotinylation increased the β‐sheet structuring propensity in B3: indeed 30% of peptides were involved in β‐sheets structures (Figure 2 ). B24 showed the highest propensity to β‐sheet structuring. Indeed, 50% of B24 peptides fell within β‐sheets (Figure 2 ), and the large part of pairs of β‐strands were preferentially aligned in parallel out‐of‐register with neighboring pairs by one and two residues as shown in Figure 1 and Figure S7 (Supporting Information). These increments were ascribable to both the N‐terminal biotinylation and the substitution of Pro and Ser with Ala. The substitution of Btn with Trp decreased the formation of stable β‐sheet structures in peptide 30 in comparison with B24 (Figure 2 ): even if they showed similar preferential alignments (Figure 3 ; Figure S7, Supporting Information) just 40% of peptides were involved in β‐sheet structures. In the case of peptide 31 the first Ser of the BMHP1 motif was mutated with Ala, but apparently, when compared to SAP 30, did not alter the β‐structuring propensity of the system (Figure 2 ). The introduction of biotin at the N‐terminal position did not improve β‐structuring propensity in B26 (Figure 2 ): only 10% of peptides were part of β‐sheets. As shown in Figures S6 and S7 of the Supporting Information B26 peptides were preferentially aligned in parallel out‐of‐register by one residue and nine residues, likely because of preferential pairings between Lys backbone and Biotin amide groups. 60, 61 2. 5 Peptide Oligomer Identification To more efficiently describe the onset and subsequent arrangement of “seeds of self‐assembling” within the molten globules we combined our recently introduced methodology with classical analyses, such as radius of gyration and nematic order parameter. 42 However, in big system simulations locally ordered aggregates may not be described by cumulative parameters of the overall system: therefore, it was necessary to track the oligomers formed during the self‐assembling process. Indeed, others proposed a nucleation‐dependent polymerization model to describe fibril formation from monomeric peptides to heterogeneous nuclei (or peptide micelles) and finally mature fibrils. 62, 63, 64 Oligomer identification required the implementation of a “nearest neighbor algorithm” on binary entries obtained by thresholding distances among peptide center‐of‐masses. Threshold distance was set at 1. 1 nm as in XRD spectra it represents the typical equatorial distance of cross‐β structures. 64 In the so‐obtained contact matrix 1 value points at a distance falling within the chosen range, thus giving a “forest” of ones. 65 The algorithm first explores each subtree representing neighboring center‐of‐masses, then backtracks and provides the peptides constituting each oligomer. 66 2. 6 Tracking Oligomers Arrangement Dynamics and β‐Structures Organization The combined use of the Morphoscanner and of the oligomer identifier algorithm allowed to detect oligomeric species with different local structural features. Then, we obtained a more detailed analysis of the β‐structuring propensity of the simulated systems. As previously mentioned peptide B24 displayed good β‐structuring propensity (Figure 2 ) and the total amount of β‐interactions stabilized after 500 ns. However, Figure 4 A II shows that β‐sheets changed their internal organization (see also Figure S8, Supporting Information); at 2500 ns β‐strands were preferentially aligned in parallel out‐of‐register by one residue, but this was not the case at 3000 ns and at 4500 ns. Oligomers showed limited variations in global morphology and mutual disposition (Figure 4 B IV–VI ) over time. As shown in Figure 4 C, the nematic order parameter P2 of almost all identified oligomers at different time‐points fluctuated below 0. 5: this was indicative of a modest (but still developing) peptide arrangement over time. Figure 4 Structural characterization of B24 molten particles at different timeframes. B24 showed good β‐sheet propensity (A I ) characterized by parallel out‐of‐register β‐strands (A II ). Parallel β‐sheets shift profiles became wider between 2500 and 4500 ns: this was matched by changings in β‐sheet topology (B I–III ) and influenced the identification of oligomers (B IV–VI ). P2 was calculated for the identified oligomers (C I–III ). Same colors between B IV–VI and C I–III point at the same oligomers identified at the selected timeframes. P2 values of the identified oligomers were calculated for all timeframes. The identified oligomers ranged from 8‐mer to 25‐mer aggregates. Interestingly, oligomers (B IV–VI ) featuring higher order (or P2 values in C I–III ) showed a large presence of β‐sheets (B I–III ). The same analysis workflow was applied to peptide 30. In agreement with previous empirical studies, 17 the mutation of Biotin with Trp in peptide 30 yielded to a more stable β‐structuring ( Figure 5 ; Figure S9, Supporting Information) than in B24. Indeed, from the early stage of SAP 30 self‐assembling, β‐strands were preferentially aligned in parallel out‐of‐register by one residue (Figure 5 A; Figure S9, Supporting Information). In addition, more similar oligomers of peptide 30 were identified at different time‐points (Figure 5 B IV–VI ) and their calculated order was mostly higher than in B24 (Figure 5 C). Figure 5 Structural characterization of 30 molten particles at different timeframes. SAP 30 had a good β‐structuring propensity (A I ) and peptides were mutually aligned in parallel out‐of‐register by one residue within β‐sheets (A II ). Shift profiles of parallel β‐sheets became sharper after 2500 ns but did not vary as extensively as in B24. The topology of β‐sheets changed slightly (B I–III ): this was reflected in a modest variation of the identified oligomers at different timeframes (B IV–VI ). Same colours between B IV–VI and C I–III point at the same oligomers identified at the selected timeframes. P2 values of the identified oligomers were calculated for all timeframes. More ordered oligomers (C I–III ) were characterized by stronger presence β‐sheet structures (B I–III ). The oligomers identified at 4500 ns were more heterogeneous and with higher P2 values (C I–III ): big oligomers identified in previous timeframes were here split in two or more subgroups. CAPs, thanks to their alternated opposite charged side‐chains, self‐assembled into stable β‐sheets ( Figure 6 A), forming a “patchwork” of bilayered aggregates. 16 CAPs formed bilayered oligomers (Figure 6 B IV–VI ) characterized by a significant presence of β‐sheets (Figure 6 B I–III ) and by an high internal alignment (Figure 6 C). This behavior was dictated by strong electrostatic interactions among oppositely charged side‐chains fostering the formation of well‐defined β‐structures. Nonetheless, variable oligomer identification and unstable P2 values revealed a still ongoing arrangement of the system given by a persistent “sliding” of the patches composing the two layers. The same tendencies were also observed for LDLK 3 peptides. Indeed, as reported in our previous work, such peptides formed β‐structured and highly ordered bilayered aggregates. 67 Figure 6 Structural characterization of CAPs (LDLD) 3 + (LKLK) 3 molten particles at different timeframes. CAPs established less ß‐interactions (A I ) than in BMHP1‐derived SAPs and β‐strands were preferentially aligned in parallel out‐of‐register by one residue throughout the simulations (A II ). CAPs formed stable β‐sheet structures (B I–III ) mainly matching oligomers distribution (B IV–VI ). β‐sheets paired into bilayered aggregates but with different orientations. Same colours between B IV–VI and C I–III point at the same oligomers identified at the selected timeframes. P2 values of the identified oligomers were calculated for all timeframes. The identified oligomers displayed a superior order (values of P2) and a slow but ongoing trend of increments toward more ordered assemblies (C I–III ). 3 Conclusions Morphoscanner, a novel software developed for secondary structure analysis of differently coarsened simulations of proteinaceous structures, combines into a single “suite” the advantages of both MD analysis and secondary structures pattern recognition tools. 39, 40, 41, 42 On previously characterized protein structures Morphoscanner recognized their β‐sheet organization and provided new information about their relative orientation and alignment. In Morphoscanner we also included a new high‐throughput workflow to investigate different facets of self‐assembly: its graphical and quantitative analyses provided new insights of SAP systems evolution over time. It was possible to more efficiently elucidate the self‐assembly process of BMHP1‐derived peptides, (LDLK) 3, and CAPs. BMHP1‐derived SAPs self‐assembled into molten particles mostly composed of peptides aligned in parallel out‐of‐register, a thermodynamically stable alignment that however may prevent any subsequent evolution toward well‐structured nanofibers detected in previous experimental works. 1, 17, 19 On the other hand, (LDLK) 3 and CAPs formed patches of anti‐parallel β‐rich aggregates evolving toward cross‐β packings, yielding to highly ordered systems compatible (at longer timeframes) with empirical observations. 1, 16, 17 Lastly, we developed a software suite useful for the analysis of molecular assembly, easily adaptable to other chemical species and coarsening levels. Indeed it can be potentially applied to the study of biological processes such as DNA, RNA hybridization or abnormal protein assembly. 62, 68, 69 Thus, the achieved level of characterization may turn useful in nanotechnology but also in biomolecular and astrobiological studies focused on the emerging properties of self‐assembling systems. 2, 3, 4, 5, 6 4 Experimental Section MD of BMHP1‐Derived SAPs : The sequences of the simulated BMHP1‐derived SAPs are listed in Table 1. Peptide monomers have the C‐terminus amidated and the N‐terminus biotynilated (or acetylated). Lysine residues are in the protonated state. Extended conformations of monomers were built with Pymol software by imposing all‐trans geometry on the backbone dihedrals. Molecular dynamics were run using version 4. 5. 5 of the GROMACS simulation package and the GROMOS53a6 force field: systems comprised eight monomers each as reported previously and explicit aqueous solvent. 18 Coarse‐grained molecular dynamics simulations have been conducted on octameric or 100‐meric systems using MARTINI force field version 2. 2. The choice of secondary structure parameters for 100‐meric systems was made by comparing UA‐MD and CG‐MD simulations of 8‐meric systems of BMHP1‐derived SAPs. Choosing the Secondary Structure Parameters in CG‐MD Simulations of BMHP1‐Derived SAPs : To select the most appropriate secondary structure parameters for 100‐mer system simulations, UA and CG simulations of 8‐mer systems were compared. Gyration radius, nematic order parameter, and the aggregation curves were used to assess the agreement between UA and CG molecular models. 18 Starting configurations of the systems modeled in UA and CG simulations consisted of extended ( E ) or UA sampled (SAM) monomers comprising the octameric systems. Three different choices of SS parameters were proposed in CG‐MD simulations: fully extended, coil or UA‐sampled secondary structures (see Tables S8–S10, Supporting Information). The SAM secondary structures parameters are monitored on UA‐MD simulations by means of the DSSP algorithm as reported in the previous work. 18 CG‐MD simulations (using the abovementioned sets of SS parameters) and UA‐MD simulations were then compared. CG‐MD simulations of UA‐sampled conformers with fully extended SS parameters resulted in gyration radii, nematic order parameters, and aggregation orders in higher agreement with the UA simulations, as shown in Figures S2–S4 of the Supporting information. CG‐MD Simulations of 100‐mer Systems of BMHP1‐Derived SAPs : The boxes containing unsolvated peptides were built using the PACKMOL software. 70 UA‐sampled monomer conformations were inserted in random orientations and positions, so that the atoms belonging to different peptides were at least at 10 Å away from each other. 18 Boxes, filled with MARTINI CG water beads, were chosen so as to mimic the 3% (w/v) concentration of SAPs typically used in empirical tests. As mentioned in Section 2. 3 fully extended SS parameters were adopted (see Table 1 ). Ions (Na+ and Cl−) were added to neutralize the systems up to 0. 015 m concentration of NaCl, in order to reproduce salt concentration of diluted PBS (1x) solution. The production phase was conducted using constant temperature, pressure, and number of molecules (i. e. , the NPT ensemble). Temperature, pressure, constraints, cut‐off value, periodic boundary conditions, and integration‐step settings were identical to 8‐mer systems simulations. Three random distributions of peptides and ions were generated and simulated for 500 ns. One of the replicas as per each SAP sequence was extended up to 2000 or 4500 ns. CG‐MD Simulations of 100‐mer Systems of (LDLK) 3 and (LDLD) 3 +(LKLK) 3 : A similar approach was adopted for simulations of (LDLK) 3 and CAPs. All‐trans configuration of the (LDLK) 3, (LDLD) 3, and (LKLK) 3 were generated by Pymol ( http://www. pymol. org/ ). The C‐ and N‐termini of peptide monomers were amidated and acetylated, respectively. At neutral pH, lysine and aspartic acid side chains, because of their weak basic and acidic nature, can be considered fully protonated and deprotonated, respectively. Peptide were distributed (using PACKMOL) in explicit water cubic boxes. Prior to production, systems underwent an equilibration phase (a 3000‐steps minimization using steepest descent method). The production phase was conducted in NPT ensemble in order to reproduce experimental conditions used in previous works. 17, 67 Strand/Peptide Alignment Analysis via Morphoscanner : Despite the structural differences observed in crystallography, β‐sheets can be described as a regular 2D lattice graph stabilized by covalent bonds (along the direction of the backbone chains) and by hydrogen‐bonds (among the backbone chains). As previously mentioned, we introduced the definition of β‐contact to define the “edges” along H‐bonds direction (Equation (3) ) (3) β − contact i j = δ r i j − r 0 where δ is the Dirac measure, 71 r ij is the distance between backbone atom‐group (grain) center‐of‐masses i and j, r 0 represents the distance between two β‐strands in cross‐β structures (range between 4. 7 and 5. 3 Å). 45 The numerical representation of β‐contacts network is provided by the BB matrix, whose elements are described in Equation (4) (4) BB i j = β − contac t i j The description of the interactions between two strands or peptides is provided by the Strand Backbone Contact matrix: a square matrix whose dimensions correspond to the number of strand/peptide backbone grains. A set of matrices, named shift matrices, was developed to be used as references for the identification of the mutual arrangements described by the strand backbone contact matrix. Shift matrices describe the different arrangements between pairs of peptides. In detail, parallel and antiparallel shift arrangements are described using the following matrix notation: (5) Positive shift parallel arrangement → P i j + = δ i + k, j (6) Negative shift parallel arrangement → P i j − = δ i − k, j (7) Positive shift anti − parallel arrangement → A i j + = δ n − i + k, j (8) Negative shift anti − parallel arrangement → A i j − = δ n − i − k, j In the previous formulas δ ij identify the Kronecker delta, n is the number of peptide backbone grains, i, j are the indexes of peptide backbone grains (varying within n ), and k is the shift value. This set of matrices describing the peptide interaction library can be represented using the following compact notation: (9) L = L i j z where z is the index of shift matrices in the library. To calculate the maximum similarity of shift matrices with peptide backbone matrix, the normalized cross‐correlation function (NCC) was used 71 (10) NCC p, q, z = ∑ i = 0 RES − 1 ∑ j = 0 RES − 1 BB i + p ∗ RES, j + q ∗ RES ∗ L i, j, z ∑ i = 0 RES − 1 ∑ j = 0 RES − 1 BB i + p ∗ RES, j + q ∗ RES ∗ ∑ i = 0 RES − 1 ∑ j = 0 RES − 1 L i, j, z Z is the index of the shift matrix L ijz that maximize the value of NCC function. P and q denote the area of the BB matrix corresponding to the peptide backbone contact matrix. Each element of the strand potential β‐interaction matrix, describing mutual alignment between couples of strands, is defined as follows (11) P p q = Z β‐Sheet Reconstruction via Morphoscanner : Flat and twisted β‐sheet structures are detected by Morphoscanner by using the backbone contact and the strand interaction matrices, i. e. , handling the peptidic system as 2D‐lattice graph. In detail, the algorithm identifies a triplet of strands making a β‐structure. It calculates the area of the system backbone contact matrix, corresponding to the interaction between the first pair of strands, and reduces this area to a row vector, as shown below (12) v r = ∑ i = 0 n A i, 1, ∑ i = 0 n A i, 2, …, ∑ i = 0 n A i, n Morphoscanner identifies the area corresponding to the other pair of strands, giving another column vector (13) v c = ∑ j = 0 n A 1, j, ∑ j = 0 n A 2, j, …, ∑ j = 0 n A n, j Finally, the projection of v r on v c is calculated as a dot product (14) v p = v r ∗ v c = v r 1 ∗ v c 1, v r 2 ∗ v c 2, …, v rn ∗ v cn The number of consecutive residues defining a structuring β‐sheet along the covalent bonds direction is calculated as the maximum number of elements included between two non‐null elements. In this way, β‐sheet structures are identified as curved rectangular 2D‐lattices whose dimensions are defined by strands and by the number of backbone grains. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800506
| 2,018
|
Advanced Science
|
Oscillatory Strain Promotes Vessel Stabilization and Alignment through Fibroblast YAP‐Mediated Mechanosensitivity
|
Abstract Endothelial cells form the interior layer of blood vessels and, as such, are constantly exposed to shear stress and mechanical strain. While the impact of shear stress on angiogenesis is widely studied, the role of mechanical strain is less understood. To this end, endothelial cells and fibroblasts are cocultured under oscillatory strain to create a vessel network. The two cell types show distinctly different sensitivities to the mechanical stimulation. The fibroblasts, sense the stress directly, and respond by increased alignment, proliferation, differentiation, and migration, facilitated by YAP translocation into the nucleus. In contrast, the endothelial cells form aligned vessels by tracking fibroblast alignment. YAP inhibition in constructs under mechanical strain results in vessel destruction whereas less damage is observed in the YAP‐inhibited static control. Moreover, the mechanical stimulation enhances vessel development and stabilization. Additionally, vessel orientation is preserved upon implantation into a mouse dorsal window chamber and promotes the invading host vessels to orient in the same manner. This study sheds light on the mechanisms by which mechanical strain affects the development of blood vessels within engineered tissues. This can be further utilized to engineer a more organized and stable vasculature suitable for transplantation of engineered grafts.
|
1 Introduction During vascularization, endothelial cells (ECs) sense two forms of mechanical stimulation, shear stress, as a consequence of blood flow, and mechanical strain, as a consequence of rhythmic heart beating and tissue contraction. 1, 2, 3, 4 While many studies have focused on the effect of shear stress on vessel mechanotransduction, there is much less information on the effect of mechanical strain on vascularization‐related processes. 4 Recently, we have shown that vessels orient under mechanical stretch conditions. 5 While control of vessel orientation is important for vascularization of aligned tissues, such as ligaments, muscle, and the nervous system, the mechanism underlying this orientation under mechanical loading has yet to be established. A better understanding of these processes can enhance the integration progression of vascularized engineered tissues and affect the morphogenesis of anastomosing host vessels upon graft implantation. YAP, a transcription cofactor involved in Hippo signaling, which regulates organ size by controlling cell proliferation and apoptosis, is also known to be involved in the mechanotransduction of mechanical forces within different cell types. 6 YAP can be localized in the cytoplasm or translocated to the nucleus, where it associates with TEAD transcription factors. 7 Previous studies have shown that various forms of mechanical stimulation, such as substrate stiffness, strain, and shear stress, affected YAP localization by actin fiber remodeling. 6, 8, 9, 10 YAP was shown to be involved in the effect of shear stress on ECs and blood vessels, 10 however its role in the effect of mechanical strain on vascularization is yet to be established. In this study, we aimed to investigate the mechanism that regulates organization and remodeling of engineered vessels, and the ECs and mural cells composing them, into a stabilized and organized vessel network in vitro under oscillatory strain. To do so, ECs and fibroblasts were cocultured in gelfoam, a collagen stretchable scaffold, and subjected to oscillatory stretch for 21 d. Our results demonstrated that fibroblasts were more responsive to the stress stimulus as compared to ECs. The mechanical strain induced YAP translocation to fibroblast nuclei, which, in turn, increased their proliferation and differentiation. Moreover, distinct alignment and migration of the fibroblasts cells, alongside vessel development, stabilization, and alignment and a subsequent increase in cell secretion of pro‐angiogenesis cytokines, were observed. Since oscillatory strain was previously shown to affect YAP expression, 9 we hypothesized that it has a role in transmitting the mechanical stretch signal and affects vessel network development. YAP inhibition, by the addition of verteporfin to the culture media, caused higher damage to the vessels in the mechanically stimulated constructs compared to the vessels in the static control. Subsequently, the implantation of these stable, aligned vascularized structures into a mouse dorsal window chamber resulted in the organization of the penetrating host vessels. 2 Results 2. 1 Vessel Alignment under Oscillatory Strain Is Mediated by Fibroblasts To determine which cell type senses the stretch stimulus, cell alignment of monocultured ECs and fibroblasts on a gelfoam scaffold exposed to oscillatory strain was examined. On day 14, fibroblasts displayed a distinct orientation and alignment, whereas ECs were randomly arranged ( Figure 1 A, B). However, when fibroblasts were cultured at low densities, no alignment was observed (Figure S1, Supporting Information). When coculturing the two cell types together and applying oscillatory strain, fibroblasts displayed clear alignment on day 7, whereas the endothelial vessel network did not show significant alignment compared to the unstretched control. On day 14 of culturing, the endothelial vessel network displayed the same alignment as the fibroblasts surrounding it (Figure 1 C, D, G), hence, ECs only aligned in response to the mechanical strain when fibroblasts were present. To assess whether newly forming endothelial sprouts orient under oscillatory stretch conditions, ECs and fibroblasts were seeded separately at either end of the scaffold. Confocal imaging showed that the endothelial cells formed sprouts only at the point where they met the fibroblasts; the vessel sprouts were aligned and oriented in the same angle as the fibroblasts (Figure 1 E, H). Figure 1 Oscillatory stretch affects cell and network patterning. A) Endothelial cells and fibroblasts were seeded separately on two different scaffolds, and subjected to oscillatory strain. Scale bar = 100 µm. B) Quantification of cell alignment of (A). C) Coculturing ECs and fibroblasts under oscillatory stretch result in aligned vessels, where fibroblasts aligned before (day 7) the forming vessels (day 14). Scale bar = 200 µm. D) SEM images of the oscillatory stretched and control coculture constructs. Scale bar = 2 µm (upper panel) and 10 µm (lower panel). E) ECs and fibroblasts were seeded at either end of a single scaffold and subjected to oscillatory stretch. F) Fibroblasts were subjected to oscillatory stretch for 12 d, after which, ECs were seeded on top of the fibroblasts and seeded under static conditions. ECs formed aligned vessels, following the fibroblasts alignment. Scale bar = 100 µm. G–I) Quantification of cells (C, F), vessel (C, F), and sprout alignment (E). * p < 0. 05, ** p < 0. 01, *** p < 0. 001, **** p < 0. 0001. When culturing fibroblast‐seeded gelfoam constructs for 12 d under oscillatory stretch or static conditions and then adding ECs for an additional 5 d without applying stretch, the endothelial vessels followed the fibroblast alignment, and were significantly more oriented when compared to the unstretched control group (Figure 1 F, I). 2. 2 EC and Fibroblast Migration Patterns under Oscillatory Strain The differences in cell alignment in response to oscillatory stretch drove us to examine whether cell migration is also affected by oscillatory stretch stimulus. To this end, ECs and fibroblasts were seeded at the center of separate scaffolds and subjected to oscillatory stretch for 14 d. The ECs migrated radially throughout the entire scaffold area, whereas the fibroblasts showed a preference toward migration along the x ‐axis of the scaffold ( Figure 2 A, B, D). When culturing the cells together on the same scaffold, ECs in the middle and fibroblasts at both ends, or at the opposite ends, ECs spread radially from the middle and evenly from the side, whereas fibroblast distribution took on a triangle‐shaped pattern (Figure 2 C, and Figure S2, Supporting Information). However, when seeding the cells together in the same area of the scaffold, endothelial vessels formed and followed the same migration pattern as the fibroblasts (Figure 2 E, F). This phenomenon demonstrated again that under stretch conditions, within a 3D environment, fibroblasts sense and respond to mechanical stimulus, whereas ECs show less responsiveness to this stimulus. Figure 2 Cell migration throughout the scaffold. A, B) EC (green) and fibroblast (red) migration patterns, as observed on days 3, 7, and 14 of oscillatory strain. Fibroblasts were concentrated in the middle, whereas ECs were spreading throughout the entire scaffold. C) Migration pattern of ECs and fibroblasts when seeded at different scaffold locations. D) Quantification of monocultured EC and fibroblast migration patterns under oscillatory strain. Scale bar = 1000 µm. E) Migration pattern of ECs (green) cocultured with unlabeled fibroblasts (black) and seeded at one scaffold end and of fibroblasts seeded alone (red) at the other end and subjected to oscillatory strain. F, G) Cell migration under oscillatory stretching and compression, white dashed lines represent the scaffolds boundaries. Scale bars = 1000 µm (left panel) and 200 µm (right panel). H) Quantification of EC and fibroblast migration on day 7 under oscillatory stretching and compression. I) Quantification of cell migration on days 7 and 14. J) Vessel, formed under oscillatory stretching and compression, alignment quantification. ** p < 0. 01, *** p < 0. 001. 2. 3 Oscillatory Stretch and Compression Differentially Affect Cell Migration and Alignment To further understand the influence of stress on cellular behavior, ECs and fibroblast were coseeded on the scaffold ends and subjected to either oscillatory stretch (Figure 2 E) or oscillatory compression (Figure 2 F) for 14 d. On day 7, ECs and fibroblasts exposed to stretch, migrated together, whereas, under compression, fibroblasts migrated slower than the ECs (Figure 2 G). Moreover, at day 14 of culturing under tension condition, both cell types migrated over longer distances when compared to cells in the compressed cultures (Figure 2 H); the stretched scaffold was fully covered with cells, whereas the compressed scaffold was half empty. Additionally, vessels were more oriented under stretch conditions (Figure 2 I). 2. 4 Oscillatory Strain Effects on Cell Growth and Network Development Is YAP‐Dependent In efforts to determine if oscillatory strain affects other characteristics as well, cell proliferation was monitored following exposure of constructs to strain. Under oscillatory strain conditions, both ECs and fibroblast cultures were denser compared to the unstretched control ( Figure 3 A, D, E). In line with these findings, more Ki67‐positive cells were observed in the oscillatory stretched group (Figure 3 B, F). In addition, when coculturing the cells under oscillatory strain conditions, the average vessel length was significantly higher on day 14, vessels were more elongated and the vessel network was more branched, with more vessel junctions (Figure 3 C, G–I). When examining the effect of different stretch amplitudes, a positive correlation between stretch amplitude and vessel quality was observed (Figure S3, Supporting Information). To test the involvement of YAP in transmitting the mechanical stretch signal, constructs were stained with YAP antibodies. YAP nucleus/cytoplasm expression ratios in fibroblasts exposed to 14 d of oscillatory stretch conditions were significantly higher when compared to the static control (Figure 3 J, M). We previously demonstrated that vessels are localized within the scaffold interior and YAP nuclear expression decreases in the cells located within these areas, 11 EC YAP expression within the stretched constructs, on day 14, was cytoplasmic and did not show a significant difference from the static control (Figure S4, Supporting Information). In addition, at the same time point, higher cytoplasmic β‐catenin staining was observed in the stretched as compared to the control group (Figure 3 K, N). Angiomotin (Amot) is part of the motin family of angiostatin‐binding proteins; it has two isoforms, Amot‐p80 and Amot‐p130. Amot expression is spatially and temporally dependent. 12 Amot expression within blood vessels was found to be during initial vessel formation, facilitating ECs migration, and is also expressed in mature and stabilized vessels. 13 Cytoplasmic Amot levels on day 14 were higher within vessels subjected to oscillatory stretching compared to static conditions (Figure 3 L, O). Figure 3 Oscillatory strain effects on cell growth and network development. A) Endothelial cells and fibroblasts were seeded separately on two different scaffolds and cultured for 14 d under oscillatory strain or static control conditions. Cell density was then quantified. Scale bar = 100 µm. B) Ki67 staining of a coculture of ECs and fibroblasts cultured for 14 d under oscillatory strain showed increased cell proliferation when compared to the static control. Scale bar = 50 µm. C) Coculture of ECs and fibroblasts under oscillatory stretch resulted in higher vessel development when compared to the static control. D–I) Quantification of single‐cell density, Ki67 staining, vessel network length vessel elongation, and number of vessel junctions. J–O) Nuclear YAP, cytoplasmic β‐catenin, and AMOT expression in the vessels exposed to oscillatory stretch or static conditions for 14 d. Scale bar = 25 µm. P) YAP staining of fibroblasts within oscillatory stretched and static control vascularized constructs, with the addition of blebbistatin to the culture medium. Scale bar = 2. 5 µm. Q) Oscillatory stretched and static control vascularized constructs with the addition of blebbistatin to the culture medium. Scale bar = 100 µm. R) Nuclear YAP and S) vessel quality quantification with and without blebbistatin addition to the culture medium. T) VE‐cadherin staining of vessels within oscillatory stretched and static control vascularized constructs cultured for 7 d with the addition of verteporfin to the culture medium, scale bar = 25 µm. U, V) Quantification of elongation and alignment of VE‐cadherin‐positive elements within verteporfin‐treated constructs. W, X) Quantification of elongation and alignment of DAPI‐positive cells within verteporfin‐treated constructs. * p < 0. 05, ** p < 0. 01, *** p < 0. 001, **** p < 0. 0001. When adding blebbistatin to the culture medium, nuclear YAP levels decreased (Figure 3 P, R), blood vessels became disrupted and vessel and fibroblast alignment was lost (Figure 3 Q, S). To explore the role of YAP in blood vessel formation, we added verteporfin, which inhibits YAP‐induced transcription, 14 to the oscillatory stretched constructs and unstretched controls on day 4 of culturing, which led to disruption of vessel network, as indicated by vascular endothelial‐cadherin (VE‐cadherin) disruption, loss of vessel alignment, and loss of nuclear alignment. Less vessel damage was observed in the unstretched control constructs (Figure 3 T–X). 2. 5 Oscillatory Strain Enhances Vessel Development and Stabilization through Secretion of Pro‐Angiogenic Cytokines Following the observed increase in vessel formation and fibroblast growth under stretch conditions, we set out to examine the growth factors secreted during these processes. Cells in the unstretched static controls showed slightly higher secretion of FGF‐2, VEFG, and PIGF on day 7 compared to the cells within the oscillatory stretched constructs. This trend was reversed on day 14 and further increased by day 21, when angiogenin, angiopoitin2, EGF, HGF, leptin, VEGF, PLGF, and angiopoitin1 secretion were higher in the cells within oscillatory stretched constructs compared to the static control. As sprouting initiation is mainly stimulated by VEGF 15 and PLGF, 16 the elevated levels of these cytokines in the static control group at earlier time points led us to hypothesize that the vessel networks form faster under static conditions. However, this vasculature was not stable and degraded faster when compared to the oscillatory stretched constructs (Figures 1 C and 3 C). In contrast, the higher levels of vessel‐initiating cytokines at later time points and the stabilizing cytokines such as angiopoitin1, which activates mural cells to attach to the forming vessels and which has an important role in maintaining vessel quiescence, 17 indicate that these vessels reach higher maturation levels and thus are stable over longer time periods ( Figure 4 ). Figure 4 Oscillatory strain affects vessel development and stabilization through secretion of pro‐angiogenic cytokines, fibroblasts differentiation, and collagen VI production. A) Cytokine secretion array was performed on medium collected from oscillatory stretched and static control constructs at days 7, 14, and 21 of culture. Pro‐angiogenic factors were secreted over longer time periods under oscillatory stretch conditions. B) αSMA, PDGFRβ, NG2, and collagen 4 staining of oscillatory stretched and static control constructs cultured for 14 d. C–E) αSMA, PDGFRβ, and NG2 colocalization with ECs vessels quantification. F–I) Markers and blood vessels alignment quantification. Scale bar = 10 µm; * p < 0. 05, ** p < 0. 01, *** p < 0. 001, **** p < 0. 0001. 2. 6 Mural Cell Differentiation and Collagen IV Secretion under Oscillatory Tensile Forces Cytokine secretion analysis showed that under oscillatory stretch, vessels develop at a slower rate and are more stable. As mural cell characteristics have been shown to have a major impact on vessel stabilization, 17 we then examined mural cell differentiation within oscillatory stretched constructs. αSMA and PDGFR‐β are markers for fibroblasts differentiation into myo‐fibroblasts, and are indicators for vessel maturation. 18 NG2, however, is a marker for fibroblasts, premature cells. 18 Under stretch conditions, unlike in the unstretched control, both αSMA and PDGFRβ colocalized with the vessel network (Figure 4 B–D), whereas NG2 localization was identical in both stretched and control samples (Figure 4 E). In addition, the highly aligned morphology of the markers, correlated with vessel alignment (Figure 4 F–H). Moreover, under oscillatory stretch conditions, collagen IV alignment was significantly higher and correlated with blood vessel alignment when compared to the static control, likely creating a more stabilized vessel structure (Figure 4 B, I). 2. 7 Implantation of Oscillatory Stretched Constructs Induces Vessel Penetration Alignment In Vivo To determine whether implantation of these mature, organized constructs impact host vessel penetration in vivo, oscillatory stretched and unstretched control constructs were implanted into a mouse dorsal window chamber and then tracked over time. The implantation location was chosen due to its isotropic characteristics, to show whether a highly stable and aligned vasculature can influence the penetration orientation of the host vessels. Vessel penetration into the construct was observed at day 11 after implantation, when the host vasculature anastomosed to and perfused the implanted vessels. By day 18 postimplantation, the graft was fully covered with host blood vessels, which replaced the implanted vessels ( Figure 5 A). The oscillatory stretched graft showed highly organized vessel penetration on day 11, which further continued by day 18, when the host vasculature was seen aligned in the same orientation as the implanted vasculature, whereas static control graft vessels were organized in a random orientation (Figure 5 A, B). Figure 5 Implantation of oscillatory stretched vascularized grafts into a mouse dorsal window chamber. A) Confocal images of the penetrating host vessels into the implanted oscillatory stretched and static control constructs cultured for 21 d in vitro, stained with rhodamine‐dextran (red) and anti‐mouse CD31 (blue) that were injected via the mouse tail vein. Scale bar = 500 µm. B) Quantification of vessel orientation; * p < 0. 05, ** p < 0. 01, *** p < 0. 001. 3 Discussion The spontaneous formation of blood vessels in cocultures of endothelial cells and fibroblasts within 3D constructs bears great promise as a solution for more effective tissue integration upon implantation. 19, 20, 21 Many recent strategies have concentrated on enhancing vessel stabilization in vitro, such as encapsulating different growth factors within the scaffold, or use of different mural cell types. 22, 23, 24 In this study, we present application of oscillatory strain as a simple means of enhancing vessel stabilization and morphogenesis within engineered tissues. The mechanically stimulated vessels were clearly more developed, stabilized, and aligned. Moreover, following their implantation, aligned penetration of the host vasculature was observed. Application of oscillatory strain on ECs cultured without fibroblasts resulted in randomly oriented and radially spreading cells, in contrast to other works showing that exposure of ECs to oscillatory strain in a 2D culturing context leads to their alignment perpendicular to the stretching direction. 25, 26 In the current work, ECs presented aligned morphology only when cultured with fibroblasts, and may be explained by a differential response to stretch in 3D versus 2D environments. Moreover, when looking at cell migration patterns, ECs spread radially and evenly throughout the entire scaffold, whereas fibroblasts showed a specific migration pattern. Overall, in our system, ECs were less responsive to the stress stimulus as compared to fibroblasts. We hypothesize that the ECM secreted by the cells plays a central role in their responsiveness to strain, since fibroblasts cultured at low densities showed random alignment; only when they secreted sufficient ECM, did they align and the ECs tracked these aligned ECM fibers and align accordingly. A recent study has shown that 6 h of oscillatory stretch of human mammary epithelial cells resulted in increased nuclear YAP levels, which was shown to correlate with cell proliferation. 9 Another work has shown that culturing cells on stiff substrates resulted in a flat‐spread morphology with higher proliferation rates and an increase in nuclear YAP, as opposed to soft substrates, which resulted in a rounded cell shape, decreased proliferation rates, and more cytoplasmic YAP. 27 These results are in correlation with our findings, which demonstrated a significant increase in cell proliferation rates and in nuclear YAP under oscillatory stretch. Moreover, we showed that addition of verteporfin, an inhibitor that abolishes YAP and TEAD interaction, thereby inhibiting YAP‐induced transcription, 7 was more destructive in the oscillatory‐stretched group, as shown by damaged VE‐cadherin within these constructs and loss of vessel alignment. Vessel stabilization is achieved by recruitment of mural cells and the formation of the basement membrane. During this stage, EC proliferation decreases, and mural cells differentiation increases. 18, 28, 29 These processes are facilitated by secretion of various cytokines. Angiopoietin (Ang) 1 stabilizes the vessel by tightening the contacts between the recruited mural cells and the ECs, thereby decreasing vessel leakiness. 17, 30 Our results show higher secretion of Ang1 in the oscillatory stretched group, suggesting that this mechanical stimulation increases vessel stability. This conclusion is also supported by the higher number of αSMA‐ and PDGFRβ‐positive cells seen surrounding the oscillatory stretched vessels. In the coronary artery system, αSMA expression was found near large and mature arteries, 18 and is considered a marker of mural cells differentiation into smooth‐muscle cells and hence, vessel maturation. 29 In addition, collagen IV deposition was aligned in the same manner as the cells, leading us to conclude that endothelial microvessels forming under oscillatory‐stretch are more mature and stable, and surrounded by aligned ECM. In order to engineer a complex tissue, there is a need to pattern the vascular system within it. This has been previously achieved using micropatterning techniques, which was applied to fabricate endothelial cords, which have facilitated host penetration alignment. 31, 32 In our in vivo model designed to assess whether the implanted construct can influence host vessel penetration orientation, the oscillatory‐stretched constructs were implanted into the dorsal area, which contains randomly aligned vasculature. Oscillatory‐stretched constructs were grown for 21 d in vitro, a time point which showed the most developed vasculature upon implantation. We hypothesized that upon implantation, the construct components, aligned vessels, surrounding fibroblasts, and ECM serve as tracks for the invading vessels, which by day 18 postimplantation covered the entire graft and took on the same orientation as the implanted vessels. 4 Conclusion These results led us to propose a new mechanism for vessel network development under oscillatory strain ( Figure 6 ): application of oscillatory stretch enhances mural cell alignment and proliferation, translocation of YAP into the nucleus, and higher expression levels of β‐catenin. In turn, more ECM is produced, and is oriented in the same direction as the fibroblasts. ECs, which are less responsive to stretch stimulus, form aligned vessels by following the fibroblast and ECM alignment. Mechanical uncoupling of the cells, by adding blebbistatin to the culture medium, results in loss of fibroblast alignment and hence, loss of vessel alignment. Moreover, when inhibiting cell proliferation and ability to respond to the mechanical stimulus, by inhibiting YAP, both fibroblasts and vessels are affected. The stretch stimulus also increases mural cell differentiation, thereby increasing vessel stability and alignment, as indicated by the secretion of stabilizing growth factors. Upon implantation of the vascularized construct, invading vessels align in the same direction as the implanted vessels. These findings provide a better understanding of the role of mechanical strain on cell development, differentiation, and morphogenesis during the vascularization process. This will enhance engineered vascularized tissue stability and utility, and significantly contribute to the field of tissue engineering and regenerative medicine. Figure 6 Model of vessel stabilization and alignment under oscillatory stretch. A) Oscillatory strain is first sensed by the fibroblasts cells (green cells), which respond by translocating YAP to the nucleus, and by an increase in cytoplasmic β‐catenin which results in alignment and increased proliferation. B) Next, endothelial cells form aligned vessels according to the fibroblast alignment; this vasculature is more complex and longer when compared to the static control and associated with AMOT upregulation. C) Then, vessel stabilization occurs: fibroblasts differentiate and express more αSMA and are recruited to the vessels (green cells); the ECM aligns and stabilizes the vessels (light blue fibers); Ang1, a vessel stabilization growth factor, is secreted. 5 Experimental Section Scaffolds and Mechanical Stimulation : Gelfoam scaffolds (Gelfoam compressed, Pfizer) were cut to 1 cm × 0. 5 cm. Mechanical stimulation was applied on the constructs 1 d postseeding, using an EBERs TC‐3 bioreactor, with uniaxial oscillatory stretching or compression (sinusoidal) of 20% strain and 1 Hz frequency, was applied for 21 d. Cell Culture and Inhibitors : Human adipose microvascular endothelial cells (HAMECs; ScienceCell) lentivirally transduced with ZsGreen fluorescent protein, were grown in endothelial cell medium (ScienceCell) supplemented with 5% fetal bovine serum (FBS) (ScienceCell) and endothelial cell growth supplement (ScienceCell), and were used for 5–9 passages. Neonatal normal human dermal fibroblasts expressing red fluorescent protein (HNDFs‐RFP) (Angio‐Proteomie) were grown in Dulbecco's modified Eagle medium (DMEM) (Gibco), supplemented with 10% FBS (HyClone), 1% nonessential amino acids (NEAAs), 0. 2% β‐mercaptoethanol (Sigma‐Aldrich), and 1% penicillin‐streptomycin solution (PEN STREP) (Biological Industries). 3D vascularized constructs were obtained by coseeding endothelial cells (HAMECs, 3 × 10 5 cells) and support cells (HNDFs, 0. 6 × 10 5 cells) on the gelfoam scaffolds, by mixing the cells, seeding them in a small volume of medium (20 µL), and incubating them for 15 min before adding medium. Verteporfin (1 × 10 −6 m, Biotest) was added to the culture medium 3 d postseeding, and then constructs were cultured for 7 d. Alternatively, blebbistatin (50 × 10 −6 m, Sigma‐Aldrich) was added to the culture medium 3 d postseeding, constructs were cultured for 14 d. Whole‐Mount and Cryosection Immunofluorescence Staining : Constructs were fixated in paraformaldehyde (4%) for 20 min, and then permeabilized with 0. 3% Triton X‐100 (Bio Lab Ltd. ) for 10 min. Constructs were then washed with PBS and immersed in BSA solution (5%; Millipore) overnight. Samples were then incubated with the following primary antibodies overnight at 4 °C: mouse anti‐human Ki‐67 (1:20; DAKO), mouse anti‐human αSMA (1:50; DAKO), goat anti‐human PDGFRβ (1:50; R&D), mouse anti‐human NG2 (1:100; Santa‐Cruz), mouse anti‐human collagen IV (1:500, Sigma‐Aldrich), goat anti‐human VE‐cadherin (1:100; Santa Cruz), mouse anti‐human YAP (1:100; Santa Cruz), rabbit anti‐human β‐catenin (1:100; Sigma‐Aldrich), or mouse anti‐human AMOT (1:100; Santa Cruz). Constructs were then treated with Cy3‐labeled (1:100; Jackson Immunoresearch Laboratory), Cy5‐labeled (1:100; Jackson Immunoresearch Laboratory), or Alexa‐488‐labeled (1:400; ThermoFisher Scientific) secondary antibodies and DAPI (Sigma‐Aldrich), for 2 h, at room temperature. Scanning Electron Microscopy (SEM) Imaging : Scaffold morphology was examined using a SEM (Zeiss Ultra‐Plus FEG‐SEM). Scaffolds without cells were carbon‐coated using a Polaron carbon coater (Quorum Technologies). Cell‐embedded scaffolds were fixed in 2. 5% (vol/vol) glutaraldehyde in 0. 1 m cacodylate buffer (Sigma‐Aldrich) for 5 min, followed by dehydration in a gradient of 70, 85, 95, and 100% ethanol, with a 5 min incubation in each solution. Scaffolds were then immersed in hexamethyldisilazane (Sigma‐Aldrich) for 5 min and air‐dried at room temperature, before being coated with a gold‐palladium mixture using a Polaron gold coater. Cytokine Array and ELISA : Medium was collected on days 7, 14, and 21 postseeding. Cytokine quantification was measured using the Human Angiogenesis Array GS1 (RayBiotech). Ang1 was quantified using the Human ANGPT1 ELISA kit (RayBiotech). Construct Imaging and Image Analysis : Whole vascularized constructs were imaged with a confocal microscope (LSM700, Zeiss), using 2. 5×, 5×, 20×, and 63× oil immersion lenses. Imaris software (BITPLANE) was used to detect YAP, PDGFRβ, and NG2 localization in the 3D image. All other image analyses were quantified using self‐written algorithms in MATLAB: all images were transformed into a binary image and then analyzed using different algorithms. Cells, nuclei, vessels, and collagen IV orientation were determined using the orientation module in the regionprops function in MATLAB software. Cell migration pattern, VE‐cadherin morphogenesis, and vessel quality under blebbistatin treatment were measured by the eccentricity parameter using regionprops algorithm in the MATLAB software. Ki67 pixel density was calculated by dividing Ki67 staining‐positive pixels by DAPI staining‐positive pixels. AMOT vessel expression was calculated by determining the number of pixels with both a positive vessel (ZsGreen) signal and AMOT‐positive signal. Animal Experiments : Animal studies and surgical procedures were conducted according to protocols approved by the Technion animal ethics committee. Dorsal window chamber assembly followed the protocol by Palmer et al. 33 Window parts were soaked in 70% ethyl alcohol for 1 h and washed three times with PBS. Athymic nude mice (≈30 g, 7–9 weeks; Harlan Laboratories) were anesthetized via intraperitoneal injection with a mixture of ketamine (100 mg kg −1 )‐xylazine (10 mg kg −1 ), delivered via a 30gauge needle. On the following 2 d, mice were subcutaneously injected with buprenorphine (0. 05 mg kg −1 ) every 12 h. To create an aseptic working area, the dorsal region of the mouse torso and the entire tail were cleaned with chlorhexidine solution and iodine. To assemble the window, mouse skin was stretched and sutured to a stabilizing device using 4‐0 silk sutures, after which, two window frames were sutured to each other using 4‐0 silk sutures. The skin was then cut and removed along the circumference of the window and then the area was washed with warm saline and covered with a 13 mm cover slip glass (Electron Microscopy Sciences, PA, USA), and fixed with a snap ring. The mice were monitored daily to assess general health. Scaffold Implantation : 2 or 3 d after implantation of the dorsal skinfold window chamber, mice were anesthetized via intraperitoneal injection of a mixture of ketamine (100 mg kg −1 )‐xylazine (10 mg kg −1 ), delivered via a 30 gauge needle. The snap ring and the cover slip glass were removed, followed by a saline wash of the tissue. The scaffold was then placed in the middle of the window groove and then covered with a new cover slip. The window was then closed with the snap ring. Mice were monitored daily and imaged using intravital microscopy. Intravital Imaging : Intravital microscopy (LSM700 confocal microscope, Carl Zeiss) was performed on days 11 and 18 postimplantation. To track the host vasculature in the mice within the graft area, Alexa flour647 anti‐mouse CD31 (Biolegend) was intravenously injected via the tail vein. The antibody was allowed to circulate for 30 min before mice were anesthetized via intraperitoneal injection of ketamine (100 mg kg −1 )‐xylazine (10 mg kg −1 ), delivered using a 30 gauge needle. Tetramethylrhodamineisothiocyanate‐dextran, (average MW 155 000, Sigma‐Aldrich) was then intravenously injected through the tail vein. Mouse body temperature was maintained during the entire imaging session with a heating chamber. Statistical Analysis : Presented data include the mean ± standard deviation. Two‐way analysis of variance (ANOVA) was performed to examine the influence of two independent categorical variables, followed by Bonferroni's multiple comparison tests. Results were considered significant for p < 0. 05. Statistical analysis was performed using GraphPad Software, a computerized statistical program. All analyses were performed in at least biological triplicates. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800518
| 2,018
|
Advanced Science
|
Insights into 2D MXenes for Versatile Biomedical Applications: Current Advances and Challenges Ahead
|
Abstract Great and interdisciplinary research efforts have been devoted to the biomedical applications of 2D materials because of their unique planar structure and prominent physiochemical properties. Generally, ceramic‐based biomaterials, fabricated by high‐temperature solid‐phase reactions, are preferred as bone scaffolds in hard tissue engineering because of their controllable biocompatibility and satisfactory mechanical property, but their potential biomedical applications in disease theranostics are paid much less attention, mainly due to their lack of related material functionalities for possibly entering and circulating within the vascular system. The emerging 2D MXenes, a family of ultrathin atomic nanosheet materials derived from MAX phase ceramics, are currently booming as novel inorganic nanosystems for biologic and biomedical applications. The metallic conductivity, hydrophilic nature, and other unique physiochemical performances make it possible for the 2D MXenes to meet the strict requirements of biomedicine. This work introduces the very recent progress and novel paradigms of 2D MXenes for state‐of‐the‐art biomedical applications, focusing on the design/synthesis strategies, therapeutic modalities, diagnostic imaging, biosensing, antimicrobial, and biosafety issues. It is highly expected that the elaborately engineered ultrathin MXenes nanosheets will become one of the most attractive biocompatible inorganic nanoplatforms for multiple and extensive biomedical applications to profit the clinical translation of nanomedicine.
|
1 Introduction The rapid progress of clinic biomedicine and nanobiotechnology has stimulated the generation of diverse novel inorganic nanosystems, which offers multiple theranostic modalities as potential alternatives in combating various diseases by synergistic therapy and multimodal imaging, especially in cancer theranostics. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 Currently, great efforts for multidisciplinary research have been focused on biomedical applications of 2D nanomaterials, a newly emerging subtype of nanomaterials with ultrathin layer‐structured topology, including mostly explored graphene and its derivatives, 12, 13 hexagonal boron nitrides (h‐BN), 14 transition metal dichalcogenides (TMDCs), 15 transition metal oxides (TMOs), 16 palladium (Pd) nanosheets, 17 black phosphorus (BP), 18, 19 and transition metal carbides (MXenes). 20 Their multifaceted properties, such as high specific surface area and intriguing physiochemical natures, make them able to satisfy the strict demands in theranostic nanomedicine such as drug delivery, phototherapy, diagnostic imaging, biosensing, and even tissue engineering. 21, 22, 23, 24, 25, 26, 27, 28, 29 MXenes, a new family of multifunctional 2D solid crystals containing a large class of transition metal carbides, nitrides, and carbonitrides with metallic conductivity and hydrophilic nature, as well as excellent mechanical properties, were developed by Barsoum and co‐workers. 20, 30, 31, 32 M n +1 X n layer (named as MXene) was fabricated by the selective extraction of A‐element from layered ternary carbides of M n +1 AX n phases ( n = 1–3), where M is an early transition metal, A is an A group element, and X is C or N. 33 MXenes typically have three different formulas: M 2 X, M 3 X 2, and M 4 X 3. The versatile chemistry of MXenes has found numerous applications in energy storage, 34, 35, 36, 37, 38, 39 water purification, 40 chemical sensors, photo‐ or electro catalysis, 41 and electromagnetic interference shielding. 42, 43 They also hold great potentials in the biomedical field. On one hand, the high specific surface areas enable the MXene nanosheets to be potential drug or protein carriers with abundant anchoring sites and reservoirs. The ultrathin layered structure with almost single‐atomic thickness endows MXenes with fascinating physiochemical properties (e. g. , photothermal conversion, 44, 45 electron transparency, X‐ray attenuation, 46, 47 and localized surface plasmon resonance 48 ) and biological behaviors (e. g. , enzyme‐triggered biodegradation, 49 cellular endocytosis, 50 distinct biodistribution, and metabolism pathway 49 ). On the other hand, the controllable component and tunable in‐plane structure of MXenes can be precisely designed and synthesized in the pristine structure of MAX phases, creating flexible/extensive multifunctionalities of MXenes in promising theranostic nanomedicine. To date, the MXenes with various attractive physicochemical properties and biological effects, and the cutting‐edge researches for emerging 2D materials, have attracted increasing attention in scientific community of nanomedicine. In this review, we summarize and discuss the current state‐of‐the‐art of 2D MXenes as a robust nanoplatform on the basis of synthetic methods, surface chemistry, and biomedical applications, as well as the related challenges and perspectives for future developments ( Figure 1 ). To be specific, the derivatives of emerging research of 2D MXenes in nanomedicine could be categorized into therapeutic modality, 45, 46, 47, 48, 49, 51, 52 diagnosis imaging, 46, 47, 49, 51, 53 biosensing, 54, 55, 56, 57 antimicrobial, 58, 59 and biosafety evaluation. 50 The bigger picture is that by gaining deeper insights into the material science and biological behavior of 2D MXene nanosheets for existing and emerging biomedical modalities, we will be able to facilitate immense and promising applications with clinical‐translation potential in benefitting the human health. Figure 1 Summary of emerging 2D MXenes used in nanomedicine. Summative scheme of emerging 2D MXenes for biomedical applications, and schematic illustration of the 2D MXene‐based nanomedical applications, including therapeutic practice, diagnostic imaging, biosensing, antimicrobial, and biosafety evaluations. 2 Synthetic Methods and Surface Chemistry The synthetic methodologies of 2D layered nanomaterials can be divided into two distinct routes: (i) top‐down approach, and (ii) the bottom‐up method. 60, 61, 62, 63 Both strategies have been performed on the fabrication of single‐, few‐layer, or multilayer nanostructure of MXenes. 2. 1 Top‐Down Synthesis The top‐down method is based on the direct exfoliation of bulk crystals, which employs various driving forces including mechanical and chemical exfoliations. To date, the general focus of MXenes' fabrication is on liquid‐phase exfoliation, a facile and high‐yield process, which has been proven to be of high efficiency in the production of ultrathin, nanoscale MXenes ( Figure 2 a). In brief, the transformation from parent MAX‐phase ceramics (Figure 2 b–d) to nanoscale 2D MXenes undergoes the following two steps: delamination by hydrofluoric acid (HF) etching to obtain the multilayer‐stacked MXenes (Figure 2 e–g), and disintegration by organic base molecules intercalation or probe sonication breakage to acquire few‐layer or single‐layer MXenes (Figure 2 h–j). Benefiting from this methodology, nearly all types of MXenes could be obtained with diversified morphologies of few‐ or single‐layered nanosheets, featuring nanoscale‐lateral size and atomic‐scale thickness. Figure 2 Synthetic methods of MXenes for biomedicine. a) Schematic diagram for the synthesis of biocompatible MXenes, including HF etching (delamination), organic base molecules intercalation (disintegration), and surface functionalization with organic molecules or inorganic nanoparticles (surface modification). b) 2D ball‐and‐stick models and SEM images of precursor MAX phase for M 2 X, e) SEM image of multilayer M 2 X, and h) 3D ball‐and‐stick models and TEM image single‐layer M 2 X‐based MXenes. Reproduced with permission. 49 Copyright 2017, American Chemical Society. c) 2D ball‐and‐stick models and SEM images of precursor MAX phase for M 3 X 2, f) SEM image of multilayer M 3 X 2, and i) 3D ball‐and‐stick models and TEM image single‐layer M 3 X 2 ‐based MXenes. Reproduced with permission. 45 Copyright 2017, American Chemical Society. d) 2D ball‐and‐stick models and SEM images of precursor MAX phase for M 4 X 3, g) SEM image of multilayer M 4 X 3, and j) 3D ball‐and‐stick models and TEM image single‐layer M 4 X 3 ‐based MXenes. Reproduced with permission. 47 Copyright 2018, Wiley‐VCH. Since the first MXene, multilayered Ti 3 C 2, was developed in 2011, a series of 2D MXenes such as multilayered Mo 2 C, V 2 C, Nb 2 C, Zr 3 C 2, and Ta 4 C 3, have been widely investigated and extensively explored of their performances and versatile applications. 30, 31, 64, 65, 66, 67 These multilayered MXenes with large lateral size of sheets were produced by traditional synthesis via liquid exfoliation approach like HF etching, which might lead to poor therapeutic efficacy and potential biosafety issue. In particular, biomedical applications necessitate small‐enough lateral size and ultrathin thickness of these 2D nanoagents to facilitate their effective transport within the blood‐circulatory system, guarantee the high accumulation/penetration into lesion location, and enable the easy excretion out of the living body. 25, 27, 68 Very recently, the typical multilayered Ti 3 C 2, as well as the emerging MXene compounds of Nb 2 C and Ta 4 C 3, has been successfully delaminated and disintegrated to produce single‐layer MXene with nanoscale‐lateral sizes (around 100 nm) and single‐atomic thickness (almost 1 nm), making the biomedical uses of MXene possible. 45, 49 2. 2 Bottom‐Up Synthesis The bottom‐up approaches are the alternative method to produce 2D nanomaterials via atomic level control of their composition and morphology. It could be applied to fabricate 2D nanomaterials which were not easy to obtain though direct exfoliation from bulk. Generally, chemical vapor deposition (CVD) growth 69, 70, 71, 72 and wet‐chemical synthesis 73, 74, 75 are the major synthetic approaches based on the bottom‐up method to develop layered materials with high production quality. Compared to traditional family of 2D materials, the current advances of bottom‐up approach to synthesize MXene show yet to be successful, perhaps owing to the complex atoms and structures of MXenes' in‐layer morphology. As a paradigm, Ren and co‐workers report the growth of 2D ultrathin α‐Mo 2 C crystals with large area and high quality via CVD in 2015. 76 Utilizing methane as the carbon source, Cu/Mo foils used as the substrate could grow into 2D ultrathin α‐Mo 2 C crystals at temperatures above 1085 °C, which features the thickness of few nanometers and lateral size of over 100 µm. Such a high growth temperature guarantees the melting of metal Cu and the formation of Mo‐Cu alloy at the liquefied Cu/Mo interface. Mo atoms could then diffuse onto the surface of the liquid Mo‐Cu alloy to grow 2D Mo 2 C crystals by reacting with the decomposed methane. Importantly, such a CVD growth strategy developed toward transition metal carbides is facile and versatile, which offers a general strategy for synthesizing varied types of high‐quality 2D ultrathin transition metal carbides crystals. Moreover, other high‐quality 2D MXene‐structural crystals, such as WC and TaC crystals with a large lateral size and few defects, were also obtained by replacing the Mo foils with the W or Ta foils during the similar CVD process. 76 2. 3 Surface Modification and Multifunctionalization MXenes reported so far commonly feature specific surface terminations including hydroxyl (—OH), oxygen (—O), or fluorine (—F), imparting hydrophilic nature to their surfaces, which, together with the high surface charge (negative zeta potential exceeding −40 mV), lead to high stability of surfactant‐free water‐based colloidal solutions. 20, 77 However, suffering from rapid aggregation and precipitation in biological mediums similar to most nanomaterials in biomedicine, the delaminated ultrathin MXenes are usually unstable in complex physiological conditions and lack of multifunctionalization. 25, 27 Thus, surface engineering is critical to endow these nanosystems with high stability and dispersity in a physiological environment, as well as multipurpose strategy by decoration of other functional materials. In general, the surface modification and engineering of MXenes have been focused on the following two approaches: one is polymer‐based strategy of surface chemistry, modifying surface with certain molecules or polymers based on noncovalent interaction. For instance, the Nb 2 C nanosheets were decorated with polyvinylpyrrolidone (PVP) molecules ( Figure 3 a). 49 Moreover, the PEGylation in Ti 3 C 2 nanosheets surface by electrostatic adsorption was achieved to maintain the stability of MXene in physiological conditions (Figure 3 b), 48 and the Ta 4 C 3 nanosheets were successfully modified with soybean phospholipid (SP) (Figure 3 c). 47 The other is inorganic nanoparticle‐based surface chemistry, decorating MXenes with multifunctional inorganic nanoparticles which could further broaden their potential applications. For instance, the integration of Ta 4 C 3 nanosheets with Fe 3 O 4 nanoparticles (Figure 3 d, h) and the combination of Ti 3 C 2 and MnO x nanoparticles (Figure 3 e, i) are typical paradigms of integrating therapeutic platforms with MRI contrast agents (CAs), endowed them with the capabilities of concurrent therapeutics and diagnostic imaging. 51, 78 Similarly, the integration of 2D Ti 3 C 2 MXene nanosheets with GdW 10 ‐based polyoxometalates (POMs) provides phototherapeutic nanoplatform with magnetic resonance (MR) and/or computed tomography (CT) imaging guidance toward xenograft tumor (Figure 3 f, j). 79 Another typical paradigm is the integrating Ti 3 C 2 MXene with the mesoporous silica nanoparticles (MSNs), a classic drug delivery system (DDS). In comparison to traditional 2D MXenes for tumor‐specific phototherapy, this surface nanopore engineering on Ti 3 C 2 MXene integrates several unique features for broadening the MXene‐based biomedical applications, including sufficient mesopore structure with confined capacity for drug delivery (DOX), enhanced hydrophilicity/dispersity/biocompatibility (PEGylation), and multiple surface chemistry for targeting modification (RGD conjugation) (Figure 3 g, k). 80 In short, such two multifunctionalization strategies are still in infancy in comparison to the extensively studied graphene/graphene oxide (GO) family owing to the synthetic and integrating complexities. It is highly expected that the rapid expansion on synthesis and applications of 2D MXenes will promote the emergence of diverse 2D MXene‐based multifunctional nanoplatforms for biomedical applications. Figure 3 Surface chemistry of MXenes for biomedicine. Schematic illustrations of polymer‐based surface chemistry of MXenes. a) PVP modification of M 2 X (e. g. , Nb 2 C nanosheets). b) PEGylation of M 3 X 2 (e. g. , Ti 3 C 2 nanosheets). c) SP modification M 4 X 3 ‐based MXenes. Schematic representations of inorganic nanoparticle‐based surface chemistry of MXenes. d) Superparamagnetic iron oxide (Fe 3 O 4 ) nanoparticles grew onto the surface of Ta 4 C 3 MXene by an in situ redox reaction. Reproduced with permission. 78 Copyright 2018, Ivyspring International Publisher. e) In situ growth of small MnO x nanosheets on the surface of Ti 3 C 2 according to a facile redox reaction. Reproduced with permission. 51 Copyright 2017, American Chemical Society. f) Integration of GdW 10 ‐based polyoxometalates (POMs) onto the surface of Ti 3 C 2 MXene though an amide bond. Reproduced with permission. 79 Copyright 2018, Springer. g) Surface nanopore engineering of silica (SiO 2 ) on the 2D Ti 3 C 2 MXene based on a process of sol‐gel chemistry. Reproduced with permission. 80 Copyright 2018, Wiley‐VCH. The STEM images and the corresponding element mappings of h) Fe 3 O 4 @Ta 4 C 3, i) MnO x @Ti 3 C 2, j) POMs@Ti 3 C 2, and k) SiO 2 @Ti 3 C 2 composite MXenes. 3 Therapeutic Applications of MXenes The unique transformable 2D in‐layer nanostructure and controllable chemical compositions endow 2D MXenes with versatile properties in benefiting biomedical applications. To date, these 2D multifunctional MXenes and their composites have been developed for theranostic applications including typical phototherapy of photothermal therapy (PTT), photothermal/photodynamic/chemo synergistic therapy, diagnostic imaging, antimicrobial, and biosensing. In this section, we introduce various therapeutic applications of diverse MXenes 2D materials. 3. 1 Photothermal Therapy Light is an external stimulus, furnished with extensive profits in efficacy for cancer phototherapeutic modality. Typically, PTT employs photothermal agents accumulated within tumors as internal energy absorbers to convert near‐infrared (NIR) light energy into heat, producing necrosis and/or apoptosis of cancer cells. 81, 82 NIR laser‐based PTT for cancer eradication has attracted considerable interest, which promotes local hyperthermia to ablate tumor tissues with poorly vascularized tumor microenvironment (TME). The exploitation of NIR light as an outer and remote trigger brings high spatiotemporal regulation of local heating effect while minimizes adverse side effects. 83, 84 Photothermal performance of a phototherapeutic agent used for photothermal conversion are determined by two fundamental parameters: the extinction coefficient (ε) and photothermal conversion efficiency (η). The extinction coefficient reveals the light‐absorption capability while the photothermal conversion efficiency indicates the performance of the agent in converting the light into heat. 49 The photothermal performance parameters of the classic inorganic photothermal agents and 2D inorganic photothermal agents have been summarized in Figure 4 a, which indicates the remarkable advantages of emerging 2D MXenes used as photothermal agents, even superior to most of the inorganic photothermal agents in the literatures. Very recently, we and other research groups have achieved a series of breakthroughs on exploiting ultrathin MXenes nanosheets for PTT, including effective photothermal ablation of tumors in a mouse model (e. g. , Ti 3 C 2 MXene), 45, 48 theranostic nanoplatform combining dual‐mode photoacoustic (PA)/CT imaging and in vivo photothermal ablation of mouse tumor xenografts (e. g. , Ta 4 C 3 MXene), 47 and highly efficient in vivo photothermal tumor eradication and tissue penetration capability in both NIR‐I and NIR‐II biowindows (e. g. , Nb 2 C MXene). 49 Figure 4 MXene‐based photothermal therapy. a) The photothermal performance parameters, including mass extinction coefficient (ε) and photothermal conversion efficiency (η), of various 2D inorganic nanomaterials in the literatures. Each symbol indicates a set of material category. b) Absorbance spectra of Ti 3 C 2 nanosheets dispersed in water at varied concentrations (30, 15, 8, 4, and 2 µg mL −1 ). Inset is the mass extinction coefficient (ε) of Ti 3 C 2 MXene. c) Photothermal‐heating curves of Ti 3 C 2 nanosheet‐dispersed aqueous suspension at varied concentrations (72, 36, 18, and 9 µg mL −1 ) by using an 808 nm irradiation (1. 5 W cm −2 ). d) Confocal laser scanning microscopy (CLSM) images of Ti 3 C 2 ‐SP‐induced photothermal ablation after various treatments. Reproduced with permission. 45 Copyright 2017, American Chemical Society. e) Temperature elevations and f) the corresponding IR thermal images at the tumor sites of 4T1 tumor‐bearing mice in groups of different treatments. g) Photographs of 4T1 tumor‐bearing mice after different treatments. h) Time‐dependent tumor growth curves after different treatments. Reproduced with permission. 47 Copyright 2017, Wiley‐VCH. As a paradigm of biocompatible 2D MXenes, monolayer or few‐layer Ti 3 C 2 nanosheets (MXenes) were first fabricated based on a liquid exfoliation method of MAX phase Ti 3 AlC 2 combining stepwise HF etching and TPAOH intercalation procedures. 45 Especially, the ultrathin Ti 3 C 2 nanosheets exhibit an especially high extinction coefficient (ε) of 25. 2 L g −1 cm −1, which is remarkably higher than that of GO nanosheets (3. 6 L g −1 cm −1 ), 85 and also notably larger than that of traditional Au nanorods (13. 9 L g −1 cm −1 ), 86 indicating a favorable NIR laser‐absorption property of Ti 3 C 2 nanosheets (Figure 4 b). The photothermal conversion efficiency (η) was calculated to be 30. 6%, higher than that of Au nanorods (21%), 87 and Cu 9 S 5 nanocrystals (25. 7%). 88 Typically, at a low Ti 3 C 2 concentration (72 µg mL −1 ) of suspension solution, the temperature reached 57 °C upon NIR irradiation in 6 min. In contrast, the control group shows nearly no temperature elevation, implying that the employing of Ti 3 C 2 nanosheets can rapidly convert NIR light into hyperthermia (Figure 4 c). Moreover, the in vitro cell apoptosis after photothermal ablation was further confirmed by confocal laser scanning microscopy (CLSM) imaging, which showed that the majority of 4T1 cells were killed by the photothermal ablation after treatment with Ti 3 C 2 ‐SP under NIR laser irradiation, exhibiting the noticeable in vitro photothermal effect of the Ti 3 C 2 ‐SP nanosheets in promoting cancer cell ablation (Figure 4 d). Furthermore, Ti 3 C 2 nanosheets were revealed to be a highly effective PTT agent for tumor hyperthermia, which enables excellent NIR light‐induced tumor ablations without recurrence by either intravenous injection of Ti 3 C 2 ‐SP (20 mg kg −1 ) or localized intratumoral injection of PLGA/Ti 3 C 2 phase‐changeable implant (2 mg kg −1 ). 45 Though Ti 3 C 2 nanosheets, as the first reported MXene‐based photothermal agent, exhibited a considerable in vivo PTT efficacy, the relative low photothermal conversion efficiency would largely hinder their further broad applications. Considering the flexible element candidates as the in‐layer component of MXenes, the species of early translation metal, M element, could be selected specifically to enhance the element‐based multifunctionalities of MXenes. The 2D Ta 4 C 3 MXene, a rarely studied type of MXenes possessing biocompatible Ta element, with ultrathin sheet‐like morphology and lateral size of ≈100 nm, has been developed for PTT of tumors. 47 The absorption spectra acquired on Ta 4 C 3 MXenes feature a board, strong absorption band, which is of similarity to those of classic 2D nanomaterials, such as graphene 81 and MoS 2, 89 providing a desirable photoabsorption property for further photothermal‐transduction process. Attractively, these ultrathin 2D Ta 4 C 3 nanosheets exhibit an extraordinarily high photothermal conversion efficiency of 44. 7%, superior to most of the inorganic photothermal agents (Figure 4 a). Subsequently, the fast temperature elevation was verified in animal models. Upon intravenous or intratumoral administration with biocompatible Ta 4 C 3 ‐SP, the temperatures of mouse tumor regions rapidly increased from ≈30 to ≈60 °C or from ≈30 to ≈68 °C in 6 min of laser irradiation, respectively (Figure 4 e, f). Thus, highly effective in vivo photothermal tumor eradication by Ta 4 C 3 ‐SP nanosheets has been successfully demonstrated (Figure 4 g, h). Specifically, the majority of previous study focused on the first NIR (NIR‐I) biologic window (noted as biowindow) (750–1000 nm), but the second NIR (NIR‐II) biowindow (1000–1350 nm) has been rarely explored. Compared to the well‐studied NIR‐I biowindow, employing NIR‐II biowindow offers two merits such as desirable depth of NIR laser‐responsive penetration and enhanced maximum permissible exposure (MPE) of biological tissues. 90 Receiving rapidly increasing attentions, the NIR‐II biowindow, a scientific frontier for biological optical imaging 91, 92 and the phototherapeutics, has aroused great interest in the scientific community of cancer theranostics. 49, 93, 94, 95 Recently, a novel ultrathin 2D Nb 2 C MXene, as a new phototherapeutic agent, has been exploited in our group for in vivo photothermal ablation of mouse tumor xenograft in both NIR‐I and NIR‐II biowindows. The single‐atomic thickness and lateral‐nanosized Nb 2 C nanosheets feature an extraordinary photothermal performance of broad NIR spectrum with an extraordinarily high photothermal conversion efficiency ( Figure 5 a, b). 49 Given that many other types of nanomaterials do show extended absorbance bands in the NIR‐II biowindow, but the absorbances of these photothermal agents, such as MoS 2, 96 black phosphorus, 97 and graphene, 81, 85 evidently decline in the biowindow, which is not favorable for the efficient NIR laser absorption. In contrast, the Nb 2 C nanosheets maintain high‐efficient light absorption in both NIR‐I and NIR‐II biowindow. In contrast, the Nb 2 C nanosheets maintain high light absorption in both NIR‐I and NIR‐II biowindows. In addition, the enhanced tissue penetration depth in NIR‐II window compared to NIR‐I window deserves detailed investigations of its photothermal tumor therapy, which is expected to attract the attention of scientific community for the investigation of photothermal agents with light responsiveness in NIR‐II biowindow. First, the in vitro toxicities of Nb 2 C−PVP to cells showed that with the increase of laser energy, more cells were killed in both NIR‐I and NIR‐II biowindows (Figure 5 c). In addition, effective photothermal conversions of the Nb 2 C nanosheets (NSs) were performed in the depth of ex vivo tissue under the penetrated NIR laser irradiation without inducing significant heating of the ex vivo tissue itself. The temperature variations at varied depth intervals of the ex vivo tissues upon NIR‐I and NIR‐II laser radiations indicate that NIR‐II biowindow promotes the photothermal heating with diminished attenuation, in contrast to that of NIR‐I biowindow (Figure 5 d). For further investigating the effective tissue penetration of photothermal tumor ablation in NIR‐I and NIR‐II biowindows in vivo, Nb 2 C‐PVP‐treated tumor‐bearing mice were exposed to 808 or 1064 nm laser radiations (Figure 5 e). It could be found that both NIR‐I and NIR‐II biowindows allowed the effective photothermal tumor eradication deep to around 4 mm inside subcutaneous tumor xenograft in nude mice (Figure 5 f). Furthermore, these surface‐engineered Nb 2 C NSs exhibit highly efficient in vivo photothermal eradication of tumor xenografts in both NIR‐I and NIR‐II biowindows (Figure 5 g, h). The aforementioned ex vivo and in vivo evaluations suggest the great promise of using Nb 2 C‐PVP NSs for deep‐tissue PTT in both NIR‐I and NIR‐II biowindows. Figure 5 MXene‐based photothermal therapy. a) Absorbance spectra of well‐dispersed aqueous Nb 2 C NSs at varied concentrations. b) Photostability profiles of an aqueous Nb 2 C NSs solution in NIR‐I and NIR‐II biowindows for five laser on/off cycles. c) Relative viabilities of 4T1 cell line after Nb 2 C‐PVP‐induced (40 µg mL −1 ) photothermal eradication at various power densities (0, 0. 5, 1. 0, 1. 5, and 2 W cm −2 ) of laser ( n = 5, mean ± SD). d) Temperature elevations of Nb 2 C NSs‐dispersed aqueous suspensions upon exposure to tissue‐penetrating NIR‐I and NIR‐II laser via photothermal conversion. e) Schematic diagram of in vivo tumor tissue penetration for photothermal conversion based on NIR‐I and NIR‐II. f) Cancer cellular proliferation at varied depths of tumor tissues by antigen Ki‐67 immunofluorescence staining (scale bar: 50 µm). g) Scheme of synthetic procedure and in vivo photothermal tumor ablation process of 2D biodegradable Nb 2 C (modified with PVP). h) Time‐dependent tumor growth curves after various treatments. Reproduced with permission. 49 Copyright 2017, American Chemical Society. 3. 2 Synergistic Therapy The typical 2D planar topology endows the MXenes with high surface area, allowing easy anchoring of versatile therapeutic agents on the surface of layered structure, similar to delivery feature of typical organic nanoplatforms, 98 inorganic mesoporous carriers, 99, 100 and other 2D nanoplatforms. 101, 102, 103 In consideration of the high photothermal conversion performance of 2D Ti 3 C 2 MXenes, they were expected to be further employed for highly efficient tumor ablation by synergistic MXene‐assisted photothermal eradication and DOX‐loaded chemotherapy ( Figure 6 a, b). 104 The as‐synthesized Ti 3 C 2 MXenes show desirable photothermal stability during the four‐cycle processes of heating and cooling, which potentially guaranteed the continuous photothermal ablation of tumor and the hyperthermia‐enhanced drug delivery process (Figure 6 c). Especially, the Ti 3 C 2 MXenes as drug delivery nanoplatform not only possess the high capability of drug loading as 211. 8%, but also enables both pH‐responsive and NIR laser‐triggered drug releasing (Figure 6 d). CLSM was further employed to demonstrate that the integration of Ti 3 C 2 ‐promoted PTT with chemotherapy caused the death for almost complete 4T1 cells (Figure 6 e). Notably, the integration of Ti 3 C 2 ‐assisted PTT with chemotherapy has achieved the complete tumor eradication without reoccurrence in therapeutic period on 4T1 tumor‐bearing mice model, exhibiting the desirable synergistic outcome of PTT and chemotherapy (Figure 6 f). Figure 6 MXene‐based synergistic multitherapies. a) Schematic illustration of surface modification of Ti 3 C 2 MXene, further surface engineering of drug loading, and stimuli‐triggered drug releasing under inner and external triggers. b) In vivo process of Ti 3 C 2 MXene‐based drug delivery system (DDS) for synergistic photo‐chemotherapy of tumor. c) The DOX‐releasing curves from DOX@Ti 3 C 2 composite nanosheets in buffer with varied pHs of 4. 5, 6. 0, and 7. 4. d) The DOX‐releasing curves with 808 nm laser irradiation on or off at varied pH values (4. 5, 6. 0, and 7. 4). e) CLSM images of cancer cells after various treatments, including control, laser, DOX, DOX@Ti 3 C 2 ‐SP, Ti 3 C 2 ‐SP + laser, and DOX@ Ti 3 C 2 ‐SP + laser groups. Scale bar: 50 µm. f) Time‐dependent tumor growth profiles after different treatments. Reproduced with permission. 104 Copyright 2018, Wiley‐VCH. g) Schematic diagram of the construction of Ti 3 C 2 ‐based nanosystem and photothermal/photodynamic/chemo synergistic therapy of tumor. h) Accumulative drug release curves with or without NIR laser irradiation at pH 4. 5, 6. 0, and 7. 4. i) CLSM images of HCT‐116 cells treated with Ti 3 C 2 ‐DOX (top), and DCFH‐DA‐stained HCT‐116 cells treated with Ti 3 C 2 ‐DOX under NIR laser irradiation for the intracellular ROS detection (bottom). Reproduced with permission. 52 Copyright 2017, American Chemical Society. Photodynamic therapy (PDT) features the merits of minimal invasiveness and spatiotemporal specificity for oncologic intervention. Critically, the excited photosensitizer (PS) will transfer its excited‐state energy to molecular oxygen species for generating reactive oxygen species (ROS), inducing tumor‐cell death upon the key cellular substances being oxidized. 105, 106 Upon integrating the cargo delivery with phototherapeutic modality such as PTT and PDT, a synergistic therapy combining chemotherapy with phototherapy could be established. MXene‐based nanotherapeutic agent is highly expected to introduce the concurrent PTT/PDT/chemotherapy‐based multitherapeutic modality to realize the high‐efficient synergistic therapeutics. As a paradigm, a new MXene‐based nanoplatform has been constructed via layer‐by‐layer surface modification of Ti 3 C 2 nanosheets with doxorubicin (DOX) as the chemotherapy drug and hyaluronic acid (HA) as tumor‐targeting agent (Figure 6 g). 52 This Ti 3 C 2 nanosheet‐based multifunctional nanoplatform (noted as Ti 3 C 2 ‐DOX) shows an as high as 84. 2% drug loading capacity. As expected, the Ti 3 C 2 ‐DOX exhibits an efficient pH‐responsive drug‐releasing behavior (in mild acidity), benefiting from the effective protection by the HA shell in neutral environment. Meanwhile, the introduction of NIR laser‐induced photothermal effect of Ti 3 C 2 (temperature up to 50 °C) could enhance the drug‐delivery efficiency in acid TME (Figure 6 h). After the intracellular endocytosis of Ti 3 C 2 ‐DOX, a NIR laser of 808 nm was irradiated to induce the photothermal effect for hyperthermia and photodynamic process for ROS generation (Figure 6 i). It is highly attractive that the effective synergistic therapeutic outcome has been achieved by the facile therapeutic drug loading on the 2D MXenes beyond the tedious material structure design, providing a paradigm of synergistic tumor therapy based on other 2D nanosystems. Active‐targeting technique is a promising pathway to enhance the accumulation of drug carriers or therapeutic agents toward disease region. A nanopore engineering has been performed on the surface of MXenes, as‐synthesized RGD‐targeted Ti 3 C 2 @mMSNs‐RGD, endowing synergistic multitherapies of MXenes with well‐defined mesoporous nanostructure for drug delivery and release, tumor‐specificity for active‐targeting response, and inherent photothermal conversion performance for PTT ( Figure 7 a, b). 80 The RGD‐targeted photothermal‐ablation efficacy of Ti 3 C 2 @mMSNs‐RGD under NIR laser exhibited that with the increase of laser energy, more hepatocellular carcinoma (HCC) cells incubated with Ti 3 C 2 @mMSNs‐RGD were killed compared to Ti 3 C 2 @mMSNs‐PEG without RGD modification (Figure 7 c). Simultaneously, Ti 3 C 2 @mMSNs‐RGD exhibited evident superiority in suppression of HCC cells owing to the active‐targeting capability between RGD peptides anchored on Ti 3 C 2 @mMSNs surface and α v β 3 ligands expressed on HCC cell membranes, facilitating more synergistic therapeutic agents to target the HCC cells via an efficient endocytosis as showed by CLSM (Figure 7 d). Systematic in vivo evaluations also indicated the high active‐targeting outcome (contributed by RGD) of Ti 3 C 2 @mMSNs‐RGD into tumor, and the synergistic chemotherapy (derived from MSNs) and hyperthermia (contributed by Ti 3 C 2 MXene) have thoroughly eradicated the tumor without further reoccurrence (Figure 7 e, f). Figure 7 MXene‐based synergistic multitherapies. a) Schematic diagram for the fabrication of ultrathin Ti 3 C 2 nanosheets and the synthetic procedure for Ti 3 C 2 @mMSNs‐RGD. b) Scheme for synergistic multitherapies on HCC cell line as assisted by DOX‐loaded mMSNs@Ti 3 C 2 ‐RGD. c) Schematic representation of pH/photothermal‐responsive drug release from DOX‐loaded Ti 3 C 2 @mMSNs‐RGD (top), and relative viabilities of SMMC‐7721 cells under NIR irradiation of different power densities (0, 0. 5, 0. 75, 1, 1. 25, and 1. 5 W cm −2 ) (bottom). d) CLSM images of SMMC‐7721 cell line coincubated with FITC‐labeled mMSNs@Ti 3 C 2 for a varied incubation times (2, 4, and 8 h). Scale bar: 50 µm. e) Time‐dependent tumor growth curves after various treatments. f) Photographs of SMMC‐7721 tumor‐bearing mice and its tumor regions in 28 d after various treatments. H&E staining for pathological changes in tumor tissues from each group. Scale bar: 100 µm. Reproduced with permission. 80 Copyright 2018, Wiley‐VCH. 4 Contrast‐Enhanced Diagnostic Imaging of MXenes The versatile physicochemical properties of 2D nanosheets enable great potentials for diagnostic imaging, implying that they could act as the CAs to enhance the diagnostic‐imaging performance. 21 In comparison to conventional 2D CAs, 2D MXene‐based agents feature quantum size effects for photoluminescence (PL) cellular imaging, intrinsic photothermal performance for PA imaging, element‐enhanced contrast for X‐ray CT imaging, and effective loading of functional CAs for MR imaging. 4. 1 Fluorescent Imaging Compared to traditional organic fluorescein, inorganic 2D nanomaterials and their corresponding quantum dots (QDs) have shown intriguing fluorescent property for bioimaging such as tunable wavelength, high photostability, and desirable quantum yields. For instance, the graphene QDs, as the first developed 2D metal‐free fluorescent nanomaterial, have found their potential utilization for efficient fluorescent imaging. 107 The ultrathin liquid‐exfoliated g‐C 3 N 4 and corresponding g‐C 3 N 4 QDs also exhibited high photoluminescence quantum yield for intracellular bioimaging. 108, 109 The luminescent Ti 3 C 2 MXene‐based quantum dots (noted as Ti 3 C 2 MQDs) with hydrophilic nature and monolayer structure were synthesized by a facile hydrothermal process. The physiochemical property and PL principle of the as‐prepared Ti 3 C 2 MQDs under different synthetic conditions have been extensively studied ( Figure 8 a). 53 These Ti 3 C 2 MQDs fabricated at 100 °C (MQDs‐100) feature excitation‐dependent PL behavior of a high quantum yield up to 9. 9%, which possibly derives from the strong quantum confinement effect (Figure 8 b). As indicated in the merged CLSM image, the MQDs were easily taken up via the endocytosis pathway (Figure 8 c). The application of MQDs as multicolor cellular‐imaging reagents was demonstrated via labeling RAW264. 7 cell line, which exhibit the high potential of MXene‐based QDs for applications in areas of optics, biomedicine, and cellular imaging. As another example, Geng and co‐workers reported a strategy for the fabrication of ultrasmall MXenes by an intralayer cutting and delamination approach under a mild condition with aqueous tetramethylammonium hydroxide (TMAOH). 110 The as‐obtained ultrasmall Ti 3 C 2 MXenes feature strong optical absorption and excitation‐dependent emission. Importantly, such a synthetic approach could also be extended to fabricating ultrasmall dots of other members in MXene family, such as Ti 2 C and Nb 2 C. More recently, Wang and co‐workers investigated the luminescent property from solvothermal‐treated Ti 3 C 2 T x MXene in dimethylformamide (DMF), and demonstrated their application in cellular imaging. 111 Figure 8 MXene‐based diagnostic imaging modalities. a) Schematic diagram of synthesizing Ti 3 C 2 MXene QDs (MQDs). b) UV–vis absorbance spectra (solid line), PLE (dashed line), and PL spectra (solid blue line, excitation wavelength of 320 nm) of MQD‐100 in aqueous solutions. c) Merged images of the bright‐field and the confocal images (488 nm) for MQD‐100 incubated with cells. Reproduced with permission. 53 Copyright 2017, Wiley‐VCH. d) In vitro CT images and f) CT contrasts of Ta 4 C 3 ‐SP nanosheet solutions and iopromide solutions at varied concentrations. e) In vivo 3D reconstruction CT images (left) and CT contrast images (right) of mice before and after intravenous administration (10 mg mL −1, 200 µL) for 24 h. g) In vivo CT contrasts before and after intravenous administration. h) PA images of Ta 4 C 3 ‐SP solutions with varied concentrations. i) In vitro PA values as a function of a series of concentrations. j) In vivo PA value temporal evolution and k) PA images of the tumor locations at varied time intervals postinjection. Reproduced with permission. 47 Copyright 2017, Wiley‐VCH. l) In vitro T 1 ‐weighted MR imaging, and m) the corresponding 1/ T 1 versus Mn concentration of MnO x /Ti 3 C 2 ‐SP nanosheets in buffer solution at different pH values after soaking for 3 h. n) T 1 ‐weighted imaging of tumor‐bearing mice after postinjection of MnO x /Ti 3 C 2 ‐SP at varied time intervals. Reproduced with permission. 51 Copyright 2017, American Chemical Society. 4. 2 Computed Tomography Imaging X‐ray CT has been widely used for medical imaging modality because of its 3D tomography of the anatomic structure based on the differential X‐ray absorptions between the lesions and tissues. 112 Nanomaterials containing high atomic number ( Z ) elements have frequently been explored as CT CAs. WS 2 ‐PEG as the CA for CT imaging was well developed owing to the higher Z of W ( Z = 74) superior to the clinical use of I element ( Z = 53). 113 In another case, bismuth ( Z = 83) was expected as a desirable candidate CT CA based on its much higher atomic number ( Z ) than most metal elements and the favorable biocompatibility. Thus, 2D topological Bi 2 Se 3 nanosheets feature strong X‐ray attenuation and exhibit enhanced CT imaging of tumor tissue in vivo. 114 In this regard, tantalum (Ta)‐based compound, Ta 4 C 3 MXenes, could no doubt function as a CT CA due to its high atomic number of 73. 47 The Hounsfield unit (HU) value and corresponding CT images of Ta 4 C 3 ‐SP dispersed in Xanthan gum exhibited a strong signal enhancement along with the increased Ta 4 C 3 ‐SP contents. The in vitro CT imaging of Ta 4 C 3 ‐SP shows enhanced contrast in comparison with the commercial iodine‐based CT CAs (Figure 8 d, f). Moreover, the tumor‐bearing mice received intravenous administration of Ta 4 C 3 ‐SP for in vivo evaluation. CT images collected in 24 h postinjection indicates distinct tumor contrast enhancement ranging from 83 ± 8. 8 to 232. 3 ± 25. 2 HU (Figure 8 e, g). Therefore, these Ta 4 C 3 MXenes could function as promising CAs for potential CT imaging. 4. 3 Photoacoustic Imaging Generally, the efficient photothermal conversion agents could function as CAs for PA imaging. PA imaging, an emerging technology overcoming the high degree of scattering of optical photons in biological tissue by making use of the photoacoustic effect, offers living subjects higher spatial resolution and allows deeper tissues to be imaged compared with most optical imaging techniques. Light absorption by molecules creates a thermally induced pressure jump that launches ultrasonic waves, which are received by acoustic detectors to form images. 115, 116 As many diseases do not exhibit a natural photoacoustic contrast, especially in their early stages, it is necessary to administer a photoacoustic contrast agent. Many conventional 2D nanosystems such as graphene, MoO x, and WS 2 have been developed to offer strong contrasts under PA imaging. 96, 117, 118 Notably, 2D MXenes are also expected to play this role. Owing to the desirable photothermal conversion efficiency of Ta 4 C 3 MXene in the NIR biowindow, PA imaging was investigated by using the Ta 4 C 3 ‐SP nanosheets as CAs. 47 In detail, in vitro PA images and the corresponding values of Ta 4 C 3 ‐SP at varied concentrations obviously indicate their contrast‐enhancement functions (Figure 8 h, i). Then, tumor‐bearing mice were intravenously injected with Ta 4 C 3 ‐SP, and the corresponding in vivo PA images were collected at varied time intervals (Figure 8 j, k). In comparison with the pretreatment image, the PA signal of the tumor region exhibits a time‐dependent lightening process featuring intensity increase from 0. 39 to 1. 0 a. u. , a highly significant value in about 24 h postinjection possibly resulting from the passive accumulation of Ta 4 C 3 MXene nanosheets via the enhanced permeability and retention (EPR) effect. 119 So far, we and other groups have reported on various MXene and their compounds for in intro and in vivo PA imaging, including Ti 3 C 2, Nb 2 C, Ta 4 C 3, and MnO x /Ti 3 C 2, under an NIR light excitation. 46, 47, 49, 51 4. 4 Magnetic Resonance Imaging (MRI) Due to the high spatial resolution, excellent contrast difference of 3D soft tissues, and noninvasive feature, MRI has been extensively applied for clinically diagnostic imaging. 120, 121 Integration of functional components with ultrathin 2D nanosheets paves a new way for multifunctionalization of 2D nanosystems, which largely broadens their multiple applications in diagnostic‐imaging fields. For example, the integration of MnO x with Ti 3 C 2 profits the Mn‐based MRI, one of the clinic‐used effective imaging modalities because of its advanced spatial resolution and desirable tissue contrast, 10 which has been investigated for diagnostic tumor MR imaging. The MnO x component in MnO x /Ti 3 C 2 features the specific pH‐responsive T 1 ‐weighted MRI capability upon arriving at the acidic TME. An evident concentration‐dependent enhancement was observed in T 1 ‐weighted MRI, and an acidic‐induced positive MRI signal intensification was clearly revealed (Figure 8 l, m). 51 Further in vivo T 1 ‐weighted MRI evaluation of MnO x /Ti 3 C 2 ‐SP nanosheets was conducted on tumor‐bearing mice. A significant enhancing effect of MRI contrasts was demonstrated in tumor sites, owing to the efficient accumulation of the MnO x /Ti 3 C 2 ‐SP nanosheets through the EPR effect and the Mn ion release in the mild acidic TME (Figure 8 n). In another case, we developed MnO x /Ta 4 C 3 composite nanosheets acted as the high‐performance contrast agents for TME‐responsive T 1 ‐weighted MR imaging, which also derived from the MnO x component. 46 5 Biosensing Beyond the versatile applications of 2D nanomaterials for therapeutics and diagnostic imaging, they could also be employed as novel biosensing systems for the detection of biomacromolecules and bio‐effects. Monolayer MoS 2 nanosheets have been used as a biosensor for DNA detection based on their strong fluorescence‐quenching effect. 122 In another example, based on 2D g‐C 3 N 4 nanosheet, a DNA biosensor was designed by utilizing the affinity changes of g‐C 3 N 4 to DNA probes upon their targeting the analyte and the related positron emission tomography (PET)‐based fluorescence‐quenching effect. 123 In comparison with traditional nanoparticle‐based biosensors, 2D nanostructures feature two prominent merits in biosensing. One is the high surface‐to‐volume ratio of 2D layered structure facilitating large‐area immobilization of sensing targets, and the other is the fascinating performances such as light‐absorption capability and fast electron transfer fluorescence‐quenching effect based on the unique physicochemical property of 2D nanosheets. 29 Indeed, there is a continuous demand for the development of highly sensitive, selective, efficient, and cost‐effective biosensing platforms. On this ground, MXenes, with a hydrophilic surface nature, metallic conductivity property, and planar atomic structure, could be expected as a promising candidate in manufacturing biologically compatible field‐effect transistors (FET) for fast, direct, and label‐free analysis of biological events. 54 Recently, a highly sensitive device based on MXene‐FET‐based biosensor with high sensitivity was developed for performing label‐free detection of dopamine, and monitoring spiking activity in primary hippocampal neurons. 57 The as‐synthesized MXene stripes of multilayered Ti 3 C 2 offer an opportunity for the active surface to be in contact with small biomolecules, achieving the conductivity signal changing in turn. The high sensitivity of response‐to‐signal perturbation of electrochemistry enables MXenes to be potential candidates for biosensing alternatives ( Figure 9 a, b). The firing of action potentials results in the release of neurotransmitters, inducing the electric signal changing by the subsequent binding between neurotransmitters and MXene surface (Figure 9 c). The neurons display normal, elaborated neurites on the surface of MXene substrates, indicating that surface‐modified MXenes have a high biocompatibility with neuron cells (Figure 9 d). The spike curve of action potentials originating from electrical signals and optical recording of fluorescence changes exhibit a high correlation coefficient of 0. 82, which endows MXene‐FET device with high temporal resolution (Figure 9 e). These results clearly demonstrate the strong correlation between electrical signals and the release of action potentials, and confirm that the MXene‐based biosensors are promising alternatives for real‐time probing of neural activities. In addition, MXene composite‐based biosensor also exhibits high sensitivity, broad sensitive range, and excellent stability. For instance, the Au/MXene nanocomposite platform was developed for the detection of sensitive enzymatic glucose. 55 A general trend of increasing current at increased glucose concentration has been observed in the measurement of GO x /Au/MXene/Nafion/GCE biosensors (Figure 9 f), which exhibits a linear amperometric response in the glucose concentration range from 0. 1 to 18 × 10 −3 m with a relatively high sensitivity of 4. 2 µA mM −1 cm −2 and a detection limit of 5. 9 × 10 −6 m (Figure 9 g). Figure 9 MXene‐based applications in biosensing. a) SEM image of multilayered Ti 3 C 2 MXene. b) Device schematics for biosensing based on Ti 3 C 2 MXene field‐effect transistor (FET). c) Schematics of real‐time recording for neuronal‐spiking activities by employing MXene‐FET device. d) Confocal image of neurons immunofluorescence staining for βIII‐tubulin (left). Merged image of the bright‐field and fluorescence channels (scale bar: 100 µm). e) The derivation of spiking activities from primary neurons by utilizing MXene‐FET device via recording the current and fluorescence changes. Reproduced with permission. 57 Copyright 2017, Wiley‐VCH. f) Amperometric current–time ( i – t ) curves for GO x /Au/MXene/Nafion/GCE biosensor under a constant voltage of −0. 402 V. g) Steady‐state calibration curves for contrast recordings between GO x /MXene/Nafion/GCE and GO x /Au/MXene/Nafion/GCE biosensors. Reproduced with permission. 55 Copyright 2016, Nature Publishing Group. 6 Antimicrobial Activity Various 2D nanomaterials have been explored for their potential antibacterial activities following different strategies. For instance, after the addition of Zn‐Ti layered double hydroxides (LDHs) to bacteria suspension under visible light, the growth of microbial species such as S. cerevisiae, S. aureus, or Escherichia coli was strongly inhibited because of the LDH size effect and generation of ROS by Ti 3+ under visible light. 124 The antimicrobial behaviors of chemically exfoliated MoS 2 (ce‐MoS 2 ) were attributable to both membrane and oxidation stress, which were investigated by addition of a ce‐MoS 2 suspension to the bacterial culture. 125 In this respect, to extensively exploit the environmental and health impacts of these emerging 2D MXenes, the antimicrobial activities of typical single‐ and few‐layer Ti 3 C 2 MXenes toward two bacterial models, E. coli and Bacillus subtilis, were investigated and compared with other carbon‐based nanomaterials ( Figure 10 a). 58 In detail, the antibacterial activities of Ti 3 C 2 MXenes were evaluated against E. coli and B. subtilis by employing bacterial growth curves and colonies growth assays (Figure 10 b). Ti 3 C 2 exhibits a higher antimicrobial activity toward both E. coli and B. subtilis in comparison with GO, a well‐studied antimicrobial agent. The concentration‐dependent antibacterial activity was revealed, and Ti 3 C 2 MXenes incubated with both bacterial cells within 4 h in a colloidal solution (200 µg mL −1 ) led to >98% viability losses of the bacterial cells (Figure 10 c). Furthermore, LDH release assay was employed to quantitatively investigate the extent of cell damage. The cytotoxicity of Ti 3 C 2 MXenes was measured by LDH release from the bacterial cells exposed to varied concentrations of Ti 3 C 2 T x nanosheets for 4 h (Figure 10 d). LDH release elevated remarkably when bacterial cells were of exposure to 200 µg L −1 of Ti 3 C 2 T x solution, exhibiting cytotoxicity of 38. 41 and 55. 24% for E. coli and B. subtilis, respectively. Such dose‐dependent cytotoxicity indicates that both the walls and the inner contents of the bacterial were damaged, demonstrating that membrane disruption perhaps function as a major cell suppressive mechanism. In order to visually characterize the antibacterial effect of Ti 3 C 2 T x MXene, changes of morphology and membrane integrity of E. coli and B. subtilis cells due to the interaction with Ti 3 C 2 T x were further evaluated by scanning electron microscopy (SEM) (Figure 10 e, f). With elevating concentration of Ti 3 C 2 T x comes more damage of bacteria, such significant change in the cell morphology/structure could be contributed by detachment of the cytoplasmic membrane from the bacterial cell wall, which was well‐matched with the outcome of LDH release assay. Figure 10 MXene‐based applications in antimicrobials. a) Schematic illustration of antibacterial activity of Ti 3 C 2 T x MXene. b) Concentration‐dependent antibacterial activities of the Ti 3 C 2 T x in aqueous suspensions: Images of agar plates onto which E. coli (top) and B. subtilis (bottom) bacterial cells were recultivated after treatment for 4 h with varied concentrations. c) Cell viability analysis of E. coli treated with Ti 3 C 2 T x and graphene oxide (GO) in aqueous suspension. d) Ti 3 C 2 T x cytotoxicity investigated by LDH‐releasing from the bacterial cells exposure to varied concentrations of Ti 3 C 2 T x. SEM images of the e) E. coli and f) B. subtilis treated with 0, 50, and 100 µg mL −1 of Ti 3 C 2 T x, at low and high magnification, respectively. Reproduced with permission. 58 Copyright 2016, American Chemical Society. g) The scheme of antimicrobial mechanism for Ti 3 C 2 MXenes. h) Cell viability studies of E. coli and B. subtilis cultivated on fresh and aged PVDF‐supported Ti 3 C 2 membranes. i) Flow cytometry analyses of E. coli and B. subtilis bacterial cells exposed to PVDF and Ti 3 C 2 MXene membranes. Reproduced with permission. 59 Copyright 2017, Nature Publishing Group. As another example, the antibacterial property of Ti 3 C 2 T x ‐modified membrane was investigated against E. coli and B. subtilis by bacterial growth on the membrane surface. 59 The antibacterial rate of fresh Ti 3 C 2 T x MXene membranes reaches more than 67% against E. coli and 73% against B. subtilis as compared with that of control samples, while aged Ti 3 C 2 T x membrane exhibited more than 99% growth suppression of both bacteria species (Figure 10 h). Flow cytometry depicted almost 70% of dead and compromised cells after 24 h exposure of both bacterial colonies in the surface of PVDF and Ti 3 C 2 T x (Figure 10 i). The demonstrated antibacterial activity of MXene‐based membranes against common bacteria species facilitates their potential application as anti‐biofouling membrane in water and wastewater treatment technology. The possible antimicrobial mechanism of Ti 3 C 2 MXenes has been proposed as follows: First, sharp edges of Ti 3 C 2 MXenes enable their effective adsorbing on the surface of microorganisms. Second, exposure of microorganisms to the sharp edges of the MXenes might induce membrane damage. Third, Ti 3 C 2 MXenes are likely to react with biomolecules in the cell wall and cytoplasm of microorganisms, breaking the cell microstructure and triggering the death of microorganism (Figure 10 g). 58 Therefore, MXenes should be introduced as a new family of 2D antimicrobial agents for their potential use in areas of health and environmental protection. 7 Biosafety/Cytotoxicity Evaluations of MXenes Traditional organic materials with high biocompatibility/biodegradation have been explored for biomedical applications, unfortunately their poor chemical/thermal stabilities and single functionality are the drawbacks impeding their clinical development. 126 In comparison, inorganic 2D MXene‐based nanomaterials show relatively high clinical translation potential, which derives from their inherent characteristics such as facile functionalization, tunable morphology/structure, desirable biocompatibility, specific biodegradation, multifunctionality, and relatively high physiological stability. These specific features are usually difficult to achieve on most organic entities. Although MXenes feature unique properties in biomedical practices, the biosafety of emerging 2D MXenes family determines their potentials in future clinical applications. The toxicity/biocompatibility evaluations of various types of 2D nanomaterials have been under progress, the impacts of biomedicine derived from crucial parameters of 2D nanomaterials, such as solubility, dispersibility, long‐term toxicity, and biodegradation, are still unknown. 127, 128 In a recent work, the potential ecotoxicological assessments of MXene‐based nanomaterials in vivo on an aquatic biota were conducted on a zebrafish embryo model, which were expected to be able to indirectly reveal the possible adverse toxicological effect on human health. The neurotoxicity and locomotion assessments showed that the neuron number in the spinal cord area of Ti 3 C 2 T x ‐treated zebrafish embryos was similar to that of dimethyl sulfoxide (DMSO) negative control ( Figure 11 a), suggesting that at the no observed effect concentration (NOEC) (50 µg mL −1 ), Ti 3 C 2 T x MXene had no adverse toxicological effect on the neuron and muscle activity of the zebrafish embryo model (Figure 11 b, c). Essentially, as the LC 50 of Ti 3 C 2 T x exceeds to 100 µg mL −1, Ti 3 C 2 T x MXene could be classified in the “practically nontoxic” group in accordance with the Acute Toxicity Rating Scale by the Fish and Wildlife Service (FWS). Figure 11 Ecotoxicological assessments, biocompatibility, and biosafety evaluations of MXenes. a) CLSM imaging of an entire embryo. b) Spinal cord adjacent to the somite 14–17 territory of the various treatments toward representative embryos. c) Average of the relative neuronal number of as‐treated embryos ( n = 14, * P < 0. 05). Reproduced with permission. 129 Copyright 2018, Royal Society of Chemistry. d) Relative viabilities of 4T1 cell line after being incubated with varied concentrations of Ta 4 C 3 ‐SP nanosheets. Reproduced with permission. 47 Copyright 2018, Wiley‐VCH. e) Relative viabilities of 4T1 and U87 cells after being incubated with Nb 2 C‐PVP nanosheets of varied concentrations. f) Histological data (H&E stained) collected from the major organs (heart, liver, spleen, lung, and kidney) of the Nb 2 C‐PVP‐treated mice in 28 d postinjection under different conditions (control, exposure to NIR‐I, NIR‐II, and daylight). Scale bars: 100 µm. g) Hematological parameters and h) biochemical blood indexes of the Nb 2 C‐PVP‐treated mice in 1, 7, and 28 d postinjection under different treatments (control, exposure to daylight, NIR‐I, and NIR‐II). Reproduced with permission. 49 Copyright 2017, American Chemical Society. In addition, there are promising progresses demonstrating their potentially high biocompatibility in biomedicine. It has been found that both MXenes and the MXene‐based composites introduce low cytotoxicities toward cells. For example, the Ti 3 C 2 ‐SP nanosheets show negligible effect on the survival of 4T1 cell line for 48 h coincubation, even at a high concentration of 400 µg mL −1. 45 The in vitro toxicities of Ta 4 C 3 ‐SP upon cells were tested by a standard Cell Counting Kit‐8 (CCK‐8) assay. Ta 4 C 3 ‐SP also shows negligible impact on the survival of 4T1 cell line, even at the concentration up to 400 µg mL −1 (Figure 11 d). Moreover, two cancer cell lines of glioma U87 cancer cells and breast 4T1 cancer cells were incubated with Nb 2 C‐PVP, indicating no significant suppression effect on both U87 cells and 4T1 cells, even at the concentration up to 200 µg mL −1 (Figure 11 e). 49 In the case of MXene‐based composite nanosheets (MnO x /Ti 3 C 2 ‐SP), negligible cytotoxicity was found even at a high concentration of 160 µg mL −1 after coincubation for 48 h, indicating the relatively high biocompatibility of MnO x /Ti 3 C 2 ‐SP composite nanosheets. 51 The further in vivo potential biosafety and biocompatibility of surface‐modified MXene nanosheets against mice were comprehensively evaluated. For instance, healthy Kunming mice with three dosages of surface‐modified MXenes (Nb 2 C‐PVP dosage of 20 mg kg −1 ) were assessed in 1, 7, and 28 d. 47 It was found that negligible abnormal behavior of mice was recorded between three treated groups (exposure to NIR‐I, NIR‐II, and daylight) and the control group, and no abnormality was observed on body weights of mice within three observation period. The hematoxylin and eosin (H&E) staining assays of major organs (heart, liver, spleen, lung, and kidney) after a 28 d feeding exhibit no obvious acute, chronic pathological toxicity, and side effects compared the control group with the treated groups (Figure 11 f). The blood tests, including hematological indexes and biochemistry parameters, were in the normal range (Figure 11 g, h). These results indicate that Nb 2 C‐PVP is potentially biocompatible for further in vivo therapeutic modalities of tumor theranostics. The same assessments were performed on Ti 3 C 2 and Ta 4 C 3 MXenes, 45, 49 which feature negligible cytotoxicity and satisfactory in vivo biocompatibility. However, in light of the current results, further investigation and optimization of in vitro and in vivo toxicity assessment such as genotoxicity 127, 128 or reproductive toxicity 130 are highly needed to employ the full potential of MXene‐based theranostic nanoplatform for versatile biomedical applications. 8 Conclusions and Perspectives This review highlights the current research progress in the design and production of pristine and functionalized MXenes and their composites, with particular emphasis on their biological and biomedical applications, including therapeutic modality, diagnostic imaging, biosensing, antimicrobial, and biosafety evaluation ( Table 1 ). It is noted that the development of 2D MXenes in nanomedicine is still in its infancy. The emerging trends in 2D MXenes are the exploitations of their intrinsic physiochemical nature such as controllable in‐plane component, transformable multidimension, and tunable terminating species that are of multifunctionality, programmability, and biocompatibility. These specific properties of 2D MXenes can be achieved and regulated by designing multicomponent or multidimensional nanosystems as well as inventing innovative fabrication strategies. Table 1 Applications of MXenes in nanomedicine Application Material Description Refs. PTT Ti 3 C 2, Ti 3 C 2 QDs, Ta 4 C 3, Nb 2 C, MnO x /Ti 3 C 2, MnO x /Ta 4 C 3 High photothermal conversion performance (Nb 2 C: η = 36. 5% (808 nm) or 46. 65% (1064 nm), Ti 3 C 2 : η = 30. 6%, and Ta 4 C 3 = 44. 7%), in vitro and in vivo photothermal ablation of tumor. 45, 49, 51, 52, 131 Synergistic PTT/PDT/chemotherapy Ti 3 C 2 ‐DOX Demonstrating stimuli‐responsive drug‐releasing behavior, superior photothermal conversion efficiency (η = 58. 3%), and effective singlet oxygen generation ( 1 O 2 ), achieving effective cancer cell killing and tumor tissue destruction both in vitro and in vivo. 52 PL imaging Ti 3 C 2 QDs A biocompatible multicolor cellular imaging probe exhibiting excitation‐dependent PL spectra with quantum yields around 10%. 53 MR imaging MnO x /Ti 3 C 2, MnO x /Ta 4 C 3 Tumor microenvironment‐responsive (pH‐responsive) contrast agents for T 1 ‐weighted MR imaging. 46, 51 CT imaging Ta 4 C 3, MnO x /Ta 4 C 3 High atomic number of Ta component endows Ta 4 C 3 MXene with contrast‐enhanced computed tomography imaging. 46, 47 PA imaging MnO x /Ti 3 C 2, Ta 4 C 3, MnO x /Ta 4 C 3, Nb 2 C High photothermal conversion performance of MXenes enables their potential PA imaging. 46, 47, 49, 51 Biosensing Ti 3 C 2 MXene‐micropattern‐based FET for probing neural activity, immobilizing hemoglobin (Hb) to fabricate a mediator‐free biosensor, amperometric glucose biosensor. 54, 55, 56, 57 Antimicrobial activity Ti 3 C 2, PVDF/Ti 3 C 2 membrane Colloidal Ti 3 C 2 T x solution exhibits antibacterial activity against E. coli and B. subtilis. 58, 59 Biosafety/cytotoxicity Ti 3 C 2, Ta 4 C 3, Nb 2 C, MnO x /Ti 3 C 2, MnO x /Ta 4 C 3, Fe 3 O 4 /Ti 3 C 2, Fe 3 O 4 /Ta 4 C 3 Ecotoxicological assessments, cytotoxicities, and in vivo potential biosafety and biocompatibility 45, 46, 47, 49, 50, 51, 78, 129, 132 John Wiley & Sons, Ltd. No doubt, numerous unprecedented challenges exist in exploiting 2D MXenes for translation applications such as large‐scale production, precise structure/composition control, targeting surface engineering, and comprehensive and long‐term evaluation on the biocompatibility/biosafety. In addition, the biomedical applications of 2D MXene in other biomedical field should be further explored and promoted, such as stem‐cell engineering, immunological therapy, regenerative medicine, gene therapy, tissue engineering, and possibly other novel cancer treatments (see detailed summary in Figure 12 ). It is highly expected that the great and multidisciplinary‐joined efforts among biology, chemistry, physics, and engineering will promote mechanistic understanding of MXene‐based systems for nanomedicine and allow currently succeeded paradigms more rapidly and extensively applied in versatile biomedicine. Figure 12 Conclusions and perspectives for 2D MXenes used in nanomedicine. Summary of current research developments and future perspectives of ultrathin 2D MXenes for versatile biomedical applications. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201800560
| 2,018
|
Advanced Science
|
Enhanced Raman Investigation of Cell Membrane and Intracellular Compounds by 3D Plasmonic Nanoelectrode Arrays
|
Abstract 3D nanostructures are widely exploited in cell cultures for many purposes such as controlled drug delivery, transfection, intracellular sampling, and electrical recording. However, little is known about the interaction of the cells with these substrates, and even less about the effects of electroporation on the cellular membrane and the nuclear envelope. This work exploits 3D plasmonic nanoelectrodes to study, by surface‐enhanced Raman scattering (SERS), the cell membrane dynamics on the nanostructured substrate before, during, and after electroporation. In vitro cultured cells tightly adhere on 3D plasmonic nanoelectrodes precisely in the plasmonic hot spots, making this kind of investigation possible. After electroporation, the cell membrane dynamics are studied by recording the Raman time traces of biomolecules in contact or next to the 3D plasmonic nanoelectrode. During this process, the 3D plasmonic nanoelectrodes are intracellularly coupled, thus enabling the monitoring of different molecular species, including lipids, proteins, and nucleic acids. Scanning electron microscopy cross‐section analysis evidences the possibility of nuclear membrane poration compatible with the reported Raman spectra. These findings may open a new route toward controlled intracellular sampling and intranuclear delivery of genic materials. They also show the possibility of nuclear envelope disruption which may lead to negative side effects.
|
1 Introduction The cellular membrane is an extremely complex and dynamic environment that represents the gateway for all cell reactions and exchanges with the surrounding biological environment. Nutrients as well as molecules that drive the interactions with other cells and tissues pass through the plasma membrane. Therefore, membrane processes and mechanisms are of fundamental importance for cell functioning. 1 In many scientific fields, gaining access to the cell interior is an essential requirement, as in the testing of new drugs, the investigation of the electrical activity within a neuronal network or the manipulation of genes to treat diseases. 2, 3 Technological progress has made electroporation—that is, the application of a transmembrane voltage on the cell walls that transiently permeabilize the cellular membrane—an established method to gain access to the intracellular compartment. 4, 5, 6 More recently, micro‐ and nanotechnology has intersected with 3D nanofabricated substrates and micro‐ or nanofluidic devices to improve control over the membrane poration. 7, 8 After permeabilization, the cell membrane heals within a few minutes, shrinking and closing the electrically opened nanopores. 9 However, the mechanisms of plasma membrane repair are still under investigation. 10 In such a dynamic landscape of cells interfacing with different shapes of 3D nanostructured substrates, 11, 12, 13, 14, 15 a deeper understanding of the cellular membrane dynamics may shed light on the yet‐unexplored behaviors of lipids, proteins, vesicles, and other constituents. In addition, it may be helpful in designing next‐generation substrates for tissue engineering and cell manipulation. 16, 17 Raman microspectroscopy has been exploited to address these issues, but the large lateral resolution (1 µm) and the even larger axial resolution (7 µm) need to be improved in order to study the local behavior of the plasma membrane. 18 Within the recent past, it has been shown that surface‐enhanced Raman spectroscopy (SERS) by means of 3D plasmonic nanostructures can provide very sensitive, localized, label‐free and noninvasive chemical analysis of living cells, enhancing the vibrational modes of molecules adsorbed onto or close to specific nanostructure hot spots. In particular, the potential of the 3D plasmonic nanostructure configuration has been shown by acquiring single Raman spectra of cells at rest in their physiological conditions. 19, 20 Similar 3D nanostructured devices have been adopted to electroporate cells in vitro by using a low voltage, 21 and the nanofluidic properties of some devices have been exploited to prove the injection of small molecules into the electroporated cells. 22 In the present study, we exploit the ability of 3D vertical nanostructures combined with multielectrode arrays (MEAs) to work at the same time as nanoelectrodes for in situ electroporation and as plasmonic antennas for SERS studies of the cell membrane dynamics (see sketch in Figure 1 ). 19, 20 Using the NIH‐3T3 cell model, we investigated the cell membrane dynamics when the cells interacted with the 3D nanostructures at rest and after electroporation. We monitored in real time different molecular species, such as lipids and proteins, both on the cell membrane and in the cytoplasm. We noticed that there is a characteristic time of 10 min on average, in which the Raman signal increases drastically, that can be related to the opening of nanopores on the cellular membrane. Importantly, during this time window, the 3D plasmonic nanoelectrodes are in contact with the intracellular environment where additional molecular varieties can be investigated, including nucleic acids. The presence of genetic material in the cytosol, together with focus ion beam/scanning electron microscopy (FIB/SEM) cross‐sectioning, suggested poration of the nuclear envelope. This achievement may pave the way to the investigation of nuclear membrane processes as well as controlled intranuclear delivery and sampling. However, it also imposes careful design and exploitation of 3D pillars to prevent potential damage of the nucleus. Figure 1 Sketch of the system with inset showing the magnification at the 3D nanostructure tip. On top of the 3D nanostructures (yellow), cells (in orange) were tightly sealed to the substrate. The plasmonic modes of the 3D nanoelectrode were excited by a 785 nm laser, and the enhanced Raman signals coming from the molecules close to it were collected. The different colors of the substrate represent bulk quartz (salmon), gold nanoelectrode (yellow), and an SU8 passivation layer (green). 2 Cells Electroporation and Raman Analysis The fabrication process of our device is based on milling by focused ionic beam (FIB) of an optical resist and its consequent inversion. 23 In contrast to established methods, 21 this technique allows the fabrication of ordered arrays of 3D nanostructures with a high velocity of milling. Moreover, it is possible to spatially arrange the nanostructures with high precision and to create devices that mix 3D and 2D features, as in MEAs with 3D nanostructures fabricated on each planar electrode. 24, 25 Because of the fabrication process, the 3D nanostructures are hollow. However, being milled on a bulk quartz MEA, the inner nanochannel is closed at the bottom and not through‐hole. We used a device with a MEA‐like configuration consisting of 24 electrodes arranged on a 4 mm 2 surface. The 3D plasmonic nanostructures fabricated on them were used for both in situ electroporation and SERS spectroscopy (see Figure 2 a–c). To have more than one 3D plasmonic nanoelectrode in contact with the same cell, thus enhancing the probability of having contact with a central portion of the cell (for more details, see Figure S5, Supporting Information), the 3D nanostructures were fabricated with a pitch of 10 or 5 µm between each other. The 24 planar electrodes could be addressed independently applying an electrical pulse train, thus porating only the cells under investigation without affecting the rest of the culture. Figure 2 a) SEM image of a single 3D plasmonic nanoelectrode embedded in the SU8 passivation layer. b, c) Magnification of six 3D nanofabricated flat electrodes and the entire MEA‐like device, respectively. e) Tilted SEM image of a fixed and resin‐infiltrated NIH‐3T3 cell cultured on the 3D plasmonic nanoelectrodes. In correspondence with the dotted line, g) the SEM image of the FIB cross section with inverted colors reveals the cell interface with the 3D plasmonic nanoelectrodes. The SU8 passivated flat substrate that is clearly visible (in white, below the cell). d) Inset of the cross section in which the 3D plasmonic nanoelectrode is close to the nuclear envelope (indicated with the starred arrow), and the cell membrane is in tight adhesion with the device (arrows without star). f) Inset of the cross section that shows the plasma membrane tightly wrapped all around the 3D plasmonic nanoelectrode (arrows) and to the flat SU8 passivation layer. To avoid cell electroporation from nonspecific sites due to irregularities in gold deposition, the flat surface of the electrodes was passivated leaving only the tips of the 3D nanoelectrodes exposed to the cell culture (see Figure 2 a, Experimental Section, and Figure S1, Supporting Information, for more details). 22 The 3D nanostructure tips also had the highest plasmonic enhancement. 19 NIH‐3T3 cells were plated at a concentration of 1. 5 × 10 4 cells cm −2 and grown for 36 h in controlled conditions. During this time, the cells strongly adhered to the substrate, showing tight sealing with the 3D plasmonic nanoelectrodes (see Figure 2 d, f, g and Figure S3, Supporting Information, on the staining and cross‐sectioning procedure), thus allowing the Raman signals coming from the membrane to be enhanced and detected. After 36 h in culture, we replaced the cell medium with phosphate buffer solution (PBS) for performing Raman spectroscopy of the in vitro living cells in liquid. When electroporation is performed, nanopores are created at the interface with the 3D nanoelectrodes, and the cellular membrane begins to settle in an attempt at healing. 26, 27 The capability of the cells to perform mitosis is preserved as well as their viability (see also Figure S2, Supporting Information, for viability tests). 28 Cells adapt to the substrate and are free to move and replicate because of the short height of the 3D plasmonic nanoelectrodes that protrude from the flat SU8 passivation layer. 28 The time scale for major displacement of the cells is on the order of hours, 29 but minor movements are faster, and for this reason, the portion of the plasma membrane in adhesion with the 3D plasmonic nanoelectrodes is not always the same and can change within the experiments. Moreover, each 3D nanoelectrode may be in proximity to a different part of the cell, being closer to the nucleus (inset in Figure 2 d) or farther from it, in proximity to a mitochondrion (Figure 2 f) or to other organelles. Pioneer works from the last years demonstrated that 3D nanoelectrodes are able to access the intracellular environment for electrophysiological measurements as well as intracellular delivery. 7, 21, 25, 30 However, as it can be appreciated in Figure 2 d, such an approach may have, as a dramatic drawback, the poration of the nuclear envelope. Such an event is a very delicate matter to consider when designing 3D interfaces and devices that work in tight contact with cells and tissues. In fact, it may enable intranuclear delivery of biomolecules and/or sampling of intranuclear content that would be of fundamental importance in many applications. On the other hand, very little is known about potential negative effects of nuclear poration on the overall health and viability of the cells. 31 While acquiring the SERS signal and using only standard optical microscopy for observing the cells, there is no a priori certain knowledge of the cell portion lying where the measurement is performed. Rather, only a guess can be made of the proximity to the cell nucleus or the edge based on information obtained from the optical images (for more details on the cell positions, see Figure S5, Supporting Information). For our experiments, we relied only on the optical images (see Figure S4c, f, Supporting Information), and the data presented in this study were acquired from 3D plasmonic nanoelectrodes close to the center of the cells. In this configuration, the probability of being in proximity to the nucleus is high, making accessible, in theory, the nuclear envelope and its content. The experimental procedure consisted of the laser excitation of a single 3D nanoelectrode with a cell lying on it. The Raman excitation was obtained with a monochromatic laser (λ = 785 nm) focused on the 3D plasmonic nanostructure creating intense hot spots in correspondence of its tip. From the literature, it is well‐known that the electromagnetic field of localized plasmonic modes decays very fast (after few tens of nm) in space. 32, 33, 34 By characterizing the radial profile of our 3D plasmonic nanoelectrodes, we showed that the optical distribution vanishes within 20 nm from the tip surface, allowing us to detect the Raman signal coming only from a small volume around the 3D nanoelectrode, as shown in Figure S2 (Supporting Information) (more details about the optical distribution can be found in Figure S4, Supporting Information). We performed in situ electroporation by applying a potential difference between the 3D plasmonic nanoelectrode and a reference platinum electrode immersed in the cell culture. The in situ permeabilization in correspondence of the plasmonic hot spots allows the enhancement of the Raman signals where the nanopores are opened, leading to the detection of the changes in the plasma membrane and studying the dynamics of rearrangement of the lipid bilayer. We used the parameters optimized for similar substrates in previous works. 21, 22 In detail, the in situ electroporation was performed using a pulse train with a 20 Hz repetition rate, an amplitude of 3 V (offset at +1. 5 V, to have pulses from 0 to 3 V), a pulse length of 100 µs, and a train pulse duration of 10 s. We also acquired Raman spectra with different electroporation parameters. However, the spectra acquired after electroporation with higher or lower voltages did not show significant or valuable information when averaged over several experiments. In fact, higher voltages could lead to bubble formation and to higher degree of permeabilization (bigger nanopores, difficulty in membrane resealing, cell death). Lower applied voltages could be less effective for local permeabilization of the cell membrane. 3 Real‐Time Monitoring of Membrane Poration and Intracellular Environment The typical SERS spectrum of a cell lying on the 3D plasmonic nanoelectrode before the electrical pulse train application is shown as a black line in Figure 3 a, while the red line represents the typical SERS spectrum acquired from the same Raman electrode just after electroporation, and the blue spectrum is the recorded signal after 20 min from the permeabilization. As can be seen, right after the electroporation we observed a strong increase of the Raman signals. To better highlight this behavior, we monitored the Raman spectra for 30 min with and without electroporation (Figure 3 c, b, respectively). In these graphs, the average time‐resolved SERS spectra are presented as color maps ( N = 6 for the graph in Figure 3 b and N = 10 cells for the graph in Figure 3 c on at least five different cell cultures). For more detail on how the data have been processed, see the Experimental Section and Figure S4 (Supporting Information). Figure 3 a) Enhanced Raman spectra recorded from the same 3D plasmonic nanoelectrode before (black), 2 min after (red), and 20 min after (blue) the application of the electroporation pulse train. Three regions of the Raman shift are highlighted in which most of the peaks are related to lipids and proteins. The background has been subtracted and the intensities coherently shifted to improve the visualization. Data are from a single experiment. b, c) Colored maps of the average SERS signals of cells lying on top of 3D plasmonic nanoelectrodes excited by a λ = 785 nm laser during 30 min of acquisition. b) The spectra do not show particular features or changes in time when there are no external stimuli applied. c) Average SERS signals of electroporated samples at 10 min from the beginning of the experiments. After electroporation (at t = 600 s, marked by the dotted line), new vibrational modes appear, and the average peaks intensities increase. Rapid shifts of new and old peaks occur, meaning that the plasma membrane and the rest of the cell undergo rearrangement. Slowly over time, the signals come back to resemble the signal before the electroporation occurred. Due to the plasmonic enhancement of the 3D nanoelectrodes, a good signal‐to‐noise ratio was obtained using a time resolution of only 6 s (accumulation of five acquisitions, each lasting 1 se plus the processing time of ≈1 s), reaching a high time detail in respect to the long observation time of 30 min. Before or in absence of electroporation, the Raman spectra were rather stable, and minor changes appear probably due to the physiological movement of the cell, namely, the natural dynamics of the cell membrane as it moves around the 3D plasmonic nanoelectrode. In contrast, when electroporation was applied (Figure 3 c at t = 600 s), the signal changed dramatically, showing the appearance of new peaks with much higher intensities than the baseline together with an increase in intensity or temporary disappearance of old peaks. The intensity increase is mainly due to the sudden change in the environment as it became intracellular or partially intracellular, making the cytoplasm and all its content accessible for detection through the nanopores in the cell membrane. By analyzing the spectra in details, we noticed that the majority of changes that appeared after electroporation occurred in the Raman shift regions corresponding to lipids (780–890 and 1400–1550 cm −1 ) 35 or proteins (1240–1310 cm −1 ). 36 This is in agreement with the fact that when the plasma membrane undergoes a process of permeabilization, a period of rearrangement of the lipid bilayer follows with the aid of several membrane proteins and protein complexes at the interface with the membrane. 10, 27, 37 Additionally, the orientation of the molecules with respect to the electric field can contribute to the peak shifts and the changes in intensity. 38 These drastic changes in the Raman spectra lasted for ≈10 min, after which the Raman spectra began to settle to the initial values. This behavior can be ascribed to the plasma membrane healing and the resealing of the nanopores that were created through electroporation. 39, 40 After 20 min from application of the electrical pulse train, the Raman peak intensities returned to values comparable to those before the electroporation, and most of the vibrational modes that appeared with the electrical pulses vanished. As this time range is comparable to that found in several other scientific works, 21, 26 this behavior can be associated to the nanopores closure attempts at the interface with the 3D nanoelectrode and to the reformation of the membrane to the pre‐electroporation conditions. Our analysis aimed to study the averaged behavior rather than a single cell response to the external stimulus, because we could not be sure on what portion of the cell membrane were in contact with the 3D plasmonic nanoelectrode. For the interpretation of the results in fact, we need to keep in mind that live biological systems are governed by extremely complex rules and are affected by many factors, and we assumed that each different region of the plasma membrane would present a slightly different response (see, e. g. , Figures S3b and S4a, Supporting Information). In other words, results from single experiments may cause misleading interpretations. In the following, we provide a more detailed analysis of peaks related to molecule of interest. We followed the time trace of the peaks throughout the experiments to compare their behavior in the presence and in the absence of the electroporating event (for more details on how the peaks were chosen, see the temporal average analysis in Figure S6, Supporting Information). We noticed that some of the lipid‐related vibrational modes were present throughout the whole measurements, such as the peaks centered at 954 and 975 cm −1, assigned, respectively, to cholesterol 19 and to fatty acid vibrational modes ( Figure 4 a, b). 35 Looking to their time trace, the peaks intensities increase with the application of electroporation (pink temporal traces) but only by few hundreds of units with respect to the behavior in the absence of electroporation (black temporal traces). On the other hand, other lipid‐related peaks showed a very different behavior in time after permeabilization of the plasma membrane. Figure 4 c, d shows the time trace of the 875 cm −1 peak, assigned to the C—C stretching of phospholipids, 35 and the peak at 1464 cm −1 related to CH 2 \CH 3 deformations in cholesterol and triacylglycerols. 41 The two peaks presented a dramatic increase in intensity after electroporation, suggesting a drastic change in the presence and orientation in the plasma membrane of the corresponding molecules. In Figure 4 e, the peaks represented in the time traces above are highlighted in the average SERS spectra for the sake of clarity. Figure 4 f depicts some of the possible configurations of the lipid bilayer. In top left, the cell membrane is sealed, while in the top right and bottom left, the membrane is broken by hydrophobic pores, and in the bottom right, the pore is hydrophilic. 42 Figure 4 a–d) Average temporal behaviors of lipid related peaks throughout the 30 min of experiments in the absence (black) and in the presence (pink) of the electroporation, which is identified by the dotted line. The peak dynamics are shown on the average spectra. In particular, the dynamics of the a) 954 cm −1 peak assigned to cholesterol, b) 975 cm −1 peak assigned to fatty acid, c) 875 cm −1 peak assigned to the C—C stretching of phospholipids, and d) 1464 cm −1 peak assigned to CH 2 \CH 3 deformations in cholesterol and triacylglycerols. e) Highlighted peaks from the global colored map of electroporated samples. Scale bar from 0 a. u. (black) to 5000 a. u. (red). f) Sketches of possible lipid membrane configurations. Top left: intact lipid bilayer, top right and bottom left: hydrophobic pores in the cell membrane, bottom right: hydrophilic pore. We suggest that these configurations may explain the transient appearance and variations of the lipids peaks right after electroporation. We noticed that the variations reported in Figure 4 a–d were not synchronized, suggesting that each molecule had a specific role (or not) in the rearrangement of the membrane after permeabilization in a certain temporal order. Importantly, such a lack of synchronization among the temporal dynamics of the different peaks and the different behavior in respect to the control samples confirmed that the detected changes were not related to the measurement (i. e. , laser power oscillations, environment conditions) but only to the electroporation event. After the membrane reformation, the Raman spectra still present some variations in respect to the initial spectra, possibly due to the membrane fluidity, which is still affected by the electroporation protocol. 3. 1 Observation of Aromatic Amino Acid and Amide Vibrational Modes The temporal evolution of the protein vibrational modes was studied ( Figure 5 ), and in particular, the dynamics of three peaks assigned to the aromatic amino acids were investigated, including tyrosine, identified at 830 cm −1, 43 the more intense peak at 1002 cm −1 related to the phenylalanine, 44 and the peak at 1552 cm −1 assigned to tryptophan. 43 In addition, the dynamics of the two peaks at 1302 and at 1545 cm −1, assigned to the C—N stretching and N—H bending of, respectively, amide III 36, 44, 45 and amide II 45, 46 vibrational modes, were studied. Different behaviors were noticed when electroporation was induced (purple lines in Figure 5 a–e) relative to measurements made without external stimulation (black spectra in Figure 5 a–e). Figure 5 Average temporal behaviors of protein‐related peaks throughout the 30 min of experiments in the absence (black) and in the presence (purple) of electroporation, identified by the dotted line. The peak dynamics are shown on the average spectra. In particular, dynamics of the a) 830 cm −1 peak, associated with the tyrosine amino acid, b) 1002 cm −1 peak assigned to phenylalanine, c) 1302 cm −1 peak that identifies the amide III vibrational mode, d) 1545 cm −1 peak assigned to the amide II vibrational mode, and e) 1552 cm −1 peak, assigned to the tryptophan amino acid. f) Highlighted peaks from the colored average map. Scale bar from 0 a. u. (black) to 5000 a. u. (red). In particular, the amino acid modes have a very stable and low intensity in the absence of electroporation, while these modes appeared afterward with a very strong intensity (Figure 5 a, b, e). In contrast, the amide II and amide III vibrational modes (Figure 5 c, d) also presented some intense peaks before the application of the electrical pulse train. However, the effect of the plasma membrane permeabilization was visible in these vibrational modes as well. These dynamics, and in particular the amino acid vibrational modes, suggested the formation of hydrophilic nanopores on the plasma membrane that allowed cytoplasmic proteins to get closer to the 3D plasmonic nanoelectrode, leading to the consequent detection of their enhanced Raman spectra. The results reported so far show that the use of 3D multifunctional nanostructures combined with MEA and Raman spectroscopy enable following spontaneous physiological changes of the plasma membrane and reorganization processes occurring after electroporation. The presented approach can also give insights into the composition and evolution of intracellular compounds. 3. 2 Nuclear Poration From the SEM cross section shown in Figure 2 inset d, one can notice that the nuclear membrane can be in close proximity to the 3D nanoelectrode, thus making nuclear poration feasible. In that scenario, DNA can exit from the safety of the nucleus and move close to the plasmonic enhancer, where its SERS signal can be detected. For this reason, the temporal behavior of the peaks assigned to nucleic acid (DNA and RNA) was analyzed in the presence and in the absence of electroporation. Figure 6 a shows the time evolution of the peak at 790 cm −1 that has been assigned to the C′ 3 —O—P—O—C′ 5 phosphodiester bond of DNA and RNA, 47, 48 while the vibrational modes at 1120 cm −1 have been generically assigned to nucleic acid (Figure 6 b). 43 The peak centered at 1252 cm −1 has been assigned to the NH 2 vibrational modes of cytosine and guanine (Figure 6 c), 49 and the peak at 1573 cm −1 has been assigned to guanine and adenine vibrational modes (Figure 6 d). 50 The four Raman enhanced peaks showed a similar behavior in time throughout the experiments, with minimal changes in intensity in the absence of electroporation (black traces and red traces before the dotted line) and larger intensities after application of the electroporating pulse train. Figure 6 Average dynamic behavior in time of DNA‐ and nucleic acid‐associated peaks. a) 790 cm −1 peak associated with the O—P—O stretching of DNA and RNA backbone, b) vibrational mode assigned to nucleic acid at 1120 cm −1, c) peak centered at 1252 cm −1 associated with vibration of cytosine and guanine, and d) peak at 1573 cm −1 related to guanine and adenine vibrational modes. Red spectra are the average of electroporated samples at 600 s (dotted line), while black spectra are the reference to which electroporation has not been applied. e) Highlighted peaks from the global colored map that indicated electroporation. Scale bar from 0 a. u. (black) to 5000 a. u. (red). f) SEM cross section (with inverted colors) of a 3D plasmonic nanoelectrode tip with a cell grown on it. The sample was fixed and analyzed after the application of electroporation. The starred arrows indicate the nuclear envelop, clearly broken close to the edge of the 3D nanostructure. The arrows without the star indicate the plasma membrane. Being able to detect RNA offers a new method to sequence and investigate the whole RNA pool of a single cell within a large culture, offering a new detection method to the challenging field of transcriptomics. 51, 52 Because the Raman signals could also be originated from molecules that are present in the cytoplasm, such as mRNA or ribosomes, we performed a more detailed cross‐sectional SEM investigation of the nuclear membrane after electroporation to corroborate the nuclear poration hypothesis. Figure 6 f shows an SEM cross‐section image of an electroporated cell, suggesting the possibility of directly porating the nuclear envelope to gain access not only to the cytoplasmic compartment, but also to the nuclear environment. In particular, the cross section presents a 3D plasmonic nanoelectrode that was close to the nucleus, and the nuclear membrane appears porated. From the SEM image, it can be seen that the tip of the 3D plasmonic nanoelectrode results in contact with the intranuclear compartment, allowing access to the information stored inside (see also Figure S6, Supporting Information). Thus, although the presented Raman fingerprints cannot be unequivocally attributed to nuclear DNA, the SEM cross section supports the hypothesis that the nuclear content may become accessible for Raman spectroscopy after in situ permeabilization. More experiments are required to investigate this intriguing possibility. We did not notice any change in the viability of the cells after the electroporation. This result means that, assuming we are in fact assisting to a nuclear poration, the tested experimental conditions are not toxic for the cells. Interestingly, the protein‐related peaks and the RNA‐ or DNA‐related peaks behaved differently in time. The nucleic acid and backbone vibrations, in fact, appeared a few minutes after electroporation (Figure 6 a, c), while the peaks associated with protein vibrational modes presented an increase in their intensity just after application of the electrical pulse train (Figure 5 a–c, e). This kind of behavior indicates that the DNA and nucleic acid molecules were not present on the cell membrane, as expected, and they needed some time to diffuse close to the 3D plasmonic nanoelectrodes and finally become detectable after permeabilization of both the plasma and nuclear membranes. In general, these results suggest a variety of different membrane reforming mechanisms because of the different recruitment times of different molecules. 4 Conclusion In this work, we presented a multifunctional platform based on MEA refined with 3D plasmonic nanostructures, which acts both as a plasmonic enhancer and as nanoelectrodes, allowing local permeabilization of the cell membrane exactly in correspondence with the plasmonic hot spot. Using enhanced Raman spectroscopy, we studied the local permeabilization of the cellular membrane and recorded the fingerprints of the molecules involved in the subsequent rearrangement for several minutes. The results of the molecular rearrangement following permeabilization of the cell membrane are in accordance with data in the literature. In fact, the experiments suggested an average closure time of the nanopores on the order of 10 min after electroporation, 39, 40 followed by a settlement period in which the membrane still results more fluid than it was before the electroporation. Remarkably, during this time window, the 3D nanoelectrode was in direct contact with the cytosol, thus providing insight into the intracellular compounds in close proximity to the 3D plasmonic nanoelectrode. Raman peaks related to lipids, proteins, and nucleic acids can be monitored. Surprisingly, DNA‐related peaks could also be found, thus suggesting nuclear envelope poration. The latter was supported by SEM cross sections, and it may represent a straightforward approach to sample or to deliver biomolecules into the nucleus. The presented method can also be a powerful approach to investigate how cells behave or proliferate on 3D nanostructures, which are promising substrates for next‐generation tissue engineering and prostheses. 5 Experimental Section Fabrication and Passivation : MEAs were fabricated by standard lithographic methods on quartz wafers. 3D vertical nanostructures were fabricated by milling through an optical resist by means of FIB. The device was passivated with an epoxy polymer (SU8), leaving only the tip of the 3D nanostructures exposed to the environment. More details on the fabrication are in the Supporting Information. Cell Culture : Before seeding of the cells, the devices were treated with plasma oxygen (60 s, 100% O 2, 100 W) to improve their wettability and sterilized through UV exposure (20 min) in a laminar flow hood. NIH‐3T3 cells were seeded on the devices with (1. 5 × 10 4 cells cm −2 ) in Dulbecco's modified Eagle's medium (DMEM) cellular medium with penicillin/streptomycin (1% pen/strep) antibiotic and fetal bovine serum (10% FBS, all by Sigma‐Aldrich) and grown at controlled humidity, CO 2 concentration, and temperature for 36 h prior to performing experiments. Cell Staining, FIB Cross Section, and SEM Imaging : Cells were fixed with glutaraldehyde solution (2. 5%) in Na cacodylate buffer (0. 1 m ) for at least 1 h on ice. Samples were extensively washed in buffer solution and incubated with glycine (20 × 10 −3 m ) in buffer solution on ice. Then, a recently developed RO‐T‐O staining protocol was performed, 53 and samples were embedded in a thin layer of Spurr resin. For more details on the procedure, see the Supporting Information. The cell cross sections were performed using a dual‐beam Helios Nanolab 650 by ThermoFisher. Slicing of the specimens was performed using high ionic currents (9. 3 or 0. 79 nA) after Pt deposition using the gas injection system (GIS) of the instrument. More details can be found in the Supporting Information. Imaging was performed with the samples tilted at 52° with respect to the electron beam, and backscattered electrons were collected using a TLD detector in immersion mode. The acceleration of the primary electrons was 3 kV, and the current used in the imaging process was i = 0. 40 nA. The cross‐sectioned images are presented with inverted colors to emphasize the cellular membranes. Electromagnetic Field Enhancement Simulation : A finite element method analysis implemented in COMSOL Multiphysics software was performed, by simulating a 1. 8 µm high 3D nanoelectrode covered with 60 nm of gold and embedded in SU‐8 polymer, leaving 700 nm of 3D plasmonic nanostructure exposed to the interface with water. The incident light is a monochromatic (λ = 785 nm) linearly polarized plane wave. Electroporation and Raman Measurement : In situ electroporation was performed by applying a pulse train with a 20 Hz repetition rate, an amplitude of 3 V, a pulse length of 100 µs, and a train pulse duration of 10 s. Raman spectra were measured by a Renishaw inVia Raman spectrometer with a Nikon 60× water immersion objective and a 1. 0 NA delivering a 785 nm laser with a power of ≈0. 6 mW. Each spectrum was the result of 1 s acquisition accumulated five times. Data Analysis : All of the Raman signals acquired during laboratory activities were processed with custom‐made programs in MATLAB 2017 or C++, and then visualized in Origin 9. 0 prior to a second analysis using this software. The data acquired by Wire3. 4 Renishaw software were exported and saved in. txt format in order to facilitate their transfer. For more details, see the Supporting Information. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800638
| 2,018
|
Advanced Science
|
Adaptable Fast Relaxing Boronate‐Based Hydrogels for Probing Cell–Matrix Interactions
|
Abstract Hydrogels with tunable viscoelasticity hold promise as materials that can recapitulate many dynamic mechanical properties found in native tissues. Here, covalent adaptable boronate bonds are exploited to prepare hydrogels that exhibit fast relaxation, with relaxation time constants on the order of seconds or less, but are stable for long‐term cell culture and are cytocompatible for 3D cell encapsulation. Using human mesenchymal stem cells (hMSC) as a model, the fast relaxation matrix mechanics are found to promote cell–matrix interactions, leading to spreading and an increase in nuclear volume, and induce yes‐associated protein/PDZ binding domain nuclear localization at longer times. All of these effects are exclusively based on the hMSCs' ability to physically remodel their surrounding microenvironment. Given the increasingly recognized importance of viscoelasticity in controlling cell function and fate, it is expected that the synthetic strategies and material platform presented should provide a useful system to study mechanotransduction on and within viscoelastic environments and explore many questions related to matrix biology.
|
Hydrogels have been widely used as matrices to culture and grow primary cells in 3D with applications in tissue engineering and regenerative medicine. 1 By incorporating various biochemical, biophysical, and mechanical cues into a single system, hydrogels provide a powerful in vitro model to probe and understand the fundamentals of cell‐niche interactions 2 and to provide information for testing hypotheses related to developmental and disease‐related processes. 3, 4 Accumulating evidence in the literature suggests that the modulus of synthetic extracellular matrices (ECMs) can have a significant influence on cell function and fate. 2, 5 However, more recently, it has been realized that other time‐dependent mechanical properties, such as viscoelasticity, can also substantially impact cellular responses. 6, 7, 8, 9 For example, Cooper‐White and co‐workers varied the crosslinking density in poly(acrylamide) gels to control the gels' creep response and showed that the spreading area and proliferation of human mesenchymal stem cells (hMSCs) increased with viscous modulus (or the energy dissipation capacity). 6 Later, Anseth and co‐workers reported on covalent adaptable networks with hydrazone bonds to prepare poly(ethylene glycol) (PEG) hydrogels with tunable rates of stress relaxation. 7 Encapsulated C2C12 cells exhibited robust cytoskeletal growth in fast relaxing gels but remained rounded in the slowly relaxing counterparts. Very recently, Chaudhuri et al. controlled the relaxation rate of alginate hydrogels by changing the polymer molecular weights and binding affinity of the calcium chelating domains. Their results suggested that stress relaxation promotes the formation of stress fibers, focal adhesions, nuclear localization of a mechanosensitive transcription regulator yes‐associated protein (YAP), and even osteogenic differentiation of MSCs. 8, 9 Collectively, these studies demonstrate that in addition to matrix modulus, the time‐varying mechanical properties (viscoelasticity) of synthetic ECMs can alter cellular mechanotransduction processes. Because many native tissues exhibit viscoelasticity, viscoelastic synthetic ECMs more closely mimic the dynamic aspects of the cellular microenvironments. This allows experimenters to probe dynamic cell–matrix interactions and provide information that is otherwise difficult to obtain using traditional elastic material systems. While previous studies have begun to unravel how cells sense their surrounding environment in viscoelastic materials, the field's knowledge of the viscoelastic effect on cellular response remains limited, especially on the short time scales where many important biological processes take place. For instance, on the subcellular level, mechanical signals can propagate from the cytoplasm to the nucleus on the order of seconds, resulting in a concomitant mechanochemical conversion. 10 However, it is unclear how, and to what extent, these mechanical signals on the small timescales may be integrated to elicit cellular behavior across a broad time spectrum. At the tissue level, many soft tissues such as adipose, 11 liver, 12 and brain 13 relax more than 50% of the imposed stress within 10 s or less. Therefore, synthetic materials with a broader range of fast matrix relaxation kinetics may allow one to probe more connections between time‐dependent material mechanics and cell behavior. Furthermore, based on findings from previous studies, 6, 7, 8, 9, 14 it is reasoned that increasing the rate of matrix relaxation better promotes cell–matrix interactions and potentially amplifies cellular responses to the viscoelastic properties of synthetic ECMs. From a polymer physics perspective, the rate of stress relaxation, or the characteristic time scale of the viscoelastic properties, in a hydrogel is inherently dictated by the type of physically associative or reversible chemical bonds that create the network crosslinks. 15, 16, 17, 18 In search of faster relaxation motifs, our investigation focused on dynamic boronate bonds, which started to gain attention in autonomous self‐healing materials, 19, 20 functional bioconjugates, 21 and complex chemical systems. 22, 23 Boronic esters are often formed between boronic acids and cis ‐1, 2‐diols ( Figure 1 a). Compared to other reversible bonds or associating motifs, boronates exhibit much faster association and dissociation dynamics. 22 Moreover, unlike aldehydes that are potentially cytotoxic at high concentrations, 7 careful selection of the boronic acids and cis ‐1, 2‐diols can circumvent such toxicity concerns. Finally, the formation of boronates occurs with high reaction specificity, with less interference from many functional groups that are present in complex biochemical environment. 22, 23 These features make boronate an attractive candidate for creating fast relaxing materials for studying cellular mechanotransduction processes. Figure 1 Design of covalent adaptable networks based on dynamic boronate bonds. a) Top: General reaction scheme showing the reversible formation of boronates between boronic acids and cis ‐1, 2‐diols. Bottom: Chemical structures of 2‐fluorophenylboronic acid (FPBA), boroxole (BX), Wulff‐type‐like boronic acid (WBA), catechol (CAT), and nitro‐dopamine (ND) studied in the experiments. b) A schematic of a dynamic network based on boronates from boronic acids and cis ‐1, 2‐diols functionalized octa‐arm PEGs. Unfunctionalized chain ends, dangling chains, and (secondary) loops are depicted as possible defects in the networks. Although boronate‐based hydrogels have been investigated extensively, few reports demonstrate their use as stable 3D cell culture platforms, 19, 24, 25, 26, 27 likely due to the weak binding strength between boronic acids and diols, particularly under physiological condition. To address this problem, this work rationally selects boronic acid variants and diols that can form dynamic boronate bonds, even in complex biochemical environments (e. g. , serum containing media). Specifically, 2‐fluorophenylboronic acid (FPBA), 1‐hydroxy‐1, 3‐dihydrobenzo[c]oxaborole ( m ‐boroxole, BX), and a Wulff‐type o ‐aminomethylphenylboronic acid (WBA) (Figure 1 a) are chosen as the boronic acid derivates. Previous studies have shown that these motifs can form dynamic boronate bonds near physiolocial pH, 27, 28, 29 by lowering the p K a of boronic acids through an electron withdrawing group, intramolecular B—O coordination, and intramolecular B—N coordination, respectively. In addition, catechol is selected as the diol to attain strong binding between boronic acids and cis ‐1, 2‐diols because the association constants of the catechol‐boronic acid pairs are often two orders of magnitude larger than the complexes formed by conformationally more flexible saccharides. 30 However, our preliminary experiments using the catechol moieties showed that oxidation of catechol is inevitable over several days, which changes the mechanical properties of the gels with time (Figure S1, Supporting Information). Inspired by a study from Ding et al. , 31 nitro‐dopamine (ND) with an electron withdrawing nitro group is used to improve the oxidative stability of the catechol. These rational design considerations are expected to tailor the boronate‐based hydrogel properties, rending them particularly suitable for long‐term 3D cell culture. Polymer networks based on reversible boronate bonds show robust gel formation and a nearly ideal network structure as inferred from rheological characterization. The selected boronic acid and nitrodopamine motifs were covalently conjugated to star‐shaped octa‐arm PEG polymers ( M w = 20. 0 kg mol −1, Đ = 1. 03) via efficient coupling reactions (synthetic details in the Supporting Information). The end group functionalization efficiencies determined by nuclear magnetic resonance (NMR) were about 90% for all the polymers synthesized. When equal parts of the boronic acid and nitro‐dopamine macromolecular gel precursors were mixed in 100 × 10 −3 m phosphate buffer at pH 7. 4, gelation took place within several seconds. Time sweep experiments suggest that gelation was rapid, and the time required to reach the gel point was shorter than the sample preparation time needed to setup the rheological experiment (Figure S3, Supporting Information). Frequency sweep tests show that boronate‐based hydrogels exhibit typical mechanical responses similar to other dynamic covalent adaptable networks 7 or physically associating networks, 17, 18 including a high‐frequency plateau modulus, a G ′ − G ′′ crossover in the intermediate frequency range indicating transition from a solid‐like to a liquid‐like behavior, and a transition to the terminal relaxation behavior where G ′ and G ′′ approach to the terminal scaling ( G ′ ∼ ω 2 and G ′′ ∼ ω) ( Figure 2 a and Figure S2, Supporting Information). The mechanical spectra of these dynamic hydrogels were further modeled with stretched exponential functions (fitting details in the Supporting Information). The stretched exponents of the fits are around 0. 90, suggesting that the rheological behavior of these materials is nearly Maxwellian and the gels behave nearly as ideal networks, despite the presence of the small defects created by incomplete polymer functionalization and some small length‐scale inhomogeneities that exist the network structure (e. g. , loops and clusters). The well‐defined molecular and network structures make these materials attractive and suitable model systems for exploring their structure–property relationships using polymer physics theories. Figure 2 Tunable viscoelastic properties of boronate hydrogels. a) A representative frequency sweep spectrum of boronate hydrogels at 10% (w/v). The rheological master curve with closed symbols is obtained from empirical time‐temperature superposition referenced at 37 °C. The black dashed lines are fits to a stretched exponential function, described in the Supporting Information. b) Comparison of the high‐frequency plateau moduli G ′ ∞ of gels made from WBA, BX, and FPBA bonding to ND. The black dashed lines show the theoretical moduli estimated from the modified phantom network theory that relates the small molecule equilibrium constant with the network modulus. c) Comparison of the network relaxation time τ = 2π/ω c of different gels plotted on a logarithmic scale. Error bars in (b) and (c) represent standard error of the mean (S. E. M. ) from repeated experiments ( n ≥ 3). Statistical analysis from one‐way analysis of variance (ANOVA) test, * and **** indicating p < 0. 05 and 0. 0001, respectively. The viscoelastic rheological properties of the boronate‐based hydrogels can be effectively controlled through small molecule design. Two rheological signatures are quantitatively compared to inform the functional group dependence in material properties. The first one is the high‐frequency elastic network modulus G ′ ∞ (Figure 2 b). It is observed that gels synthesized from FPBA motifs give the largest G ′ ∞, 20. 2 ± 3. 8 kPa for 10% (w/v) gels, whereas ones made from WBA form the softest materials, with a G ′ ∞ of 9. 1 ± 0. 8 kPa. Because these dynamic gels are built on reversible bonds, it is reasoned that the G ′ ∞ is closely related to the small molecule binding constants. Following the method from Lascano et al. , 22 UV–vis spectra of ND‐containing solutions were measured in the presence of varying concentrations of different boronic acids (Figure S4, Supporting Information) and the changes in absorption were used to derive the small‐molecule dissociation constants ( K D ) of various boronate pairs (Figure S5, method details described in the Supporting Information). Consistent with the hydrogel rheological characterization, the small molecule measurements show that the FPBA‐ND pair is the strongest bond among the three investigated, with a K D of 0. 23 × 10 −3 ± 0. 02 × 10 −3 m, and the binding affinity to ND weakens with BX and WBA (Table S1, Supporting Information). To further elucidate the correlation between the small molecule K D and the high‐frequency elastic network modulus G ′ ∞, phantom network theory was modified to account for the effect of the dissociation/association of the boronates on network mechanics (derivation in the Supporting Information). The theoretical modulus provided by the modified theory matches the experimentally measured results and increases with small molecule K D (Figure 2 b), further confirming that binding energetics controls the high‐frequency plateau modulus of the dynamic networks. The discrepancy between theory and experiments can be attributed to the presence of molecular defects and network imperfections, such as loops and dangling chains, and non‐negligible network structural inhomogeneity for gels near the overlap concentration. Variation in the structure of boronic acids not only changes the network modulus but also significantly alters the relaxation behavior of the materials. Here, the relaxation dynamics of boronate gels is characterized by frequency sweeps instead of direct stress relaxation tests to avoid confounding effects from instrument inertia when characterizing fast‐relaxing materials in the short‐time regime. The relaxation time of these dynamic hydrogels is inferred from the G ′ − G ′′ crossover frequency, τ = 2π/ω c. For all the gels studied here, the relaxation time constants are on the order of 1 s or below (Figure 2 c), suggesting fast network relaxation compared to many other dynamic hydrogels reported in literature. 7, 9, 17 Interestingly, τ and G ′ ∞ share the same dependence on the boronic acid structure: the strongest FPBA‐ND bonds show the slowest relaxation and the weakest WBA‐ND bonds show the fastest. For S N 1 type dissociation mechanisms, many studies have concluded that the network relaxation time is governed by the small‐molecule dissociation time. 15, 18 While it is challenging to directly measure the dissociation rate constants of the boronate bonds in aqueous solution, the timescale on which the boronates dissociate is in a similar range as reported by others. 22, 32 Overall, our results demonstrate that the viscoelastic properties of these covalent adaptable networks can be directly related to the properties of the reversible small‐molecule boronate pairs, which provides a foundational basis for the rational design of dynamic gels for cell studies. A generalizable strategy was further developed to prepare hydrogels that would be stable under cell culture condition, yet still exhibiting significant stress relaxation. While the aforementioned experiments show that the dynamic mechanical properties of gels can be tuned by the chemical structure of the boronate bonds, the fast association–dissociation dynamics renders these materials completely soluble when placed in a sink of cell culture media, often containing excess monosaccharides that facilitate dissociation. Indeed, all the boronate hydrogels completely dissolved in low glucose Dulbecco's modified Eagle's medium (DMEM) within less than 4 h at 37 °C. To overcome the limitation of rapid gel erosion, a secondary, orthogonal crosslinking mechanism was introduced into boronate hydrogels, where the chemistry is based on permanent bonds formed by strain promoted azide‐alkyne cycloaddition (SPAAC) ( Figure 3 a). When the amount of permanent bonds is above the network percolation threshold, it is expected that the majority of macromolecular gel precursors are covalently bound to a stable, three dimensionally crosslinked network, thus minimizing mass loss over time. Figure 3 Design of stable hydrogels that exhibit fast relaxation dynamics. a) A schematic showing hybrid networks containing both reversible boronate bonds and permanent SPAAC bonds. b) Representative frequency sweep of hydrogels after being swollen in hMSC growth media (low glucose DMEM, supplemented with 10% fetal bovine serum (FBS), 100 U mL −1 penicillin, 100 µg mL −1 streptomycin, and 1 µg mL −1 fungizone) for 7 d. Measurement was performed at 37 °C. The inset photograph shows gel appearance after being swollen in media for 7 d, demonstrating gel stability. Scale bar 1 cm. To prove the feasibility of this design concept, a fraction of the end groups of the octa‐arm PEGs were first derivatized into azides and then functionalized with FPBA or ND, respectively (synthetic details in the Supporting Information). The functionality of the azide moieties was confirmed by NMR, showing that each polymer had approximately two azide groups on average. By mixing octa‐arm PEG‐dibenzylcyclooctyne, octa‐arm PEG‐ND/azide, and octa‐arm PEG‐FPBA/azide, gels with hybrid crosslinks formed very quickly, on the order of seconds. Rheological characterization of hybrid gels shows distinctive features compared to the purely adaptable networks (Figure 3 b). In the low frequency regime where ω < 0. 1 rad s −1, G ′ is always larger than G ′′ in the hybrid gels and G ′ approaches a second plateau modulus. On these relatively large timescales, all the reversible boronate bonds are relaxed and therefore do not contribute to the mechanical properties. In contrast, the SPAAC motifs are irreversible, and the hybrid networks maintain its elastic response, even at long times. The network relaxation time, estimated from the frequency at which G ′′ max occurs, 33, 34, 35, 36 is approximately twice that of the purely adaptable boronate gels. This slight increase in the relaxation time may be due to topological constraints imposed by the permanent network, and similar phenomena have been reported in engineered protein hydrogels with covalent crosslinks 36 or chain entanglement. 35 Compared to protein‐polymer hydrogels developed by Dooling and Tirrell, 37 the boronate‐based hydrogels developed in this work can access similar fast relaxation rates but achieve a larger degree of stress relaxation (over 90% rather than 32–59%, Figure 3 b), a characteristic more reminiscent of soft tissues. Furthermore, equilibrium calculations (Supporting Information) suggest that the majority of the boronate bonds retains their adaptability even in the presence of glucose, which only causes a 3% decrease in the total amount of boronate bonds. This is attributed to the dissociation constant of the FPBA‐ND pair, which is approximately two orders of magnitude smaller than that of the FPBA‐glucose pair. This analysis also emphasizes the importance of using ND moieties to attain strong binding to boronic acids. More importantly, these materials are stable under typically used experimental conditions for culturing primary cells and the hMSCs used herein. The initial modulus, relaxation time, dry mass, and volumetric swelling ratio of the gels do not show statistically significant differences over 7 d in media (Figures S6 and S7, Supporting Information). These observations also suggest that the oxidation of ND is negligible in cell culture media within the experimental time window. The absence of degradation and the fast relaxation dynamics make these materials useful for in vitro cell culture in either two or three dimensions (i. e. , cells seeded on the surface or encapsulated in the hydrogels). Hydrogels with the mixed types of crosslinks show exceptional cytocompatibility. Here, hMSCs were encapsulated at a density of 1 × 10 6 cells mL −1 and cultured in growth media for 7 d. For a control, hMSCs were encapsulated at the same density in a purely elastic hydrogel prepared from the SPAAC chemistry alone (formulation in Table S2, Supporting Information), whose elastic modulus ( E = 14. 1 ± 2. 7 kPa, assuming a Poisson's ratio of 0. 5) closely matched the initial modulus of the hybrid gels ( E = 16. 5 ± 2. 4 kPa), despite a small difference in the polymer concentration of gels. Both gels had the fibronectin‐mimicking peptide GRGDS incorporated at a concentration of 3 × 10 −3 m to promote cell–matrix adhesion. On day 1, cells in both gels showed high viability post encapsulation, 97. 0% ± 0. 4% and 93. 3% ± 2. 8% for stress‐relaxing and elastic gels, respectively ( Figure 4 ). On day 7, cells remained over 90% viable in stress‐relaxing gels, and a significant fraction of the cells had spread within these stress‐relaxing materials. In contrast, cell viability decreased to just 81. 3% ± 2. 0% in the elastic control, and the cells remained rounded. These results show that stress relaxation better maintains cell viability and permits cell–matrix interactions and spreading, even without the presence of any permanent degradation mechanism, such as enzymatically or hydrolytically cleavable linkers. Furthermore, cell proliferation was quantified by directly counting the nuclei density in images of cell‐laden gels. Although the cell density showed a slight increase in the stress‐relaxing gels from day 1 to day 7, the small difference was not statistically significant (Figure S8, Supporting Information). In comparison, the cell density decreased to ≈82% in the elastic gels on day 7 (Figure S8, Supporting Information), which suggests that the aforementioned differences in cell viability are less likely to be related to differences in proliferation. Figure 4 Cytocompatibility of boronate‐based hydrogels. a) Representative maximum intensity projection images of cells in elastic and stress‐relaxing hydrogels on day 1 and day 7. Cells were stained with calcein AM (green, live) and ethidium homodimer (red, dead). Scale bar = 100 µm. b) Quantification of cellular viability. Error bars represent the standard errors of the mean (S. E. M. ). Statistical analysis from one‐way ANOVA test, * indicating p < 0. 05. Intrigued by the apparent differences observed in the cell morphology between the stress‐relaxing and elastic gels indicated from the calcein staining, the effect of matrix stress relaxation on hMSC morphology and cytoskeletal organization was systematically investigated. Cells were stained for F‐actin with rhodamine phalloidin and for nuclei with 4′, 6‐diamidino‐2‐phenylindole ( Figure 5 a). In the elastic controls, cell volume and sphericity did not change appreciably from day 1 to day 7 (Figure 5 b, c). Strikingly, hMSCs were significantly larger in the stress‐relaxing gels and showed even early stages of spreading on day 1, as indicated by the calculated decrease in sphericity (the definition of sphericity is provided in the Supporting Information). During the time course of culture from day 1 to day 7, cell volume continued increasing (Figure 5 b and Figure S9, Supporting Information), eventually reaching ≈23 × 10 3 µm 3, nearly four times that in the elastic control. The increase in cell volume is reminiscent of previous finding in enzyme‐degradable gels; 38 however, in the adaptable, stress relaxing gels the changes in cell shape is only permitted through physical remodeling of the surrounding networks. The mechanistic cause for the increase in cell volume might possibly be attributed to an increased density in intracellular content such as proteins and DNAs, and/or mechanical activation of ion channels regulating water influx. 39 In addition, it is observed that a large fraction of cells extended processes in the stress‐relaxing gels (Figure S10, Supporting Information), reinforcing the fact that the reversible boronate bonds provide a unique means for cell matrix mechanical signaling. Figure 5 Comparison of cell morphology and YAP/TAZ subcellular localization in elastic and stress‐relaxing gels. a) Representative immunofluorescence staining for F‐actin (orange), nucleus (blue), and YAP/TAZ (magenta) in hMSCs cultured in elastic and stress‐relaxing gels for 1 and 7 d. Images are shown as one cross‐sectional slice of a z ‐stack with a step size of 1. 5 µm. Scale bar = 5 µm. b–f) Quantification of cell volume, cell sphericity, nuclear volume, nuclear sphericity, and YAP/TAZ nuclear to cytosolic intensity ratio at the corresponding condition. Statistical analysis from one‐way ANOVA test, *, **, and **** indicating p < 0. 05, 0. 01, and 0. 0001, respectively. Experiments were performed in three replicates per condition, and over 30 cells analyzed per replicate. To further investigate how cells respond to the fast stress‐relaxing microenvironment, temporal changes in subcellular localization of the transcriptional coactivators YAP/TAZ (yes‐associated proteins and PDZ binding domains) were studied. Recent investigations have shown that YAP/TAZ behaves as cellular rheostats, which largely influence mechanosensing and affect cell function and fate. 2, 40 For hMSCs in elastic hydrogels, YAP/TAZ mainly remained in the cytoplasm throughout the 7 d culture (Figure 5 a, f and more representative images in Figure S11, Supporting Information), reminiscent of previous findings in slowly relaxing 9 or nondegradable gels. 38 In contrast, significant changes in the YAP/TAZ nuclear to cytoplasmic intensity ratio were observed for cells in the stress relaxing gels. Specifically, YAP/TAZ was primarily in the cytoplasm on day 1, as the case in purely elastic controls; however, quantitative analysis suggested a significant difference between the two gels (Figure 5 f). While the mechanistic cause for this finding remains elusive, these results indicate that stress relaxation changes the rate at which cells respond to their surrounding matrices. Even though the relaxation of adaptable gels is fast, on the very short time scales (seconds), significant matrix remodeling is a consequence of numerous cycles of cellular applied stress/strain and matrix relaxation/creep. Therefore, fast viscoelastic mechanics can influence cellular behavior when integrated across a broad time spectrum, even on the long time scales (hours to days) where cell shape and YAP/TAZ subcellular distribution continuously evolve. By day 7, a large population of hMSCs showed nuclear localized YAP/TAZ and the YAP/TAZ nuclear to cytosolic intensity ratio significantly increased (Figure 5 a, f). Collectively, these results indicate that YAP/TAZ subcellular localization might be directly related to a cell's ability to remodel its surrounding environment. Interestingly, the aforementioned changes in cell volume and sphericity, as well as YAP/TAZ subcellular localization, are associated with changes in nuclear volume and shape (Figure 5 d, e). This observation implicates a possible role of the nucleus in mechanotransduction in viscoelastic microenvironment, as recently reviewed by Discher and co‐workers 41 and Szczesny and Mauck. 42 In addition, correlations were found between cell volume, nuclear volume, and YAP/TAZ nuclear to cytoplasmic intensity ratio (Figure S12, Supporting Information). Given the importance of YAP/TAZ in controlling gene expression, it is expected that these fast‐relaxing gels will likely provide a useful tool to manipulate important mechanical signals for directing MSC fate and current work on this topic is under way. Last, it should be noted that the experiments above were repeated using hMSCs from a different donor, and all the findings on the effect of viscoelastic properties on stem cell morphology, nuclear morphology, and YAP/TAZ subcellular localization hold (Figure S13, Supporting Information). While admittedly there are some subtle differences in the results between the two biological replicates, these discrepancies are likely due to the differences in donor age and race, but the effect of stress relaxation on cellular responses is general. In summary, this work has demonstrated that the viscoelastic properties of boronate‐based hydrogels can be readily tuned by changing the small‐molecule structure and properties through rational design. In addition, a robust strategy has been developed to prepare fast stress‐relaxing hydrogels that maintain structural and mechanical properties under typical conditions for culturing primary cells (e. g. , hMSCs). While the current study only uses reversible boronate bonds as the stress relaxation motifs, the generality of this approach enables facile incorporation of other dynamic bonds to systematically control the viscoelastic properties of hydrogels over a broad range of timescales. Finally, these fast‐relaxing hydrogels are shown to be cytocompatible, promote cell–matrix interactions, and induce cell spreading and YAP/TAZ nuclear localization. Given the relevance of fast relaxation with many biological processes from subcellular structures to a whole tissue, it is expected that these materials will find use as versatile platforms to further probe the fundamentals of cell–matrix interactions. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201800776
| 2,018
|
Advanced Science
|
Silicon‐Enhanced Adipogenesis and Angiogenesis for Vascularized Adipose Tissue Engineering
|
Abstract The enhancement of adipogenic differentiation of bone marrow mesenchymal stem cells (BMSCs) and sufficient vascularization remain great challenges for the successful reconstruction of engineered adipose tissue. Here, the bioactive effects of silicon (Si) ions on adipogenic differentiation of human BMSCs (HBMSCs) and the stimulation of vascularization during adipose tissue regeneration are reported. The results show that Si ions can enhance adipogenic differentiation of HBMSCs through the stimulation of the expression of adipogenic differentiation switches such as peroxisome proliferator‐activated receptor γ and CCAAT/enhancer‐binding protein α. Furthermore, Si ions can enhance both angiogenesis and adipogenesis, and inhibit dedifferentiation of cocultured adipocytes by regulating the interactions between HBMSC‐derived adipocytes and human umbilical vein endothelial cells, in which the promotion of the expression of insulin‐like growth factor 1 and vascular endothelial growth factor plays vital roles. The in vivo studies further demonstrate that the designed composite hydrogel with the ability to release bioactive Si ions clearly stimulates neovascularization and adipose tissue regeneration. The study suggests that Si ions released from biomaterials are important chemical cues for adipogenic differentiation and biomaterials with the ability to release Si ions can be designed for adipose tissue engineering.
|
1 Introduction Adipose tissue engineering strategies have shown great potential in the treatment of soft tissue defects originated from severe burns, trauma, tumor resection, and congenital deformities. 1 Bone marrow mesenchymal stem cells (BMSCs) are a promising cell type for tissue engineering due to their strong proliferation ability and multiple differentiation potentials, which are able to differentiate into different cell linages such as osteoblasts, adipocytes, chondrocytes under appropriate stimuli conditions. 2 Pittenger et al. first reported that BMSCs can differentiate into adipogenic linage by treatment with l‐methyl‐3‐isobutylxanthine (IBMX), dexamethasone, insulin, and indomethacin, which is considered as the universal method for adipogenic induction. [[qv: 2b]] In recent years, some improved methods have been developed to accelerate adipogenic differentiation of MSCs. [[qv: 1c, 3]] However, many studies have shown that reconstruction of adipose tissue generated by adipogenic‐differentiated MSCs is limited and the amount of newly formed adipose tissue was insufficient to meet the requirement of adipose tissue reconstruction. [[qv: 1a, 4]] One problem may be the low adipogenic differentiation efficiency of MSCs, and gradual decrease of adipogenic differentiation capability of MSCs during extensive in vitro expansion and long‐term inductive differentiation by using the conventional chemical induction, which usually takes 2–3 weeks for MSCs to differentiate into adipocytes. [[qv: 3a, 5]] Therefore, how to increase the efficiency of adipogenic differentiation of MSCs is of the utmost importance. [[qv: 1c, 3b, d]] In addition to the insufficient adipogenic differentiation potential of MSCs, it has been widely recognized that the lack of sufficient vascularization is also one of the key problems resulting in failure of long‐term survival of regenerated adipose‐like tissue. [[qv: 1a, 6]] Native fatty tissue is highly vascularized and requires rich vascular supply to support its highly metabolic activity. [[qv: 1b, c]] Therefore, for the tissue engineering scaffolds whose size is larger than the physiological diffusion limit of oxygen and nutrients, the newly formed adipose tissue is at risk to suffer from reabsorption and necrosis due to the lack of vascular supply. [[qv: 1c, 7]] In order to solve the problem of vascularization, different methods have been investigated. A tissue engineering chamber (TEC) model has been reported, in which vascularized adipose tissue was generated by implanting a hollow chamber containing an adipose tissue flap subcutaneously in the groin of the rat. 8 Based on the animal study, a clinical study has been conducted, which confirmed the effectiveness of the application of the TEC in human, but limitation still exists, including complicated surgical procedure and low success rate. 9 Furthermore, considering the coexistence of adipocytes and endothelial cells in native fatty tissues, previous studies have demonstrated that the coculture of MSC‐derived adipocytes and endothelial cells might be an effective strategy to reconstruct engineered adipose tissue with better vascularization. [[qv: 1c]] However, the problem with the coculture strategy is that the differentiated adipocytes often undergo rapid dedifferentiation during in vitro coculture. [[qv: 1c, 10]] This is also an important issue that affect adipose tissue construction, and up to now, there seems no more effective strategies inhibiting the dedifferentiation of differentiated adipocytes. [[qv: 1b]] In recent years, more and more evidences show that bioactive ions play important roles in regulating cell behaviors including cell differentiation. 11 Some studies have shown that zinc promoted adipogenesis of rat adipocytes and stimulated the conversion of glucose to lipids in 3T3‐L1 fibroblasts and adipocytes in vitro and in vivo, 12 suggesting that some bioactive ions indeed can function as stimuli in promoting adipogenic differentiation. It is known that silicon (Si) as an essential trace element participates in bone and skin tissue regeneration. [[qv: 11a, 13]] Many studies have demonstrated that Si ions have the activity to enhance osteogenic differentiation of human BMSCs (HBMSCs). [[qv: 13a]] In addition, it has been reported that Si ions can stimulate angiogenesis of endothelial cells, and silicate‐based bioactive materials can enhance wound healing by regulating cell–cell interactions, stem cell migration and stimulating blood vessel formation. [[qv: 13b, d, 14]] These results suggest that Si ions may play a multifunctional role by enhancing stem cell differentiation, activating cell–cell interactions and stimulating angiogenesis. Based on the consideration of the multiple roles of Si ions in tissue regeneration, we hypothesize that 1) Si ions may enhance adipogenic differentiation of HBMSCs, thereby accelerating adipose tissue regeneration; 2) Si ions may enhance adipose tissue regeneration by stimulating angiogenesis; 3) Si ions may regulate the interactions between HBMSC‐derived adipocytes and endothelial cells and inhibit the dedifferentiation of cocultured adipocytes, and therefore promote both adipogenesis and angiogenesis. To confirm our hypotheses, in this study, we first investigated the bioactive effect of Si ions on adipogenic differentiation of HBMSCs, followed by exploring the effect of Si ions on the interactions of HBMSC‐derived adipocytes and human umbilical vein endothelial cells (HUVECs), and specifically synergistic effect of Si ions on enhancing angiogenesis of cocultured HUVECs and inhibiting the dedifferentiation of cocultured adipocytes. Finally, based on the bioactive concentration of Si ions identified, we reconstructed an engineered adipose tissue in a nude mice subcutaneous implant model by combining the cocultured HBMSC‐derived adipocytes and HUVECs in calcium silicate/alginate composite hydrogel with the ability to release bioactive Si ions. 2 Results 2. 1 Effects of Si Ions on Cell Proliferation and Morphology of HBMSCs and HUVECs The effects of ions released from calcium silicate (CS) on cell proliferation were evaluated and the results are shown in Figure 1 a, b. It is clear to see that at dilution ratios from 1/4 to 1/256 (Si‐ion concentration: 0. 5–29. 27 µg mL −1 ), CS extracts showed no cytotoxicity for both HBMSCs and HUVECs, but CS extracts without dilution (CS1) showed a certain degree of cytotoxicity for both type of the cells. More interestingly, CS extracts at a certain dilution range revealed a stimulatory effect on cell proliferation, whereas the active concentrations for the stimulation were different for different cell types (Figure 1 a, b). The CS extracts diluted from 1/2 to 1/128 (Si‐ion concentration: 0. 95–59. 57 µg mL −1 ) significantly stimulated HBMSC proliferation on day 7 with 13–28% increase as compared to the control group (Figure 1 a). For HUVECs, CS extracts at dilution ratios from 1/32 to 1/128 (Si‐ion concentration: 0. 95–3. 67 µg mL −1 ) stimulated cell proliferation by 7–13% as compared to the control group. At dilution ratios of 1/4 to 1/16 (Si‐ion concentration: 7. 36–29. 27 µg mL −1 ), CS extracts maintained the viability of HUVECs which indicated that these concentrations were not cytotoxic to HUVECs (Figure 1 b). In order to determine the bioactive concentrations of the Si ions for the stimulation of cell proliferation, CS extracts diluted with Dulbecco's modified Eagle medium (DMEM) by a series of gradient dilution at the ratios from 1 to 1/256 were measured and the results were listed in Table 1. It can be seen that Si concentrations in all the dilutions of CS extracts were much higher than those in the control medium (DMEM). In contrast, the Ca and P ions' concentrations of most dilutions were at similar level as that in the control medium except the original extract and 1/2 and 1/4 dilutions, which were lower than the control medium. These results suggest that at the dilution range from 1/8 to 1/256, Si ions play a key role in regulating cell proliferation, and the bioactive Si‐ion concentration for the stimulation of HBMSC proliferation is in the range between 0. 95 and 59. 57 µg mL −1, and that for the stimulation of HUVEC proliferation is in the range of 0. 95–3. 67 µg mL −1, which is much narrow than that for HBMSCs. The fluorescence images of actin cytoskeleton staining revealed the morphology of HBMSCs and HUVECs cultured with different concentrations of Si ions (7 and 14 µg mL −1 ) for 7 days (Figure 1 c). It can be seen that HBMSCs demonstrated an elongated, fibroblastic appearance, and HUVECs showed a typical cobblestone‐like appearance, indicating that Si ions did not affect the morphology of HBMSCs and HUVECs (Figure 1 c). Figure 1 Cell proliferation and morphology of HBMSCs and HUVECs cultured with different concentrations of Si ions. a, b) CCK‐8 assay indicated that Si ions in certain concentration range can stimulate the proliferation of HBMSCs and HUVECs ( n = 6 for both the groups; * p < 0. 05). c) Actin cytoskeleton staining assay showed that both concentrations of Si ions (7 and 14 µg mL −1 ) did not affect the morphology of HBMSCs and HUVECs. Scale bar, 75 µm for HBMSCs and 50 µm for HUVECs. The actin cytoskeletons were stained with FITC–Phalloidin (green), and nucleus was counterstained by using DAPI (blue). Table 1 Ion concentration of calcium silicate extracts diluted at ratios from 1 to 1/256 using DMEM medium Si [µg mL −1 ] Ca [µg mL −1 ] P [µg mL −1 ] DMEM 0. 02 ± 0. 00 58. 15 ± 0. 60 25. 86 ± 0. 32 CS1 120. 52 ± 0. 43 a) 5. 97 ± 0. 77 5. 96 ± 0. 26 CS1/2 59. 57 ± 0. 15 a) 33. 09 ± 0. 55 15. 44 ± 0. 32 CS1/4 29. 27 ± 0. 53 a) 45. 76 ± 0. 42 20. 75 ± 0. 02 CS1/8 14. 53 ± 0. 21 a) 56. 54 ± 0. 89 24. 54 ± 0. 30 CS1/16 7. 36 ± 0. 13 a) 57. 82 ± 0. 24 25. 20 ± 0. 21 CS1/32 3. 67 ± 0. 23 a) 58. 01 ± 0. 05 25. 15 ± 0. 25 CS1/64 1. 90 ± 0. 19 a) 58. 17 ± 0. 33 25. 85 ± 0. 10 CS1/128 0. 95 ± 0. 10 a) 58. 30 ± 0. 55 25. 97 ± 0. 15 CS1/256 0. 50 ± 0. 03 a) 58. 55 ± 0. 81 26. 11 ± 0. 07 a) Indicates that the Si‐ion concentrations in CS extracts were significantly higher than those in DMEM ( p < 0. 05). John Wiley & Sons, Ltd. 2. 2 Effects of Si Ions on Adipogenic Differentiation of HBMSCs The effect of Si ions on adipogenic differentiation of HBMSCs was analyzed and the results are shown in Figure 2. The Oil Red O staining showed that when HBMSCs were cultured with adipogenic differentiation medium in presence of Si ions for 21 days, the adipogenic differentiation of HBMSCs was clearly enhanced as compared with cells only cultured in adipogenic medium without Si ions (Figure 2 a, b). As shown in Figure 2 b, the lipid accumulation in cells treated with media containing 7 and 14 µg mL −1 Si ions (AM21‐Si7, AM21‐Si14) was increased by 21% and 41%, respectively, as compared to that in the cells cultured without Si ions (AM21‐Si0). The results suggest that Si ions have significant contribution to adipogenic differentiation, although their relative effect was smaller than the effect of adipogenic medium. Under normal growth condition without adipogenic differentiation inducer, Si ions could not stimulate lipid accumulation of HBMSCs (Figure 2 a, b GM group). To further confirm whether Si ions could enhance adipogenic gene expression during differentiation, two transcriptional factors peroxisome proliferator‐activated receptor γ (PPARγ) and CCAAT/enhancer‐binding protein α (C/EBPα), which play key roles in driving adipogenic differentiation, 15 and several markers of terminal adipogenic differentiation including fatty acid binding protein 4 (FABP4), leptin, and adiponectin were detected by quantitative real‐time polymerase chain reaction (qRT‐PCR) (Figure 2 c). Similar to the result of Oil Red O staining, Si ions were observed to significantly upregulate the gene expression of PPARγ, C/EBPα, FABP4, leptin, and adiponectin in adipogenic medium, which indicated that Si ions do promote adipogenic differentiation of HBMSCs (Figure 2 c). In contrast, in normal growth medium, no upregulation of adipogenic gene expression was observed both with and without Si ions, suggesting that Si ions are not an inducer, and rather an enhancer of the adipogenic differentiation of HBMSCs. Figure 2 Oil Red O staining and adipogenic gene expression of HBMSCs cultured in different mediums with or without Si ions. a) Oil Red O staining of cells showed that Si ions stimulated the increase of lipid accumulation. Lipid droplets were stained in red. Scale bar, 20 µm. b) Quantitative analysis of Oil Red O staining showed that Si ions significantly stimulated lipid accumulation ( n = 3 for both the groups; * p < 0. 05). c) Quantitative PCR analysis showed that Si ions enhanced the adipogenic marker gene of PPARγ, C/EBPα, FABP4, leptin, and adiponectin expression of HBMSCs ( n = 3 for both the groups; * p < 0. 05). GM‐Si0, GM‐Si7, GM‐Si14: HBMSCs cultured with growth medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ) without differentiation inducer for 21 days, respectively; AM21‐Si0, AM21‐Si7, AM21‐Si14: HBMSCs cultured with adipogenic differentiation medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ), respectively, for 21 days. In order to further explore the effects of Si ion as a differentiation enhancer on the stimulation of adipogenesis, HBMSCs were first cultured with adipogenic differentiation medium alone for 4, 8, and 12 days, respectively, and then further cultured up to 21 days in normal growth medium containing different concentrations of Si ions (7 and 14 µg mL −1 ). The results showed that when cells have been induced to adipogenic differentiation for certain time, then even without differentiation inducer, Si ions could also enhance lipid accumulation and expression of adipogenic genes of HBMSCs ( Figure 3 ). This result indicates that once the cells are pushed into the direction of adipogenic differentiation, Si ions can push the cells to continue adipogenic differentiation even without the differentiation inducer. More interestingly, after treating cells with adipogenic differentiation medium (AM) alone for 12 days, then replacing AM with growth medium containing Si ions and culturing for up to 21 days, the degree of adipogenic differentiation enhanced by Si ions could reach or even exceed the differentiation level of the cells cultured in adipogenic medium without Si ions for 21 days (Figure 3 c, AM12‐Si7, AM12‐Si14, AM21‐Si0). Figure 3 The effects of Si ion as a differentiation enhancer on the stimulation of adipogenic differentiation. a) Oil Red O staining of cells showed that Si ions stimulated the increase of lipid accumulation. Lipid droplets were stained in red. Scale bar, 20 µm. b) Quantitative analysis of Oil Red O staining showed that Si ions significantly stimulated lipid accumulation ( n = 3 for both the groups; * p < 0. 05). c) Quantitative PCR analysis showed that Si ions enhanced the adipogenic marker gene of PPARγ, C/EBPα, FABP4, leptin, and adiponectin expression of HBMSCs ( n = 3 for both the groups; * p < 0. 05). AM4‐Si7, AM4‐Si14: HBMSCs cultured with adipogenic differentiation medium alone for 4 days, and then further cultured up to 21 days in growth medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ), respectively; AM8‐Si0, AM8‐Si7, AM8‐Si14: HBMSCs cultured with adipogenic differentiation medium alone for 8 days, and then further cultured up to 21 days in growth medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ), respectively; AM12‐Si0, AM12‐Si7, AM12‐Si14: HBMSCs cultured with adipogenic differentiation medium alone for 12 days, and then further cultured up to 21 days in growth medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ), respectively; AM21‐Si0: HBMSCs cultured in adipogenic differentiation medium without Si ions for 21 days. 2. 3 Effects of Si on Angiogenesis of Cocultured HBMSC‐Derived Adipocytes and HUVECs To explore the effects of Si ions on cell–cell interactions and angiogenesis, an in vitro coculture model with HBMSC‐derived adipocytes and HUVECs was applied. von Willebrand factor (vWF) immunostaining showed clear capillary‐like networks formation in cocultured cells, while there was no clear network formed in monocultured adipocytes or HUVECs. Furthermore, the addition of Si ions in the coculture medium remarkably stimulated the capillary‐like network formation, which indicated that Si ions enhanced angiogenesis of the cocultured cells ( Figure 4 a). Enzyme‐linked immunosorbent assay (ELISA) analysis also showed that although Si ions also promoted vascular endothelial growth factor (VEGF) secretion of monocultured adipocytes and HUVECs as compared to cells cultured without Si ions, a significantly higher increase of the VEGF secretion was observed in cocultured cells in the presence of Si ions as compared to monocultured cells (Figure 4 b). This result indicates a possible activation of the interactions of the cocultured adipocytes and HUVECs by Si ions which contribute to angiogenesis. Figure 4 Immunofluorescence staining of vWF and VEGF secretion of monocultured cells and cocultured cells treated with or without Si ions. a) vWF staining showed that clear capillary‐like networks were formed in the cocultured cells but not in the monocultured HUVECs and adipocytes. In addition, Si ions stimulated the network formation. vWF was stained in green and nuclei in blue with DAPI. Scale bar, 150 µm. b) Quantitative ELISA analysis showed that Si ions enhanced the VEGF secretion of monocultured cells and cocultured cells in culture medium ( n = 3 for both the groups; * p < 0. 05). ECs‐Si0, ECs‐Si7, ECs‐Si14: HUVECs (ECs) cultured with growth medium containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 3 days; ACs‐Si0, ACs‐Si7, ACs‐Si14: HBMSCs cultured in adipogenic differentiation medium with different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 15 days for the development of the adipocytes' (ACs) phenotype; CO‐Si0, CO‐Si7, CO‐Si14: HBMSCs first cultured in adipogenic differentiation medium with different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 12 days for adipogenic differentiation, and then cocultured with HUVECs containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 3 days. In order to elucidate possible mechanism of Si ion activated angiogenesis and cell–cell interactions, the two type of cells were separated after coculturing for 3 days and the expression of angiogenic genes such as VEGF and its receptor VEGFR2, insulin‐like growth factor 1 (IGF1) and its receptor IGF1R from the separated cells were measured ( Figure 5 ). Similar as observed in VEGF protein secretion, compared with monocultured HUVECs and adipocytes, co‐HUVECs and co‐adipocytes showed clear upregulation in VEGF expression, respectively (Figure 5 a). More interestingly, the VEGF expression in co‐adipocytes was much higher than that in co‐HUVECs in Si containing medium (Figure 5 a). In contrast, Si ions only stimulated VEGFR2 expression in co‐HUVECs, and did not stimulate the VEGFR2 expression in co‐adipocytes (Figure 5 b). Figure 5 Angiogenic and adipogenic gene expression of monocultured cells and cocultured cells treated with or without Si ions. HBMSCs were first cultured in adipogenic differentiation medium for 12 days for the development of the adipocyte phenotype, and then cocultured with HUVECs containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 3 days. a–d) Quantitative PCR analysis showed that Si ions stimulated the gene expression of VEGF, VEGFR2, IGF1, and IGF1R of cocultured HUVECs and cocultured adipocytes ( n = 3 for both the groups; * p < 0. 05). ECs‐Si0: monocultured HUVECs without Si ions; CO‐ECs‐Si0, CO‐ECs‐Si7, CO‐ECs‐Si14: cocultured HUVECs incubated with different concentrations of Si ions (0, 7, and 14 µg mL −1 ); ACs‐Si0: monocultured adipocytes without Si ions; CO‐ACs‐Si0, CO‐ACs‐Si7, CO‐ACs‐Si14: cocultured adipocytes incubated with different concentrations of Si ions (0, 7, and 14 µg mL −1 ). IGF1, an important growth factor, acts through its receptor IGF1R on the cytomembrane to stimulate endothelial cells migration and regulate angiogenesis. 16 Therefore, we also analyzed the messenger RNA (mRNA) levels of IGF1 and IGF1R genes (Figure 5 c, d). Similar to the result of VEGF and VEGFR2 expression, Si ions strongly upregulated the expression of IGF1 in co‐adipocytes (Figure 5 c), and IGF1R expression in co‐HUVECs (Figure 5 d). 2. 4 Effects of Si on the Inhibition of Dedifferentiation of Cocultured Adipocytes To explore the effects of Si ions on the inhibition of dedifferentiation of cocultured adipocytes, we further compared the degree of adipogenic differentiation between the cocultured cells with or without Si ions. Oil Red O staining and ELISA analysis of adiponectin secretion of mono‐ and cocultured cells were shown in Figure 6. It is clear to see that cocultured adipocytes showed reduced Oil Red O staining and adiponectin secretion as compared to monocultured adipocytes, indicating a drastic dedifferentiation of cocultured adipocytes (Figure 6, CO‐Si0). Interestingly, we found that Si ions significantly stimulated lipid accumulation and adiponectin secretion of cocultured adipocytes, and the highest effective Si‐ion concentration for stimulation was 14 µg mL −1, which resulted in a remarkable increase of lipid accumulation and adiponectin secretion even higher than that of the monocultured adipocytes in adipogenic medium without Si ions (Figure 6, CO‐Si14). In addition, IGF1, known also as an important adipogenic factor[[qv: 3c, 17]] besides the role in angiogenesis, was found clearly downregulated in cocultured adipocytes (ACs) as compared to monocultured adipocytes (Figure 5 c, CO‐ACs‐Si0), which also indicates the dedifferentiation of cocultured adipocytes. However, Si ions at the concentration of 14 µg mL −1 also significantly stimulated IGF1 expression in cocultured adipocytes as compared to the cocultured cells without Si, which was even higher than that of the monocultured adipocytes in adipogenic medium without Si ions (Figure 5 c, CO‐ACs‐Si14). These results suggest that Si ions are able to inhibit the dedifferentiation of cocultured adipocytes by stimulating lipid accumulation and adiponectin secretion, and promoting the expression of the adipogenic factor IGF1. Figure 6 Oil Red O staining and adiponectin secretion of monocultured cells and cocultured cells treated with or without Si ions. a) Oil Red O staining of cells showed that Si ions stimulated the increase of lipid accumulation in the coculture system and inhibited the dedifferentiation of cocultured adipocytes. Lipid droplets were stained in red. Scale bar, 20 µm. b) Quantitative analysis of Oil Red O staining ( n = 3 for both the groups; * p < 0. 05). c) Quantitative ELISA analysis showed that Si ions enhanced the adiponectin secretion of cocultured cells in culture medium ( n = 3 for both the groups; * p < 0. 05). ACs‐Si0: HBMSCs cultured in adipogenic differentiation medium without Si ions for 12 days for the development of the adipocytes' (ACs) phenotype; CO‐Si0, CO‐Si7, CO‐Si14: HBMSCs first cultured in adipogenic differentiation medium with different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 12 days for adipogenic differentiation, and then cocultured with HUVECs containing different concentrations of Si ions (0, 7, and 14 µg mL −1 ) for 3 days. 2. 5 Effects of Si Ions on In Vivo Adipose Tissue Regeneration In order to further investigate the stimulatory effects of Si ions on vascularized adipose tissue engineering in vivo, we reconstructed an engineered adipose tissue in a nude mice subcutaneous implant model by combining the cocultured HBMSC‐derived adipocytes and HUVECs in calcium silicate/alginate composite hydrogel with the ability to release bioactive Si ions. Figure 7 a showed the appearance of engineered adipose tissue after 8 weeks transplantation. It is clear to see that the implanted hydrogel with monocultured adipocytes or cocultured cells all showed obvious formation of adipose‐like tissues as compared with the pure hydrogel implantation, and Si ion–releasing hydrogel with cocultured cells showed a larger and more complete adipose tissue morphology than that with monocultured adipocytes (Figure 7 a). Oil red O staining further confirmed the optical observation that Si ions remarkably enhanced the formation of adipose‐like tissue in hydrogels with both mono‐ and cocultured adipocytes, and coculture group showed higher adipose tissue formation than monoculture group (Figure 7 b, c). Figure 7 Tissue appearance and Oil Red O staining of engineered adipose tissue in a nude mice subcutaneous implant model by combining the monocultured cells or cocultured cells in calcium silicate/alginate composite hydrogel with the ability to release bioactive Si ions. a) Appearance of engineered adipose tissue showed that Si ions stimulated adipose tissue growth and vascular network (black arrows) formation. High magnification corresponded to boxed area (red) in the low magnification images. b) Oil Red O staining of implants showed that Si ions promoted adipose tissue regeneration in vivo. Lipid droplets were stained in red and nucleus were counterstained in blue with hematoxylin. Scale bar, 100 µm. c) Quantitative analysis of Oil Red O staining of the implants indicated that Si ions stimulated adipose tissue formation ( n = 5 for both the groups; * p < 0. 05). SA: hydrogels without cells; Si‐SA: Si ion–released hydrogels without cells; ACs‐SA: hydrogels with monocultured adipocytes; ACs‐Si‐SA: Si ion–released hydrogels with monocultured adipocytes; CO‐SA: hydrogels with cocultured adipocytes and HUVECs; CO‐Si‐SA: Si ion–released hydrogels with cocultured adipocytes and HUVECs. 2. 6 Effects of Si Ions on In Vivo Neovascularization In addition to the adipose like tissue appearance, the Si ion–releasing sodium alginate hydrogels with cocultured cells (CO‐Si‐SA) showed clear formation of blood vessel networks as demonstrated by massive blood vessels ingrowth and blood vessel branches with several millimeters in length, significantly different as compared with other groups (Figure 7 a). In order to further confirm enhanced vascularization, hematoxylin and eosin (H&E) staining and cluster of differentiation 31 (CD31) immunohistochemical staining were performed, and the results are shown in Figure 8. H&E staining showed that the Si ion–releasing hydrogels with cocultured cells (Figure 8 a, CO‐Si‐SA) strongly stimulated the formation of big blood vessel, as demonstrated by red blood cells distribution in lumen, while no obvious angiogenesis in groups of pure hydrogels without cells were observed (Figure 8 a, SA, Si‐SA). CD31 staining revealed that the CO‐Si‐SA group showed the similar vascular structure to native fatty tissue characterized by numerous microvessels connected to adipocytes (Figure 8 b). 18 Quantitation of blood vessel density and vessel diameter distribution revealed that CO‐Si‐SA group showed larger number and diameter of blood vessels than CO‐SA and ACs‐Si‐SA (Figure 8 c, d). For hydrogels with cocultured cells, the majority of blood vessels formed in CO‐Si‐SA group were bigger than 25 µm in diameters, while the diameters of the vessels in CO‐SA group was smaller than 25 µm (Figure 8 d). For hydrogels with monocultured adipocytes, it is clear to see that the diameter distribution of blood vessels in AC‐Si‐SA group was between 10 and 20 µm, while that in AC‐SA group was smaller than 10 µm (Figure 8 d). Figure 8 H&E staining and CD31 immunohistochemical staining of engineered adipose tissue implants. a) H&E staining showed that Si ions stimulated neovascularization in vivo. Blood vessels (black arrows) was stained in red and nuclei in blue. Scale bar, 100 µm. b) CD31 immunohistochemical staining showed that Si ions stimulated numerous microvessels' formation. Blood vessels (red arrows) was stained in brown and nuclei in blue. High magnification corresponded to boxed area (red) in the low magnification images. Scale bar, 100 µm for low magnification groups, 20 µm for the high magnification groups. c) Quantitative analysis of blood vessel density of the implants ( n = 5 for both the groups; * p < 0. 05). d) Quantitative analysis of the vessel diameter indicated that Si ions stimulated the increase of vessel diameters ( n = 112 for both the groups). SA: hydrogels without cells; Si‐SA: Si ion–released hydrogels without cells; ACs‐SA: hydrogels with monocultured adipocytes; ACs‐Si‐SA: Si ion–released hydrogels with monocultured adipocytes; CO‐SA: hydrogels with cocultured adipocytes and HUVECs; CO‐Si‐SA: Si ion–released hydrogels with cocultured adipocytes and HUVECs. 3 Discussion The efficiency of adipogenic differentiation of BMSCs and sufficient vascularization are key issues for successful adipose tissue engineering. [[qv: 1c, 3a, 5, 6]] Based on the multifunctional role of Si ions on the regulation of stem cell differentiation, activation of cell–cell interactions, and stimulation of blood vessel formation, [[qv: 13a, b, 14]] we hypothesized that Si ions may participate in the regulation of adipogenic differentiation of BMSCs and stimulation of vascularization during adipose tissue regeneration, in particular, Si ions may affect the interactions between BMSC‐derived adipocytes and endothelial cells which contribute to angiogenesis and adipogenesis. Our results confirmed our hypothesis that Si ions indeed enhance adipogenic differentiation of HBMSCs, and significantly activated angiogenesis and adipogenesis in adipose tissue regeneration by regulating the interactions between HBMSC‐derived adipocytes and HUVECs. In vivo studies further confirmed that Si ion releasing hydrogels with the cocultured HBMSC‐derived adipocytes and HUVECs stimulate vascularized adipose tissue regeneration. The efficient stimulation of adipogenic differentiation of HBMSCs is one of the key steps in adipose tissue engineering. Although the conventional chemical inducing based on Pittenger's method for the adipogenic differentiation of HBMSCs has been proved to be effective, the differentiation degree and efficiency is not high enough. [[qv: 2b, 3a, c, 5]] One of the core scientific hypothesis of the current study is that Si ions can regulate adipogenic differentiation of HBMSCs. Our results indicated that under normal growth condition without adipogenic differentiation inducer, Si ions could not stimulate adipogenic differentiation of HBMSCs. However, when HBMSCs were cultured with adipogenic differentiation medium in presence of Si ions for 21 days, the adipogenic differentiation of HBMSCs was clearly enhanced as compared with cells only cultured in adipogenic medium without Si ions. The lipid accumulation and expression of adipogenic marker genes such as PPARγ, C/EBPα, FABP4, leptin, and adiponection in cells cultured with Si ions were significantly higher than that in cells cultured without Si ions. This result suggests that Si ions are not an inducer, and rather an enhancer of adipogenic differentiation of HBMSCs. More interesting is that when HBMSCs have been induced with AM alone for 12 days, then replacing AM with normal growth medium containing Si ions and culturing further for up to 21 day, the adipogenic differentiation was further enhanced even without adipogenic medium, and the enhancement was even higher than that of the cells cultured in adipogenic medium for the same time period. This means that, once the cells are already on the way to go adipogenic differentiation, Si ions can promote continuous differentiation of HBMSCs along the adipogenic differentiation direction without adipogenic inducer and with higher efficiency than adipogenic medium. It is known that many genes are expressed in along the signaling pathway of adipogenic differentiation. It has been reported that PPARγ and C/EBPα are the key transcription factors that act as molecular switches to drive adipogenic differentiation. These molecular switches control the fate of MSCs to differentiate into adipocytes. 15 Once the differentiation switches were activated and opened, MSCs then initiated adipogenesis by activating expression of adipogenic genes such as FABP4, leptin, and adiponection. 19 Many studies confirmed that most growth factors or stimuli that upregulated adipogenesis ultimately exerted their effects through regulation of PPARγ and C/EBPα expression. 20 In the present study, we found that Si ions strongly enhanced expression of PPARγ and C/EBPα in adipogenic‐induced cells. This result suggests that although Si ions are not able to induce the initiation of adipogenic differentiation, but once the differentiation has been initiated, Si ions can significantly enhance the molecular switches such as PPARγ and C/EBPα to drive the adipogenic differentiation. Therefore, the role of Si ions in the enhancement of adipogenic differentiation is possibly due to the enhanced activation of upstream molecules of adipogenic signaling pathway. In addition to the insufficient adipogenic differentiation efficiency of HBMSCs, the lack of effective vascularization is another problem to overcome in adipose tissue engineering. [[qv: 1a, 6]] Previous studies have demonstrated that the use of MSC‐derived adipocytes and endothelial cell coculture model might be an effective strategy to reconstruct engineered adipose tissue with better vascularization. [[qv: 1c]] Our previous study has confirmed that Si ions significantly stimulated angiogenesis of endothelial cells by regulating cell–cell interactions in BMSCs/endothelial cells and fibroblast/endothelial cells' coculture system. [[qv: 13a, 14]] Therefore, one of our hypothesis is that Si ions may stimulate angiogenesis by regulating the interactions between HBMSC‐derived adipocytes and endothelial cells. Our results indeed demonstrated that Si ions significantly enhanced capillary‐like network formation of cocultured HUVECs by activating the cocultured adipocytes to express high level of angiogenic growth factors VEGF, which then upregulated the expression of the VEGF receptors (VEGFR2) on the cytomembrane of cocultured HUVECs. Previous studies have demonstrated that VEGF secreted by adipocytes plays an important role in enhancing angiogenesis. 21 Lai et al. found that the blockade of VEGFR2 abolished neovascularization of HUVECs in a coculture system. 22 In our experiments, we found that in the presence of Si ions, not only the expression of angiogenic growth factors VEGF in cocultured adipocytes was upregulated, but also the expression of VEGF receptors such as VEGFR2 in cocultured HUVECs was significantly enhanced, while cocultured adipocytes almost did not express VEGFR2. These results indicate that the stimulation of angiogenesis in HBMSC‐derived adipocyte/HUVEC coculture system by Si ions is mainly through a paracrine pathway. In addition to VEGF, another important growth factor IGF1 has been found involved in both angiogenesis and adipogenesis by interacting with its receptor IGF1R, followed by the activation of the IGF1/IGF1R signaling pathway. [[qv: 16, 17, 20b]] Similar to VEGF expression, our results showed that Si ions strongly upregulated the expression of IGF1 in cocultured adipocytes in the coculture system, while stimulated the expression of IGF1 in cocultured HUVECs in a lower level than cocultured adipocytes. In contrast, the IGF1 receptor (IGF1R) was not only significantly upregulated in cocultured HUVECs, but also upregulated in cocultured adipocytes by Si ions at certain degree, indicating that IGF1‐related signaling was activated by Si ions not only through paracrine, but also through autocrine pathways, which contribute to the enhancement of both angiogenesis and adipogenesis in the coculture system. In recent years, growth factors like IGF1 and VEGF have been widely used for tissue engineering applications including adipose tissue regeneration for enhanced angiogenesis and adipogenesis. [[qv: 3c, 23]] However, applications of additive growth factors are still limited due to the challenge in controlled delivery and activity maintenance. Our results showed that Si ions significantly enhanced the expression of IGF1 and VEGF from cells, which indicates a new possibility to induce and utilize the intrinsic growth factors of the seeding cells for enhanced tissue reconstruction, and this approach may significantly reduce the complicity and cost of tissue engineering applications. However, although the interactions of HBMSC‐derived adipocytes with endothelial cells can enhance angiogenesis, a rapid dedifferentiation of cocultured adipocytes has been observed, which may negatively affect adipose tissue reconstruction. 10 In the present study, we also observed the phenomenon in the coculture experiments, in which cocultured adipocytes reduced lipid accumulation and adiponectin secretion after culturing with HUVECs. Interestingly, when Si ions were added in the coculture medium, the decrease of the lipid accumulation and adiponectin secretion was clearly reduced, and at the Si‐ion concentration of 14 µg mL −1, lipid accumulation and adiponectin secretion were even significantly higher than that in monocultured adipocytes. These results suggest that Si ions can inhibit the dedifferentiation of cocultured adipocytes. As an important adipogenic factor for accelerating adipose tissue regeneration, IGF1 has been found to play an important role in the regulation of adipogenesis through its interaction with its receptor IGF1R. 17 In the present study, we found that without addition of Si ions, the IGF1 and IGF1R expression in cocultured adipocytes was clearly reduced as compared with that in monocultured adipocytes which further confirmed the dedifferentiation effects. Interestingly, when Si ions were added in coculture medium, the IGF1 and IGF1R expression in cocultured adipocytes was recovered, and was even higher than that in monocultured adipocytes at the Si‐ion concentration of 14 µg mL −1. These results suggest that Si ions can inhibit the dedifferentiation of cocultured adipocytes through stimulating the expression of the adipogenic factor IGF1 and its receptor IGF1R. Taken together, the activation of the interactions of the HBMSC‐derived adipocytes and HUVECs by Si ions not only enhanced angiogenesis through paracrine mechanisms, but also inhibited dedifferentiation of cocultured adipocytes, in which the activation of IGF1/IGF1R signaling pathway may play a vital role. Considering the in vitro findings of the stimulatory effect of Si ions on adipogenesis and angiogenesis, and to further prove the applicability of the bioactive Si ions for adipose tissue engineering, we prepared a composite hydrogel with the ability to release Si ions in the bioactive concentration range. When the hydrogel was loaded with cocultured HBMSC‐derived adipocytes and HUVECs and implanted in vivo, adipose‐like tissue was formed with significantly higher vascularization than the control. These results indicated that Si ions released from biomaterials strongly stimulated vascularized adipose tissue regeneration by enhancing the interactions between cocultured adipocytes and HUVECs in vivo. Histological analysis of H&E staining and CD31 immunohistochemical staining revealed that the newly formed adipose tissue stimulated by Si ions had a structure similar to the natural fatty tissue, as demonstrated by the connection of every adipocyte to at least one capillary, which suggested possible functional adipose tissue formation. 18 In addition, Si ions significantly promoted the formation of big blood vessels with length of several millimeters, and the self‐assembly into a network structure, which indicated the enhancement of blood vessel formation in the tissue constructs. The sufficient vascularization in turn ensured the supply with nutrients for the growth of newly formed adipose tissue. Based on the in vitro and in vivo results, the possible mechanisms of the bioactive Si ions on adipose tissue reconstruction may be described as the following, as shown in Figure 9. Si ions convert the chemical signal into biological signal by stimulating the interactions between HBMSC‐derived adipocytes and HUVECs, simultaneously activate adipogenesis and angiogenesis through the intracellular signal transmission, and finally enhance vascularized adipose tissue regeneration. On the one hand, Si ions strongly stimulate the expression of IGF1 mainly from cocultured adipocytes, which subsequently acted on its receptor IGF1R on the cytomembrane of co‐adipocytes, resulting in the activation of adipogenesis and inhibition of dedifferentiation, in which autocrine effects played a key role. On the other hand, Si ions remarkably enhance the expression of VEGF and IGF1 mainly from cocultured adipocytes, which subsequently act on their receptors VEGFR2 and IGF1R, respectively, on the cytomembrane of co‐HUVECs, resulting in the initiation of angiogenesis, in which paracrine effects played a key role. In summary, Si ion–enhanced adipogenesis and angiogenesis for vascularized adipose tissue regeneration may function through three possible pathways. First, Si ions enhanced adipogenesis through the enhancement of driving force of differentiation, as demonstrated by the increase of the expression of differentiation switches PPARγ and C/EBPα. Second, Si ions exerted its effects on angiogenesis of cocultured HUVECs and on the inhibition of dedifferentiation of cocultured adipocytes via stimulating the interactions of cocultured adipocytes and HUVECs, in which the promotion of the expression of growth factors IGF1 and VEGF played vital roles. Third, Si ions significantly enhanced blood vessel formation which contributed to adipogenesis and subsequently adipose tissue regeneration. Figure 9 Illustration of the mechanisms of Si ions enhancing angiogenesis and adipogenesis for better vascularized adipose tissue regeneration. 4 Conclusion In this study, the bioactive effects of Si ions on adipogenic differentiation of HBMSCs were first reported. Results suggest that Si ion is an effective adipogenic differentiation enhancer of HBMSCs. Once the cells were pushed into the direction of adipogenic differentiation by adipogenic inducing medium, Si ions further pushed the cells to continue adipogenic differentiation even without the differentiation inducer, and this enhancement of adipogenic differentiation was even higher than the adipogenic medium. The possible mechanism of Si ions on the enhancement of adipogenic differentiation might be due to the activated expression of adipogenic differentiation switches such as PPARγ and C/EBPα, which were responsible for the initiation of adipogenic differentiation and further activated the expression of adipogenic genes such as FABP4, leptin, and adiponection. Furthermore, Si ions not only enhanced angiogenesis of cocultured HUVECs, but also inhibited dedifferentiation of cocultured adipocytes via affecting the interactions between HBMSC‐derived adipocytes and HUVECs. For the activation of angiogenesis, Si ions mainly stimulated the VEGF/VEGFR2 and IGF1/IGF1R signaling pathways, in which paracrine effects played a key role. For the activation of adipogenesis and inhibition of dedifferentiation, Si ions mainly stimulated the IGF1/IGF1R signaling pathway, in which autocrine effects played a key role. Finally, based on the in vitro findings of the stimulatory effect of Si ions on adipogenesis and angiogenesis, a composite hydrogel with the ability to release bioactive Si ions was designed and in vivo study further demonstrated that the bioactive hydrogel loaded with HBMSC‐derived adipocytes and HUVECs evidently stimulated the formation of vascularized adipose tissue. 5 Experimental Section Si Ion–Containing Extract Preparation and Ion Concentration Determination : CS powders were prepared by a chemical coprecipitation method. 24 Si ion–containing media was prepared by using CS powders based on the fact that CS releases ions gradually when soaked in cell culture medium. CS extracts were prepared according to the protocol reported in the literature, 25 which was adapted from the standard procedure in ISO10993‐1. Briefly, CS powder was added into serum‐free DMEM (Gibco, USA) medium at a solid/liquid ratio of 200 mg mL −1 and incubated in a humidified incubator containing 5% CO 2 at 37 °C for 24 h. The supernatant was collected, centrifuged (4000 rpm, 10 min), and sterilized through a filter membrane (Millipore, USA, 0. 22 µm). The collected CS extracts were defined as stock solutions. Then, CS extracts were diluted with DMEM (Gibco, USA) by a series of gradient dilution at the ratios from 1 to 1/256. The Si, Ca, and P ion concentrations were measured by inductively coupled plasma atomic emission spectroscopy (ICP‐AES, Vista AX, Varian, Palo Alto, USA). Cell Culture : HBMSCs were obtained from Cyagen Biosciences Inc. (China) and cultured in growth medium consisting of low glucose DMEM (Gibco, USA), 10% fetal bovine serum (FBS) (Gibco, USA), 1% penicillin–streptomycin (Gibco, USA). For adipogenic differentiation, HBMSCs were grown to postconfluence for 2 days, and then the growth medium was replaced with adipogenic differentiation A medium consisting of high glucose DMEM (Gibco, USA), 10% FBS, 1% penicillin–streptomycin, 0. 5 × 10 −3 m IBMX (Sigma‐Aldrich, USA), 10 µg mL −1 insulin (Sigma‐Aldrich, USA), 100 × 10 −6 m indomethacin (Sigma‐Aldrich, USA), 1 × 10 −6 m dexamethasone (Sigma‐Aldrich, USA) for 3 days with or without Si ions. The Si‐ion concentration used for this experiment was determined first in a preliminary screening experiment with CS extracts in the concentration range from 1. 75 to 14 µg mL −1, and the Si concentrations of 7 and 14 µg mL −1 were selected due to the observed bioactive effect of these two concentrations on stimulating PPARγ expression (data not shown). Thereafter, the A medium was replaced with adipogenic differentiation B medium consisting of high glucose DMEM, 10% FBS, 1% penicillin–streptomycin, 10 µg mL −1 insulin for 1 day with or without Si ions. A + B represented a cycle of differentiation. 3 cycles later, cells were treated with B medium to 21 days for full development of the adipocyte phenotype. HUVECs were isolated as the previously described method. 26 For cocultures, HBMSCs were first cultured in adipogenic differentiation medium containing Si ions for 12 days for adipogenic differentiation, and then cocultured with HUVECs containing Si ions for 3 days. HBMSCs and HUVECs used in the study were all at passage 3. Cell Proliferation Assay : A Cell Counting Kit (CCK)‐8 assay was used to evaluate the effects of ionic products from CS on cell proliferation. HBMSCs and HUVECs were seeded in 96‐well plates at 1 × 10 3 cells per well for 24 h. Then, cells were treated with media containing CS extracts for different time periods. The cell viability was measured after culturing for 3 and 7 days, respectively, using a CCK‐8 kit (Dojindo Laboratories, Japan) according to the manufacturer's instructions. Briefly, at the end of each culture time point, cell culture medium was removed, and the cells were incubated with fresh medium containing CCK‐8 reagent (10:1) for 1 h at 37 °C in an incubator. The absorbance was measured spectrophotometrically using an enzyme‐linked immunoadsorbent assay microplate reader (Epoch, BIO‐TEK, USA) at a wavelength of 450 nm. Actin Cytoskeleton Staining Assay : After being cultured with different concentrations of Si ions (7 and 14 µg mL −1 ) for 7 days, cells were fixed in 4% paraformaldehyde for 30 min followed by staining with fluorescein isothiocyanate (FITC)–phalloidin (Sigma‐Aldrich, USA) for 30 min. Then, nucleus was counterstained by using 4, 6‐diamidino‐2‐phenylindol (DAPI, Sigma‐Aldrich, USA) for 5 min. The morphology of the cells was visualized using a confocal laser scanning microscopy (CLSM, TCS SP8, Leica, Germany). Oil Red O Staining : Differentiated adipocytes were fixed with 4% paraformaldehyde for 30 min at room temperature. Oil Red O working solution was prepared by diluting 3 mL of 0. 5% Oil Red O stock solution (Sigma‐Aldrich, USA) in 2 mL of distilled water and filtered through a 0. 22 µm pore‐size filter (Millipore, USA) before use. Then, cells were stained with Oil Red O working solution for 1 h and washed with distilled water to remove excess dye for photographing. For quantitative analysis, cells were destained in 100% isopropanol for 15 min and absorbance was measured at 492 nm. qRT‐PCR Analysis : Total RNA was isolated using TRIzol reagent (Invitrogen, USA) according to the manufacturer's protocol. mRNA was reverse‐transcribed into complementary DNA (cDNA) using a PrimeScript RT Master Mix kit (Takara, Japan). Quantitative PCR analysis was performed by a StepOnePlus Real‐Time PCR System (Applied Biosystems, USA) using a SYBR Premix EX Taq kit (TIi RNaseH Plus, Takara, Japan) according to the instructions. Primer sequences used for qRT‐PCR analysis are shown in Table 2. Table 2 Primers used for qRT‐PCR Target gene Forward primer sequence (5′–3′) Reverse primer sequence (5′–3′) Glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) ACGGATTTGGTCGTATTGGGCG CTCCTGGAAGATGGTGATGG PPARγ GATACACTGTCTGCAAACATATCACAA CCACGGAGCTGATCCCAA C/EBPα AAGAAGTCGGTGGACAAGAACAG TGCGCACCGCGATGT FABP4 GCTTTGCCACCAGGAAAGTG ATGGACGCATTCCACCACCA Leptin TCACACACGCAGTCAGTCTC GAGGTTCTCCAGGTCGTTGG Adiponectin CCTAAGGGAGACATCGGTGA CAATCCCACACTGAATGCTG IGF1 ATGGGAAAAATCAGCAGTCTTC CTACATCCTGTAGTTCTTGTTT IGF1R ACCCGGAGTACTTCAGCGC CACAGAAGCTTCGTTGAGAA VEGF TATGCGGATCAAACCTCACCA CACAGGGATTTTTCTTGTCTTGCT VEGFR2 CCCAGGCTCAGCATACAA AAAGAC CCAGTACAAGTCCCTCTGTCCC John Wiley & Sons, Ltd. vWF Staining : After coculturing for 3 days, cells were fixed for 30 min with 4% paraformaldehyde, and permeabilized with cold methanol for 5 min at room temperature, followed by the blockage with 10% goat serum for 1 h at 37 °C. Then, cells were incubated overnight with rabbit anti‐vWF antibody (Abcam, UK, 1:200) at 4 °C and then with Alexa Fluor 488 goat anti‐rabbit secondary antibody (Invitrogen, USA, 1:500) for 1 h at 37 °C. Nuclei were stained with DAPI (Sigma‐Aldrich, USA) for 5 min at room temperature. Images were then taken with confocal laser scanning microscopy (CLSM, TCS SP8, Leica, Germany). ELISA Analysis : After coculture for 3 days, the culture medium was collected and centrifuged for 5 min at 10 000 rpm at 4 °C. The content of VEGF and adiponectin in the supernatant was quantified by human VEGF and adiponectin ELISA kit (Absci, USA) according to the manufacturer's protocol. Separation of HBMSC‐Derived Adipocytes and HUVECs Using Magnetic Beads : To investigate the interactions between HBMSC‐derived adipocytes and HUVECs, the two cells were separated after coculturing for 3 days using magnetic beads according to the method established by Guillotin et al. 27 The magnetic beads coupled with an antibody against CD31 (Invitrogen, USA), which is a specific protein of endothelial cells, therefore could specifically recognize and separate HUVECs from HBMSC‐derived adipocyte. The separated two cells were named co‐adipocytes and co‐HUVECs, respectively. Si‐SA Hydrogel Preparation : Si‐SA hydrogel was prepared according to the previous report. [[qv: 13b]] Briefly, CS powders (1% weight) were homogeneously dispersed into a stirred 1. 5% w/v alginate solution (SA) with a syringe, followed by the addition of 0. 5% aspartic acid (Asp) into the alginate solution with CS. Released Ca ions from CS caused by the hydrolysis of Asp further crosslinked SA, and the gelation subsequently occurred. Pure SA hydrogel was prepared by the addition of 0. 1 m CaCl 2 solution into alginate solution as the control. Nude Mice Subcutaneous Implantation : The animal experimental protocols in this study were approved by the Ethics Committee of the Shanghai Jiao Tong University School of Medicine. Before implantation, HBMSCs were first cultured in adipogenic differentiation medium containing Si ions for 21 days to achieve the adipocyte phenotype. There were 6 different groups in the implantation experiment, including SA without cells, Si‐SA without cells, SA with monocultured adipocytes, Si‐SA with monocultured adipocytes, SA with cocultured adipocytes and HUVECs, Si‐SA with cocultured adipocytes and HUVECs. For SA groups, 100 µL pure SA hydrogel encapsulating of 1 × 10 6 cells or not was in vitro crosslinked by 0. 1 m CaCl 2. For Si‐SA groups, 100 µL pure SA hydrogel encapsulating of 1 × 10 6 cells or not was first mixed before the addition of CS. After the gelification, hydrogels encapsulating with or without cells were implanted into subcutaneous pockets of 6 week old female nude mice which were purchased from Shanghai Laboratory Animal Center. A total of 18 mice were assigned randomly into 6 groups with 3 animals in each group. Two subcutaneous pockets on the two sides of armpits were made and only one kind of hydrogel was placed on each mouse. So for each hydrogel, there were 6 parallel samples placed on 3 animals. All mice were housed in the specific pathogen free (SPF) animal facility of the Laboratory Animal Center of the Shanghai Jiao Tong University. At 8 weeks, mice were sacrificed and implants were taken out and analyzed for blood vascular formation and adipose tissue formation. Histological and Immunohistochemical Analysis : Implants were retrieved at 8 weeks and first photographed for the appearance analysis followed by the preparation for histological and immunohistochemical analysis. For adipose tissue formation analysis, the implants were fixed in 4% paraformaldehyde, embedded in optimal cutting temperature (OCT) compound, and frozen at −80 °C. Then, the samples were cut into 10 µm sections, stained with Oil Red O working solutions (Sigma‐Aldrich, USA), and counterstained with hematoxylin. For neovascularization analysis, the implants were fixed in 4% paraformaldehyde and embedded in paraffin and then cut into 5 µm sections. The sections were then subjected to deparaffinizing, rehydrating and stained with H&E (Sigma‐Aldrich, USA) as the manufacturer's instructions. For immunohistochemical staining, after deparaffinizing and rehydrating, the sections were boiled in 0. 01 mol L −1 sodium citrate buffer solution (pH 6. 0) for 10 min for antigen retrieval, and then treated with 3% H 2 O 2 in methanol for 10 min to block the endogenous peroxidases activity. After being blocked with 5% goat serum for 1 h at room temperature, the sections were incubated at 4 °C overnight with primary antibody against CD31 (Abcam, UK, 1:300). According to the manufacturer's instructions, biotinylated secondary antibodies were then applied for 1 h, followed by the treatment with streptavidin–horseradish peroxidase conjugates. The immunoreaction was observed using 3, 3 N ‐diaminobenzidine tertrahydrochloride (DAB) staining (Gene Tech, China). Finally, the samples were counterstained with hematoxylin and dehydrated and covered with coverslips. Images were observed and photographed with an optical microscope coupled with charge‐coupled device (CCD) (DMI 4000, Leica, Germany). Quantitative analysis of the adipose tissue positive stained area was performed by Image‐Pro Plus software. Quantitative analysis of the diameter of blood vessels was analyzed by Image J software. Statistical Analysis : All data shown in this study were presented as means ± standard deviation (SD). Statistical analysis for determination of differences between groups was accomplished using two‐tailed analysis of variance, performed with a computer statistical program (Student's t ‐test), and * p < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201800808
| 2,019
|
Advanced Science
|
Biomimicking Fiber Scaffold as an Effective In Vitro and In Vivo MicroRNA Screening Platform for Directing Tissue Regeneration
|
Abstract MicroRNAs effectively modulate protein expression and cellular response. Unfortunately, the lack of robust nonviral delivery platforms has limited the therapeutic application of microRNAs. Additionally, there is a shortage of drug‐screening platforms that are directly translatable from in vitro to in vivo. Here, a fiber substrate that provides nonviral delivery of microRNAs for in vitro and in vivo microRNA screening is introduced. As a proof of concept, difficult‐to‐transfect primary neurons are targeted and the efficacy of this system is evaluated in a rat spinal cord injury model. With this platform, enhanced gene‐silencing is achieved in neurons as compared to conventional bolus delivery ( p < 0. 05). Thereafter, four well‐recognized microRNAs (miR‐21, miR‐222, miR‐132, and miR‐431) and their cocktails are screened systematically. Regardless of age and origin of the neurons, similar trends are observed. Next, this fiber substrate is translated into a 3D system for direct in vivo microRNA screening. Robust nerve ingrowth is observed as early as two weeks after scaffold implantation. Nerve regeneration in response to the microRNA cocktails is similar to in vitro experiments. Altogether, the potential of the fiber platform is demonstrated in providing effective microRNA screening and direct translation into in vivo applications.
|
1 Introduction MicroRNAs (miRs) are short noncoding RNAs that effectively regulate protein translation by enabling mRNA degradation and/or repressing protein translation. 1, 2 This process plays significant roles in modulating protein expression by RNA interference (RNAi). Correspondingly, miRs have been recognized as powerful therapeutic agents with high efficacy in modulating cell fate and numerous efforts have been placed on screening differential expressions of miRs in disease contexts in hope of identifying promising therapeutics. 3, 4 While the potential of miRs as promising therapeutics for tissue regeneration (cartilage, 5 skeletal muscle, 6 and bone 7 ) is well known and numerous libraries of promising miR candidates have been created, 8, 9 several bottlenecks remain and have been preventing the successful utilization of miR therapy. 10 First and foremost, is the lack of robust nonviral platforms to deliver miRs effectively in vitro and in vivo. 11, 12 Although nanoparticles have been utilized to deliver miRs, 13, 14 repeated administrations were almost always required to achieve long‐term gene silencing in vivo. 15 Given this limitation, biodegradable hydrogel is increasingly employed as a platform for sustained delivery of miRs to facilitate tissue regeneration. 16, 17, 18 However, hydrogels are often isotropic in architecture and hence lack the ability to direct tissue regrowth. 19 Second and equally important, is the fact that despite the plethora of in vitro drug screening platforms available, such as micropillars 20 and microfluidic devices, 21 it remains difficult to translate in vitro outcomes to in vivo applications with good correlations. 22, 23, 24, 25 Consequently, the development of effective miR therapeutics, particularly targeting tissue regeneration, remains slow. Here, we introduce a biomimicking aligned fiber platform which allows the delivery of miRs both in vitro and in vivo in a sustained and nonviral manner. The easy incorporation of different miRs into this platform also makes it possible to conduct extensive miR screening. Furthermore, this fiber‐miR delivery system can be easily translated into a 3D configuration by incorporating it with collagen hydrogel, which further extends its application for in vivo miR screening and directing tissue regeneration. To evaluate the robustness of this platform, we used a stringent criterion by targeting difficult‐to‐transfect primary neurons of different ages from both the central and peripheral nervous systems (CNS and PNS, respectively). We further evaluated the in vivo performance of the fiber system by using a complete transection spinal cord injury (SCI) model as a proof of concept. In particular, we modulated the intrinsic growth ability of neurons by targeting axon local protein synthesis. Within the body, axons can extend for long distances of up to 3 m. For efficient cell function, protein synthesis occurs locally at the terminal and growth ends of axons so that the transport of biomolecules from the main cell body (i. e. , cell soma) is not critical. Such controlled local protein synthesis at the ends of axons (a. k. a. distal axons) 26 has allowed severed axons to undergo regeneration within hours after injuries. 26, 27 This local protein synthesis is finely controlled by miRs, and miR‐21, 28 miR‐222, 29 miR‐132, 30 and miR‐431 31 have been identified to enhance nerve regeneration when they are overexpressed in cultured neurons independently. 28, 29, 32 Specifically, MiR‐431 enhances Wnt signaling, which is required for neurogenesis and axon growth by decreasing the expression of a Wnt antagonist, Kremen1. 31 In addition, miR‐431 could regulate motor neuron neurite length by targeting chondrolectin, one of several molecules that plays a role in motor neuron axon outgrowth. 33 MiR‐21, on the other hand, targets Sprouty2, which in turn inhibits the Ras/Raf/extracellular signal‐regulated kinase (ERK) pathway, and antagonizes fibroblast growth factor (FGF) signaling. 28 Mir‐222 targets phosphatase and tensin homolog (PTEN), an inhibitor of the PI3K pathway that is important to central axon growth. 29 Finally, Ras GTPase, p120RasGAP (Rasa1), which is involved in cytoskeletal regulation, is targeted by miR‐132. 34 Therefore, in contrast to the conventional ways of simply modulating the microenvironment of injured neurons with neurotrophic factors 35 or neutralizing growth inhibitory molecules, 36 we recognize that mature neurons possess diminished regenerative capacity, which is a major cause of regeneration failure. Hence, the mere modulation of the cell/tissue microenvironment may not be sufficient to achieve the desired regeneration outcomes. 27 Specifically, we hypothesized that the incorporation of miRs into biomimicking polycaprolactone (PCL) fiber constructs would enhance gene silencing by localized and prolonged availability of miRs. Furthermore, by taking advantage of the biomimicry nature of fibers, physiologically relevant behaviors could be elicited in cells in vitro; hence allowing good correlation between in vitro and in vivo biological behaviors. We show that such fiber‐mediated miR delivery platform may serve as effective in vitro and in vivo drug testing systems. 2 Results 2. 1 Characterization, Cellular Response, and Cy5‐RNA Distribution on PCL Aligned Fibers The PCL fibers were produced using an electrospinning technique. Figure 1 A shows the morphology of the aligned fibers before and after poly‐3, 4‐dihydroxy‐ l ‐phenylalanine (DOPA) coating. Poly‐DOPA coated fibers had an average diameter of 948 ± 127 nm while uncoated fibers had an average fiber diameter of 790 ± 152 nm. This suggests that poly‐DOPA coating did not alter the topography of these aligned fibers ( p > 0. 05). Figure 1 Characterization of aligned fiber substrates. A) Uncoated fibers (fiber diameter φ = 790 ± 152 nm) and poly‐DOPA coated fibers (φ = 948 ± 127 nm). B) E14 cortical neurons cultured for 3 d on aligned fibers versus tissue culture plates (TCPS). C, D) Average total length of neurite and average length of the longest neurite, showing that the aligned topography significantly promoted neurite extensions. * p < 0. 05, Student's t ‐test. E) Fluorescent microscopy image shows a uniform distribution of Cy5‐RNA on aligned fiber substrates. F) Cy5 signals were clearly detected in the cultured P1 cortical neurons, indicating successful cellular uptake. G) Release profile of microRNAs from the aligned fiber substrates. To evaluate the effect of aligned fiber topography on neurite outgrowth, E14 cortical neurons were cultured on tissue culture plates (TCPS) and aligned fibers for comparison. Three days after culture, aligned neurite extensions were detected on the fiber substrates. Conversely, cells that were cultured on TCPS grew toward different directions (Figure 1 B). As indicated in Figure 1 C, D, cortical neurons cultured on aligned fibers exhibited significantly longer neurite outgrowth in terms of the total length and the longest length of neurites as compared to conventional 2D cultures. A uniform distribution of Cy5‐labelled double‐stranded RNA (Cy5‐RNA), which was of similar size as miRs, was clearly identified on the aligned fiber substrates after 2 h of coating (Figure 1 E). Three days after cell seeding onto these Cy5‐RNA absorbed scaffolds, Cy5‐RNA signals were detected in the neurons, suggesting that the miRs were taken up successfully by the cortical neurons (Figure 1 F). The average loading efficiency of miRs on the aligned fiber scaffolds was around 77%. Specifically, 15% of miRs were released at Day 7. As can be seen from the release curve, a sustained delivery of miRs was achieved by our aligned fiber platform (Figure 1 G). 2. 2 Systematic In Vitro Screening for Optimal miR Combinations to Promote Neurite Outgrowth Here, we analyzed all possible combinations of the four miRs to identify the optimal cocktails that could maximize neurite outgrowth. In order to ensure the robustness of the miR mixtures in enhancing regeneration from neurons, we utilized neurons of different ages (embryonic day 14 (E14) vs postnatal day 1 (P1) vs adult) and origin (CNS vs PNS). Figure 2 summarizes the top five cocktails that were identified using the aligned fiber screening platforms. In all samples, neurite projections were found in both directions, parallel to the aligned fibers (Figure 2 A, D, G, white arrows). Such aligned neurites enabled easy tracing and quantification of neurite outgrowth lengths. Figure 2 In vitro miR cocktail screening using E14, P1 cortical neurons and adult DRG neurons. A) Representative fluorescent microscopy images of E14 cortical neurons. B, C) Average total neurite length and average length of longest neurite of the top five groups of E14 cortical neurons. D) Representative fluorescent microscopy images of P1 cortical neurons. E, F) Average total neurite length and average length of longest neurite of the top five groups of P1 cortical neurons. G) Representative fluorescent microscopy images of adult DRG neurons. H, I) Average total neurite length and average length of longest neurite of the top five groups of adult DRG neurons. * p < 0. 05, ** p < 0. 01, *** p < 0. 001, One‐way ANOVA/Kruskal–Wallis test and Mann–Whitney post hoc test. 2. 2. 1 Embryonic Cortical Neurons As shown in Figure 2 A–C, scrambled Neg miR (Neg miR) induced a baseline total length of 233. 7 ± 4. 711 µm and a longest length of 140. 8 ± 3. 744 µm. These measurements were not significantly different as compared to untreated cells and cells that were treated with transfection reagent, TransIT‐TKO (TKO) (Figure S1A–C, Supporting Information). As shown in Figure S1A, D, E in the Supporting Information, for treatment with individual miRs, neurite outgrowth was similar for miR‐21 (285. 3 ± 6. 721 µm) and miR‐132 (287. 2 ± 7. 064 µm). These miR treatments, in turn, resulted in longer neurite outgrowths than miR‐222 (233. 5 ± 5. 352 µm) and miR‐431 (246. 6 ± 5. 352 µm). For treatment with two‐miR cocktails, miR‐222/miR‐431 showed the best result followed by miR‐132/miR‐431 and miR‐21/miR‐132. Among these six groups, miR‐21/miR‐431 resulted in the smallest extent of neurite outgrowth (Figure S1A, F, G, Supporting Information). As depicted in Figure S1A, H, I in the Supporting Information, for three‐miR cocktails, the withdrawal of miR‐21 resulted in the longest neurite extension followed by the withdrawal of miR‐222. Finally, the combination of four miRs did not generate the best outcome (Figure S1J, K, Supporting Information). Taken together, the top five groups that induced the longest neurite outgrowth in E14 cortical neurons were: miR‐132/miR‐222/miR‐431 > miR‐21/miR‐132/miR‐431 > miR‐132 > miR‐21 > miR‐222/miR‐431 (Figure 2 A–C). 2. 2. 2 P1 Cortical Neurons As shown in Figure 2 D–F, similar to the embryonic cortical neurons, Neg miR treatment induced a baseline in terms of total length (231. 3 ± 6. 577 µm) and the longest length of neurites (125. 4 ± 3. 997 µm). These values obtained were similar to the embryonic cortical neurons and did not show significant difference as compared to untreated cells (249. 6 ± 6. 188 µm) and cells treated with TKO only (228. 2 ± 6. 776 µm) (Figure S2A–C, Supporting Information). For treatment with individual miRs (Figure S2A, D, E, Supporting Information), similar trends were observed as compared to embryonic cortical neurons. Specially, miR‐21 (321. 3 ± 8. 264 µm) resulted in the longest neurite outgrowth, followed by miR‐132 (295. 3. 2 ± 6. 348 µm). MiR‐222 (251. 7 ± 5. 411 µm) and miR‐431 (269. 1 ± 7. 135 µm) were similar and more inferior. As shown in Figure S2A, F, G in the Supporting Information, for two‐miR cocktails, miR‐222/miR‐431 showed the best result followed by miR‐132/miR‐431 and miR‐21/miR‐132. Among these six groups, miR‐21/miR‐431 was the worst. These trends were consistent with those observed in embryonic cortical neurons. For three‐miR cocktails, the withdrawal of miR‐21 exhibited the longest neurite extension followed by the withdrawal of miR‐222 (Figure S2A, H, I, Supporting Information). Finally, the combination of four miRs did not generate the best outcome (Figure S2J, K, Supporting Information). Altogether, the top five groups that induced the longest neurite outgrowth in P1 cortical neurons were: miR‐222/miR‐431 > miR‐132/miR‐222/miR‐431 > miR21 > miR132 > miR‐21/miR‐132/miR‐431 (Figure 2 D–F). These groups were similar to those obtained when embryonic cortical neurons were evaluated, although the ranking of the extent of neurite outgrowth is slightly different. 2. 2. 3 Adult Dorsal Root Ganglion (DRG) Neurons As shown in Figure 2 G–I, Neg miR induced a baseline total length of 591. 0 ± 132. 7 µm and a longest length of 309. 2 ± 57. 4 µm. Similar to cortical neurons, these measurements were not significantly different as compared to untreated cells and cells that were treated with TKO only (Figure S3A–C, Supporting Information). As shown in Figure S3A, D, E in the Supporting Information, for treatment with individual miRs, miR‐21 (1091. 9 ± 6. 0 µm) resulted in longer neurite outgrowths again, an observation that was similarly made with the cortical neurons. However, in contrast to cortical neurons, miR‐431 (1080. 8 ± 161. 7 µm) resulted in similar and longer neurite outgrowths as miR‐21, followed by miR‐132 (1057. 5 ± 161. 8 µm). The worst was observed with miR‐222 treatment (956. 6 ± 97. 0 µm). For treatment with two‐miR cocktails, only slight similarity to cortical neurons was observed. Specially, miR‐222/miR‐431 showed the best result. However, this was then followed by miR‐21/miR‐222 and miR‐21/miR‐431, which was different from cortical neurons. Among these six groups, miR‐132/miR‐431 resulted in the smallest extent of neurite outgrowth (Figure S3A, F, G, Supporting Information). As demonstrated in Figure S3A, H, I in the Supporting Information, for three‐miR cocktails, the withdrawal of miR‐21 resulted in the longest neurite extension again. This was then followed by the withdrawal of miR‐132. Finally, consistent with the results from cortical neurons, the combination of four miRs did not generate the best outcome (Figure S3J, K, Supporting Information). Taken together, the top five groups that induced the longest neurite outgrowth in adult DRG neurons were: miR‐222/miR‐431 > miR‐21/miR‐222 > miR‐21/miR‐431 > miR‐132/miR‐222/miR‐431 > miR‐21 ( Figure 3 G–I). Correspondingly, miR‐21, miR‐222/miR‐431, and miR‐132/miR‐222/miR‐431 appeared to enhance neurite outgrowth in neurons, regardless of the age and origin of neurons. Figure 3 Target gene knockdown in P1 cortical neurons cultured on aligned fiber substrates at day 3 in vitro. A) Fold change of Sprouty2 after treatment with miR‐21. B) Fold change of PTEN after treatment with miR‐132. C) Both PTEN ( p = 0. 3145) and Kremen1 ( p = 0. 2287) were knocked down after treatment with miR‐222/miR‐431. D) Rasa1, PTEN ( p = 0. 0635) and Kremen1 ( p = 0. 0645), were knocked down following miR‐132/miR‐222/miR‐431 treatment. All comparisons were normalized with Neg miR‐treated group. * p < 0. 05, ** p < 0. 01, Student's t ‐test. 2. 3 Aligned Fiber‐Mediated miR Delivery Induced Target Gene Silencing The target gene silencing effect was evaluated 3 d after scaffold‐mediated miR transfection in P1 cortical neurons. Considering cell viability and in vivo relevance, P1 cortical neurons were chosen. Based on the screening data of P1 cortical neuron, the top four groups (miR‐21, miR‐132, miR‐222/miR‐431, and miR‐132/miR‐222/miR‐431) were selected and the gene expression levels of their downstream targets were determined. As compared to Neg miR treatment, the transfection of miR‐21 and miR‐132 resulted in significant knockdown of their downstream targets, Sprouty2 (Figure 3 A) and Rasa1 (Figure 3 B), respectively. The gene silencing effect of miR‐222/miR‐431, however, was not as obvious as the treatment with single miRs (Figure 3 C). On the other hand, the knockdown effect in miR‐132/miR‐222/miR‐431 treated group was more significant, especially for Rasa1 (Figure 3 D). As compared to 2D bolus delivery (Figure S4, Supporting Information), the miRs in each group exhibited significantly better gene silencing effects when cells were transfected using the fiber scaffolds. 2. 4 MiR Treatment Enhanced the Formation of Growth Cones Besides evaluating target gene knockdown in the top four groups of miRs, we also analyzed the effects of these miRs on growth cone formation by immunofluorescent staining on 2D cultures so as to gain insight to the possible mechanisms that may be involved in this miR‐related enhanced neurite outgrowth. As shown in Figure 4 A, three days after cell seeding, the growth cone structures were located at the tips of the neurites (pointed by the arrows). The number of growth cones and the growth cone area were quantified. Correspondingly, miR‐132 and miR‐132/miR‐222/miR‐431‐treated neurons exhibited more growth cones as compared to Neg miR‐treated neurons, followed by the miR‐21‐treated group (Figure 6 B). Statistical analysis of the growth cone area showed that miR‐132‐and miR‐132/miR‐222/miR‐431‐treated neurons also exhibited larger growth cone areas as compared to Neg miR treatment. This was then followed by miR‐21‐and miR‐222/miR‐431‐treated groups (Figure 4 C). Figure 4 Growth cone quantification following miR treatment. A) Representative confocal images of immunofluorescent‐stained neurons. White arrows represent growth cones. Yellow boxes depict high magnification images of growth cones within white boxes. B) Number of growth cones and C) area of growth cone per neuron, suggested that the treatment of miR/miR cocktails significantly promoted growth cone formation. * p < 0. 05, ** p < 0. 01, *** p < 0. 001, Shapiro–Wilk normality test followed by Kruskal–Wallis test and Mann–Whitney post hoc test. 2. 5 In Vivo Scaffold Characterization To translate the axon‐growth promoting aligned fibers to implantable constructs, we utilized collagen to support aligned fibers to formulate a 3D structure. Figure 5 A shows the cross‐section view of fiber‐hydrogel scaffold. This scaffold comprised of electrospun aligned fibers that were supported by a collagen matrix. The presence of collagen maintained the fiber orientation and alignment without inducing fiber fusion. The aligned fibers possessed an average diameter of 1. 245 ± 0. 13 µm, as shown in Figure 5 B. Figure 5 In vivo scaffold design. A) SEM image of cross‐section of fiber‐hydrogel scaffold, showing that aligned fibers were uniformly distributed and well‐oriented inside the collagen matrix. B) SEM image of aligned fibers with an average diameter of 1. 245 ± 0. 13 µm. 2. 6 MiR Treatment Improved Nerve Ingrowth into the Fiber‐Hydrogel Scaffold after SCI To evaluate the in vivo efficacy of our 3D aligned fiber scaffold, we used a complete transection SCI model as a proof of concept. Based on our in vitro screening data from Section 2. 2, we chose the top single, double and triple miR combinations for in vivo study. These miR combinations were coupled with Neurotrophin‐3 (NT‐3) as previous SCI studies have suggested that the presence of NT‐3 is crucial for promoting neuronal survival, axonal sprouting, and regeneration. 37, 38, 39, 40 However, we found that NT‐3 alone was insufficient to stimulate robust neurite ingrowth into the scaffold after complete spinal cord transection ( Figure 6 ). Figure 6 MiRs stimulated extensive neurite ingrowth into scaffolds 2 weeks after SCI. A) Representative fluorescent microscopy images of NF‐200 (red) expression at injury sites in the various groups. White dotted boxes represent the region of enlargement (green boxes). B) Quantification of the percentage of scaffold stained with NF‐200 suggested that the presence of miRs significantly promoted nerve regeneration after SCI. * p < 0. 05, Shapiro–Wilk normality test followed by Kruskal–Wallis test and Mann–Whitney post hoc test. On the other hand, when NT‐3 was coupled with the selected miR combinations (Figure 6 ), extensive neurofilament ingrowth into the scaffolds was observed, especially for the triple miR treatment. Specifically, neurofilaments were evenly distributed inside the scaffolds. Furthermore, the aligned fibers provided topographical guidance to direct regenerated neurofilaments in the direction that was parallel to the spinal cord. Similar to in vitro observations, there was very limited neurofilament ingrowth into the scaffolds that incorporated Neg miR (Figure 6 A, B). 2. 7 MiR Treatment Did Not Induce Obvious Glial Scar Formation To evaluate any possible effects of the miR cocktails on glial scarring, we analyzed glial scar formation by glial fibrillary acidic protein positive (GFAP + ) signals. As shown in Figure S5 in the Supporting Information, the implantation of fiber‐hydrogel scaffolds showed no difference in glial scarring regardless of miR combinations. 3 Discussion Neural tissue engineering approaches, such as using nerve grafts and biomaterial scaffolds, or using cells and neurotrophic factors have emerged as alternatives for nerve injury treatment. 41 Evidences have indicated that the combination of these therapies could better mimic the properties of the microenvironment that is required for nerve repair. 42 Electrospun fibers have been shown to mimic the architecture of the natural extracellular matrix and provide the necessary topographical cues to modulate cell fate. 43, 44, 45 Specifically, electrospun fibers could mimic the size scale and architecture of axons, which allows controlled and sustained delivery of biomolecules to direct oligodendrocyte progenitor cell (OPC) differentiation, maturation 43 and myelination. 46 As small noncoding RNAs that effectively modulate cellular behaviors, miRs have emerged as one of the most prominent biomolecules with powerful therapeutic potential. Correspondingly, numerous screening studies have established libraries of dysregulated miRs in the context of diseases and injuries in hope to develop effective treatments. 9, 47, 48 Unfortunately, the use of miRs for therapeutic purposes remain limited. We speculate that this is partly due to the lack of robust nonviral platforms to deliver miRs efficiently in vitro and in vivo; and the shortage of effective drug screening devices that correlate in vitro outcomes with in vivo performances. With these requirements in mind, we developed a robust biomimicking fiber platform, which could deliver miRs in a sustained (Figure 1 G) and nonviral manner using PCL electrospun fibers. PCL was utilized due to its biocompatibility and is also a U. S. Food and Drug Administration (FDA) approved polymer. As such, PCL has been widely used for both in vitro cell culture 49 and in vivo implantation. 50 By applying this screening platform, we showed that even sensitive and notoriously difficult to transfect cells, like neurons, could be modulated. In addition, the efficacy of this platform was further evaluated in a rat spinal cord transection model as a proof of concept. The biomimicry nature of the system provided good biological correlation between in vitro and in vivo outcomes. As compared to conventional 2D cultures, our aligned fiber platform provided contact guidance to orientate and promote the growth of neurites. Consequently, significantly longer neurite outgrowth lengths were achieved in E14 cortical neurons (Figure 1 B–D). Notably, this is the first time that CNS neurons were tested on aligned fibers versus 2D cultures. Regardless, this result is consistent with the observations that have been reported by other groups when DRG neurons from the PNS were cultured on aligned fibers versus 2D film controls. 51, 52 Furthermore, the configuration of our aligned fiber platform is convenient for imaging and quantification due to the obvious growth direction of the neurites. By coating the fibers with poly‐DOPA, various biomolecules such as laminin and miRs, which facilitate neuron attachment and neurite extension respectively, can be adsorbed onto the scaffolds. As a mussel‐inspired bioadhesive, 53 poly‐DOPA has been used for sustained delivery of miRs 43, 54 and siRNAs. 55 Poly‐DOPA coated substrates allowed the efficacious absorption of the miRs onto the aligned fibers without altering fiber topography (Figure 1 A) or the bioactivity of the miRs. Moreover, the sustained delivery of miRs from the aligned fibers reduced cytotoxicity 43 and facilitated better gene silencing outcomes as compared to conventional bolus delivery (Figure 3 vs Figure S4, Supporting Information). To identify the amount of miRs that is required to exert effective downstream gene silencing in primary neurons, we carried out a titration process and compared 0. 25, 0. 5, and 1 µg of miRs. From these preliminary studies, 1 µg of miRs was found to be an effective amount for gene regulation in primary neurons and was hence utilized for our actual experiments despite that the loading efficiency on the fibers was around 77%. Besides establishing a novel miR‐screening substrate that correlates well in vitro and in vivo, we also introduced a new approach to enhance nerve regeneration after SCI. SCI lead to devastating outcomes of paralysis and functional impairment and are the major causes of morbidity and mortality, particularly in young adults and children. 56 Specifically, instead of the conventional methods of modulating the microenvironment after injury with overexpression of growth factors 35 and neutralizing any growth inhibitory molecules, 36 we targeted the intrinsic growth ability of neurons by enhancing local protein synthesis at the growth cones of axons using miRs. Here, the miRs of interest, miR‐21, miR‐222, miR‐132, and miR‐431, have been shown to increase significantly at the growth cones of neurons after nerve injuries. 28, 29, 30, 31 The upregulation in expression of these miRs modulates different pathways in neurons, resulting in robust axon regeneration. 29, 57 In particular, they either enhance pathways that regulate neurogenesis and axon growth or directly remove molecular brakes that prevent regrowth. 28, 29, 31, 33, 34 Based on our previous work with OPCs where miR‐219/miR‐338 cocktail transfection enhanced differentiation outcomes more than miR‐219 treatment alone, 43 we speculated that a combination of miRs may also benefit axon regeneration. Therefore, to further understand the capabilities of these miRs and their effects on neurons, we proceeded to investigate the possible synergistic or antagonistic effects of these miRs by using our aligned fiber miR screening platform. Our miR screening data showed that there are many similar miR combinations for all three types of neurons, regardless of their age and origin. Adult DRG neurons exhibited more robust neurite outgrowth and greater enhancement with miR treatment than CNS neurons. Interestingly, the enhancement in neurite length of DRG neurons after miR treatment, as observed in this work, was stronger (1. 8–2. 2 times) as compared to the literature. 28, 34 Such observation may be due to the synergistic effects of topographical cues and the sustained miR signaling using our fiber substrates. It should be noted, however, that for miR cocktails, there might also be antagonistic effects, which could affect neuronal outgrowth. This might explain why the four‐miR cocktail did not yield the best outcome. To further evaluate the effects of these miRs/miR cocktails in promoting neurite outgrowth, the number of growth cones and growth cone area were quantified on 2D cultures. The choice of 2D cultures for the analyses of growth cone formation stems from the fact that as compared to aligned fiber substrates, the growth cone sprouts were completely extended on 2D cultures, hence facilitating visualization and quantification. So far, real‐time polymerase chain reaction (RT‐PCR) and growth cone evaluations were carried out to reveal the possible mechanisms of these miR cocktails in promoting neurite outgrowth. However, this work did not illustrate related pathways that may have been affected by these miR combinations and their possible antagonistic effects. Future works, such as immunostaining, western blot, and RNA sequencing, would be suitable to further elucidate these mechanisms in greater detail. Based on our in vitro screening data, the top single, double, and triple miR cocktails were evaluated for their efficacy in vivo using SCI as a proof of principle. In doing so, we also analyzed if the biomimicry nature of the aligned fibers could provide good correlation between in vitro and in vivo outcomes. To enable the direct implantation of aligned fibers within the spinal cord, we utilized the fiber‐hydrogel system. This platform consisted of a collagen matrix that enabled the delivery of multiple biochemical factors (miRs, NT‐3). 58 Most importantly, it also supported and retained aligned fibers in a 3D configuration to guide the direction of axon growth liken the in vitro situation. Consequently, we found that the trends of in vivo nerve regeneration closely resembled the in vitro results, particularly those involving CNS neurons. Specifically, miR‐132/miR‐222/miR‐431 generally provided the best regeneration outcomes as compared to miR‐21, miR‐222/miR‐431, and Neg miR treatments. Comparing in vivo versus in vitro outcomes, we noted that the best in vitro candidate, miR‐222/miR‐431, revealed similar results as miR‐21, in terms of NF200 positive signal intensity inside the scaffolds after SCI (Figure 6 A). On the other hand, the triple miRNA combination, miR‐132/miR‐222/miR‐431, showed significantly higher neurite density inside the scaffolds as compared to all the other groups (Figure 6 A, B). One possible reason may be due to the enhanced growth cone formation and growth cone area in response to the triple miR cocktail as observed in Figure 4. The potential of NT‐3 in promoting neuronal survival, axonal sprouting, and regeneration following SCI is well documented. 37, 38, 39, 40 However, in our case, when NT‐3 was used alone to treat SCI, nerve ingrowth was observed but not robust (Figure 6 A). In contrast, axonal growth was greatly improved when NT‐3 was coupled with miRs. These observations suggest that the mere modulation of the microenvironment by neurotrophins may be insufficient to achieve prominent nerve regeneration after nerve injuries. Conversely, by adding miRs to increase the intrinsic growth ability of axons, a significant regeneration was obtained (Figure 6 B). Through our systematic in vitro and in vivo studies, we have extended our knowledge on neuron intrinsic growth property and axon regeneration by using our aligned fiber screening platforms. Although SCI was used as a proof of principle to evaluate the in vivo efficacy of our aligned fiber construct, the promising outcomes observed at early time points of recovery suggests that this method holds tremendous potential for SCI treatment. Therefore, in‐depth studies such as tracing and functional evaluations will be conducted to further investigate the efficacy of our platform and the effects of these miRs on promoting functional recovery in the long term. Altogether, we have developed a robust nonviral gene delivery platform for both in vitro and in vivo studies, which addresses the problems of the lack of a delivery system and weak in vitro and in vivo correlations. 4 Conclusion In this study, we established a biomimicking aligned fiber substrate that provides sustained and effective nonviral delivery of miRs for in vitro and in vivo miR screening. As compared to conventional 2D cultures, our aligned fiber substrates promoted longer neurite outgrowth and enhanced gene silencing in primary neurons. In addition, this platform could also be implantable for direct in vivo miR screening. As a proof of concept, complete transection SCI was carried out to evaluate the efficacy of this aligned fiber‐hydrogel system in vivo. Robust nerve ingrowth was observed as early as two weeks after scaffold implantation after SCI. Given the ability to deliver miRs nonvirally, our scaffolds may be translated to clinical applications without raising biosafety concerns associated with virus transfections. Studies are now ongoing to further evaluate the capability of our fiber‐hydrogel scaffolds in promoting nerve regeneration, remyelination, and functional recovery after SCI. 5 Experimental Section Materials : Polycaprolactone ( M w: 45 000 (45k PCL) and 80 000 (80k PCL)), DOPA, 2, 2, 2‐trifluoroethanol (TFE, ≥99. 0%), poly‐ d ‐Lysine (PDL) (P0899), cytosine arabinoside (Ara‐C), 5‐fluoro‐2′‐deoxyuridine and Heparin sodium were purchased from Sigma‐Aldrich. Alexa‐Fluor 555 goat anti‐Mouse, Alexa‐Fluor 488 Phalloidin, Scrambled Neg miR, miR‐21‐5p (PM10206), miR‐132‐3p (PM10166), miR‐222‐3p (PM11376), and miR‐431‐5p (PM10091), laminin (23017015), DAPI (4′, 6‐diamidino‐2‐phenylindole), paraformaldehyde (PFA, 7230681), phosphate buffered saline (PBS; pH7. 4), SYBR Select Master Mix, Trizol Reagent, B‐27 supplement, N‐2 supplement, goat serum, neurobasal medium, Glutamax supplement, penicillin–streptomycin, bovine serum albumin (BSA, A1000801), and Quant‐iTTM RiboGreen RNA reagent kit (Invitrogen) were obtained from Life Technologies, USA. Cy5‐labelled double‐stranded RNA(Cy5‐RNA) of similar size as miR (i. e. , 21–23 base pairs) and TKO were purchased from Integrated DNA Technologies (IDT) and MirusBio respectively. Mouse anti‐βIII Tubulin (Tuj‐1) (801202) and Chicken anti‐NF200 (822601) was purchased from Biolegend. Rabbit anti‐GFAP (Z0334) was obtained from DAKO. Rat‐tail Collagen type I was purchased from Corning. NT‐3 and nerve growth factor (NGF) were purchased from PeproTech. Dulbecco's modified Eagle medium/F12 (DMEM/F12) medium was purchased from Lonza, Switzerland. Fetal bovine serum (FBS) was acquired from Research Instruments. In Vitro Studies—Fabrication and Characterization of Aligned Fibers : In vitro aligned fiber scaffolds were fabricated using the electrospinning process (Figure S6A, Supporting Information). Briefly, 50 mg of PCL ( M w : 45 000, a. k. a. 45k PCL) was melted in an 18 mm × 18 mm mold at 60 °C before being cooled down to room temperature. The resulting block of 45k PCL polymer was then cut into pieces of 0. 5 cm × 1. 0 cm × 2. 0 cm before they were sectioned into 20 µm thick films. These 45k PCL sheets were then washed three times in distilled water and placed in 1 × PBS at room temperature for immediate use. 45k PCL sheets were dried at 37 °C for 2 h prior to the electrospinning of fibers. These dried sheets were then sterilized with 70% ethanol before placing on glass coverslips (diameter: 18 mm) as shown in Figure S6B, C in the Supporting Information. Once placed in position, the coverslips were heated at 50 °C to designate a 10 mm × 10 mm cell seeding area before six of these coverslips were placed on a rotating wheel for the collection of electrospun fibers (Figure S6A, Supporting Information). For electrospinning, PCL ( M w : 80 000, a. k. a. 80k PCL) was dissolved in TFE to obtain a 14 wt% solution. The homogenous solution was then loaded into a syringe and dispensed at a fixed rate of 1. 0 mL h −1 by a syringe pump (New Era pump Systems Inc. , USA). Positive 8 kV (Gamma High Voltage, USA) and −4 kV were then applied to the polymer solution and the rotating collector (2400 rpm), respectively. The syringe and the collector were separated at 22 cm apart. Electrospun fibers were then collected on the coverslips and held in alignment by the melted 45k PCL sheets. Here, two different molecular weights of PCL were used due to their slight difference in melting point. Specifically, 45k M w PCL sheets melted at 50 °C, which allowed them to be premelted onto the glass coverslips to serve as supports. At this temperature, the 80k M w PCL fibers remained intact and were, hence, collected and stuck onto the 45k M w PCL sheets. The morphology of the aligned fiber substrates was evaluated by scanning electron microscopy (SEM) (JOEL, JSM‐6390LA, Japan) under an accelerating voltage of 10 kV after sputter coating with platinum for 100 s at 10 mA. The average fiber diameters were then quantified by measuring 100 fibers from high magnification images (2500 ×) using Image J software (National Institutes of Health (NIH), USA). In Vitro Studies—Preparation of miR‐Loaded PCL Fibers : Suspended PCL fibers on coverslips were fitted into 12‐well plates and sterilized with 70% ethanol for 30 min. Thereafter, the fibers were immersed into 0. 5 mg mL −1 DOPA that was dissolved in poly‐DOPA coating buffer (10 × 10 −3 m bicine and 50 × 10 −3 m NaCl, pH = 8. 5) before placing on an orbital shaker (120 rpm) for 4 h. DOPA‐coated fibers were then washed with deionized water and lyophilized overnight. Subsequently, the lyophilized fibers were coated with 70 µg cm −2 of PDL for 1 h at 37 °C before rinsing off any unbound PDL with distilled water and coating these fibers with laminin at 7 µg cm −2 for 2 h at 37 °C. 1. 5 µL of TransIT‐TKO was diluted in DMEM before complexation with 1 µg of miRs. The complexation was performed at room temperature for 10 min. After removing laminin, these complexes were placed on the 10 mm × 10 mm cell seeding area to allow complete adsorption at 37 °C for 2 h. All suspended PCL fibers were then divided into 18 groups, as shown in Table S1 in the Supporting Information. In Vitro Studies—In Vitro Characterization of miR‐Loaded Aligned Fibers : To determine the distribution of miRs on the aligned fibers, Cy5‐RNA was complexed and coated using the same protocol as highlighted in Section 5. 2. 2. Two hours after coating, the scaffolds were rinsed once with PBS. The drug distribution was then observed with a fluorescent microscope (LeicaDMi8). To evaluate the extent of cellular uptake, P1 cortical neurons were seeded onto Cy5‐RNA absorbed aligned fibers and cultured for 3 d. Following that, the colocalization of Cy5‐RNA, DAPI, and neuronal marker, βIII‐Tubulin, were examined under a confocal microscope (Zeiss LSM710). To calculate microRNA loading efficiency, aligned fiber scaffolds ( n = 3) with neg miR complex were washed once with 1 × PBS. The amount of unbound miR was then decomplexed by heparin (250 µg mL −1 ) before being determined by RiboGreen Assay. The fluorescence intensity was then measured by a microplate reader (Tecan, Infinite 200). MicroRNA loading efficiency was calculated based on the following equation (1) Loading efficiency % = Total mass of miRs − Mass of unbound miRs Total mass of miRs × 100 % Thereafter, the miR‐loaded aligned fiber scaffolds were completely submerged in 1 mL of 1 × PBS and incubated at 37 °C. At each time point, 1 mL of supernatant was collected and an equal volume of fresh PBS was then added. The amount of miRs that were released at each time point were then determined using RiboGreen assay. Cumulative release profile was plotted as a percentage of the actual mass of miRs loaded on the scaffolds. In Vitro Studies—Primary Rat Neuron Isolation and Culture : All animal experiments were approved by the Institutional Animal Care and Use Committee, Nanyang Technical University (IACUC, NTU). Cortical Neurons : Dissociated cultures of rat cortical neurons were generated from either time‐mated embryonic day 14 (E14) rats or pups (P1). The cortices from either E14 or P1 were digested with 0. 25% trypsin at 37 °C for 15 min. After homogenization, the suspension was passed through a 70 µm cell strainer (BD, Bioscience, USA) and the cells were seeded onto the miR‐loaded PCL scaffolds at a density of 25 000 cells scaffold −1. 2D PDL and laminin coated coverslips were set in parallel, onto which the neurons were transfection by 100 × 10 −9 m miRs at 1:1 (v/v) of miR: TKO one day after cell seeding. All cultures were then maintained in Neurobasal media, which contained 10% FBS, 2% B‐27 Supplement, penicillin–streptomycin (10 µg mL −1 ), and 1% GlutaMAX. The medium was half changed 2 d after cell seeding and the culture process was maintained for 3 d in total. Neurons on scaffolds and coverslips were collected for immunofluorescent staining or real‐time PCR. Adult DRG Neurons : Three adult Sprague‐Dawley rats were utilized and DRGs were collected in DMEM/F12 medium in a petri dish. All the meninges from the DRGs were taken off to minimize culture contamination with other cells. Dissociated cells were then seeded onto the miR‐loaded PCL scaffolds at a density of 5000 cells scaffold −1. The cultures were maintained at 37 °C and 5% CO 2 in the DMEM/F12 medium containing 1:100 penicillin/streptomycin, 10% horse serum, 1% of N 2 and 50 ng mL −1 of NGF. Half of the medium was changed 24 h after seeding and 10 µM per well of Ara‐C and 20 µM per well of 5‐fluoro‐2'‐deoxyuridine were added to the culture. Cultures were kept for 3 d postseeding. In Vitro Studies—Real‐Time PCR : After 3 d of culture, P1 cortical neurons seeded on the scaffolds were lysed by TRIzol reagent and the RNA was extracted. The seeding density was 25 000 cells per scaffold and six scaffolds were pooled together. 500 ng of RNA was used for reverse transcription. Real‐time PCR was then carried out using SYBR Green Supermix in a StepOnePlus system (Applied Biosystems, USA). The sequences of the primers are shown in Table S2 in the Supporting Information and RNA18S was used as the housekeeping gene. All the primers showed similar amplification efficiency, hence the ΔΔ C t method was used for fold change analysis. All results were normalized by the C t value of neurons that were treated with Neg miR. In Vivo Studies—Fiber‐Hydrogel Scaffold Fabrication: Fibers : Poly(caprolactone‐ co ‐ethyl ethylene phosphate) (PCLEEP) copolymer ( M w = 59 102, M n = 25 542) was a gift from Dr. Yucai Wang's lab. It was synthesized as reported previously. 59 PCLEEP was dissolved in TFE at 33% w/w and settled overnight before use to ensure homogeneity. A two‐pole air‐gap electrospinning technique was adopted to fabricate aligned PCLEEP fibers (Figure S6D, Supporting Information). Briefly, the electrospinning solution was loaded into a 3 mL syringe that was subsequently capped with a 21‐gauge blunt‐tipped needle. This needle tip was then charged with +8 kV and the two‐pole air‐gap collector was charged at −4 kV. The electrospinning solution was released at a flow rate of 1. 5 mL h −1 by a syringe pump. PCLEEP fibers were then deposited within a 5. 0 cm air gap area that was between the stationary poles. Each set of fibers was obtained after 6 min and 30 s of spinning and combined into set sets. Fibers were sterilized under UV light for 30 min before stacking the layers and rolling them into a bundle of fibers. A sterilized cylindrical mold (8. 0 mm in length and 3. 5 mm in inner diameter) was used to set the fiber bundles in the core region prior to the addition of collagen matrix. In Vivo Studies—Fiber‐Hydrogel Scaffold Fabrication: Collagen Matrix : Rat‐tail type 1 collagen was used to fabricate the hydrogel matrix according to the manufacturer's protocol. Briefly, 10 × PBS, 1. 0 N NaOH, deionized (DI) water, and collagen type 1 were added into a sterile 600 µL micro‐tube in the listed order and mixed gently to get a final collagen concentration of 3. 0 mg mL −1. To promote infiltration of neurofilaments, growth factors were incorporated into this matrix. NT‐3 was reconstituted in 0. 1% BSA and 400 µg µL −1 heparin at 1:1 v/v to arrive at a stock concentration of 2 µg µL −1. 4 µL of NT‐3 stock solution was then used to substitute 4 µL of DI water in the 250 µL collagen mixture per 8 mm mold. Hence, a total of 8 µg of NT‐3 was loaded into this mixture. 20 µg (7. 5 µL of 100 × 10 −6 m miR stock) of selected miRs were complexed with TKO (1:1 v/v) and subsequently loaded. A total of 2 µg of NT‐3 and 5 µg of miRs were used per animal. This collagen mixture was kept on ice until dispensed into the mold that contained the electrospun fiber bundle in the core region. Hydrogel formation took place at room temperature for 30 min before placing the scaffold at −20 °C for 4 h prior to overnight lyophilization. Scaffolds were cut into 2 mm long under sterilized conditions before implantation into each animal. In Vivo Studies—Spinal Cord Transection and Scaffold Implantation : Female Sprague‐Dawley rats (7–9 weeks, 200–250 g) were obtained from In Vivos Pte Ltd (Singapore). Rats were anesthetized with an intraperitoneal injection of ketamine (73 mg kg −1 ) and xylazine (7. 3 mg kg −1 ). All animals were injected with buprenorphine subcutaneously (0. 05 mg kg −1 ) before the surgery. The surgical field was shaved and cleaned with 70% ethanol and treated with betadine. The skin was incised above the thoracic level, and the muscles were moved apart to expose the vertebra at level T8–T11. A dorsal laminectomy was performed on T9–T10. Dura was cut open and 2 mm of the spinal cord was removed using fine micro scissors. A 2. 0 mm long fiber‐hydrogel scaffold was then implanted to reconnect the rostral and caudal parts of the resected spinal cord. Afterward, the dura was sutured, and a 50 µm thick PCL film was put above the spinal cord to cover the injury area. The muscles were then sutured and the skin was closed with wound clips. Animals were randomly divided into five treatment groups as presented in Table S3 in the Supporting Information. In Vivo Studies—Immunohistochemistry: In Vitro : After culturing for 3 d, neurons were fixed with 4% PFA for 30 min. After washing in 0. 1 m PBS for three times (5 min each), the cells were permeabilized in 0. 1% Triton X‐100 in 0. 1 m PBS for 15 min. Thereafter, the samples were incubated in nonspecific blocking solution (5% goat serum) for 1 h at room temperature, followed by incubation with primary antibody, mouse‐anti βIII‐Tubulin (Tuj‐1, 1:1000), overnight at 4 °C. Following that, the cells were washed and detected with Alexa Fluor 546 fluorescent secondary antibodies (1:1000) at room temperature for 2 h. The nuclei were counterstained with DAPI. Cortical neurons cultured on 2D glass coverslips were stained with Tuj‐1, Phallodin‐488 (1:500), and DAPI. For cortical neurons cultured on the scaffolds, 80 cells were quantified in each trial and 240 neurons were counted in total. For adult DRG neurons cultured on the scaffolds, 50 solitary cells from each group were imaged and quantified. Neurons were counted in terms of total length and the longest neurite length in each group. Three biological repeats were carried out for each experimental group of both cortical neuron and DRG neuron cultures. For cortical neurons cultured on 2D, 40 cells imaged from three glass coverslips for each experimental group were quantified in terms of growth cone numbers and the growth cone area per neuron. All the quantifications were done using the ImageJ software. In Vivo Studies—Immunohistochemistry: In Vivo : At 14 d postinjury, animals were perfused with 0. 9% saline followed by 4% ice‐cold PFA. After perfusion, 1. 5 cm of spinal cords containing the injury site were dissected and postfixed for 2 h before transferring to 15% sucrose for 24 h followed by 30% sucrose at 4 °C until cryosectioned. Spinal cord samples were sectioned into 20 µm thick horizontal sections and directly mounted on glass slides. The frozen sections were blocked with 10% goat serum and incubated for at least 1 h in a humidified box. The following primary antibodies were used: chicken anti‐NF200 (1:1000) and rabbit anti‐GFAP (1:1000). Samples were subsequently washed three times with PBS and incubated with the following secondary antibodies: Alexa Fluor 555‐conjugated Goat Anti‐Chicken (1:1000) and Alexa Fluor 488‐conjugated Goat Anti‐Rabbit (1:700) for 1. 5 h. Nuclear staining was performed by incubating the sections with DAPI (1:1000) at room temperature for 10 min after the secondary antibodies. All samples were finally examined using a fluorescent inverted microscope (Leica DMi8). For tissue samples, stitched images of the injury site were taken under 10x magnification. For nerve ingrowth measurements, the percentage of neurofilament area occupied in the entire scaffold was quantified. For glial scar measurement, the percent area of GFAP + signal within 250 µm from interface of the injury site was quantified. The GFAP + signals were correlated to pixel intensity of fluorescent images. The images were converted to eight‐bit and thresholded to segregate GFAP + signals and background. All images were taken under the same setting. All quantifications were done using the ImageJ software. Statistical Analyses : One‐way analysis of variance (ANOVA) and Tukey post hoc test was used when the data were normally distributed and had equal variances. For data that were not normally distributed or had unequal variances, Kruskal–Wallis and Mann–Whitney U ‐test was used for comparison between more than two groups. For comparison between two groups, Student's t ‐test was used. All values, unless mentioned otherwise, were represented as mean ± S. E. M. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201801037
| 2,018
|
Advanced Science
|
Bio‐Inspired Micropatterned Platforms Recapitulate 3D Physiological Morphologies of Bone and Dentinal Cells
|
Abstract Cells exhibit distinct 3D morphologies in vivo, and recapitulation of physiological cell morphologies in vitro is pivotal not only to elucidate many fundamental biological questions, but also to develop new approaches for tissue regeneration and drug screening. However, conventional cell culture methods in either a 2D petri dish or a 3D scaffold often lead to the loss of the physiological morphologies for many cells, such as bone cells (osteocytes) and dentinal cells (odontoblasts). Herein, a unique approach in developing a 3D extracellular matrix (ECM)‐like micropatterned synthetic matrix as a physiologically relevant 3D platform is reported to recapitulate the morphologies of osteocytes and odontoblasts in vitro. The bio‐inspired micropatterned matrix precisely mimics the hierarchic 3D nanofibrous tubular/canaliculi architecture as well as the compositions of the ECM of mineralized tissues, and is capable of controlling one single cell in a microisland of the matrix. Using this bio‐inspired 3D platform, individual bone and dental stem cells are successfully manipulated to recapitulate the physiological morphologies of osteocytes and odontoblasts in vitro, respectively. This work provides an excellent platform for an in‐depth understanding of cell–matrix interactions in 3D environments, paving the way for designing next‐generation biomaterials for tissue regeneration.
|
Cells in the body reside in a distinct 3D microenvironment – highly structured extracellular matrix (ECM) that is a nanofibrous network and possesses well‐organized hierarchic architecture ranged from nano to macro scales. When cells are removed from the in vivo microenvironment and are cultured on an artificial matrix in vitro, the cells usually cannot retain their 3D physiological morphologies. For instance, osteocytes are star‐shaped bone cells with dendritic processes extending from a small chamber called a lacuna to many minute channels called canaliculi. 1 When the osteocytes were extracted from a bone tissue and cultured on a 2D tissue culture plastic, the cells rapidly lost the characteristic dendritic shapes, leading to the changes of gene expressions. 2, 3 Similarly, an odontoblast in the pulp chamber of a tooth has a long process that extends deeply into a dentinal tubule, and the odontoblast loses its long process when cultured on a petri dish. 4, 5 Since cell morphology (or cell shape) is one of the crucial factors to regulate many biological processes, including stem cell commitment and selective differentiation, 6, 7 recapitulation of the physiological cell morphology in vitro is essential to elucidate these fundamental biological questions, leading to the development of new approaches for tissue regeneration and drug screening. To do that, a bio‐inspired 3D platform that is capable of precisely mimicking both the hierarchic architecture and the compositions of the ECM needs to be developed to accommodate the cells extracted from a tissue or organ. Bone and dentin are mineralized tissues and their ECMs have a hierarchic structure in which the well‐defined tubules/canaliculi, which have a diameter ranged from several hundred nanometers to a few micrometers, are embedded in a highly interconnected nanofibrous 3D network. 1, 8 Reconstructing such a hierarchic architecture using biodegradable materials and integrating it into a biomimetic 3D platform is a considerable challenge. Conventional approaches using synthetic hydrogels and collagen‐based matrix can form 3D fibrous network, 9, 10, 11 however; those scaffolding materials are not capable of mimicking the tubular microstructures of the ECMs of bone and dentin tissues. Consequently, to date, there are no suitable platforms that are capable of recapitulating the morphologies of osteocytes and odontoblasts in vitro. Micropatterning approaches, which include photolithography, ink‐jet printing, microcontact printing, soft lithography, and self‐assembly, are widely used to control cell–material interactions within a microdomain, 12, 13, 14, 15, 16 and is a potential tool to fabricate bio‐inspired platforms. Currently, most of the micropatterning methods are limited to fabricate 2D substrates using non‐biodegradable materials, such as polydimethylsiloxane, polyacrylamide, and polystyrene, and cannot mimic the 3D microstructure of the ECM. 17, 18, 19 More importantly, none of the micropatterned substrates can truly recapitulate the components and the hierarchic architecture of the ECMs of bone or dentin. Herein, we report a unique approach to developing a 3D ECM‐like micropatterned matrix as a physiologically relevant 3D platform to recapitulate the bone and dentin cell morphologies in vitro. The bio‐inspired micropatterned matrix precisely mimics the hierarchic 3D tubular/canaliculi architecture as well as the compositions of the ECM of mineralized tissues. In addition, the synthetic micropatterned matrix is capable of precisely controlling one single cell in a microisland, providing an excellent platform to study cell–matrix interactions. Using this bio‐inspired 3D platform, for the first time, we successfully manipulated individual bone and dental stem cells to recapitulate the physiological morphologies of osteocytes and odontoblasts, respectively. Scheme 1 illustrates the steps for the preparation of the bio‐inspired 3D platform. It started from the fabrication of a nanofibrous matrix using an electrospinning process, which is a remarkably simple, robust, and versatile technique to generate ECM‐like nanofibrous structure (Scheme 1 A). We selected methacrylate‐modified gelatin (GelMA) as the substrate biomaterials for three reasons. First, gelatin is derived from collagen (the main component of natural ECM) by hydrolysis; therefore, is an ideal biomaterial to mimic the composition of natural ECM. 20, 21 Because gelatin is a denatured protein, the denaturing hydrolysis process eliminates the potential risk of pathogens transmission associated with collagen, making it an excellent biomaterial for tissue engineering. Second, gelatin is transparent in aqueous solution and allows us to conveniently monitor cells that are embedded in the gelatin matrix using optical microscopies, including confocal laser scanning microscopy. 22, 23 Third, GelMA is chemically modified gelatin with double bonds, which facilitates chemical crosslinking with polyethylene (glycol) diacrylate (PEGDA) to create a stable cellular non‐adhesive background. After the electrospining process, the GelMA nanofibrous matrix was crosslinked with carbodiimide in a solvent mixture (acetone/water = 9 5 /05 v/v) to preserve the nanofibrous structure (Scheme 1 B). In the next step, PEGDA was cast onto the surface of the crosslinked GelMA matrix (Scheme 1 C), followed by a UV‐induced photolithography process to create a micropatterned matrix (Scheme 1 D). The alkene groups on the PEGDA and GelMA were initiated and crosslinked together to stabilize the micropattern. Finally, a computer‐aided laser ablation technology was carried out using a Leica LMD 7000 system to create a 3D nanofibrous micropatterned tubular matrix (Scheme 1 E). Scheme 1 Illustration of the fabrication of a nanofibrous micropatterned tubular 3D platform. As shown in Figure 1 A–C, nanofibrous micropatterned gelatin microislands were successfully created by combining the electrospinning and photolithography processes. The microislands were composed of gelatin nanofibers with an average diameter of approximately 200 nm, which is at the same range of collagen fibers in natural ECM. The shape and size of microislands were precisely controlled by the photomasks, and Figure 1 A shows well‐organized circular microislands with a diameter of 60 µm. The PEGDA hydrogel interpenetrated with the nanofibrous gelatin matrix and covered the substrate surfaces that surrounded the microislands. A unique laser ablation process was performed to form 3D tubular microislands, which mimic the hierarchical structures of natural dentin and bone matrix. Specifically, Figure 1 D, E is the synthetic tubular matrix with one tubule in a microisland, which had the same architecture and similar composition of natural dentin matrix (Figure 1 F). Furthermore, the size of the tubule in the synthetic tubular matrix was approximate 3 µm, which is the same as that of a natural dentin tubule. Figure 1 G, H is the synthetic tubular matrix with multiple tubules in a microisland, which resembled the architecture and composition of natural bone matrix (Figure 1 I). Figure 1 A) SEM image of a nanofibrous micropatterned gelatin matrix with circular microislands of 60 µm in diameter. B) The magnified image of (A), showing the nanofibrous microisland surrounded with cell‐repellent PEG. C) The enlarged image of (B), showing gelatin nanofibers in the microisland. D) SEM image of a 3D nanofibrous micropatterned matrix with one tubule in each microisland. E) The magnified image of (D), showing the microisland has a 3D nanofibrous tubular architecture, which precisely mimics the natural tubular structure of F) dentinal ECM. G) SEM image of a 3D nanofibrous micropatterned matrix with multiple tubules in each microisland. H) The magnified image of (G), showing the microisland has a 3D nanofibrous tubular architecture, which aims at resembling the natural structure of I) bone ECM. J) A photographical image of a micropatterned gelatin matrix, showing the matrix is transparent and easy to handle. K) The Young's moduli of the nanofibrous gelatin matrices with different crosslinking times (12 and 24 h) and with the addition of PEGDA onto the micropatterned gelatin matrix. L) The elongation at break of the gelatin nanofibrous matrix with different crosslinking time (12 and 24 h) and with the addition of PEGDA onto the micropatterned matrix ( n = 5, * p < 0. 05). We developed this innovative laser‐guided ablation approach, which is a noncontact, high‐precision, and computer programming of machining process, to introduce 3D microstructure into nanofibrous matrices. Using this approach, the tubular size, distribution, and density were precisely controlled to match those of natural dentin and bone ECMs. The laser power, laser writing speed, and pulse frequency were the major factors to control the tubular structure. Specifically, the laser power was used to control the size and the depth of the tubular pores, and the laser writing speed and laser pulse frequency were used to adjust the distance between the tubules. Small tubules (<10 µm) were regenerated with low laser power (<20 µJ), and narrow spaces between the tubules were obtained from high pulse frequencies and low writing speed (Figure S1, Supporting Information). However, when a high laser frequency or a low writing speed was used, the tubules would be interconnected and form microgrooves (Figure S1D, Supporting Information, where the writing speed = 1200 µm s −1 and the frequency = 40 Hz). Because the laser ablation approach was precisely modulated by a computer programing, the depth of the tubules could be readily controlled by the repetition of the laser ablation process. In addition, the orientation of the tubules to the matrix was conveniently adjusted via the angle between the laser and the matrix plane (Figure S2, Supporting Information). Within the range of the power scale in our experiments, the surface chemistry of the tubular matrix mostly remained intact, as indicated by the element compositions of carbon, nitrogen, and oxygen in the matrix (Figure S3, Supporting Information). The 3D tubular gelatin matrix had excellent mechanical properties (Figure 1 J). Both the Young's modulus and the elongation at break of the micropatterned matrix increased with the crosslinking time (Figure 1 K, L). More strikingly, the incorporation of PEGDA with the gelatin matrix increased the mechanical strength from 88 ± 5 to 184 ± 27 MPa, which was due to the crosslinking of the PEGDA with GelMA. While the crosslinking of the PEGDA with GelMA reduced the elasticity of the matrix, the elongation at break was still more than 21%, which was appropriate to be used as a cell culture substrate. Since the 3D tubular gelatin matrix had almost the same compositions to those of collagen (the major organic component of the ECM in bone and tooth tissues), the mechanical property of the gelatin matrix was similar to that of the decalcified bone/tooth tissues. To further increase the mechanical strength of the tubular gelatin matrix, a simulated body fluid incubation process, which was developed in our previous study, 20 can be adopted to incorporate bone‐like apatite onto the surface of the biomimetic 3D tubular gelatin matrix. We chose human dental pulp stem cells (DPSCs) as a model cell type and examined how the DPSCs interacted with the nanofibrous micropatterned matrix. The DPSCs quickly attached to the microislands after they were seeded onto the micropatterned matrix. Within 1 h, the cell started to spread in the microisland, and reached to a stable stage 24 h after cell seeding (Movie S1, Supporting Information and Figure S4, Supporting Information). Regardless of the size of the microisland, the micropatterned matrix strictly confined the DPSCs within the nanofibrous microislands, confirming the strong cell‐repellent effect of the PEG on the micropatterned matrices ( Figure 2 A, C). While the DPSC in a smaller microisland was less spreading, it had a higher cell height on the microisland (Figure 2 D). Figure 2 A–C) Typical morphologies of human DPSCs on the microislands with different sizes. A) 25, B) 40, and C) 60 µm. D) The average cell height on the microislands with different sizes. The cell heights were calculated from the z stack of the confoal images ( n = 50, * p < 0. 05). E–H) Confocal images of a DPSC cultured on the FITC‐labeled microisland (60 µm in diameter). E) Gelatin nanofibers in the microisland. F) F‐actin of the DPSC. G) The nucleus of the DPSC. H) The confocal image that merges the nanofibers, F‐actin and nucleus. I–K) Focal adhesion images of a DPSC adhesion on a microisland (60 µm in diameter). I) F‐actin of the DPSC, showing the cytoskeleton of the DPSC. J) The distribution of vinculins of the DPSC on the microisland. K) The nucleus of the DPSC. L) The confocal image that merges the F‐actin, vinculin, and nucleus. Fluorescent images were used to further examine DPSC adhesion on the bio‐inspired microislands. To make the nanofibrous microisland visible under fluorescence microscopy, the fluorescein isothiocyanate (FITC) labeled gelatin was added during preparation of the nanofibrous matrix. Since the UV light quenched the FITC molecules surrounding the microislands during the process of the photolithography, the fluorescent microislands were obtained (Figure 2 E). F‐actin, which was stained with red color to present the cytoskeleton, clearly showed the widespread morphology of the DPCS within the microisland (Figure 2 F), consistent with the SEM observation (Figure 2 A–C). Vinculin, a key protein of focal adhesion complex, is an indicator to evaluate the formation of focal adhesion. As shown in Figure 2 –L, the expression of vinculin was detected both at the edge and in the middle of the microisland, revealing the strong interaction between the DPSC and the nanofibers. The microislands with different sizes had considerably high cell occupation ratios ( Figure 3 A–F). Furthermore, the cell occupation ratio on the microislands increased with the microisland size. In addition, the addition of seeding times increased cell numbers on the microislands (Figure 3 G). For example, more than 70. 7% of the microislands with the size of 60 µm were occupied by the DPSCs when they were seeded twice. Figure 3 Confocal images of DPSCs after cultured on the microislands of different sizes for 1 week. The DPSCs were stained with phalloidin (red) to show the cells occupying the circular microislands with the diameters of A) 25, B) 40, and C) 60 µm. D–F) The DPSCs were stained with DAPI to show single or multiple cells in one microisland with the diameter of D) 25, E) 40, and F) 60 µm. G) The effects of microisland diameters and seeding times on the ratios of micropatterns occupies with cells. H) The cell number distribution in each microisland with different diameters. For (G) and (H), five region of interests (ROIs) and at least 100 microislands in each ROI were selected in each group for the analyses. The cell numbers on each microisland were controlled by the size of microislands. For instance, among the microislands resided by DPSCs, over 80. 2% of the microislands with the size of 25 µm was occupied by single cells (Figure 3 H). In contrast, only 37. 5% of the microislands with the size of 60 µm were occupied by single cells. Accordingly, the ratios of two cells and more than two cells in each microisland increased with the size of the microisland. The micropatterned matrices had excellent capability to constrain stem cells within the microislands. Moreover, even the cells were restricted to proliferate, most of the cells in the microislands stayed alive after they were cultured for 4 weeks, confirming the high biocompatibility of the micropatterned gelatin matrix (Figure S5, Supporting Information), which was difficult to achieve when using other micropatterned surfaces. 24 One of the advantages of the biomimetic micropatterned gelatin matrix was the flexibility of being functionalized with bioactive molecules. Besides having cell adhesion motif, the GelMA possesses free amino groups, which allows the GelMA microislands to readily conjugate with proteins or peptides using a simple carbodiimide crosslinking chemistry. We selected bone morphogenetic protein 2 (BMP‐2) as an example, and grafted BMP‐2 onto the nanofibers of the microislands using a one‐step process with a heterobifunctional crosslinker that contains both N ‐hydroxysuccinimide ester and maleimide groups (Scheme S1, Supporting Information). To detect the distribution of the coupled BMP‐2 on the microislands, the BMP‐2 antibody was added for immunofluorescent staining. It was shown in Figure 4 A–D that the BMP‐2 was strictly confined in the microislands. The BMP‐2 was evenly distributed in the microislands with a density of 80 ng cm −2, which could be easily modulated by the reactant concentration during the crosslinking process. The cell adhesion ratio on the microislands was enhanced after the conjugation of BMP‐2 onto microislands (Figure S6, Supporting Information). Alkaline phosphatase (ALP) is an odontogenic differentiation marker of DPSCs, and the ALP assay was performed to evaluate the bioactivity of the BMP‐2 incorporated in the microislands. As shown in Figure 4 E–G, the DPSC in the microisland exhibited a much stronger ALP expression than the control group, indicating the high bioactivity of the BMP‐2 conjugated onto the microislands. It should be noted that other proteins and peptides can also be incorporated onto the nanofibrous micropatterned matrix using the same conjugating approach. Figure 4 Conjugation of BMP‐2 onto the nanofibers of the microislands. A) FITC‐labeled nanofibers on the microisland of a diameter of 60 µm. B) The BMP‐2 (red) on the microisland. C) The merged image of (A) and (B). D) The distribution of BMP‐2 on the micropatterned gelatin matrix. The values of the relative intensity indicate that the BMP‐2 can only be detected inside the microislands. E) The typical ALP staining image of a DPSCs cultured on a microisland in an odontogenic differentiation medium for 3 d. F) The typical ALP staining image of a DPSC cultured on a BMP‐2 coupled microisland in an odontogenic differentiation medium for 3 d. G) The relative intensity of the ALP expression of the DPSCs cultured on the microislands with/without BMP‐2 conjugation ( n = 10, * p < 0. 05). Scale bar: 20 µm. The micropatterned gelatin matrix was transparent and could be further combined with cytological section techniques to obtain high‐quality images at the lateral view of cells on the microislands, which was very difficult, if not impossible, for the micropatterns prepared from other materials, such as glass, silicon, or polystyrene. Figure 5 A–C showed that all the cell bodies of the DPSCs remained on the surface of the microislands, while the short pseudopodia inserted into the nanofibrous matrix. Overall, the nanofibrous microislands were a 2D matrix and did not allow the DPSC to exhibit in vivo like cell morphology. Figure 5 A) The cross‐section overview of DPSCs cultured on a micropatterned matrix. B) The cross‐section view of a DPSC cultured on a microisland. C) The enlarged confocal image of B), showing the short pseudopodia of the DPSC inserted into the nanofibrous matrix. D) The typical SEM image of a DPSC cultured on a 3D microisland with a single tubule. E) The 3D reconstructed morphology of the DPSC on the 3D microisland. F) The morphology of odontoblasts in vivo. G) The typical SEM image of a BMSC cultured on a 3D microisland with multiple tubules. H) The 3D reconstructed morphology of the BMSC on the 3D microisland. I) The morphology of an osteocyte in vivo. J) The cellular process diameters of DPSC and BMSC on the 3D microislands, and the cellular process diameters of human odontoblast and mouse osteocyte. K) The cellular process lengths of DPSC and BMSC on the 3D microislands. To stimulate the formation of in vivo like morphologies of odontoblasts and osteoblasts, 3D nanofibrous tubular microislands were prepared and seeded with DPSCs and BMSCs. Specifically, the DPSC was seeded on a microisland that had a single tubule with a diameter of 3. 0 µm, and the BMSC was seeded on a microisland that had 28 tubules with an average diameter of 1. 8 µm. After being cultured on the biomimetic 3D microislands for 48 h, the morphologies of the DPSC and BMSC were showed in Figure 5 D, I. The DPSC was highly polarized with a long process inserting into the tubule, while the cell body displayed spherical shape aligning on the surface of the microisland (Figure 5 D, E), which resembled the 3D morphology of the odontoblast in the body (Figure 5 F). Similarly, the cell body of the BMSC attached to the surface of tubular microisland, while a number of dendrite‐like structures were inserted into the tubules of the microisland (Figure 5 G, H). Overall, the BMSC on the microisland had the same 3D morphology of the osteocyte in the body (Figure 5 I). Semiquantitative analyses further showed that the DPSC on the tubular microisland had the cellular process with a diameter of 2. 4 µm, similar to the process of human odontoblast in vivo (2. 2 µm, Figure 5 J). Also, the BMSC on the multicanaliculi microisland had the cellular dendrites with an average diameter of 0. 7 µm, similar to the dendrites of the human osteocyte (0. 5 µm, Figure 5 J). In addition, the average process lengths of the DPSC and BMSC on the microislands were 11. 5 and 7. 1 µm, respectively (Figure 5 K). It is expected that the process lengths of the DPSC and BMSC would further increase with extended culturing time and longer tubules in the micropatterned matrix. Those results clearly show the pivotal role of the biomimetic 3D platform in guiding the formation of physiological morphologies of osteocytes and odontoblasts. It should be noted that the bio‐inspired micropatterned 3D matrix is also an excellent platform to recapitulate the 3D physiological morphologies of many other types of cells, such as neuron and dendritic cells. In summary, we developed a unique approach to build bio‐inspired 3D micropatterned matrices that precisely mimic the hierarchic architecture and compositions of natural ECM. Using the ECM‐like micropatterned tubular matrices as a template, we, for the first time, recapitulated the physiological morphologies of bone and dentinal cells in vitro. The micropatterned matrix is an excellent platform for an in‐depth understanding of cell–matrix interactions in 3D environment, which will not only help elucidate many fundamental biological processes, but also guide the design of next‐generation biomaterials for tissue repair and regeneration. Experimental Section Fabrication of Nanofibrous Gelatin Matrix : GelMA and FITC‐labeled gelatin were synthesized as previously described. 25, 26 An electrospinning process (Spraybase platform, Ireland) was carried out to prepare nanofibrous gelatin matrix at room temperature. The electrospinning solution was prepared by dissolving the GelMA into the mixed solvents of hexafluoroisopropanol/acetic acid/ethyl acetate/water (5/2. 5/1. 5/1) with a concentration of 20% (w/v). A voltage of 12 kV and a feeding rate of 0. 5 mL h −1 were used to fabricate the gelatin nanofibers, which were collected onto a drum of 30 cm in diameter with a rotation rate of 80 rpm. The distance between the drum collector and the spray tip was 10 cm. After collecting a thickness of approximately 100 µm, the nanofibrous gelatin matrix was crosslinked with carbodiimide in a solvent mixture (acetone/water = 95/5 (v/v)). To visualize nanofibrous matrix under fluorescent microscopy, the FITC‐labeled gelatin (2%) was added into the GelMA solution. Fabrication of Patterned Microislands : The gelatin matrix was cut to a size of 1 cm × 1 cm and was mounted on a glass slide. Ten microliters of PEGDA aqueous solution (20%) including 1% of 2‐hydroxy‐4′‐(2‐hydroxyethoxy)‐2‐methylpropiophenone) was added to the gelatin matrix. Next, a photomask (Digidat, Inc. CA, USA) was covered on the gelatin matrix, and was exposed to a UV light with a power of 15 mW cm −2 for 1 min. The micropatterned matrix was incubated in distilled water for 1 h to remove the unreacted PEGDA, and was dehydrated in ethanol, and vacuum dried for later use. Conjugation of BMP‐2 onto the Nanofibers of the Microislands : 4‐( N ‐Maleimidomethyl) cyclohexane‐1‐carboxylic acid 3‐sulfo‐ N ‐hydroxysuccinimide ester sodium salt (Suflo‐SMCC) was used to conjugate BMP‐2 onto the gelatin nanofibers. After 4 mg of Suflo‐SMCC was dissolved in 1 mL of phosphate‐buffered saline (PBS), the gelatin matrix was added into the Suflo‐SMCC solution for 1 h at room temperature. After the activation process, the gelatin matrix was washed with PBS for three times, and incubated in 100 mL of BMP‐2 solution (50 µg mL −1 ) at 4 °C for 1 h. The resulting nanofibrous matrix was washed with PBS and air‐dried. Fabrication of 3D Nanofibrous Micropatterned Tubular in Microislands : To generate 3D tubular microislands, the micropatterned matrix was paved and dried on a glass slide (Figure S7A, Supporting Information). The microislands were visible under a LMD 7000 system (Figure S7B, Supporting Information). The array of the tubules was programmed with the software of the LMD 700 (Leica micro‐dissection V7. 5. 1). The tubules within the microislands were generated via a laser ablation process, and the sizes of the tubules were controlled by the laser power and laser frequency (Figure S7C, Supporting Information). Cell Experiments : Human dental pulp stem cells (DPSCs) were a gift from Dr. Songtao Shi, The University of Pennsylvania School of Dentistry. The cells were isolated from surgical waste (extracted human wisdom teeth) that was approved by IRB (Protocol# USC IRB #HS‐07‐00701). Informed signed consent was obtained from the volunteer. Human bone marrow stem cells (BMSCs) were purchased from Lonza. Both the DPSCs and BMSCs were cultured in an ascorbic acid‐free α‐modified essential medium (a‐MEM; GIBCO, Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS; Invitrogen) and 1% penicillin–streptomycin (Invitrogen) in a humidified incubator with 5% CO 2 at 37 °C. The culture medium was changed every 2 days, and the DPSCs and BMSCs of passages 3–5 were used for the experiments. To test cell selectivity on microislands, 2 × 10 4 of cells were seeded onto 1 cm 2 of microislands. One hour after cell seeding, the unattached cells were washed off the matrix by gently pipetting up the culture medium. When necessary, a secondary seeding process was conducted 1 h after the first seeding step. An odontogenic medium was used to test the ALP activity of the DPSCs on the microislands. 27 For F‐actin staining, the cell‐microisland construct was rinsed with Dulbecco's phosphate‐buffered saline (DPBS) and fixed with 4% paraformaldehyde in PBS for 30 min. The staining process was operated following with the manufacturer's instruction (CF633 phalloidin, Biotium, 00046), then followed by the staining of nuclei with Hoechst 33342 (1 µg mL −1 ) for 15 min and washed with PBS for three times. For immunofluorescence staining of vinculin, the samples were blocked in 5% goat serum (Gibco) for 4 h at room temperature, and were reacted with antivinculin antibody (1:150, Abcam ab129002) and CF633 phalloidin (10 U mL −1, Biotium, 00046) overnight. After being washed with PBS for three times, the samples were stained with Alexa Fluor Plus 555 secondary antibody (1:200, Invitrogen, A32732) for 2 h, followed by 1 µg mL −1 Hoechst 33342 for 20 min. The samples were washed and mounted with coverslip (CoverWell Imaging Chamber Gasket). For confocal observation, directly mounted samples were used for taking top‐view images. Three samples including at least 300 microislands were collected to measure the cell occupation ratio (the percentage number of microislands that are occupied by cells) and cellular spreading area within each microisland. To obtain high‐quality images of the lateral view of cells, the stained samples were embedded and processed with freezing microtome section. The sections with a thickness of 30 µm were harvested and immediately mounted with coverslips for confocal observations. At least 30 lateral images from three samples were collected and measured for the lengths and diameters (the diameters at the half lengths) of the cellular processes. High‐resolution images were taken using the stack mode with each step of 400 nm. The image files were exported for 3D reconstruction in Imaris 9. 0. ALP staining was operated following the manufacturer's instruction. Briefly, the samples were washed with PBS for 2 min, and fixed with 60% citrate buffered acetone for 5 min. The samples were immersed into an ALP staining solution for 30 min. After that, the samples were washed and stained with Hoechst 33342 for 20 min. The relative activity of the ALP was measured and analyzed following a previous report. 17 For SEM observation, each sample was dehydrated with graded ethanol solutions (50%, 70%, 95%, 100%, 30 min for each) and dried in a vacuum oven. The dried samples were coated with gold using a sputter coater (SPI‐module Sputter Coater Unit, SPI Supplies/Structure Probe, Inc. ) and observed under a SEM machine (JSM6010, JEOL). Statistical Analysis : Quantitative results were presented as mean ± standard deviation. The unpaired Student's t ‐test was used to test the significance between two groups, and the analysis of variance test was applied for multiple group comparisons. A value of P < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201801039
| 2,018
|
Advanced Science
|
Engineering Precision Medicine
|
Abstract Advances in genomic sequencing and bioinformatics have led to the prospect of precision medicine where therapeutics can be advised by the genetic background of individuals. For example, mapping cancer genomics has revealed numerous genes that affect the therapeutic outcome of a drug. Through materials and cell engineering, many opportunities exist for engineers to contribute to precision medicine, such as engineering biosensors for diagnosis and health status monitoring, developing smart formulations for the controlled release of drugs, programming immune cells for targeted cancer therapy, differentiating pluripotent stem cells into desired lineages, fabricating bioscaffolds that support cell growth, or constructing “organs‐on‐chips” that can screen the effects of drugs. Collective engineering efforts will help transform precision medicine into a more personalized and effective healthcare approach. As continuous progress is made in engineering techniques, more tools will be available to fully realize precision medicine's potential.
|
1 Introduction Each person responds differently to certain drugs. When prescribed with ineffective drugs, the patients may not only pay for the financial cost, but also suffer from the physiological side effects that could be catastrophic in certain circumstances. 1 This calls for the tailoring of therapies to fit each individual. Customizing healthcare services for an individual or a group of people have the potential to reduce the economic burdens and enhance quality of life. The “Precision Medicine Initiative” spearheaded by the Obama Administration proposed to apply the technological advances in genomics to improve healthcare outcomes. 2 In 1990, the “Human Genome Project, ” aimed at mapping an entire human genomic sequence, took about 13 years to complete, even with the collective efforts of scientists and engineers all around the world, and cost up to 1 billion US dollars. 3 Since then, the “1000 Genomes Project” has mapped genomic sequences in 2504 individuals from 26 different global areas and identified over 88 million common genetic variations, including single nucleotide polymorphisms, insertions/deletions, and structural variants. 4 With the tremendous progress made in sequencing techniques, genomic sequencing has become readily available to the general public, costing about $1000. 5 There is no doubt that progress in next‐generation sequencing and genome‐wide association analysis will lead to the prospect of “precision medicine” where health care plans will take individual genomic variations into consideration. 6 So far, over a million of human genomes have been sequenced in research settings, 7 and this access to personal genomic data as well as our understanding of the genetic mechanisms of diseases has fueled the current concept of “precision medicine. ” 8 Genetic profiling, once an anecdotal technique in molecular biology, has become an easily accessible tool for the general public. Commercial genomic profiling companies, such as 23andMe that use saliva as the genetic source, 9 not only provide the public with bioinformatic services but also inform people about genomic‐guided healthcare knowledge. Genomics‐based precision medicine can be applied to people of all ages, ranging from newborn infants to the elderly. 10 Oncology is currently the main target for precision medicine due to the prevalence and lethality of cancer as well as the damaging side effects of anticancer therapies. 11 Individual genetic mutations could be taken into account for predicting risk factors for cancer, guiding diagnostic tests, and designing treatments. Genomic profiling also casts light on the probability of cancer metastasis 12 and tumor relapse 13 where the doctor can take a biopsy from the patient for analysis. Precision medicine is a powerful approach and can benefit all types of cancers, and has been used in many cancers including pancreatic cancer, 14 breast cancer, 15 glioblastoma, 16 anaplastic thyroid carcinoma, 17 bladder cancer, 18 colorectal cancer, 19 biliary tract cancer, 20 and adrenocortical tumors. 21 Additionally, genetics‐based precision testing has also been incorporated into treating other illnesses such as newborn screening, 22 pediatric rheumatology, 23 cardiovascular disease, 24 diabetes, 25 hypertension, 26 allergies, 27 anaphylaxis, 28 kidney disease, 29 Parkinson's disease, 30 multiple sclerosis, 31 inflammatory bowel disease, 32 and psychological diseases like suicidality 33 and schizophrenia. 34 Genomic sequencing has the potential to provide us with the information about genetic variations between people, though our understanding of the association between genetic sequences and diseases is far from being complete. 35 In the simplified case of monogenic diseases, where a mutation in one gene causes a disease, the accuracy of diagnosis based only on genetic sequence information is unsatisfactory. 36 This is due to our poor understanding of gene regulatory mechanisms. 37 In the case of cancer, the ability of the cancer cells to mutate continuously makes it even more difficult to predict genotypes. 38 Even with an improved understanding of the correlation between genotypes and phenotypes, the origin of many diseases cannot be explained by the information we can extract from genetic sequences. While genomic profiling has been hailed as the driving force for precision medicine, surprisingly, little attention has been paid to engineering approaches. Even in the absence of genetic information, engineering can contribute to precision medicine by harnessing other aspects of personal health‐related information. For examples, without genomic profiles of a patient at hand, induced pluripotent stem cells (iPSCs) derived from the individual contain all the person's genomic information and they can be derived into various lineages to show patient‐specific responses to therapies. Without sequencing an individual's leukocyte antigens, patient‐derived immune cells could be engineered to target diseased cells without harming “self” cells of a patient. For information that cannot be predicted by genomics, such as such as heart rate, blood pressure, levels of metabolites, and biomarkers, engineered biosensors could provide accurate readout and provide a timely advice for medication. Drug delivery devices and regenerative therapies based on the engineering of smart biomaterial can respond differently to an individual's physiological traits for realizing personalized therapies. In this review, we will discuss engineering approaches that consider both genetic and nongenetic factors to provide higher treatment precision ( Figure 1 ). More specifically, we will describe personalized biomaterials that could provide scaffolds for artificial tissue growth and tools for surgical intervention, biomaterial‐based drug delivery systems that could provide on‐demand drug release by integrating sensing and drug release in a closed‐loop system, wearable medical devices that could monitor the physiological status of a patient in real time, engineered immune cells that could directly harness the internal defense system of a patient to fight diseases, stem cells that could be developed using patient‐sourced cells to provide patient‐specific cells or organs, and “organs‐on‐chips” with patient‐derived cells that could provide direct information about individual responses to prescribed therapies. Furthermore, we will discuss emerging technologies that could further contribute to precision medicine and the challenges that need to be addressed for this technology to reach its full potential. Figure 1 Engineering precision medicine. Biomaterials engineering, cell engineering, organs‐on‐chips, personalized implants, and personalized devices together with genomics‐based methods to enable precision medicine. 2 Engineering Precision Medicine 2. 1 Engineering Cells for Precision Medicine 2. 1. 1 Immune Cell Engineering for Cancer Therapy Human immune system maintains the health of the body by fendering out exogenous pathogens as well as clearing out endogenously failed cells. There have been century‐long efforts to harness immune systems for human health, and vaccination has been a standard healthcare method in the last few decades. 39 Recent breakthroughs in cancer immunotherapy, like discovering immune checkpoints and engineering chimeric antigen receptor (CAR)‐T cells, 40 have reinvigorated the field of oncology. Immune therapies harnessing patients' immune systems, including phenotypically activating or genetically engineering autologous immune cells, provide another approach for tailoring precision medicine. Vaccines achieve precision medicine by using a fragment of disease‐relevant peptide to train immune cells, especially B cell and T cells, to recognize the target cells. The long‐recognized potency of vaccines in preventing infectious diseases has inspired physicians to acquire a cancer‐specific vaccine. 41 The way of training the patients' own immune cells to recognize cancers using vaccines is straightforward, but there exist many more hurdles for cancer vaccines than the ones for targeting viruses due to the fact that cancer cells have many antigens similar to normal cells. 42 Mutated proteins which are characteristics of specific cancers, namely neoantigens, are the rare markers that could distinguish one cancer from another. 43 Key to the vaccination strategy is the identification of patient‐specific cancer neoantigens, and promising results have already been attained for several skin associated cancers. 44 An emerging strategy to detour the tremendous efforts for neoantigen screening, however, is to instead use the whole cell lysate of the patient‐derived tumors for stimulating the immune system. 45 Training a patient's immune system to fight cancer is a multistep biological process where the therapeutic efficacy is difficult to predict. In comparison, synthetic biology approaches can be used to directly reprogram a patient's immune cells to recognize a signature antigen on cancer cell surfaces. This provides a faster and more controllable way of activating immune systems. 46 T cells can directly kill cancer cells when the T cells and the costimulatory receptors recognize the antigens on target cell membranes. 47 A simplified version of the T cell receptors—CAR—was engineered into patient‐derived T cells to recognize a predefined cancer antigen in that patient. In an improved construct, Eyquem et al. inserted the CAR‐encoding DNA fragment to the genomic locus of T cell receptor α constant (TRAC) with the assistance of CRISPR‐Cas9 ( Figure 2 a), leading to much higher potency than viral vector‐constructed counterparts (Figure 2 b). Robust anticancer efficacy of the CAR‐T cells has been demonstrated that significantly improved the survival of treat mice (Figure 2 c). 48 In 2017, the first two CAR‐T therapies, Yescarta and Kymriah, that target CD‐19 of B‐cell leukemias, were approved. 49 Hundreds of more CAR‐T cells are under development for many cancers, 50 among which both liquid hematological cancers and solid tumors, such as pancreatic cancer, ovarian cancer, glioblastoma, neuroblastoma, are included. 51 However, the immunosuppressive tumor microenvironment, strong stroma barrier, and genotype mutations of the cancer cells in solid tumors made them more difficult to be killed. Figure 2 Engineering CAR‐T cells using CRISPR‐Cas9 for precision cancer immunotherapy. a) The CD19 CAR was precisely inserted into the TRAC locus by CRISPR‐Cas9‐mediated genome editing. The insertion also caused the knockout of T cell receptors, delaying T cell exhaustion. b) Quantification of T cell exhaustion markers expression on the engineered T cell. Insertion of CAR into the T cell receptors locus (TRAC‐1928z) significantly reduced the expression of T cell exhaustion markers, showing huge advantage over non‐targeted construction approaches. c) The 2‐in‐1 engineering strategy generated CAR‐T cells with more robust anticancer efficacy, increasing the survival of treated mice. Reproduced with permission. 48 Copyright 2017, Springer Nature. The potency of CAR‐T therapy toward blood cancer cells is accompanied by safety concerns, such as off‐target binding to healthy cells or the occurrence of cytokine storms, that could be lethal in some cases. 52 Thus, more engineering approaches are needed to precisely control the activity of injected CAR‐T cells, like turning on CAR signaling with external signals 53 or causing the suicide of the CAR‐T cells in the case of cytokine storm. 54 The success of the engineered T cells also inspired arming nonimmune cells, such as human embryonic kidney cells, with cancer sensors and killing effectors for targeted tumor therapy. 55 This nonimmune cell engineering strategy could reduce the concern for unexpected immune cell activation. By obviating the necessity of genetically engineering T cells, immune checkpoint inhibitors can directly activate a patient's T cells for fighting cancer. Immune checkpoint molecules are expressed on immune cells as tolerance regulators to minimize the potential damage to normal tissues. 56 This pathway becomes hijacked by cancer cells or other pathogens to evade T cell‐mediated immune surveillance or even “exhaust” T cells into nonfunctional states. 57 Programmed cell death protein 1 (PD1) and cytotoxic T lymphocyte antigen 4 are two well‐characterized immune checkpoints where antibodies that target these pathways have demonstrated robust therapeutic efficacy in treating multiple types of cancer. Several PD1 or Programmed Death‐Ligand 1 inhibitors, such as avelumab, 58 nivolumab, 59 atezolizumab, 60 pembrolizumab, 61 and durvalumab, 62 have received accelerated approval from FDA for treating melanoma, non‐small‐cell lung carcinoma, head and neck squamous cell carcinoma, bladder cancer, and metastatic urothelial carcinoma. 63 This approach has the advantage of harnessing endogenous T cells in their natural state, avoiding the time consuming and expensive process of engineering T cells. In terms of precision medicine, typing the dominant immune checkpoint pathways in patients' tumors could help guide oncologists to decide the optimal inhibitor for each patient. The discovery of new immune checkpoint pathways will thereby further expand this therapy, helping more patients with precision. 2. 1. 2 Stem Cell Engineering for Precision Medicine Human pluripotent stem cells derived from embryos can self‐replicate in vitro indefinitely and differentiate into almost any desirable cell types. 64 Due to the ethical and availability issues associated with using embryonic stem cells, iPSCs are used as an alternative. iPSCs provide a more accessible source of stem cells for regenerative medicine 65 and could generate cell replacement/regeneration therapies. Stupp and co‐workers demonstrated a scaffold based on self‐assembled peptides to mimic the unidirectional structure of muscle fiber ( Figure 3 a). 66 With growth factors and muscle stem cells encapsulated in the fibers, the stem cell therapy was delivered by a retracting injection system (Figure 3 b). The peptide base scaffold significantly improved the engraftment efficacy of muscle stems cells (Figure 3 c). Furthermore, iPSCs could also be used to build a “disease‐in‐a‐dish” model for disease modeling or drug screening. 67 Since the iPSCs can be obtained from patient‐derived cells, such as skin or blood, they contain the potential of enabling personalized therapies or disease models with exactly the same genomic background as the patient. With the advances in recent genome editing technologies, cells with genetically desirable phenotypes, such as the ones expressing desired biomarkers, could be easily generated. 68 Personalized cells with rationally engineered genetic traits provide a versatile platform for generating precision medicine therapies. Figure 3 Bioscaffold‐mediated stem cell delivery for muscle regeneration. a) Nanofibers self‐assembled from amphiphilic peptides were designed to encapsulate muscle stem cells and growth factors to facilitate muscle regeneration. b) Injection of the stem cell‐loaded scaffold into the target location using a delivery device. c) Muscle stem cells delivered by the injectable scaffold showed more robust proliferation. Reproduced with permission. 66 Copyright 2017, National Academy of Sciences. Stem cell‐based precision medicine could be attained through the following scenarios: 1) healthy cells from patients with failed organs could be reprogrammed into iPSCs, and the iPSCs redifferentiate into desired cell types, such as cardiomyocytes 69 or neurons, 70 for refilling the degenerated organs. For example, iPSC‐derived eye‐associated cells (retinal pigment epithelial) were among the first that was translated to the clinical trials. 71 2) iPSC with patient‐derived genetic defects, like monogenic disorders, can be genetically corrected by gene editing technologies. The engineered stem cells could then be used as the source for growing healthy cells for cell replacement or repair therapy. 72 3) iPSCs with patient‐specific genotypes can be directly differentiated into the disease‐associated cell types for disease modeling. This application can be useful in cases where the cause of the disease is unclear, and direct screening for an effective therapy is not possible. However, the iPSC‐based engineering approach could be time‐consuming. Recent developments in cell reprogramming enabled the iPSC stage to be skipped and had one type of somatic cells directly converted into another type of somatic cells. This process is known as transdifferentiation. 73 This transdifferentiation approach has been applied to convert fibroblasts into other cell types including cardiomycytes, 74 neurons, 75 or hepatocytes. 76 Nonetheless, skipping the iPSC stage also skipped its proliferation capabilities, making scaling‐up difficult. Stem cells can be delivered in the form of scaffold‐free cells or with the assistance polymeric scaffolds. Soft biocompatible materials are generally used as scaffolds, providing a niche to mimic the mechanical properties of the natural extracellular matrix (ECM). These scaffolds can be integrated with signaling cues to regulate the proliferation, differentiation, and migration of the cells in a way similar to tissue engineering approaches. 77 2. 2 Engineering Precision Biomaterials for Tissue Engineering and Drug Delivery Advances in biomaterials contribute to precision medicine by providing a toolbox of biocompatible materials that can interact with cells and tissues in a predictable manner. 78 Studies of material–cell interactions provide us with knowledge about the effects of the environment on cells as well as give us the tools to engineer tissue scaffolds, medical devices, and therapeutic delivery carriers. 79 Precision biomaterials are engineered with specific mechanical and biochemical properties to enable the design of personalized implants or organ replacements. For example, shear‐thinning polymers could be easily injected and conform to the shape of a patient's cavity to form a patient size specific implant; 80 enzyme or pH degradable materials can be used in manufacturing bioscaffold that will be resorbed by the patient in a rate specific to the patient's physiological environment; decellularized ECM from a patient‐derived tissue provides a highly biocompatible as well as size‐fit scaffold. To build complex smart devices for biomedical engineering, precision biomaterials need to be engineered with the ability to execute numerous unit operations, such as “separator” where the material can specifically attach to molecules or cells for separation, “sensors” where the material can detect specific analytes or electrochemical signals, “responders” where the material can change its morphology or get degraded to release the payload, “controllers” where the material can affect local microenvironment for changing the behavior of the cells, “processors” where the material allows high throughput analysis. 81 Progress in polymer chemistry has improved our control over polymerization, and the versatility in polymer chemical composition tuning has generated a large library of biocompatible materials, such as polyethylene glycol, polycaprolactone, poly (N‐isopropylacrylamide), poly(glycolic acid), poly(lactic acid), and poly(lactic acid‐co‐glycolic acid). Other than synthetic materials, the modification of natural compounds, ranging from proteins to polysaccharides, has resulted in many highly biocompatible materials. 82 For example, gelatin methacryloyl is a widely used material that mimics many properties of the ECM, including protease‐assisted degradability, tunable mechanical strength, and easy functionalization, to support cell attachment, proliferation, and migration. 83 Other well‐characterized natural polymers include elastin, collagen, chitin, chitosan, alginate, and hyaluronic acid. 84 Additionally, these naturally derived materials also have the advantage of showing minimal inflammatory responses even after long‐term implantation. Development in bioconjugate chemistry has facilitated our ability to synergize the desired properties of different materials, such as the modification of nonadherent polymeric scaffolds with the RGD peptide for cell attachment 85 and inclusion of various growth factors or morphogen for directing cell growth. 86 It has been shown that tuning the physical or biochemical properties of biomaterials, such as matrix stiffness, topology, or the number and type of adhesive ligands, could direct lineage specification of stem cells. 87 By controlling these parameters, engineered biomaterials can simulate the ECM of different organs and could help the growth of respective artificial organs. For example, by using a collagen‐based scaffold, an autologous bladder has already been created using patient‐derived biopsies for transplantation. 88 Furthermore, by utilizing an injectable microporous scaffold that has a degradation rate determined by the wound‐environment‐associated metalloprotease of individual patients, patient endogenous cells have been recruited to wound sites for tissue regeneration ( Figure 4 ). 89 The porous scaffold was assembled from microparticles prepared by microfluidics, where metalloprotease‐degradable cross‐linkers were incorporated (Figure 4 a). Robust cell growth and migration could be observed within the porous structure (Figure 4 b) and rapid tissue regeneration for wound‐bed closure was observed in a mouse model (Figure 4 c). Figure 4 A porous and degradable scaffold for personalized wound treatment. a) Polymeric microparticles were prepared by microfluidics, incorporating a metalloprotease‐degradable peptide as cross‐linkers. The water‐in‐oil approach segmented the monomers and crosslinkers into monodispersed droplets, which were then crosslinked inside the droplets. b) The microparticles were further purified and annealed with FXIIIa to form porous scaffolds that can support cell growth. c) The induction of in vivo wound healing by the engineered scaffold for 7 days. Shown are the representative wound healing images from Balb/c mice. The microparticle‐based porous scaffold induced much faster wound healing than other materials. Reproduced with permission. 89 Copyright 2015, Springer Nature. In addition to supporting the growth of artificial tissues, biomaterials alone can be tailored for direct in vivo tissue interventions. For instance, an angiographic catheter made of textured nylon has been used to vascularize a subcutaneous space for human islet transplantation. 90 Although the biomaterials were removed during the transplantation, preconditioning of the transplantation site with the biomaterials improved the viability of the islet cells. In another example, a shear‐thinning material tailored from gelatin and inorganic nanoparticles was demonstrated to intervene in blood flow. 91 Delivered by a catheter, the hybrid biomaterial was used to achieve a minimally invasive method of targeted vascular embolization. Another aspect through which biomaterials could help supplement precision medicine is through the engineering of smart drug delivery systems. Drug delivery devices could improve the “precision” of precision medicine through pharmacokinetics. 92 Ideally, customized drug delivery systems could deliver the optimal dose of the drug to the intended organs at the specific time. 93 Engineering smart biomaterials for drug delivery systems that can sense the needs of an individual's physiological state and adjust its drug release profile accordingly is an attractive strategy for formulating precision pharmacokinetics. 94 Nanomaterial‐based drug delivery systems could penetrate multiple physiological barriers and provide a simple strategy for the targeted delivery of therapeutics to diseased sites. 95 Both synthetic materials, such as polymers, lipids, inorganic salts, and metals, or natural materials, such as proteins, nucleic acids, or cell membranes, 96 can be prepared as nanoscaffolds for this purpose, and with more advanced sensors and actuators, nanoscopic smart drug delivery systems could also be further engineered. 97 In addition, advances in surface chemistry have brought about nanoscale drug delivery systems that can evade immune surveillance. Armed with targeting ligands, such as antibodies, peptides, aptamers, and small molecules, nanoscopic drug carriers could get continually delivered to desired tissues after extended in vivo circulation. 98 With the availability of a pool of effective and specific antibodies for targeting various plasma‐membrane associated antigens, profiling targetable antigens on diseased sites, like tumors, allows nanoparticle based drug delivery systems (nanomedicine) to be customized for each patient. 99 To ensure that the loaded drug will only be released at the targeted site, the nanocarriers will need to hold the drug until they have reached the designed destination. To achieve this, physiological characteristics associated with the disease, such as pH change, overexpressed hydrolytic enzymes, increased reducing or oxidative environments, or the variations in oxygen levels, could be used as cues for the release of the drugs. 100 Specific drug release could also be controlled using external physical signals, such as ultrasound, magnetic fields, electric fields, or radio‐waves. 101 These cues could be used either individually or synergistically to improve the accuracy and timeliness that will efficiently deliver drugs to targeted tumors. With the progress already made in conjunction with a comprehensive list of stimuli‐responsive materials, a variety of smart drug delivery systems can be constructed. 102 The emerging field of engineering nanoscale theranostics that integrate diagnostic and therapeutic modules into one nanoformulation provides real‐time visualization of drug delivery processes. 103 This combined approach allows for the visualized optimization of nanomedicine therapeutic efficacy that can then be applied into the different physiology of each patient. When nanocarriers are combined with gene therapies, such as plasmid DNA, mRNA, iRNA, or miRNA, the nanomaterial formulation could be tailored to meet the individual needs of the patient based on their specific genetics. 104 Besides systemically administered formulations that utilize targeting ligands for achieving “precision, ” local delivery has the advantage of directly acting on the desired tissue. 105 For chronic diseases that have a constant requirement of a certain level of drug in the blood, administering repeated drug doses to the patient is important for maintaining the drug level within the therapeutic window to enhance the effectiveness of the applied medicine. Drug release depots may therefore be required for continuous local drug delivery, which would alleviate this issue by maintaining a therapeutically effective dosage over an extended time frame. 106 In terms of “precision, ” closed‐loop drug delivery systems that monitor the level of a physiological signal, such as the concentration of a molecule, and control the release of a drug accordingly fit this model. 107 Gu and co‐workers demonstrated a “Closed‐loop” device that uses integrated sensors for sensing blood glucose levels and actuated insulin drug release from stored depots that are notable “precise dosing” devices ( Figure 5 a). 108 The formulation was delivered in the format of a microneedles patch (Figure 5 b) and a glucose level‐dependent release of insulin release could be observed both in vitro and in vivo (Figure 5 c, d). Besides blood glucose monitoring, this “closed‐loop” strategy is highly desirable in treating other diseases that need real‐time regulation of the drug dosage. Figure 5 A closed‐loop drug delivery system that enables precise dosing of insulin according to the blood sugar level of the patient. a) Schematics of the drug delivery device; polymeric micelles that could release insulin in response to blood sugar levels were formulated into microneedles for painless delivery. b) Fluorescent microscope image of the microneedle patch. c) Glucose‐responsive insulin release from the microneedle patch in vitro. d) The insulin delivery patch kept blood glucose levels within a normal range for an extended range of time. Reproduced with permission. 108 Copyright 2015, National Academy of Sciences. 2. 3 Smart Diagnostic Devices 2. 3. 1 Point of Care Devices Rapid analysis of biochemical markers in an individual enables precision medicine by providing rapid readout of physiological markers. Advances in smart sensors enable “point‐of‐care” (POC) devices that allow analyzing our health status at any time and at any place. 109 It is defined as medical diagnosis performed at the site of care, i. e. , at the time and place of patient care. Fingerstick blood test is a simple example of POC devices, 110 they are designed either to give a digital result for pregnancy test or a numerical measurement for glucose concentration of the blood. 111 The global POC device market is expected to grow from US$ 23. 16 in 2016 to US$ 36. 96 billion in 2021 at the compound annual growth rate of 9. 8% from 2016 to 2021. 112 Large diagnostic facilities in hospitals, including imaging techniques such as ultrasound, 113 magnetic resonance imaging, 114 and computed tomography (CT), 115 provide valuable information about the status of organs in the patients. However, the equipment are large, and have a high time‐cost burden and require trained personnel. POC devices show significant advantages in spatial flexibility over the heavy medical equipment. Besides, POC testing has a much faster turnaround time. The spatial flexibility and timeliness of POC testing allow immediate diagnosis of diseases by rapid detection of patient‐derived samples, such as saliva, urine, or blood, and enable much quicker medical decisions. 112 In such a way, diseases can be diagnosed and treated at very early stage. It reduces the possibility of further deterioration of the pathological condition, which in certain cases may lead to fatal risks such as heart failure, stroke, and cancer. POC devices are in huge demand from patients with chronic diseases and people whose physical condition requires inspection on a regular basis, such as for the elderly and postoperative patients. Furthermore, POC devices are labeled as more user friendly. This is of great importance for disease diagnostics in developing and underdeveloped nations where healthcare services are inadequate. Current commercial POC devices are mostly designed to analyze blood or urine. 116 The recent advances in flexible electronics, bioelectronics, and biosensors are extending the application of POC devices toward testing of biomarkers in sweat, tear, saliva, and interstitial fluid using electrochemical and microfluidic sensors; mechanical movements of the human body using piezoelectric devices; physiological signals such as electrocardiography, electromyography, and electroencephalography using flexible electronics. Moreover, the emerging technologies in biocompatible sensors are rapidly enabling POC testing of diseases directly inside our body by using ingestible POC devices that are composed of fully biocompatible materials. With advances in these technologies, we can vision more revolutionary POC devices to be developed. 2. 3. 2 Wearable Devices In order to know the biochemical characteristics of each individual for precision medicine treatments, real‐time sensing is necessary. Readily available and applicable sensing devices not only improve the capability of physicians to precisely diagnose a disease, but also increase the participation of patients where the patient could easily monitor their own health. 117 In particular, the field of emerging wearable sensors allows people to monitor their health and know their specific healthcare needs in real time. 118 Portable and disposable sensors that can track physiological signals without the assistance of medical experts are thus highly desirable. Rapid monitoring of a patient's condition with simple devices provides valuable data that could be used for accurate evaluation of the patient's health status. 119 More recently, wearable sensors and flexible electronics have shown great potential in making real‐time health monitoring devices. Similar to commercially available smartwatches that contain heart rate sensors as a standard configuration, wearable electronic devices that can monitor sweat biomarkers are emerging as key players in this realm. Gao et al. developed a sensor array that could be worn as wristband or headband to monitor metabolites in the sweat ( Figure 6 a). 120 The integration of sensors and wireless communication component enabled real‐time monitoring of perspiration‐related health information on mobile phones (Figure 6 b). In a proof‐of‐concept study, the wearable sensors provide real‐time monitoring of a person's hydration status (Figure 6 c). There are many other devices that integrate several sensors and can monitor multiple physiological signals in real time to provide a better perspective about the status of human body. 121 Figure 6 A wearable device integrated with multiple sensors to enable precise monitoring of the physiological status of an individual. a) The sensor can be worn on the wrist or forehead of an individual to monitor physical activities. b) The wearable band is designed to monitor glucose, lactose, sodium, and potassium levels in sweat as well as body temperature in real time. c) A proof‐of‐principle study to monitor subject's dehydration status from physical activities. Reproduced with permission. 120 Copyright 2016, Springer Nature. With the integration of electrochemical sensors, wearable devices could detect health‐related biomarkers. These devices provide a noninvasive method that extracts health‐related information from bodily fluids. For example, monitoring biomarkers in interstitial fluid (ISF) gains comparable information with respect to blood monitoring because key biomarkers reflecting body condition such as glucose, ion concentrations (such as Na +, K + ), proteins in the ISF are very similar to blood. Hence, in recent years, developing wearable sensors monitoring biomarkers in ISF has attracted considerable attention. It has the potential to replace conventional sampling from blood in a minimally invasive manner by using microneedles or iontophoresis electrodes. 122 Saliva also contains a variety of biomarkers, including malondialdehyde, vitamin C, and proteomes, which are related to oxidative stress and thus have been used as biospecimen for the reflection of diseases such as autism, Alzheimer's disease, Parkinson's disease, atherosclerosis, heart failure, and cancer. 123 Continuous monitoring of ionized calcium and pH of sweat using a wearable sensor provides essential information on human metabolism and minerals homeostasis. 124 Proteins, salts, enzymes, and other chemical composition in the tear reflect the eye‐related diseases such as dye eye and systemic disorders of the human body. Tumor‐related biomarkers are also recently found in tear which can be used to predict breast cancer. Recording temperature, PH, and oxygen concentration in the wounded skin allows us to gain significant information on the wound healing process and helps the treatment of chronic wound by developing smart bandages that integrate with drug delivery systems. 125 In addition, physiological signals can be detected in a noninvasive manner by attaching “skin‐sensors” onto the human body. 126 For example, blood pressure is measurable by developing wearable sensors containing piezoelectric materials which are able to produce an electric signal when they are placed under mechanical stress. The pulses of the blood can hence induce a voltage signal in the piezoelectric materials and subsequently be collected by the sensor. 127 Similar piezoelectric sensors have been used to detect heart rate, breath content, and other body motions. 128 Recording electrocardiogram (ECG), electromyogram, electroencephalogram (EEG) signals can be realized by developing sensors that incorporate ionic gel electrodes or ion‐to‐electron transducers such as conducting polymers. 129 These sensors are able to detect ionic electrophysiological signals of the human body and convert these ionic signals to processable electronic signals. 130 Their low skin‐contact impedance and ability to amplify the weak ionic current of the physiological signals allow recording of the physiological signals such as ECG and EEG with high signal to noise ratios. 129, 131 Devices that monitor glucose levels from sweat and use this information to trigger insulin release are available for diabetic patients. 132 In this instance, however, the poor correlation between blood and sweat glucose levels could potentially limit the translation of this technology to skin sensors. Changes in sweat blood glucose levels have been shown to have 10–20 min delays. 133 Similar delays may also be observed in other types of secreted body fluids, which can be potentially dangerous for diabetic treatments. In addition, the secretion mechanisms of many other biomolecules, such as proteins, peptides, or hormones, into body fluids are still unclear. 134 Nevertheless, a promising solution to improve sensing accuracy is to combine these sensors with minimally invasive devices like microneedles, so that the resulting devices can directly monitor biomarkers contained in blood. Microneedles, which create small pores on the skin that cause little to no pain, provide a platform to incorporate interstitial fluid and blood biomarker‐targeted sensors. 135 Furthermore, sensors integrated onto minimally invasive surgical tools for gathering internal tissue related signals could enhance the precision of surgical operations. For example, a sensor on a biopsy needle measuring young's modulus has been employed to help surgeons to monitor the mechanical properties of the tissues around the needle in real time. 136 This way, surgery precision was improved by distinguishing between the mechanical properties of normal tissues from diseased ones, such as tumors. Personal health data can be collected by these aforementioned sensors continuously. Intelligent devices like smart phones can be employed to learn the biomolecular patterns and lifestyles of the patients through wireless communication. Personalized healthcare advice or intervention could then be generated based on this information. With advances in sensing techniques, we expect that the sensor platform can not only monitor the physical and biochemical signals from the human body but can also analyze the data collected through the sensor so that it is possible to alert individuals about their psychological ailments, such as anxiety, stress, or depression in real‐time. 137 In the future, a smartwatch may not only remind one to take the medicine, but also tell the person to calm down when one feels nervous (while giving a talk to a large crowd of people, for example) and directly alert medical staff when fatal signals, such as heart attack, are received. 2. 4 3D‐Printing Facilitated Precision Tissue Engineering As 3D printing technology advances, engineering medical implants and artificial organs with patient‐specific spatial architecture has emerged as an approach for precision medicine. 138 Like inkjet printers, 3D printers have simplified the process of designing and manufacturing products using a range of different materials as the ink, such as polymers, ceramics, or metals. 139 The 3D printing technology has been widely adopted into producing low volume end‐using parts. Numerous additive manufacturing methods, such as material jetting, binder jetting, material extrusion, vat photopolymerization, sheet lamination, powder bed fusion, and direct energy depostion, have been adopted into the production of medical implants. 140 With patient‐derived anatomy information and clinically relevant biomaterials, implants with patient‐specific shape and size could be precisely manufactured by 3D printing. 141 Customized implants that perfectly fit the defect sites of the patients can significantly shorten the time for surgical operation, reducing the need for adding fillers or removing healthy tissues. Calcium phosphate, bioactive glasses, and metals are commonly used “ink” in 3D printing of orthopedic implants. 142 The 3D printed hips with customized size and sufficient mechanical strength after printing are still strong after a decade of implantation. The 3D printing technique is also handy in tailoring replacements for craniofacial defects, which often cause physiological and psychological pains to the patient. Reconstituting the complex 3D structure of the damaged craniofacial regions needs to be done with high structural precision due to the consideration of aesthetic outcomes. 143 Besides hard material based implants, 3D printing is also powerful in manufacturing soft tissue implants, such as for lung and heart therapy. With computer tomographic image of a patient's airway, an artificial trachea splint has been laser‐printed with bioresorbable polymer. 144 The customized implant has been used in treating tracheobronchomalacia in newborns, where continuous growth of the child requires the implant to be fit as well as elastic. Using a highly haemocompatible elastomeric material (silicone/polyurethane) as the ink, a soft occluder with CT‐imaged acquired patient‐specific size has been printed for the left atrial appendage. 145 With cells included into the printing inks, 3D bioprinting emerged and adopted the advantages of the geometrical precision in conventional 3D printing. The technique of growing differentiated cells in an organized 3D manner is paving the way for bringing replacement therapies for various tissues types, such as cardiac tissue. 146 The prospect of growing tissues with patient‐derived cells could mitigate the concern of immune rejection, expanding the pool of available tissues for organ transplantation. 147 Computer‐aided 3D bioprinting has facilitated the construction of complex tissue structures through layer‐by‐layer deposition of cell‐laden scaffolds or free cells. 148 Similar printing techniques, such as inkjet, laser, extrusion, acoustic, and stereolithography, were used in bioprinting. Biomaterials with favorable printing properties, such as shear‐thinning, facile gelation after printing, and appropriate yield stress, are preferred as bioinks. In general, 3D bioprinters have the ability to inject printable biomaterials, or even cells in the form of sheets or spheroids, with micrometer precision. 149 Using a custom‐made 3D bioprinter, Lind et al. printed multiple materials sequentially ( Figure 7 a) and created devices that i) guided the alignment of cardiac cells and ii) generated sensors that can monitor the contraction of the cardiac tissue (Figure 7 b, c). 150 The 3D prototypes with elaborate internal and external structures can be rapidly built with controllable shape, porosity, and mechanical properties. 151 With the assistance of biomedical imaging, personalized tissues with customized size and shape were printed. 152 However, immune rejection of the artificial tissue is still a concern for these scaffolded tissue structures. The incorporation of patient‐derived starting materials, such as autologous cells 148 or growth factors, 153 into the bioinks helps reduce the rejection and guides printed tissues a step closer toward realizing precision medicine. Figure 7 A 3D printed cardiac micro‐physiological system integrated with 3D printed sensors. a) The automated printing of multiple materials in 7 steps to generate a platform that integrated both cell constructs and sensors. b) The contraction of the anisotropic tissue leads to deflection of the cantilever, where an electrical signal was detected. c) The optimization of the microfabricated grooves for aligning cardiac cells. Reproduced with permission. 150 Copyright 2018, Springer Nature. Injectable biomaterial scaffolds are the fundamental building blocks for creating 3D replacement for regenerative therapy. 154 Many investigations have studied the physical and biochemical effects of the scaffolds on the behavior of the cells, including stem cell differentiation, migration, or viability. 155 In the ideal case, the implanted biomaterial scaffold alone is able to attract endogenous resources from the patient to regenerate the tissue. Besides synthetic materials, natural materials such as the ECM residue resulting after decellularization of the tissue provide a structurally authentic scaffold for tissue regeneration. These decellularized scaffolds maintain their shape, size, chemical, and biological characteristics of the original tissue, which is very challenging for synthetic scaffolds to recapitulate. 156 Another challenge for engineering thick tissues is vascularization, which supplies nutrients and oxygen to the cells and removes waste from the tissues. 157 Microfabrication techniques using microfluidics are capable of generating controllable microscopic structures, making microfluidics a versatile tool for tuning the engineered tissues. 158 Hollow microfibers or hydrogels prepared with soft lithography could be lined with vascular endothelial cells on the inner wall to simulate the native vasculature. 159 Similarly, 3D printing can be integrated with a glass filament‐mediated molding to generate in vitro vasculature. 160 Additionally, 3D printing offers solutions to this challenge through either printing porous acellular scaffolds or forming vasculature by depositing layers of different stable materials like fugitive. 161 Currently, the prospect of engineering entire functional organs, such as lungs or hearts, is not yet realizable. Some architecturally simple tissues, like skin, blood vessels, cartilage, or bone, are much easier for tissue engineering to replicate. 162 With patient‐derived stem cells and somatic cells, the engineered tissues can be personalized, taking yet another step toward tissue level precision medicine. Furthermore, quality control over the risks of genetic mutations, tumorigenesis, or functional instability still needs to be realized in place for translating engineered tissues to patients. 2. 5 Organs‐on‐Chips for Precision Drug Screening Although genomics helps determine the optimal anticancer drug of choice for each individual and there have been exceptional reports along these lines, 163 pairing a genetic mutation with a drug could only occur in 2–6. 4% of the patients. 164 Information about the origin of a cancer by studying its native tissues could provide more direct information related to its unique characteristics. 165 Expanding patient‐derived cancer tissues in vitro and then directly testing their responses to available anticancer therapies provide an alternative method to optimize precision anticancer prescriptions. The key to this approach is that the “avatars” need to capture the physiological traits of patient's tissue in vitro based on their histological structures and gene expression profiles. 166 Culturing 2D established cancer cell lines is a routine method in preclinical anticancer drug development. However, the cell lines cannot faithfully capture the genotypes of a specific patient's cancer, as the heterogeneity of the cell line differs from the that of the tumor, 167 and the 2D cellular response to the tested drug is very different from the actual scenario in vivo. 168 By comparison, there are many advantages to the 3D organoid‐based approaches: only a small amount of patients' tissue is needed to generate the required construct in vitro, 169 rare tumor types specific to the patient can be recreated, cell heterogeneity of the tumor is captured, drug resistance behavior similar to the tumor is simulated, toxicity and genetic profiling analyses could be performed on the cultured tumor mass, and effective drugs can be prescreened for patients. 170 For a typical mice‐based tumor xenograft method, the organoid based approach is more suitable for high‐throughput drug screening due to its low cost, rapid in vitro tumor growth, and avoiding ethical issues regarding animal use. 171 Biopsies from liver cancer, 172 gastrointestinal cancer, 173 head and neck squamous cell carcinoma, 174 ductal pancreatic cancer, 175 and prostate cancer 176 have been used to create organoids in vitro. Genetic profiling has shown the recapitulation of the source cancer tissue, a viable strategy for the screening of anticancer drug candidates. It has been demonstrated that conserving the heterogeneity of the tumor microenvironment is a key factor to accurately predicting the efficacy of a drug. From an engineering perspective, the “Organs‐on‐a‐Chip” approach, integrating organoid culture and organoid behavior monitoring into one system, can enable high throughput drug testing platforms for precision medicine to be developed. For example, Zhang et al. demonstrated an organs‐on‐chips system that integrated computerized microfluidics, sensors, and tissue models for real‐time monitoring drug‐induced responses ( Figure 8 a, b). 177 In a heart–liver‐cancer model, anticancer efficacy of Doxorubicin (DOX) (Figure 8 c) as well as cardiac toxicity caused liver‐mediated metabolism of DOX (Figure 8 d) was simulated. However, current organoid chips are labor intensive, not high throughput and lack sensing methods. Further development should be made to focus on the miniaturization of the components on an integrated chip. In light of the increasingly high cost of developing new drugs, the use of the “Organs‐on‐a‐Chip” approach for prescreening the efficacy and toxicity of new drugs could be a time‐ and money‐saving strategy. In addition, the “Organs‐on‐a‐chip” approach can be applied to be expanded to study various types of cells, including both diseased and normal cells, for investigating metabolic and signaling pathways. 178 Figure 8 Automated sensing of cardiac toxicity of an anticancer drug using multi organoids on a chip. a) An automated system that integrates computer‐controlled fluidics, organs‐on‐chips platform, and in line sensing. b) Design of a heart–liver‐cancer‐on‐chip system to monitor the on‐target and off‐target toxicity of DOX. c) In‐line monitoring of the anticancer efficacy of DOX to liver cancer organoid. d) In‐line monitoring of the cardiac toxicity of DOX using biosensors. Reproduced with permission. 177 Copyright 2017, National Academy of Sciences. A biopsy is the standard method for acquiring tissues from patients. In the case of cancer metastasis and circulating cancers, capturing the metastatic cells from blood and concentrating them for analysis can acquire disease‐specific information about these tissues. Circulating tumor cells (CTCs) may be less invasive than the primary tumor to obtain, and they could help in staging cancer and improving prognoses. 179 Finding methods for specifically capturing the targeted cancer cells from the blood circulation is the key challenge for acquiring CTCs. Through utilizing properties specific to CTCs, such as overexpressed biomarkers or differences in density, size, or charge, many methods have been developed for CTC separations. 180 Often, a combination of immunoaffinity and magnetic‐ or fluorescence‐based separations is widely used for isolating CTCs; a representative example of this is the FDA‐approved CellSearch system that captures epithelium sourced CTCs. Microfluidic devices with size exclusion or antibody‐modified channels have been demonstrated as a convenient way of isolating CTCs as well. The capturing efficacy of these channels could be fine‐tuned by optimizing the flow rate, turbulence, and variety/density of antibodies. Furthermore, by using transparent devices, imaging and capturing could be done simultaneously. Captured CTCs can be used for directing genetic profiling or growing organoids before analysis. Besides CTCs, however, the capture of other circulating cells, such as trophoblasts, has been demonstrated to provide prenatal diagnostics. 181 Capturing other biocomponents from the blood, such as exosomes 182 or circulating DNA, 183 also provides additional personalized molecular information about a patient. However, not all target cells are circulating, for tissues that are not readily accessible by either liquid or solid biopsies, differentiating cell lines in vitro using iPSC‐based technology offers an alternative method to access them. For example, iPSC‐derived cardiomyocytes have been used for testing drug toxicity. 184 3 Future Directions and Challenges In a hypothetical scenario, when a patient with breast cancer is hospitalized, the last thing she would want is to try different anticancer therapeutics, accumulating all the toxic side effects from each treatment. To tailor therapy specifically for the patient, one can run genetic sequencing on the patient's tumor tissues to acquire her genetic information and propose an optimal therapy based on the correlations between her genotype and phenotype. Meanwhile, one can also use engineering approaches to validate what this patient needs experimentally. For example, one can use biosensing techniques to test expressed biomarkers in real time to monitor her disease's progression; one could take a biopsy of the tumor and build a tumor‐on‐chip device to rapidly test the response of the tumor to different therapies so that one does not need to rely on the accuracy of bioinformatic predictions alone; and one can engineer smart biomaterials that control the delivery and release of an anticancer drug specific to the patient's tumor microenvironment so that one dose of the drug could fight cancer for weeks or months. With more emerging advances and breakthroughs in engineering techniques and basic biology, 185 more controllable tools will become available to increase the accuracy and application areas of precision medicine. Current bioinformatics‐based precision medicine is mostly based on our capability to read DNA. With genome editing tools like CRISPR‐Cas9 commercially available, 186 it is now a routine laboratory technique to write predesigned DNA sequences into cells. The prospect of treating hereditary diseases from their genetic roots sheds light on many untreatable diseases. 187 Many exploratory studies are ongoing to integrate CRISPR into therapies to correct monogenic diseases. 188 This technology is readily integrated with regenerative medicine by editing stem cells or with cancer immunotherapy by engineering T cells, but it is not yet ready to be directly applied to humans. With the challenges of delivery efficiency 189 and the debate over off‐targeting, 190 new medicines with nucleotide‐level precision are expected. Current methods to build artificial tissues from cell culture are far from ready for clinical applications in terms of wanting to generate anatomically and functionally meaningful organs. The other “top‐down” approach that works toward directly growing artificial organs in animals to generate functional tissues is more straightforward. In one example, endogenous retroviruses in porcine were systemically knocked out with the CRISPR‐Cas9 system to obviate the concern of transmitting these viruses from porcine animals to humans. 191 Although immune compatibility issues have not been addressed in these studies yet, the approach of generating humanized pig organs are promising for producing structurally complex organs for xenotransplantation. Effects from the environment have long been recognized as important regulators of human health; recent progress has revealed insight about our interactions with the environment into greater detail. An important venue that humans interact with the environment is through microorganisms. Traditional human‐associated microbiology studies focused on pathogenic microbes; 192 moreover, recent studies have revealed that nonpathogenic microbes are also known to affect an individual's health, such as his or her immune system or metabolism. 193 The composition of microorganism inside the human body or the microbiome differs significantly among people, leading to variations in the strength of an individual's immune system as well as reactions to a given drug. Infants experience their first microbiome though mother–infant transmission in the birth canal; this may be the first training of the newborn immune system to adapt to a world full of microbes. It has been found that infants missing this training by cesarean‐mediated births show weaker immune systems. Furthermore, the activity of the local microorganisms changes the metabolic pathways of a given drug, changing its pharmacokinetics. It has been recently validated that intestinal microbiome composition can severely affect the therapeutic efficacy of anticancer drugs. 194 Health monitoring using wearable, implantable, and point of care sensors represents a future trend for achieving real‐time healthcare management. We expect that these sensors will be incorporated into every aspect of our life by bridging the “internet of things” with personalized healthcare. These sensors are already in the smartwatches we wear, and soon, they will be in other wearable gears. Perhaps we will see smart cars that not only self‐navigate but also keep an eye on the health status of the driver. They might automatically pull over to a rest area when the driver is fatigued. Furthermore, we expect smart homes with healthcare sensors integrated within furnitures or appliances. The refrigerator might give a suggestion of a grocery list to order a balanced diet for an individual, or the air conditioner might detect pathogens within a person's breath and send out warnings. Despite the promising future of precision medicine, there are still many challenges that need to be addressed before precision medicine can fully be realized and benefit everyone. 1. In terms of customized precision medicine, there is a trade‐off between convenience and cost. From a practical perspective, the higher the degree of precision, the more complex the healthcare service will be. Similar to comparing a tailored suit to off‐the‐shelf pajamas, the more bells and whistles we incorporate into designing the device, the better the device will be, but the time and cost of fabricating it will increase, restricting everyone from having the same access to that level of care. So, it is desirable to have a panel of healthcare services that have different levels of precision to fit the need for precision medicine. 2. Engineered precision medicine will need to have an extremely high versatility to meet the needs of individual patients, increasing the complexity of gaining regulatory approvals. Using the classic DOX‐loaded liposome (Doxil) as an example, 195 we may customize many personalized Doxil derivatives that target different receptors and trigger drug release in response to the various physiological environmental factors in different patients. Besides the manufacturing cost for the diverse product lines, getting each of the personalized medicines approved would be a daunting task. Furthermore, prescreening patients to fit a particular therapy might contradict traditional clinical trials that require randomized samples. 196 To address this challenge, engineers could either simplify the design or incorporate mainly FDA‐approved components into their formulation. In the meantime, regulatory innovations are needed to facilitate the approval of new drugs. 3. Our current knowledge about human physiology only applies precision medicine to a tiny portion of healthcare related issues. For example, our understanding of the correlation between biomarkers that can be detected from body fluids and related diseases is still limited. 197 More fundamental research about the mechanism of many diseases is needed, which requires us to invest more in fundamental researches. 4. Compared to traditional medicine, more well‐trained practitioners with training in genomics and engineering will be needed. In order to use devices based on engineered biomaterials, the clinicians will need to have a general understanding of various materials. All these add‐on requirements make it possible for licensing and board certification exams to include genomic‐ or engineering‐related topics. To facilitate incorporation of the new requirements, it is better to adapt current genomics and engineering trainings designed for professional geneticists and engineers to fit the need of clinicians. By focusing on essential skills or knowledge needed to apply genetic discoveries or engineered devices into the clinic, it could reduce the load on medical students but also enables them to harness the advantage of precision medicine. 5. Although most people may agree to share real‐time data from therapeutic devices, such as blood glucose monitoring, some may have concerns about sharing their static data, like their genetic information. They may also have concerns about being discriminated against for job applications based on their genetic‐information, if the information is made available to the decision maker, i. e. , the employer. To address the privacy concerns from the general public but still make the sensitive informative available to the research community, it is necessary to build a data‐share system with different levels of privacy settings. Strong network security needs to be in place to prevent unauthorized access to the information. 6. Precision medicine is based on our capability to manage “Big Data, ” 198 including genomics, sensing, imaging, and other available health records. In the long run, we should be able to keep people's health records from birth to death as well as from a molecular to societal level as well. For the data to be “big” enough, healthcare data from patients with different ethnic backgrounds are needed. For example, African descendants are currently underrepresented in terms of data collection. Increasing the diversity of collected data from people of different races could spread the benefits of precision medicine to more people. 199 The use of artificial intelligence techniques in assisting “Big Data” management for personalized medical care is promising, as it potentially holds a high level of precision and accuracy in disease diagnostics, history, treatment, and prognosis. 200 Overall, precision medicine is an ambitious approach that needs collective efforts from physicians, patients, insurance companies, information technology developers, bioengineers, and others. It requires knowledge and technologies from various fields, such as medicine, genetics, chemical engineering, materials engineering, bioengineering, and pharmaceuticals, and people all around the world need to contribute to make it a reality. In order to transform our current healthcare infrastructure into the era of precision medicine, we need to remove many of the barriers between people from different areas. New drugs, devices, etc. , are needed to take this endeavour to the next level. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201801175
| 2,018
|
Advanced Science
|
Exogenous Physical Irradiation on Titania Semiconductors: Materials Chemistry and Tumor‐Specific Nanomedicine
|
Abstract Titania semiconductors can be activated by external physical triggers to produce electrons (e − ) and holes (h + ) pairs from the energy‐band structure and subsequently induce the generation of reactive oxygen species for killing cancer cells, but the traditional ultraviolet light with potential phototoxicity and low‐tissue‐penetrating depth as the irradiation source significantly hinders the further in vivo broad biomedical applications. Here, the very‐recent development of novel exogenous physical irradiation of titania semiconductors for tumor‐specific therapies based on their unique physiochemical properties, including near infrared (NIR)‐triggered photothermal hyperthermia and photodynamic therapy, X‐ray/Cerenkov radiation‐activated deep‐seated photodynamic therapy, ultrasound‐triggered sonodynamic therapy, and the intriguing synergistic therapeutic paradigms by combined exogenous physical irradiations are in focus. Most of these promising therapeutic modalities are based on the semiconductor nature of titania nanoplatforms, together with their defect modulation for photothermal hyperthermia. The biocompatibility and biosafety of these titania semiconductors are also highlighted for guaranteeing their further clinical translation. Challenges and future developments of titania‐based therapeutic nanoplatforms and the corresponding developed therapeutic modalities for potential clinical translation of tumor‐specific therapy are also discussed and outlooked.
|
1 Introduction As one of the mostly explored multidisciplinary research frontiers, nanomedicine has attracted the broad attention of scientific community ranging from material science, chemistry, pharmacy, biology, and biomedicine. 1, 2, 3, 4, 5, 6 It has shown the intriguing performance and application prospect in molecular imaging for disease diagnosis, targeted drug delivery for enhanced chemotherapy, some physically triggered novel therapeutic modalities, diagnostic biosensing, and even tissue engineering. 7, 8, 9, 10 Various nanoparticles with their intrinsic desirable composition, nanostructure, physiochemical property, and biological effects have been explored to achieve the efficient therapeutic performance and outcome in the past decades, among which inorganic nanoplatforms have been attracting the high research interest very recently because of their unique physiochemical property, multifunctionality (e. g. , optical property, magnetism, electronic behavior and acoustic property) and relatively high biocompatibility. 11, 12, 13, 14, 15, 16, 17 Organic nanosystems have been broadly investigated and some of them have entered the clinical stage for benefiting the patients. 1, 18, 19, 20, 21 It is noted that the organic nanosystems typically lack the functionality, which means that they cannot be easily designed for some unique and specific theranostic purposes. Comparatively, inorganic nanosystems can be facilely endowed with specific properties of magnetism, fluorescence, ultrasound responsiveness, electronic conductivity, etc. They can also be designed with some intriguing nanostructures and topologies. For instance, the mostly explored mesoporous silica nanoparticles (MSNs) are fabricated with well‐defined mesoporous nanostructure, which provide the large reservoirs for the efficient loading and delivery and therapeutic guest molecules. 22, 23 Another paradigm of inorganic nanoparticles is the mostly studied superparamagnetic iron oxide nanoparticles (SPIONs) for contrast‐enhanced magnetic resonance imaging (MRI), magnetically targeted drug delivery and magnetic hyperthermia, which are all based on their intriguing magnetic properties. 24, 25, 26 Especially, the plasmonic resonance property of gold (Au) nanoparticles has been adopted for photo‐triggered hyperthermia, computed tomography (CT) imaging, and biosensing applications. 27, 28 Compared to mostly explored metal oxides such as silica, manganese oxide, and iron oxide nanoparticles, titania nanosystems have emerged as a novel inorganic nanoplatform with their intrinsic physiochemical properties suitable for biomedical applications. 29, 30, 31, 32, 33, 34, 35 Titanium (Ti) has been demonstrated as one of the biocompatible elements. For instance, titanium oxide (TiO 2 ) has been extensively used in colorant in food, 36, 37, 38 cosmetics, 39, 40, 41 and sunscreen. 42, 43, 44, 45, 46 Especially, Ti‐containing metal alloys have been employed as the medical implantation devices. 47, 48, 49, 50, 51, 52, 53 As one of the mostly explored semiconductors, traditional TiO 2 nanoparticles have a bandgap of 3. 2 eV, which can be excited by ultraviolet (UV) light. 54, 55, 56, 57 The UV light with radiative energy higher than its bandgap of 3. 2 eV (387 nm) excites an electron to the conduction band (CB) and creates an electron hole pair. 58, 59, 60, 61 The produced electrons can reduce the absorbed molecular oxygen to superoxide radicals, and the holes are capable of oxidizing the water molecules into hydrogel radicals (•OH). Therefore, TiO 2 nanoparticles can function as the inorganic photosensitizers to produce large amounts of reactive oxygen species (ROS) for photodynamic therapy (PDT). 62, 63, 64, 65, 66, 67 Especially, the semiconductor nature and unique photoresponsiveness of TiO 2 have been employed for the degradation of organic substrates 68, 69, 70 and deactivation of microorganisms/viruses. 71, 72, 73, 74 However, traditional titania nanoparticles only respond to UV light, which unfortunately has potential phototoxicity and low tissue‐penetrating depth, severely hindering their further clinical translation. 29, 75 Very recently, the fast development of theranostic nanomedicine has explored more effective exogenous physical triggers for activating titania nanoparticles for achieving some specific but intriguing therapeutic modalities, such as near infrared (NIR)‐triggered photothermal therapy (PTT), NIR‐activated PDT, X‐ray/Cerenkov radiation (CR)‐activated deep‐seated PDT, and US‐triggered SDT ( Figure 1 ). All these exogenous physical triggers are featured with their intrinsic characteristics but also with some drawbacks, which is the main topic of this review. Therefore, this review discusses the very‐recent progresses of adopting novel exogenous physical irradiations to activate titania semiconductors for some specific tumor therapies, which are mainly based on the semiconductor nature of titania nanosystems accompanied with the unique oxygen‐defect modulation for photothermal hyperthermia. In addition, the critical issue of biocompatibility and biosafety of these titania nanoplatforms has also been discussed to potentially guarantee the further clinical translation. Finally, the critical challenges and further future developments of the novel physical‐triggered titania‐based nanoplatforms and the corresponding intriguing therapeutic modalities are also deeply discussed to promote the progress of this novel inorganic nanoplatform in theranostic nanomedicine. Figure 1 Schematic illustration of exogenous physical activation of titania nanoparticles for tumor‐specific therapy. It includes NIR‐activated PDT/PTT, radiation‐activated PDT, US‐activated SDT and physical activation‐based synergistic therapy. The related research frontiers are also summarized in the figure, such as nano‐synthetic chemistry for titania fabrication, structure/composition optimization, surface engineering, and biosafety evaluation. 2 Synthesis, Multifunctionalization, and Surface Engineering of Titania Nanoparticles in Biomedicine The rational design and successful construction of titania‐based nanoplatforms are the bases for achieving the high theranostic performance in biomedicine, which is mostly based on the advances of nanosynthetic chemistry and material chemistry. 76, 77, 78, 79, 80, 81, 82, 83 We recently synthesized highly dispersed mesoporous titania nanoparticles (MTNs) with monodispersity and uniformity by a method of prehydrolysis of titanium precursors combined with solvothermal treatment, which provided both the high crystallized framework and well‐defined mesoporous nanostructure. 84 In addition, we recently synthesized oxygen‐deficient core/shell‐structured black TiO 2− x nanoparticles by a facile aluminum reduction methodology for enhanced sonodynamic therapy (SDT) and simultaneous NIR‐triggered photothermal hyperthermia at NIR‐II biowindow. 85 Especially, mesoporous black TiO 2− x nanoparticles were fabricated by a two‐step procedure where the mesoporous white TiO 2 nanoparticles were initially synthesized, followed by reduction under hydrogen atmosphere at 500 °C for 1 h to turn white TiO 2 into black TiO 2− x. 86 Especially, the advances of nanosynthetic chemistry make the precise controlling of titania's composition, nanostructure and functionality possible where these titania nanocomponents can also be integrated with other functional moieties or nanoparticles to achieve some specific purposes. For instance, the NaYF 4 :Yb, Tm nanoparticles were initially coated by a silica layer with the grafting of (3‐aminopropyl)‐trimethoxysilane, which provided the positively charged amino groups for efficient binding titanium precursors, guaranteeing the gradual epitaxial growth of uniform TiO 2 layer onto the surface of initially synthesized nanoparticles. 87 NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ @TiO 2 (UCNPs@TiO 2 ) core/shell nanoparticles were synthesized by direct in situ growth protocol. NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ were initially synthesized, followed by surface modification with polyvinylpyrilidone (PVP). 88 Then, TiF 4 acted as the Ti precursors for the direct formation of TiO 2 nanoshells on the surface of UCNPs by the hydrolysis and condensation process. This one‐step PVP‐mediated methodology is facile and generic for the construction of TiO 2 ‐based nanocomposites, especially for the construction of UCNPs@TiO 2 composite nanoplatforms. 89 For TiO 2 ‐based functionalization, we also directly grew TiO 2 nanoparticles onto the surface of graphene oxide (GO) by a hydrothermal treatment of the cosolvent solution of TiO 2 nanoparticles and GO suspension, based on which TiO 2 nanoparticles were uniformed dispersed onto the surface of 2D planar GO. 90 Au–TiO 2 nanocomposites were synthesized by a facile photoreduction of Au 3+ ions, which could be deposited on the surface of TiO 2 nanoparticles to grow Au nanoparticles under UV irradiation. 91 The particle size of deposited Au nanoparticles could be controlled by adopting the UV‐irradiation durations. Especially, the dumbbell‐like Au–TiO 2 nanoparticles were synthesized by a seed‐mediated growth approach. 92 Au nanoparticles were initially synthesized, acting as the growing sites for the TiO 2 generation in an anisotropic manner by controlling the hydrolysis degree of introduced Ti precursor. The surface chemistry of titania‐based nanoplatforms is also of high significance in biomedicine. For instance, the adequate surface modification can either improve the stability of these nanoparticles in physiological solution or achieve the positive‐targeting accumulation into tumor cells/tissues. The surface of Au–TiO 2 nanocomposites was modified with biocompatible carboxymethyl dextran (CMD) for achieving prolonged systemic circulation and subsequent enhanced tumor‐homing capability. 91 We modified the surface of black TiO 2− x nanoparticles with NH 2 –PEG 2000 molecules by a simple sonication procedure, which was based on the coordination interaction between N component of PEG molecules and Ti atoms of black TiO 2 nanoparticles. 85 We also modified the surface of TiO 2 ‐loaded GO by PVP molecules for enhanced stability in physiological condition, which could guarantee the further in vivo biomedical applications on combating cancer. 90 In addition, the precoating of SiO 2 layer onto the surface of TiO 2 ‐based nanocomposite could provide the anchoring sites of silane group of maleimide‐PEG‐silane, achieving the efficient PEGylation of TiO 2 ‐based nanocomposites. 87 Especially, anti‐cAngptl4 Ab was conjugated onto the surface of N‐TiO 2 /NaYF 4 :Yb, Tm nanocomposites for targeted cancer‐cell PDT on killing cancer cells as induced by NIR irradiation. 93 3 Light Irradiation on Titania for PDT PDT on combating cancer is featured with noninvasiveness and tumor specificity, which typically employs the external physical light source for activating photosensitizers to produce toxic ROS and consequently kill the cancer cells. 94, 95, 96, 97 Light‐excited PDT has been clinically used for the treatment of cancers on skin and other epidermal tissues. As compared to traditional organic photosensitizers, TiO 2 ‐based inorganic nano‐photosensitizers are featured with high stability and nontoxicity, which has shown broad application potentials in PDT‐based cancer treatment. 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108 Based on titania nanoparticles, a polychromatic visible light‐activated nano‐biohydrid system was constructed by covalently binding an antibody via a dihydroxybenzene bivalent linker that could selectively recognize glioblastoma multiforme (GBM) cells ( Figure 2 a). 109 This targeting strategy enhanced the intracellular uptake of TiO 2 nanoparticles and produced large amounts for ROS for damaging the cell membrane, inducing the cancer‐cell death under the visible‐light irradiation (Figure 2 b, c). The PDT efficiency of titania could also be achieved by heterogeneous atom doping. For instance, the Fe‐doping of TiO 2 nanotubes was demonstrated to realize near‐visible light‐driven (2. 30 mW cm −2, ≈405 nm) PDT on killing cervical cancer cells, and the phototoxicity of Fe‐doped TiO 2 nanotubes was much higher than that of undoped TiO 2 nanotubes. 103 Furthermore, nitrogen‐doped TiO 2 (N‐TiO 2 ) also showed visible light‐triggered PDT against HeLa cancer cells where the N‐doped TiO 2 nanoparticles were featured with higher PDT efficiency as compared to that of pure TiO 2 nanoparticles. 110 The related mechanism investigation revealed that N‐TiO 2 induced more loss of mitochondrial membrane potential and higher increase of intracellular Ca 2+ and nitrogen monoxide in HeLa cancer cells than pure TiO 2 nanoparticles. Figure 2 a) Schematic illustration of the fabrication of titania nanoparticles with surface‐linked IL13R‐recognizing antibody, and their further recognition and binding to surface IL13R of cancer cell for visible light‐activated ROS generation and subsequently inducing cancer‐cell death. In vitro therapeutic efficiency by evaluating the phototoxicity of TiO 2 ‐mAb against b) A172 GBM cells (high ILα2R expression) and c) U87 GBM cells (low ILα2R expression). Reproduced with permission. 109 Copyright 2009, American Chemical Society. To further enhance the PDT efficiency of TiO 2 ‐based photosensitizers, TiO 2 nanoparticles were conjugated with ruthenium complex (N3) for improved and synergistic production of ROS in both hypoxic and normoxic conditions. 111 By light irradiation (365 nm), the N3 injected electrons into TiO 2 nanoparticles, resulting in the production of three‐ and fourfold more hydroxyl radicals (•OH) and hydrogen peroxide (H 2 O 2 ) as compared to bare TiO 2 nanoparticles, respectively. Bare TiO 2 nanoparticles could oxidize water molecules to produce hydroxyl radical (•OH, Figure 3 a). The presence of light‐induced electron–hole pair in TiO 2 facilitated the reduction of molecular oxygen to superoxide and then transformation to single oxygen ( 1 O 2, Figure 3 b). Especially, under the hypoxic condition, the N3 facilitated the electron–hole reduction of absorbed water molecules to enhance the hydroxyl radical production with nearly threefold increase (Figure 3 c, d). This strategy could transform TiO 2 photosensitizer from a dual type I and II PDT nanoagents into a mainly type I photosensitizer independent of the oxygen level (Figure 3 d–f). 111 This work provides an efficient strategy to enhance the TiO 2 ‐based PDT efficiency in hypoxic condition by N3 hybridization. Coating a homogenous TiO 2 layer onto the surface of ZnTPyP self‐assembly nanocrystal achieved the photoelectron transfer at ZnTPyP self‐assembly/TiO 2 interfaces, which further enhanced the two‐photon PDT against HeLa cancer cells via type‐1‐like PDT process. 65 This titania‐based composite nanoplatform is very intriguing because the achieved two‐photon PDT is highly desirable for deep‐tissue disease treatment. 96, 112 Figure 3 Schematic illustration of ROS generation, including a) H 2 O 2 production and transformation into hydroxyl radical, b) singlet oxygen ( 1 O 2 ) production, and c) electron injection by N3 into TiO 2 –N3 to enhance H 2 O 2 and hydroxyl radical production. The generation rate of d) hydroxyl radical, e) H 2 O 2, and f) singlet oxygen under hypoxic conditions. Reproduced with permission. 111 Copyright 2017, WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. The major challenge of TiO 2 ‐based PDT is the light responsiveness only in the wavelength range of UV or visible light, which has the low tissue‐penetrating distance and causes the failure in the treatment of deep‐seated tumor. Upconversion nanoparticles (UCNPs) are capable of generating high energy light from the low energy light such as NIR light. 113, 114, 115 Therefore, lanthanide‐doped UCNPs can convert NIR light into UV or visible photons via an anti‐Stokes emission process, which potentially acts as the “nano‐transducers” to achieve NIR‐triggered PDT. 116, 117, 118 On this ground, core/shell‐structured UNCPs (NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ ) with enhanced upconverting UV emission were initially synthesized, followed by coating with TiO 2 shells using TiF 4 as the Ti precursor to in situ grow TiO 2 shells onto the surface of UNCPs under mild hydrolysis condition ( Figure 4 a–c). 88 The UCNPs core emitted upconverting light in UV/visible range by 980 nm NIR irradiation, which was substantially diminished by the absorbance of TiO 2 shells in such a wavelength range (Figure 4 a). Such an energy‐transferring process induced the extracellular and intracellular generation of ROS for causing the cancer‐cell death by inducing the cell apoptosis. HeLa tumor‐bearing model results showed that the intratumoral injection of NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ @TiO 2 (UCNPs@TiO 2 ) core/shell nanoparticles followed by 980 nm laser irradiation achieved the substantial tumor‐growth suppression with high therapeutic efficiency/outcome (Figure 4 d, e), which was further demonstrated by immunohistochemical staining for caspase 3. 88 Figure 4 a) Schematic illustration of NIR‐triggered PDT based on UCNPs@TiO 2 ‐based nano‐photosensitizers and the underlying therapeutic mechanism including the induced apoptosis of cancer cells. TEM images of b) NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ core/shell nanoparticles and c) NaYF 4 :Yb 3+, Tm 3+ @NaGdF 4 :Yb 3+ @TiO 2 (UCNPs@TiO 2 ) core/shell nanoparticles. d) The tumor‐volume changes of tumor‐bearing mice in varied treatment groups as shown in the figure, and e) the corresponding digital photographic images of excised tumors at the end of treatments. Reproduced with permission. 88 Copyright 2015, American Chemical Society. f) The scheme of UCNP‐mediated activation of surface‐coated TiO 2 layer for ROS production to kill the cancer cells. TEM images of g) TiO 2 ‐UCNPs and h) Mal‐PEG‐TiO 2 ‐UCNPs. i) The in vivo OSCC tumor‐growth volumes within the 35 d duration after varied treatments as indicated in the figure. Reproduced with permission. 87 Copyright 2015, American Chemical Society. Similarly, Zhang and co‐workers coated a TiO 2 layer onto the surface of SiO 2 ‐coated UCNPs (NaYF 4 :20%Yb, 0. 5%Tm) for 980 nm NIR‐triggered PDT. The UCNPs core converted NIR irradiation into UV light, which photoexcited electrons in the valence band (VB) of TiO 2 shell to the CB, forming the photo‐induced hole–electron pairs (Figure 4 f). 87 The postgenerated hole–electron pairs reacted with surrounding molecular oxygen and water molecules to generate ROS and then induce the cancer‐cell death. The TiO 2 ‐coated UCNPs were clearly characterized by TEM image (Figure 4 g, h). After the intratumoral injection of these composite nanoparticles followed by 980 nm NIR irradiation, the significant tumor‐growth suppression was achieved (Figure 4 i). In fact, the TiO 2 ‐coated UCNPs themselves are highly biocompatible, and only the NIR‐irradiated tumor region can produce toxic ROS, therefore their impact to normal cells and tissues are low, leading to high therapeutic biosafety. Furthermore, anti‐EGRF‐affibody was conjugated to PEGylated TiO 2 –UCNPs nanocomposites for targeting epithelial growth factor receptor (EGFR) overexpressing oral cancer cells and the subsequent NIR‐excited PDT with the therapeutic outcome of significantly suppressed tumor growth and improved survival rate of tumor‐bearing mice. 104 The photoresponsive wavelength range could also be controlled by rational design of the composition and nanostructure of titania‐based nanoplatforms. For instance, 808 nm NIR‐activation of black TiO 2 nanoparticles with a narrow bandgap of around 2. 32 eV was demonstrated to absorb NIR light and subsequently produce abundant ROS for photodynamic killing of bladder cancer cells. 120 In addition, Au cluster‐anchored black anatase TiO 2− x nanotubes (designated as Au 25 /B‐TiO 2− x ) were stepwise synthesized by gaseous hydrogen reduction of TiO 2 nanotubes followed by the deposition of Au clusters ( Figure 5 a). 119 These Au 25 /B‐TiO 2− x exhibited the photo‐responsiveness in NIR range (650 nm) for PDT against cancer. The surface modification of Au clusters changed the electrical distribution in the composite nanosystem, which could reduce the recombination of electrons and holes as triggered by NIR irradiation (Figure 5 b). Importantly, the hydrogen reduction generated large amount of Ti 3+ ions in the matrix of black TiO 2‐x, which extended the light response of anatase TiO 2 nanoparticles from UV light to NIR light. In vivo therapeutic evaluation on tumor‐bearing xenograft revealed that the significantly enhanced therapeutic efficacy was achieved based on the photocatalytic synergistic effect, which substantially suppressed the tumor growth after the injection of Au 25 /B‐TiO 2− x followed by NIR irradiation (Figure 5 c). Figure 5 a) Schematic illustration of the synthetic process for Au 25 /B–TiO 2− x nanotubes. b) The scheme of photocatalytic mechanism for ROS production as assisted by the introduced Au 25 /B–TiO 2− x nanotubes. c) In vivo therapeutic outcome as indicated by the relative tumor‐volume changes after varied treatments. Reproduced with permission. 119 Copyright 2017, WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. To achieve simulated sunlight‐irradiated PDT, TiO 2 –Au–graphene (designated as TAG) heterogeneous nanocomposites were designed and fabricated for employing simulated sunlight as physical triggering source to kill melanoma skin cancer cells by photodynamic effect. 121 The narrow bandgap of Au nanoclusters and staggered energy bands of Au–TiO 2 –graphene resulted in the efficient use of simulated sunlight, which also enhanced the separation efficiency of electron–hole pairs for producing large amounts of hydroxyl and superoxide radicals. 122, 123, 124, 125, 126, 127 Typically, the sunlight‐excited electrons from HOMO to LUMO of Au nanoclusters were transferred to the conductive band of titania nanoparticles and then to the graphene matrix, which further acted as the free electrons and further generated superoxide radicals by reacting with oxygen molecules. The holes from both HOMO of Au nanoclusters and valance band of titania nanoparticles accumulated on HOMO of Au nanoclusters, which further reacted with water molecules to produce hydroxyl radicals ( Figure 6 ). These TGA nanocomposites have been demonstrated to trigger a series of toxicological effects on killing B16F1 melanoma cells against B16F1 tumor xenograft, indicating high photodynamic efficiency of this prominent therapeutic modality for sunlight‐triggered PDT effect. Although above‐mentioned paradigms are effective on phototriggered PDT for cancer therapy based on TiO 2 ‐based photosensitizers, this therapeutic modality is still suffering from the low tissue‐penetrating capability of light as the irradiation source. Figure 6 Schematic illustration of the fabrication of TAG composite nanoplatform and the related PDT mechanism for cancer therapy. Reproduced with permission. 121 Copyright 2017, WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. 4 Laser Irradiation on Titania for PTT In addition to NIR‐triggered PDT for combating cancer, NIR‐induced PTT has emerged as an efficient therapeutic modality for tumor treatment. 28, 128, 129, 130, 131 Typically, the exogenous NIR laser can penetrate through the skin and activate the photothermal agents for converting NIR energy into heat and then ablating the tumor tissue by simply elevating the tumor temperature subsequently. 132, 133, 134, 135, 136 Therefore, the development of desirable photothermal‐conversion agents plays the determining role for achieving the efficient and desirable PTT outcome. 137, 138, 139, 140, 141, 142, 143, 144, 145, 146 Based on an ambient heterogeneous spark discharge, Au–TiO 2 heterodimers were fabricated by incorporating Au component into TiO 2 nanoparticles, which exhibited the visible light‐induced photothermal effect on killing HeLa cancer cells based on the localized surface plasmon resonance of integrated ultrafine Au nanoparticles. 149 Al reduction could transform white P25‐type TiO 2 nanoparticles into oxygen‐deficient black TiO 2− x (B‐TiO 2− x ) nanoparticles, 150, 151, 152 which endowed these black TiO 2− x nanoparticles with unique photothermal‐conversion capability for efficient photothermal hyperthermia of cancer. Mo et al. modified the surface of B‐TiO 2− x nanoparticles with PEG molecules for guarantee their high stability in physiological condition ( Figure 7 a). After intravenous administration into HeLa tumor‐bearing mice, these PEGylated B‐TiO 2− x nanoparticles efficiently accumulated into tumor tissue and rapidly elevated the tumor temperature by 808 nm NIR irradiation, causing the complete photothermal eradication of tumor tissue (Figure 7 b–d). 147 Besides the Al reduction to fabricate black TiO 2− x nanoparticles, the hydrogenated black TiO 2 (H‐TiO 2 ) nanoparticles also exhibited high NIR absorption, which were further developed as the photothermal‐conversion nanoagents for efficient tumor photohyperthermia based on their high photothermal‐conversion efficiency of as high as 40. 8% at the wavelength of 808 nm. 153 Figure 7 a) Schematic illustration of synthesizing PEGylated black TiO 2− x nanoparticles and their unique functionality for PA imaging‐guided photothermal hyperthermia of tumor under NIR laser irradiation. b) The relative tumor‐volume changes after varied treatments including control group, NIR group, TiO 2− x group and TiO 2− x combined with NIR irradiation group. c, d) Photographic image of tumor at the end of each treatment. Reproduced with permission. 147 Copyright 2016, Elsevier. e) TEM images of TiO 2 nanoparticles with varied Nb‐doping amount. f) UV–vis–NIR absorbance spectra of Nb‐doped TiO 2 nanoparticles in chloroform. g) The relative HeLa tumor‐volume changes as a function of feeding time after different treatments as shown in the figure. Reproduced with permission. 148 Copyright 2017, Royal Society of Chemistry. In addition to the mostly explored high‐temperature treatment strategy to obtain black TiO 2 nanoparticle for PTT against cancer, Chen and co‐workers successfully converted UV‐responsive TiO 2 nanoparticles to blue TiO 2 nanocrystals by a simple Nb‐doping approach. 148 The different Nb‐doping amount induced varied morphology of TiO 2 nanocrystals with high dispersity (Figure 7 e). Especially, the efficient Nb‐doping endowed these blue TiO 2 with the strong NIR absorbance (Figure 7 f), which was originated from the localized surface plasmon resonances because of Nb doping‐induced considerable free electrons. These blue Nb‐doped TiO 2 nanocrystals efficiently converted laser at NIR‐II biowindow (1064 nm) into heat and induced the photothermal effect on ablating the tumor tissue with high PTT efficiency (Figure 7 g). 148 The endowed targeting property of titania nanoparticles potentially enhances the tumor‐accumulation efficiency for improved cancer therapy. On this ground, the surface of NIR‐responsive TiO 2 nanoparticles as the photothermal‐conversion nanoagents was conjugated with cyclo(Arg‐Gly‐Asp‐ d ‐Tyr‐Lys) peptide c(RGDyK) for targeted photothermal hyperthermia of cancer ( Figure 8 ). 154 Based on the absorption of electron localized on Ti(III) sites and free electrons existing in the conduction bond, these TiO 2 nanoparticles showed high photothermal‐conversion efficiency of nearly 38. 5%. The surface‐modified c(RGDyK) peptide selectively targeted the α v β 3 integrin on the cancer‐cell membrane (U87‐MG human glioblastoma cells) for efficiently killing the cancer cells, demonstrating the effectiveness of targeting strategy for improving the therapeutic efficiency of PTT. 154 Figure 8 The scheme of the synthetic procedure of targeted TiO 2 –RGD nanoparticles and their unique functionality for NIR‐triggered photothermal hyperthermia against α v β 3 integrin‐overexpressed cancer cells. Reproduced with permission. 154 Copyright 2017, Springer Nature Publishing Group. 5 Radiation‐Activated Titania for PDT The traditional external laser‐activated PDT or PTT still suffers from the low tissue‐penetrating depth of laser because of the rapid light attenuation passing through tissue and difficulty for reaching the deep‐seated malignant lesions, which only confines the photointerventions for the treatment of superficial diseases. 155 Radiation therapy by using radiation sources can solve above‐mentioned critical issue because of the high tissue‐penetrating capability of these radiation sources. 156, 157, 158, 159, 160, 161 Especially, the advances of theranostic nanomedicine has demonstrated the augmenting effect of some nanoparticulate radiosensitizers for substantially enhanced radiation‐therapy outcome. 162, 163, 164, 165, 166, 167, 168, 169, 170 Therefore, recent advances have also revealed the possibility of exogenous physical radiation sources for activating titania‐based nanoplatforms to achieve efficient cancer therapy. Au and titania anisotropic nanostructure was rationally designed as radio‐sensitizers for X‐ray‐activated radiation therapy. 92 Typically, bare TiO 2 nanoparticles could generate cytotoxic hydroxyl and superoxide radicals by the activation of UV light ( Figure 9 a). Dumbbell‐like Au–TiO 2 nanoparticles (DATs) can be activated by ionizing radiation and then produce secondary photons or electrons, 171, 172, 173 which could induce the ROS production and migrate over the interface of DAT to TiO 2 component for further ROS production on the surface of TiO 2 (Figure 9 b). The anisotropic nanostructure of DATs was constructed by stepwise seed‐mediated growth (Figure 9 c, d). Based on the strong asymmetric electric coupling between Au component and dielectric TiO 2 at the interface, these DATs exhibited a synergistic therapeutic efficiency on X‐ray‐triggered radiation therapy where the production of secondary electrons and ROS from DATs substantially enhanced the radiation effect, causing the high tumor‐suppressing effect (Figure 9 e–g) and survival rate of tumor‐bearing mice (Figure 9 f). 173 This paradigm demonstrates that the rational integration of TiO 2 nanoparticles with functional nanoparticles can significantly enhance the efficiency of radiation therapy by taking the unique characteristics of each integrated component. Figure 9 The scheme of ROS production on a) photoactivated TiO 2 nanoparticles and b) X‐ray‐induced hybrid DATs. TEM images of hybrid DATs at different magnifications. c) In vivo tumor‐volume changes of tumor‐bearing xenograft after different treatments as indicated in the figure, and d) corresponding survival rate of SUM159‐tumor‐bearing mice after varied treatments. e) Photographic images of SUM159 tumor‐bearing mice before and at the end of treatments. f) The survival rate of tumor‐bearing mice after varied treatments, and g) corresponding representative photographic images of mice before and after different treatments. Reproduced with permission. 92 Copyright 2018, American Chemical Society. As an internal light source, CR is featured with high tissue‐penetrating depth, which is typically triggered when the charged particles (e. g. , β + and β − ) pass through a dielectric medium beyond the light speed. 174, 175, 176 The UV can be emitted by CR for triggering UV‐responsive photosensitizers for PDT. 174, 177 On this ground, CR‐induced therapy was achieved by the radiation of PET radionuclides for activation of TiO 2 nanoparticles to produce hydroxyl and superoxide radicals ( Figure 10 a). 174 Especially, titanocene (Tc) was further anchored onto the surface of TiO 2 nanoparticles for enhancing and complementing CR‐irradiated TiO 2 cytotoxicity because it could generate cyclopentadienyl and titanium‐centered radicals once exposure to UV light (Figure 10 b). Furthermore, apo‐transferrin (Tf) was modified onto the surface of TiO 2 nanoparticles for enhancing the positive accumulation into the tumor tissue (Figure 10 c). The results demonstrated that the intravenous administration of Tf‐anchored TiO 2 nanoparticles and clinically employed radionuclides efficiently suppressed the tumor growth (Figure 10 d) accompanied with prolonged survival rate of tumor‐bearing mice (Figure 10 e). This paradigm provides a new strategy to develop low‐radiance‐sensitive nanophotosensitizers for efficient Cerenkov‐radiation‐activated cancer therapy with the tissue‐depth impendence. Figure 10 a) Schematic illustration of CR activation of TiO 2 nanoparticles for producing cytotoxic hydroxyl and superoxide radicals by the electron–hole pair generation where CR was generated by PET radionuclides. b) The scheme of CR activation of Tc for the generation of cyclopentadienyl radical and titanium‐centered radical by photofragmentation. c) Schematic illustration of the fabrication of TiO 2 –PEG, TiO 2 –Tf, and TiO 2 –Tf–Tc, nanoparticles. d) The tumor‐volume changes of HT1080‐tumor‐bearing nude mice after varied treatments as indicated in the figure. e) The survival rate after the treatments with 0. 87 mCi/0. 1 mL FDG. Reproduced with permission. 174 Copyright 2015, Springer Nature. The efficient PDT strongly depends on the ROS production efficiency, which is significantly influenced by the local photon intensity. 179 Gallium‐68 (Ga‐68) is a promising CR source because of the 30‐time higher Cerenkov productivity as compared to fluorine‐18 (F‐18) such as 18 F‐fluorodeoxyglucose ( 18 F‐FDG). Therefore, 68 Ga‐labelled bovine serum albumin ( 68 Ga‐BSA) was employed as the CR source to activate dextran‐modified TiO 2 nanoparticles for inhibiting the tumor growth ( Figure 11 a), which could emit UV light to produce electron (e − ) and hole (h + ) from energy band of TiO 2 and generate ROS subsequently. 178 By PET imaging, it has been found that intratumoral injection of 68 Ga‐BSA and 18 F‐FDG showed the similar tumor uptake of 68 Ga‐BSA and 18 F‐FDG (Figure 11 b, c). Importantly, the tumor‐bearing mice after the treatment with 68 Ga‐BSA and TiO 2 photosensitizer exhibited significantly inhibited tumor volume (Figure 11 d) and prolonged survival time while the mice in the group of 18 F‐FDG and TiO 2 showed much lower tumor‐suppressing rate, indicating that Ga‐68 could act as the more efficient radionuclide as compared to F‐18 for CR‐induced in vivo PDT on combating cancer. The effective cancer treatment of deep‐seated tumor by radiation‐activated TiO 2 nanoparticles is highly promising for clinical use but the potential biosafety risk of radiation source should be seriously considered. Figure 11 a) Schematic illustration of CR‐activated TiO 2 photosensitizers for producing ROS to kill the cancer cells. The comparison of CR‐induced PDT by dynamic PET imaging of 4T1 tumor‐bearing mice after intratumor injection of equivalent amount of b) 68 Ga‐BSA and c) 18 F‐FDG from 30 min to 3 h. d) The tumor‐volume changes of tumor‐bearing mice after varied treatments as indicated in the figure. Reproduced with permission. 178 Copyright 2018, American Chemical Society. 6 Ultrasound Irradiation on Titania for SDT Ultrasound (US) has been broadly explored in biomedicine for decades, not only for diagnostic imaging but also for therapeutic applications. 180, 181, 182, 183, 184, 185, 186 For instance, the thermal, mechanical, and cavitation effects of high‐intensity focused ultrasound (HIFU) have been used for noninvasive cancer surgery. 187, 188, 189, 190 In addition, the sonosensitizer‐involved SDT produces ROS for inducing the cancer‐cell death for cancer‐dynamic therapy. 191, 192, 193, 194, 195, 196, 197, 198 Especially, US is featured with high tissue‐penetrating depth in human bodies, which can reach internal organs such as liver, spleen, and kidney. Therefore, both the tissue‐penetrating capability and theranostic biosafety of US make it a promising exogenous physical triggering source for versatile biomedical applications. As the mostly explored inorganic nanosonosensitizers with high biocompatibility and stability, titania nanoparticles have been extensively employed for US‐activated SDT against cancer. 199, 200, 201, 202 PEGylated TiO 2 nanoparticles have been demonstrated to be effective on inducing the cell death of U251 monolayer cells (1. 0 MHz, 1. 0 W cm −2 ), and the related therapeutic mechanism was found to be different from that of UV light‐induced PDT. 203 Avidin protein‐conjugated TiO 2 nanoparticles were designed to preferentially discriminate cancerous cells from healthy cells for targeted SDT. 204 For in vivo assessment, the combination of TiO 2 nanoparticles and US irradiation (1 MHz, 1. 0 W cm −2, 2 min) substantially inhibited the tumor growth on subcutaneously implanted C32 xenograft, demonstrating the high in vivo therapeutic efficiency of TiO 2 ‐sonosensitized SDT. 205 Especially, the introduction of dual‐frequency US for activation of TiO 2 nanoparticles as the nano‐sonosensitizers was demonstrated to be more efficient for enhancing the hydroxyl radical production, which was verified in vitro on killing HepG2 cells. 206 We recently synthesized MTNs for US‐triggered SDT ( Figure 12 a). 84 These MTNs were featured with ellipsoidal topology and high dispersity (Figure 12 b). Especially, they showed the highly single‐crystalline structure with well‐defined mesoporosity, which could enhance the SDT efficiency based on the fact that the high crystallity without defects could avoid the recombination of electrons (e − ) and holes (h + ) as triggered by US irradiation. The mesoporosity potentially facilitated the encapsulation and delivery of therapeutic agents such as anticancer drugs. After accumulation into tumor tissue of PEGylated MTNs (PEG‐MTNs) via the typical enhanced permeability and retention (EPR) effect, the US‐triggered SDT effect achieved 40% tumor‐suppression rate under the intravenous administration mode. 84 Hydrophilized TiO 2 (HTiO 2 ) nanoparticles were fabricated by anchoring CMD onto the surface of TiO 2 nanoparticles for guaranteeing the high stability in physiological condition, prolonging the blood‐circulation duration and enhancing the tumor accumulation. 207 The accumulation of HTiO 2 into tumor tissue and further US activation not only enhanced the immune response but also destroyed the tumor microvasculature (Figure 12 c), which was demonstrated by the gradually decreased tumor volume (Figure 12 d) and the decreased tumor vasculature (Figure 12 e) by US‐triggered SDT effect in bright‐field images. Figure 12 a) Schematic illustration of the accumulation of PEG–MTNs into the tumor tissue, and further US‐triggered production of ROS for killing cancer cells. b) TEM images of MTNs at low (left image) and high (right image) magnifications. Reproduced with permission. 84 Copyright 2017, Royal Society of Chemistry. c) The scheme of HTiO 2 nanoparticle‐enhanced SDT, including EPR effect‐enabled accumulation into tumor and US‐triggered ROS production to enhance the immune response and destroy tumor microvasculature. d) The tumor‐volume changes of SCC7 tumor‐bearing mice in each treatment group. e) The bright‐field images of tumor vasculature by US‐triggered SDT effect. Reproduced with permission. 207 Copyright 2016, Springer Nature. f) Schematic illustration of in vivo US‐triggered activation of HAu–TiO 2 nanoparticles for SDT and g) the underlying mechanism regarding the ROS production by US activation of HAu–TiO 2 nanoparticles. h) The comparison of tumor‐volume changes with respect of feeding time by varied treatments as indicated in the figure. Reproduced with permission. 91 Copyright 2016, American Chemical Society. It is noted that the low quantum yield of nanosonosensitizers resulting from the fast electron–hole recombination hinders the further clinical translation of TiO 2 ‐based sonosensitizers. To address this critical issue, noble metal Au was combined with TiO 2 nanoparticles to prevent the undesirable electron–hole recombination by trapping the sono‐excited electrons (Figure 12 f, g). 91 This principle has been extensively explored in the typical TiO 2 ‐based photocatalysis. In addition, CMD was also anchored onto the surface of Au–TiO 2 nanoparticles for further in vitro and in vivo evaluations. It is important to find that more ROS could be produced under US activation of Au–TiO 2 composite nanosonosensitizers as compared to pure TiO 2 without Au deposition, demonstrating the effectiveness of Au and TiO 2 combination. This enhanced SDT effect was also revealed in tumor‐therapeutic outcome where the Au–TiO 2 composite nanosonosensitizers induced the more significant tumor suppression as compared to TiO 2 nanoparticles upon US activation (Figure 12 h). 91 By learning the lessons from typical photocatalysis, we recently fabricated an oxygen‐deficient TiO 2−‐ x nanosonosensitizers for enhancing the SDT efficiency against tumor, which was achieved by Al reduction at high temperature to create an oxygen‐deficient TiO 2− x layer onto the surface of TiO 2 nanoparticles ( Figure 13 a, b). 85 Such an oxygen‐deficient TiO 2− x layer facilitated and enhanced the separation of electrons (e − ) and holes (h + ) from the energy‐band of TiO 2 semiconductor, which was activated by external physical US irradiation (Figure 13 a). This effect has been demonstrated to substantially enhance the SDT efficiency at solvent level, in vitro cellular level and in vivo tumor xenograft level (Figure 13 c). Especially, such a process to create oxygen‐deficient TiO 2− x (black TiO 2− x ) endowed this unique TiO 2 ‐based nano‐sonosensitizers with unique photothermal‐conversion capability at NIR‐II biowindow (1064 nm), which synergistically enhanced the SDT efficiency with the therapeutic outcome of complete tumor eradication (Figure 13 c, d). 85 Figure 13 a) Schematic illustration of the fabrication of PEGylated B‐TiO 2− x nanosonosensitizers and enhanced SDT by ROS production and synergistic NIR‐II‐triggered photothermal hyperthermia. b) High‐resolution TEM image of B‐TiO 2− x nanosonosensitizers and corresponding SAED patter (inset image). c) The tumor‐volume changes of 4T1 tumor‐bearing mice after varied treatments as indicated in the figure, and d) corresponding photographic images of tumor at the end of treatments. Reproduced with permission. 85 Copyright 2018, American Chemical Society. Compared to light‐triggered TiO 2 ‐based photosensitizer for PDT, US‐activated SDT based on TiO 2 nano‐sonosensitizers is more applicable for clinical use based on the high tissue‐penetrating depth of US as compared to the conventional light as the irradiation source. However, US‐activated SDT is still at the preliminary stage, which still requires the further deep understanding of the underlying mechanism on the anticancer effect, which is highly beneficial for further improving the SDT efficiency on combating cancer. 7 Exogenous Physical Irradiation on Titania for Synergistic Cancer Therapy Although above‐mentioned therapeutic modalities enabled by TiO 2 ‐based nanoplatforms have shown promising clinical‐translation potential, each of these therapeutic modalities suffers from its intrinsic drawbacks hindering further broad applications. For instance, the therapeutic efficiency of RT, PDT, and SDT is limited by the hypoxia microenvironment of tumor. The heat shock response of local phototriggered hyperthermia causes the low PTT efficiency. The continuous chemotherapy usually induces the multidrug resistance (MDR) of cancer cells. To solve this critical issue, the combination therapy with involved two or more therapeutic modalities is expected to integrate the features and advantages of each therapeutic modality to achieve synergistic therapeutic outcome, 92, 208, 209, 210, 211, 212, 213, 214 which has been broadly explored in abundant therapeutic‐modality combinations including physical‐triggering of TiO 2 nanoplatforms. Based on the high electroconductivity of graphene, we recently loaded TiO 2 nanoparticles onto the surface of graphene (designated as MnO x /TiO 2 –GR–PVP, MnO x for MR imaging) for enhanced and synergistic SDT and photothermal hyperthermia ( Figure 14 a), which was uniformly distributed onto graphene's surface (Figure 14 b, c). 90 On one hand, the high electroconductivity of graphene facilitates the separation of electron (e − ) and hole (h + ) pairs from the energy‐band structure of TiO 2 nanosonosensitizers upon external US irradiation, which could significantly enhance the SDT efficiency on killing the cancer cells. On the other hand, the graphene matrix showed the high photothermal‐conversion performance for hyperthermia, which further synergistically enhanced the SDT efficiency of loaded TiO 2 nanosonosensitizers (Figure 14 d) with high therapeutic biosafety as indicated by neglectable body‐weight changes (Figure 14 e). This paradigm demonstrates that the rational combination of nano‐TiO 2 semiconductors with high electroconductivity nanosystems could enhance the SDT efficiency by facilitating the US‐triggered separation of electron and hole pairs. Figure 14 a) Schematic illustration of the fabrication of MnO x /TiO 2 –GR–PVP composite nanosheets, and their synergistic therapy based on MR/CT/PA multiple imaging‐guided photothermal hyperthermia (808 nm) and enhanced SDT. b) The scheme of loading MnO x and TiO 2 onto the surface of graphene nanosheets, and corresponding c) TEM (left image) and SEM (right image) images. d) The tumor‐volume changes with the prolonged feeding time after varied treatments as shown in the figure, and e) corresponding body‐weight changes in each therapeutic group. Reproduced with permission. 90 Copyright 2017, American Chemical Society. To reverse the MDR of cancer cells, a “nano‐bomb” was designed for US‐triggered multiple and synergistic cancer therapy based on hollow MTNs. 215 Chemotherapeutic drug doxorubicin acting as the ammunition was loaded into MTNs as the ammunition depot, and the surface of MTNs was coated by dsDNA as the safe device to avoid the prerelease of loaded doxorubicin ( Figure 15 ). Especially, the US irradiation on drug‐loaded MTNs achieved multiple effects, including US‐triggered SDT for MTN‐sonosensitized ROS generation, US‐activated drug release, reversal of MDR, and final synergistic cancer treatment. The reversal of MDR of MCF‐7/ADR cancer cells was based on the inhibition of mitochondrial energy supply by the US‐triggered “explosion” of MTNs, causing the substantially suppressed tumor growth. 215 Figure 15 Schematic illustration of US‐triggered combinatorial therapy using a “nano‐bomb, ” including US‐triggered SDT for ROS generation, US‐activated drug release, reversal of MDR, and final synergistic cancer treatment. Reproduced with permission. 215 Copyright 2018, Elsevier. In addition, envelope‐type mesoporous titanium dioxide nanoparticles (MTN) were fabricated with the subsequent loading of docetaxel (DTX) accompanied with a high drug‐loading capacity of ≈26% ( Figure 16 a). 216 Furthermore, β‐cyclodextrin (β‐CD) was anchored onto the surface of MTN as a bulky gatekeeper, which was based on a ROS‐sensitive linker to seal DTX within the mesopores (Figure 16 b). Upon US irradiation, large amounts of ROS were produced by SDT effect to break the ROS‐sensitive linker and then trigger the DTX release from the mesopores. Therefore, the US irradiation not only induced ROS generation for SDT cancer therapy, but also triggered DTX releasing from mesopores for synergistic chemotherapy, which was demonstrated by the synergistic therapeutic outcome where the tumor growth in the synergistic group got the maximum suppression (Figure 16 c). 216 Magnetic core/shell structured Fe 3 O 4 ‐NaYF@TiO 2 nanocomposites were constructed for synergistic chemotherapy by loaded doxorubicin and SDT by the TiO 2 component. 217 The further surface engineering with hyaluronic acid (HA) enabled targeted intracellular transportation, which induced high tumor‐inhibition rate of 88. 36% in synergistic group, much higher than that of the single therapeutic modality such as chemotherapy (28. 36%) and SDT (38. 91%). Figure 16 a) Schematic illustration of stepwise synthesis of MTN@DTX‐CD composite nanosystem. b) The scheme of microstructure of MTN@DTX‐CD and the functionality of each component. c) The tumor‐volume changes of S180 tumor‐bearing mice after varied treatments as indicated in the figure. Reproduced with permission. 216 Copyright 2015, American Chemical Society. TiO 2 ‐based nanoparticles have been explored as the drug‐delivery nanosystems for chemotherapy. 218, 219, 220, 221 For instance, ZnPc@TiO 2 (ZnPc: zinc phthalocyanine) hybrid nanoparticles were employed for the intracellular delivery of anticancer drug doxorubicin based on the electrostatic interaction between drug molecules and nanocarriers. 67 Doxorubicin was also loaded into Fe 3 O 4 @TiO 2 core/shell nanoparticles for synergistic chemotherapy by loaded chemotherapeutic drug and US‐activated SDT by the TiO 2 component as sonosensitizer. 222 The typical strategy only loaded the chemotherapeutic drug molecules onto the surface of TiO 2 nanoparticles, therefore the drug‐loading amount was low and the release in uncontrollable. On this ground, Wu and co‐workers coated a mesoporous silica layer onto the surface of black TiO 2 nanoparticles for achieving synergistic chemotherapy and photothermal hyperthermia. 223 On one hand, the mesoporous silica layer provided the large reservoir for the drug loading/delivery. On the other hand, the black TiO 2 acted as the photothermal‐conversion nanoagents for PTT. By the integration with folic‐acid targeting, the mesoporous silica‐coated black TiO 2 ‐enabled produced the synergistic therapeutic outcome on suppressing the tumor growth against MCF‐7 breast cancer xenograft. 223 TiO 2 nanoparticles could integrate with other functional nanosystems for achieving some specific synergistic therapeutic purposes. For instance, praseodymium (Pr)‐doped TiO 2 on GO nanoplatforms were fabricated by a facile hydrothermal synthesis ( Figure 17 a). 224 First, the sp 2 carbonaceous framework of GO converted NIR light into heat for photothermal hyperthermia (Figure 17 b). Second, the Pr‐doped TiO 2 nanoparticles could absorb more hydroxide ions onto the surface to promote the generation of hydroxyl radicals and suppress the electron–hole recombination. Third, the 4f electron transition of doped Pr achieved the incorporation of additional energy levels in the bandgap of TiO 2, which induced the enhanced photocatalytic activity on killing cancer cells under visible light (450 nm). Fourth, this composite nanosystem could store therapeutic anticancer agents (doxorubicin) for enhanced chemotherapy. Especially, the synergistic chemotherapy, PTT and phototriggered PDT (triple‐therapeutic modality) significantly induced the cancer‐cell death as compared to either monomodal or dual‐modal therapy (Figure 17 c). 224 Figure 17 The scheme of a) synthesizing Pr‐TiO 2 /NGO nanocomposites and b) light‐activated photocatalytic process for ROS production. c) In vitro therapeutic efficacy of different treatments, including mono, dual, and triple‐modality cancer‐cell treatment such as chemotherapy (loaded Dox), PTT, and phototriggered dynamic therapy. Reproduced with permission. 224 Copyright 2015, WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. The previous discussion has mentioned that the UV wavelength range of 320 to 400 nm might cause the phototoxicity and have the low‐penetrating capability. To solve the critical issue of UV‐responsiveness of traditional TiO 2 nanoparticles, zinc phthalocyanine as the intriguing photochemical molecule with high stability, efficiently extended the light window of TiO 2 nanoparticles from UV region to NIR region for phototreatment, which was based on an intercomponent electron transfer between zinc phthalocyanine and titania nanoparticles. 225 Especially, the ROS‐sensitive compound BCBL was conjugated to zinc phthalocyanine‐modified TiO 2 nanoparticles for ROS‐triggered chemotherapy ( Figure 18 ). Upon NIR irradiation, the generated large amounts of ROS triggered the release of loaded BCBL for chemotherapy, which also acted as the toxic species for PDT, inducing the synergistic chemotherapeutic and PDT efficiency. 225 The UV‐activated TiO 2 nanoparticles have been previously demonstrated to reverse the MDR of cancer cells. 226 To further overcome critical issue of UV light for reversal of MDR, doxorubicin was loaded into NaYF 4 :Yb/Tm‐TiO 2 inorganic photosensitizers for simultaneous 980 nm NIR‐activated PDT and intracellular drug delivery. 227 The surface folic acid modification enhanced intracellular uptake of the nano‐photosensitizer and accelerated the doxorubicin release in both drug‐sensitive MCF‐7 and drug‐resistant MCF‐7/ADR cancer cells, inducing the synergistic MCF‐7/ADR tumor‐inhibition rate of up to 90. 33%, significantly higher than that of free doxorubicin. Figure 18 Schematic illustration of synthesizing mTiO 2 –BCBL@ZnPc nanoparticles and their unique therapeutic functionality for cancer therapy, which was based on ROS‐responsive drug‐releasing performance under NIR irradiation by breaking up the phenylboronic acid ether. Reproduced with permission. 225 Copyright 2018, Royal Society of Chemistry. The construction of mesoporous titania‐coated UCNPs is expected to achieve NIR‐triggered PDT and simultaneous chemotherapy for synergistic therapy, originating from UNCPs core and mesoporous titania shell. Mesoporous TiO 2 upconverting nanoparticles (abbreviated as MTUN) were synthesized by direct coating of a mesoporous TiO 2 layer onto the surface of NaGdF 4 :Yb25%, Tm0. 3% as mediated by a middle silica layer ( Figure 19 ). 228 The UCNPs converted NIR irradiation to UV light, which further activated mesoporous TiO 2 layer to generate ROS for inducing cancer‐cell apoptosis. Especially, the well‐defined mesopores of surface TiO 2 layer acted as the drug‐storage reservoirs for drug delivery and chemotherapy, inducing the synergistic NIR‐activated PDT and chemotherapy. Importantly, the HA was anchored onto the surface of this composite nanosystem for targeting cluster determinant 44 (CD44) that was overexpressed on cancer‐cell membrane and achieving controlled drug releasing as triggered by the specific enzyme in tumor region. Based on mesoporous TiO 2 ‐coated UCNPs, a photolabile o ‐nitrobenzyl derivative was incorporated to act as the gate by forming a sensitive linker for avoiding the drug releasing. 229 The NIR‐triggered ROS production not only induced the PDT effect, but also cause the breaking of the sensitive linker for on‐demand drug releasing, leading to synergistic chemotherapy and PDT against cancer cells. To further enhance the drug‐loading capability, rattle‐type UCNPs@Void@TiO 2 nanocomposites were fabricated with large voids between UCNPs core and mesoporous TiO 2 shell, producing TiO 2 ‐based PDT by NIR irradiation and doxorubicin‐induced synergistic chemotherapy. 230 Figure 19 The scheme of the fabrication of NaGdF 4 :Yb, Tm@mTiO 2 core/shell nanoparticles the novel drug‐delivery system for simultaneous and synergistic chemotherapy and NIR‐triggered PDT. Reproduced with permission. 228 Copyright 2014, WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. The rational structure design of TiO 2 ‐based nanoplatforms could endow them with more therapeutic functionalities. 231 It has been demonstrated that anticancer drug doxorubicin‐loaded TiO 2 nanoparticles overcame the MDR of breast cancer cells (MCF‐7/ADR) by bypassing the P‐glycoprotein‐mediated doxorubicin‐pumping system. 218 Furthermore, TiO 2 ‐based composite nanosystems (DOX@TiO 2− x @PAD‐Cy5. 5, PDA: poly dopamine) were stepwise synthesized for simultaneous fluorescent/PAT bimodal tumor imaging and NIR‐activated chemo/photodynamic/photothermal combinatorial therapy ( Figure 20 a). 86 Because of the high photothermal‐conversion capability of TiO 2− x matrix, the tumor temperature was rapidly elevated upon NIR irradiation (808 nm, Figure 20 b, c). Especially, the NIR irradiation of DOX@TiO 2− x @PAD‐Cy5. 5 generated ROS for efficient PDT, and the presence of mesopores in TiO 2− x matrix provided the reservoirs for the encapsulation and controllable delivery of therapeutic anticancer drugs (doxorubicin) with unique responsiveness to endogenous mild acidity of TME and exogenous NIR irradiation. The simultaneous and synergistic triple therapy induced the high tumor‐suppressing outcome with almost complete tumor eradication (Figure 20 d), which was caused by DOX‐induced DNA damage and PDT/PTT‐induced mitochondrial dysfunction/change of membrane. 86 Figure 20 a) Schematic illustration of synthesizing DOX@TiO 2− x @PAD‐Cy5. 5 nanocomposites, and their unique functionality for dual‐mode imaging‐guided synergistic chemotherapy, PTT, and PDT. b) Infrared thermal images of MDA‐MB‐231 tumor‐bearing mice after the administration of DOX@TiO 2− x @PAD‐Cy5. 5 nanocomposites followed by NIR irradiation, and c) corresponding temperature‐elevating profiles. d) The tumor‐volume change as a function of feeding time after the different treatments as indicated in the figure. Reproduced with permission. 86 Copyright 2017, American Chemical Society. Both the microstructure and functionality of TiO 2 ‐based nanoplatforms could be rationally designed for achieving synergistic cancer‐therapeutic outcome, which is highly unique in TiO 2 nanosystems because of their easy synthesis and specific semiconductor nature. It should be noted that the surface inertness of TiO 2 nanoparticles makes their surface engineering difficult, which typically seeks the help from other organic or inorganic functionality by some unique synthetic strategies, such as inorganic silica coating or hydrophobic–hydrophobic interaction. It should be noted that the multifunctionalization design of TiO 2 nanoparticles should be based on the practical requirements of clinical use because the complex design of these composite nanosystems usually causes the difficulty for potential clinical translation. 8 Diagnostic‐Imaging of Titania for Therapeutic Guidance and Monitoring It has been well demonstrated that some metal oxides nanoparticles can act as the contrast agents for enhancing the diagnostic‐imaging resolution and sensitivity of diverse imaging modalities, such as manganese oxide (T 1 ‐weighted MR imaging), 233, 234, 235, 236, 237, 238 gadolinium oxide (T 1 ‐weighted MR imaging), 239, 240, 241, 242 iron oxide (T 2 ‐weighted MR imaging), 243, 244, 245, 246, 247, 248 and tantalum oxide (CT imaging). 249, 250, 251 Titania nanoparticles have been seldom explored for enhancing the contrast of various diagnostic‐imaging modalities because of lacking the characteristic physiochemical properties. Fortunately, the fast advances of material‐synthetic chemistry and nanomedicine make it possible based on two typical strategies. On one hand, the structure of titania nanoparticles can be tuned with contrast‐enhanced imaging functionality. On the other hand, these titania nanoparticles can be integrated with some imaging contrast agents for achieving some specific imaging purposes. It is intriguing that the diagnostic‐imaging capability of titania nanoparticles can play the specific role for precise therapeutic guidance and monitoring, which is promising for enhancing the therapeutic efficiency and mitigating the damage to the surrounding normal tissue/cell. The aforementioned discussion has revealed that the oxygen‐deficient black titania nanoparticles could be endowed with photothermal‐conversion capability at NIR range. This property has been generally developed for contrast‐enhanced PA imaging, such as 2D MXene, 130 black phosphorous, 252, 253 MoS 2, 254, 255 and Au nanoparticles. 256, 257, 258 On this ground, oxygen‐deficient black titania nanoparticles were explored as the contrast agents for PA imaging after the injection into tumor‐bearing mice. 86 It has been found that the obvious contrast enhancement was observed in tumor region after intratumoral injection of these black titania nanoparticles ( Figure 21 a), demonstrating their imaging capability. As another paradigm, titania nanoparticles were integrated with manganese oxide nanoparticles to construct a composite nanosystem, where the integrated manganese oxide nanoparticles acted as the contrast agents for T 1 ‐weighted MR imaging and guided the SDT of cancer as contributed by the titania component in the composite nanosystem (Figure 21 b). 90 Especially, titania nanoparticles were simultaneously conjugated with fluorescent moieties and Gd‐based chelates for labeling HeLa cancer cells by both fluorescence microscopy and MR imaging, showing the dual‐imaging capability of the titania‐based composite nanosystem (Figure 21 c). 232 Additionally, the construction of magnetic Fe 3 O 4 –TiO 2 nanocomposites achieved simultaneous T 2 ‐weighted MR imaging and PDT against MCF‐7 cancer cells. 259 Figure 21 a) In vivo PA imaging of tumor‐bearing mice before and after intratumoral administration of DOX@TiO 2− x @PDA‐Cy5. 5 nanocomposites for prolonged durations. Reproduced with permission. 86 Copyright 2017, American Chemical Society. b) In vivo T 1 ‐weighted MRI signal‐intensity variations after intravenous administration of MnO x /TiO 2 –GR–PVP for different time intervals. Reproduced with permission. 90 Copyright 2017, American Chemical Society. c) Schematic illustration of surface conjugation of fluorescent moieties and Gd chelates for concurrent fluorescence and MR imaging, and the further UV‐activated production of hydroxyl radicals for killing the cancer cells. Reproduced with permission. 232 Copyright 2011, American Chemical Society. 9 Biocompatibility and Biosafety of Theranostic Titania The previous discussion has mentioned that titanium (Ti) element is one of the most biocompatible elements present in nature, as demonstrated by the fact that TiO 2 ‐based micro/nanoparticles have been broadly used in food, cosmetics, and sunscreen and Ti‐containing metal alloys has been used as the medical implantation devices. It is highly expected that these Ti‐based nanoparticles as discussed in the review are also biocompatible in biomedical applications. However, it is generally accepted that the particles would induce some abnormal biological behaviors and effects or even toxicity when their particle size are reduced into nanoscale. Therefore, systematic investigation of these titania nanoparticles should be further conducted to guarantee their high biocompatibility and biosafety for further clinical translation. Actually, the biological effects and biocompatibility of titania‐based compound or micro/nanoparticles have been broadly investigated in the past decade, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 274, 275, 276 which has also been summarized and discussed in some excellent reviews. 277, 278, 279 Therefore, this review herein focuses more on the biocompatibility and biosafety of some rationally designed novel titania‐based nanoplatforms with unique responsiveness to exogenous physical irradiations. Our previous work has demonstrated the SDT effect of MTNs with well‐defined mesopores for combating cancer. 84 Furthermore, we systematically assessed the in vivo biocompatibility of these MTNs on healthy mice. It has been found that either single high dose at 150 mg kg −1 or repeated dose at as high as total 400 mg kg −1 exhibited no obvious in vivo toxicity, as demonstrated by the hematology markers and blood biomedical parameters where no significant changes were monitored as compared to control groups without any treatments, indicating the high biocompatibility of these MTNs. 84 For titania‐based nanocomposites, it has been demonstrated that Au 25 /B‐B‐TiO 2− x nanotubes not only showed low hemolytic effect on red blood cells, but also revealed their low cytotoxicity to L929 cells (mouse fibroblast cell line) and HeLa cells (human cervical cancer cell line). 119 The targeting titania‐based nanocomposites anti‐EGFR–PEG–TiO 2 –UCNPs were demonstrated to have no major sub‐acute or long‐term toxicity as revealed in no significant blood biomedical, hematological or histopathological changes at the dose of 50 mg kg −1. 104 One of the unique advantages of titania‐based nanoplatforms with responsiveness to exogenous physical irradiation is the high therapeutic biosafety. These nanoplatforms can only induce the toxic effect under the tumor sites as irradiated by external diverse physical triggers while other organs or tissues without physical irradiation will not be damaged even if these titania nanoparticles are accumulated into them. This high therapeutic biosafety is expected to significantly mitigate the side effects of traditional therapeutic modalities such as chemotherapy where the toxic drugs or substances are usually introduced, causing the severe side effects. Our results have demonstrated that SDT against cancer with the assistance of black TiO 2− x nanosonosensitizers induced no obvious pathological changes of the major organs after the therapeutic process, demonstrating the high therapeutic biosafety of this SDT modality. 85 In addition, the combinatorial and synergistic SDT and chemotherapy of DTX‐loaded MTN were demonstrated to be featured with sustainably decreased side effects of loaded chemotherapeutic drug DTX by avoiding the spleen and hematologic toxicity to tumor‐bearing mice. 216 10 Conclusions and Outlook As one of the mostly explored biocompatible metal oxides in biomedicine, TiO 2 nanosystems are featured with their intrinsic physiochemical properties for some specific theranostic applications, which is mainly originated from their semiconductor nature. Traditional strategies mainly focus on the UV light irradiation of TiO 2 nanoparticles for PDT by forming the electrons (e − ) and holes (h + ) pairs from the energy‐band structure and then inducing the ROS generation for killing the cancer cells. The potential phototoxicity and low tissue‐penetrating depth of UV light severely limit the further in vivo biomedical applications of these TiO 2 nanosystems. The fast development of nanosynthetic material chemistry enables the fine tuning of the composition, nanostructure, and property of TiO 2 nanosystems possible. Importantly, the intriguing development of theranostic nanomedicine promotes the generation of diverse novel therapeutic modalities, which can be easily extended to TiO 2 nanosystems for achieving physiochemical property‐oriented bioapplications, especially for cancer treatment. On this ground, this review mainly focuses on the very‐recent development of TiO 2 ‐based nanoplatforms for cancer treatment with specific focuses on the NIR‐triggered photothermal hyperthermia, NIR‐activated PDT, X‐ray/CR‐activated deep‐seated PDT, US‐triggered sonodynamic therapy, and some synergistic therapeutic paradigms. Most of these novel therapeutic modalities are based on the semiconductor nature of TiO 2 nanoplatforms, together with the defect modulation for PTT ( Table 1 ). Table 1 Paradigms of nanotitania semiconductors for exogenous physical irradiation‐activated tumor‐specific therapy Nanotitania Irradiation source Therapeutic modality Performance Refs. Targeted TiO 2 Visible light Photodynamic therapy Enhanced intracellular uptake and visible light‐activated PDT for damaging the cell membrane 109 N3‐TiO 2 Light irradiation (365 nm) Photodynamic therapy Enhanced hydroxyl radical production under the hypoxic condition 111 UCNPs@TiO 2 NIR irradiation (980 nm) Photodynamic therapy Inducing the substantial tumor suppression with high therapeutic efficiency by NIR irradiation 88 UCNPs@TiO 2 NIR irradiation (980 nm) Photodynamic therapy Efficiently killing the cancer cells both in vitro and in vivo by NIR activation 87 Au 25 /B–TiO 2− x NIR irradiation (650 nm) Photodynamic therapy Improved tumor‐suppressing effect based on photocatalytic synergistic effect by NIR irradiation 119 TiO 2 –Au–graphene Simulated sunlight Photodynamic therapy Triggering a series of toxicological effects on killing B16F1 melanoma cells against B16F1 tumor xenograft 121 Black TiO 2− x NIR irradiation (808 nm) Photothermal therapy Elevating the tumor temperature and inducing tumor‐tissue hyperthermia 147 Nb‐doped TiO 2 NIR irradiation (1064 nm) Photothermal therapy Ablating the tumor tissue and suppressing tumor growth at NIR‐II biowindow 148 Au–TiO 2 X‐ray Photodynamic therapy Inducing a synergistic therapeutic outcome with high tumor‐suppressing effect and improved survival rate of mice 92 TiO 2 –Tc‐Tf Cerenkov radiation Photodynamic therapy Suppressing tumor growth and improved survival rate with deep tissue‐penetrating depth 174 Dextran–TiO 2 Cerenkov radiation Photodynamic therapy Efficiently killing the cancer cells and improving the survival rate of tumor‐bearing mice 178 Mesoporous TiO 2 Ultrasound Sonodynamic therapy Inducing tumor‐suppressing effect against 4T1 tumor xenograft 84 Hydrophilized TiO 2 Ultrasound Sonodynamic therapy Enhancing immune response, suppressing tumor growth and destroying tumor microvasculature 207 Au–TiO 2 Ultrasound Sonodynamic therapy Improved SDT effect against cancer by trapping the sono‐excited electrons 91 Black TiO 2− x Ultrasound and NIR (1064 nm) Sonodynamic therapy and photothermal therapy Synergistic SDT and PTT on killing the cancer cells accompanied by enhanced SDT effect by oxygen‐deficient titania layer 85 MnO x /TiO 2 ‐GR Ultrasound and NIR (808 nm) Sonodynamic therapy and photothermal therapy Decreasing the re‐combination of electrons and holes for enhanced SDT effect on suppressing the tumor growth 90 Targeted TiO 2 Visible light Photodynamic therapy Enhanced intracellular uptake and visible light‐activated PDT for damaging the cell membrane 109 John Wiley & Sons, Ltd. The unique physiochemical property of TiO 2 nanosystems has achieved high therapeutic efficacy of aforementioned cancer‐therapeutic modalities, which is difficult to be achieved in other metal oxides such as SiO 2 nanoparticles and superparamagnetic Fe 3 O 4 nanosystems. Although these TiO 2 ‐based novel therapeutic modalities are highly promising, it should be noted that they are still in infancy and at the preliminary stage. The further clinical translation is still facing some critical challenges to be resolved in near future as discussed in detail in the following subsections ( Figure 22 ). Figure 22 The summative scheme of previous work, current status, and future progress on physical irradiation‐activated titania nanoparticles for versatile biomedical applications, especially on combating cancer. 10. 1 Fabrication of TiO 2 Nanosystems The synthetic process for desirable TiO 2 nanosystems is a bit more difficult as compared to other metal oxides such as SiO 2 and Fe 3 O 4 because the hydrolysis of titanium precursors is very fast in most cases. Therefore, their morphology and nanostructure are difficult to be precisely controlled. Especially, some fabrication process requires high‐temperature treatment such as the metal reduction to synthesize oxygen‐deficient black TiO 2− x nanoparticles, which avoidably causes the aggregation of TiO 2 nanoparticles with low dispersity. In addition, there lacks the specific surface chemistry for the surface modification of these TiO 2 nanoparticles, which, however, is necessary for guaranteeing their high stability in physiological solution or achieving the high tumor accumulation by targeting modification. These fabrication difficulties might lower the therapeutic efficacy of TiO 2 ‐based therapeutic modalities or cause the biosafety issue. Therefore, two strategies are suggested to fabricate desirable TiO 2 nanosystems. On one hand, the precise controlling of the sol–gel process during the synthesis of TiO 2 nanosystems should be taken into consideration, which can fabricate TiO 2 nanoparticles with desirable composition, nanostructure, and physiochemical property. On the other hand, the surface of as‐synthesized TiO 2 nanoparticles could be endowed with some functional groups for further surface modification, which can be achieved by some specific treatments such as PEGylation or silica/mesoporous silica coating. It means that the surface chemistry can be controlled by using some “mediators” for satisfying the surface‐modification requirements. 10. 2 Biocompatibility Issue of TiO 2 Nanosystems As compared to SiO 2 and Fe 3 O 4 metal oxides, the biological effects and biocompatibility of TiO 2 nanoparticles are significantly less explored, which severely hinders their further clinical translation because of the lack of solid biocompatibility data. It is considered that TiO 2 has been used in colorant in food, cosmetics and sunscreen and Ti‐containing metal alloys has been broadly used as the medical implantation devices, therefore these TiO 2 nanosystems might possess the relatively high biocompatibility. Our previous results have demonstrated the low in vivo toxicity of either mesoporous TiO 2 nanoparticles or black TiO 2− x nanosystems. These preliminary results are encouraging, but the further systematic investigations on the biocompatibility and biosafety issue are still highly urgent and necessary, which is the following research target in the near future. 10. 3 Not Very Clear Mechanism on TiO 2 ‐Based Novel Therapeutic Modalities To overcome the drawbacks of traditional UV light irradiation for activating TiO 2 nanoparticles, NIR light, X‐ray, CR, and US have recently been explored to activate TiO 2 nanoparticles. The therapeutic performance is highly encouraging, but the underlying mechanism has not been fully revealed. Most of the results are mainly based on some phenomena in vitro. The exact in vivo therapeutic process is still highly challenging to monitor and determine because of the lack of adequate techniques and the complex in vivo environment, making it difficult to further optimize and enhance the therapeutic efficacy due to the lack of precise knowledge on the related mechanism. Therefore, the further therapeutic‐efficacy optimization requires the knowledge accumulation of the therapeutic mechanism, which is highly difficult but significantly urgent. 10. 4 Influence of Crystalline Types of TiO 2 Nanoparticles on Therapeutic Performance It has been well documented that TiO 2 nanoparticles have varied crystalline types with different physiochemical properties. However, at current stage, most therapeutic applications of titania nanoparticles did not consider the influences on the crystalline types of titania nanoparticles because the therapeutic use of titania nanoparticles as the emerging inorganic nanoplatform is still in the infancy, which still requires the following systematic investigations on the detailed underlying mechanism of the influence of crystalline types, precise structure/composition control and the following performance optimization. 10. 5 More Biomedical Applications and Close Collaborations Most of the therapeutic applications of TiO 2 ‐based nanoparticles are related to cancer treatment except the antibacterial applications in some specific conditions. The intriguing performances of TiO 2 nanoparticles with unique responsiveness to some types of external irradiations provide more opportunities for broader biomedical applications, such as tissue engineering, wound healing, gene therapy, stem‐cell therapy, etc. Therefore, the following fundamental researches should focus more on other specific bioapplications. However, it should be noted that such a multidisciplinary research requires the close collaborations of researchers/scientists with different background, which can guarantee each step of the clinical translation of TiO 2 ‐based nanosystems is involved with professional researchers/experts with adequate knowledge to deal with the critical issues during the translation. TiO 2 ‐based nanoplatforms have emerged as one of the most promising therapeutic nanplatforms for cancer treatment because of their intrinsic physiochemical property and relatively high biocompatibility, which might be potential to act as the alternative substitute for mostly explored SiO 2 and Fe 3 O 4 metal oxides. This intriguing expectation should come true after further systematic investigations on their fabrication, biosafety evaluation and performance optimization. Especially, the nanomedical applications of TiO 2 ‐based nanoplatforms also inspire the researchers to explore more functional inorganic nanosystems with some unique property‐oriented biomedical applications. Therefore, it is highly believed that these TiO 2 ‐based nanoplatforms will find their broad bioapplications in theranostic nanomedicine and personalized biomedicine based on the fast developments of material science, chemistry and the multidisciplinary theranostic nanomedicine. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201801260
| 2,018
|
Advanced Science
|
Ion Therapy: A Novel Strategy for Acute Myocardial Infarction
|
Abstract Although numerous therapies are widely applied clinically and stem cells and/or biomaterial based in situ implantations have achieved some effects, few of these have observed robust myocardial regeneration. The beneficial effects on cardiac function and structure are largely acting through paracrine signaling, which preserve the border‐zone around the infarction, reduce apoptosis, blunt adverse remodeling, and promote angiogenesis. Ionic extracts from biomaterials have been proven to stimulate paracrine effects and promote cell–cell communications. Here, the paracrine stimulatory function of bioactive ions derived from biomaterials is integrated into the clinical concept of administration and proposed “ion therapy” as a novel strategy for myocardial infarction. In vitro, silicon‐ enriched ion extracts significantly increase cardiomyocyte viability and promote cell–cell communications, thus stimulating vascular formation via a paracrine effect under glucose/oxygen deprived conditions. In vivo, by intravenous injection, the bioactive silicon ions act as “diplomats” and promote crosstalk in myocardial cells, stimulate angiogenesis, and improve cardiac function post‐myocardial infarction.
|
1 Introduction Eliminating vascular obstruction and promoting angiogenesis can reduce myocardial apoptosis and necrosis, which is critical in the treatment of acute myocardial infarction (AMI). 1 Despite the most advanced pharmacological and medical device treatment methods were used, neither of these promote cardiac angiogenesis and neovascularization nor stimulate the endogenous repair leading to steadily increasing incidence of heart failure as well as poor prognosis. 2 Although some cell‐based or noncellular biomaterial based approaches and cotransplant cells 3, 4 with synthetic biomaterials strategies are under trial, they are still facing great challenge for clinical applications. In particular, the poor retention and engraftment of transplanted cells as well as the difficulty in implantation and degradation of biomaterials discounts the therapeutic effect dramatically. 3, 5 Recent studies have shown that one remarkable mode of therapeutic action is that the injected stem cells as well as its membrane function through the secretion of paracrine factors 6, 7 and trigger intracellular protective/regenerative pathways in the host cells 8, 9 to promote endogenous repair. These results indicate that stimulating the crosstalk among cardiac host cells and promoting endogenous angiogenesis are essential in re‐establishing blood supply flow and may rescue the existing working cardiomyocytes and improve the prognosis of patients post‐AMI. Therefore, development of strategies or substances to regulate cardiac host cells, in particular the interaction between endothelia cells and cardiac cells and enhance angiogenesis are critical for myocardial infarction. Previous works including our studies have found that silicate‐based biomaterials, especially bioactive glasses (BGs) and the silicate ceramics (CS) can promote cell–cell communications by affecting gap junction associated Connexin 43 (Cx43) mediated endothelial cell behavior, 10 improving the interactions between human umbilical vein endothelial cells (HUVECs) and bone marrow stromal cells (BMSCs) or human dermal fibroblasts (HDFs) and stimulating angiogenesis both in vitro and in vivo, and further enhance tissue regeneration. 11, 12, 13 Moreover, the bioactivity of these silicate based biomaterials mainly rely on the released ions, in particular the silicon ions. 11 Inspired by the functions of these bioactive materials and stem cell therapy, we proposed a novel therapeutic strategy—“ion therapy, ” in which bioactive ions derived from silicate biomaterials are intravenously injected and act as “diplomates” to regulate cell–cell communications and angiogenesis. Our hypothesis is that the intravenously injected ions may on one side activate cardiac cells and on the other side stimulate angiogenesis, and finally inhibit cell necrosis and rebuilt the blood supply. In this study, the therapeutic effects of “ion therapy” on the endogenous myocardium repair and neovascularization were investigated, and we demonstrated for the first time, that the intravenously injected Si ions indeed have significant therapeutic effect on AMI by regulating cell behaviors and angiogenesis. 2 Results 2. 1 Ion Concentrations of CS Extracts and Effects of Silicon‐Enriched Ion Extracts on Neonatal Rat Cardiomyocytes Cell Viability To investigate the effects of bioactive ions on neonatal rat cardiomyocytes (NRCMs) cell viability, CS extracts and serial dilutions of extracts (1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, and 1/256) were prepared by phosphate buffered solution (PBS) or Dulbecco's modified Eagle medium (DMEM) as previously described. The ion concentrations of Ca, P, and Si were determined by inductively coupled plasma atomic emission spectroscopy (ICP‐AES) and the results are shown in Table 1. It is found that Ca and P ion concentrations in DMEM diluted CS extracts and 1/16CS were much lower than that of DMEM while Si ion concentrations in CS extracts and the dilutions at 1/16, 1/32, 1/64, 1/128, and 1/256 ratios were significantly higher than that of DMEM. In DMEM diluted CS at 1/32, 1/64, 1/128, and 1/256 ratios, no significant differences were found in Ca and P concentrations when compared to DMEM. Next, we investigate the ion concentrations of PBS diluted CS. No significant difference was found in Ca concentrations between all groups and P ion concentrations in PBS diluted CS extracts, 1/16CS and 1/32CS were much lower than that of PBS while no difference were found at 1/64, 1/128, and 1/256 ratios. Furthermore, Si ion concentrations at all diluted ratios were significantly higher than that of PBS. It is interesting to note that the Ca and P concentrations in original CS extracts in DMEM or PBS are not only clearly lower than that in DMEM or PBS, respectively, but also lower than that in the corresponding extracts dilutions. This is possibly due to the calcium phosphate precipitation in DMEM or PBS induced by silicates as reported previously. 14, 15 Then, in order to evaluate the protective effect of Si ions, NRCMs cell viability under normoxia and glucose/oxygen deprived conditions were examined in the presence of silicon‐enriched ion extracts. Figure 1 a reveals that NRCMs cell viability cultured for up to 5 d in 1/8, 1/16, 1/32, and 1/64 CS extracts diluted with DMEM are higher than that of control medium under normoxia condition while no difference was observed between the original CS extract and the control suggesting a stimulatory effect of Si ions in certain concentration range on NRCMs under normal culture condition. In contrast, under glucose/oxygen deprivation condition in PBS, NRCM cell viability was apparently decreased when compared to normal condition. It is interesting to note that the viability of the cells cultured in CS extracts diluted at 1/8, 1/16, 1/32, and 1/64 CS ratio are higher than that in PBS and the highest cell viability is found at the concentrations between 1/32 and 1/64 ratios (Figure 1 b). These observations indicated that glucose/oxygen deprivation harmed NRCMs and resulted in a decrease of the cell viability whereas silicon ions can alleviate this damage and have potential protective effect on NRCMs. Table 1 Ion concentrations of CS extracts diluted with DMEM medium and PBS Ca [µg mL −1 ] P [µg mL −1 ] Si [µg mL −1 ] DMEM 59. 33 ± 0. 97 25. 22 ± 0. 34 0. 02 ± 0. 01 (1)CS in DMEM 5. 11 ± 0. 25 b) 6. 37 ± 0. 09 b) 113. 59 ± 0. 30 b) CS/1/16 in(1) 56. 19 ± 0. 46 a) 23. 92 ± 0. 38 a) 6. 66 ± 0. 05 b) CS/1/32 in(1) 57. 89 ± 0. 62 24. 98 ± 0. 68 3. 36 ± 0. 04 b) CS/1/64 in(1) 58. 92 ± 0. 73 25. 24 ± 0. 70 1. 63 ± 0. 02 b) CS/1/128 in(1) 59. 47 ± 0. 42 25. 01 ± 0. 28 0. 80 ± 0. 02 b) CS/1/256 in(1) 59. 62 ± 0. 55 25. 13 ± 0. 52 0. 41 ± 0. 01 b) PBS 0. 06 ± 0. 00 274. 44 ± 1. 00 0. 12 ± 0. 01 (2)CS in PBS 0. 05 ± 0. 01 32. 52 ± 0. 44 d) 283. 88 ± 3. 43 d) CS/1/16 in(2) 0. 08 ± 0. 01 260. 22 ± 0. 55 d) 17. 42 ± 0. 37 d) CS/1/32 in(2) 0. 06 ± 0. 00 268. 08 ± 0. 63 c) 9. 08 ± 0. 75 d) CS/1/64 in(2) 0. 06 ± 0. 00 270. 91 ± 0. 81 4. 53 ± 0. 18 d) CS/1/128 in(2) 0. 05 ± 0. 00 273. 06 ± 0. 10 2. 26 ± 0. 03 d) CS/1/256 in(2) 0. 05 ± 0. 00 273. 23 ± 0. 79 1. 25 ± 0. 02 d) a) p < 0. 05 and b) p < 0. 001 when compared with concentration of ions in DMEM medium. c) p < 0. 05 and d) p < 0. 001 when compared with concentration of ions in PBS; All data are obtained from three independent experiments; mean ± SD. John Wiley & Sons, Ltd. Figure 1 Effects of silicon‐enriched ion extracts on NRCMs cell viability under normal and glucose/oxygen deprived conditions. a) NRCM cell viability cultured in different CS extracts diluted with DMEM for 1, 3, and 5 d under normoxia conditions (* p < 0. 01 vs control group for Day 1; † p < 0. 01 vs control group for Day 3; ‡ p < 0. 01 vs control group for Day 5; n = 6 for each group). b) NRCMs cell viability cultured in different CS extracts diluted with PBS for 90 and 120 min under glucose/oxygen deprived conditions. Control group was cultured in normal condition for an equivalent time (* p < 0. 01 versus control group for 90 min; † p < 0. 01 vs control group for 120 min; ‡ p < 0. 01 vs PBS group for 90 min; § p < 0. 01 vs PBS group for 120 min; n = 6 for each group). All data are obtained from three independent experiments; mean ± SD. 2. 2 Effects of Silicon‐Enriched Ion Extracts on Apoptosis of NRCMs under Glucose/Oxygen‐Deprived Conditions In Vitro 2. 2. 1 Terminal Deoxynucleotidyl Transferase dUTP Nick End Labeling Staining of NRCMs under Glucose/Oxygen Deprived Conditions In vitro, the glucose/oxygen deprivation induced NRCMs apoptosis was used to mimic the AMI condition in vivo. Cardiomyocytes apoptosis was investigated by TUNEL staining and the percentage of TUNEL positive cells were calculated. The results are shown in Figure 2 a, b. It is clear to see that under glucose/oxygen deprived conditions, NRCMs went into apoptosis after culturing for 90 min, and with the prolonged culture time up to 120 min, the apoptosis was clearly increased as seen in the increase of the numbers of TUNEL‐positive cells (Figure 2 a). In contrast, a significant decrease of TUNEL‐positive cells was observed in the cells treated with 1/64 CS extract both after 90 and 120 min culturing indicating an inhibitory effect of the bioactive ions on NRCMs apoptosis (Figure 2 a). Most interestingly, the inhibitory effect of Si ions was slightly stronger in 120 min group than that in 90 min group. This may be because the damage of the cells in glucose/oxygen deprived conditions for 90 min was much lower than that for 120 min so that the protective effect of silicon ion extracts was not that obvious. In addition, the effect of silicon‐enriched ion extracts on ROS level and proliferation of NRCMs under glucose/oxygen deprived conditions was also investigated. The results revealed that the silicon‐enriched ion extract clearly decreased the expression of ROS in NRCMs (Figure S4, Supporting Information). The gene expression of cTnT and Myh6 was increased in 1/64CS treated group in normal condition. However, under glucose/oxygen deprivation condition, only cTnT gene was significantly increased in 1/64CS group (Figure S2, Supporting Information). Figure 2 Effect of silicon‐enriched ion extract on apoptosis of NRCMs under glucose/oxygen deprived conditions in vitro. a) Representative pictures of NRCMs with TUNEL staining (green) and DAPI (blue) under glucose/oxygen deprived conditions in vitro. Red arrows show TUNEL‐positive NRCMs. Scale bar represents 50 µm and all data are obtained from three independent experiments. b) Quantitative analysis of TUNEL‐positive NRCMs (10 pictures for each group). * p < 0. 05 versus PBS‐90 min, † p < 0. 05 versus PBS‐120 min; mean ± SD. 2. 2. 2 Expression of Apoptotic‐Associated Mitogen‐Activated Protein Kinase Family Proteins and Cleaved‐Caspase 3 in NRCMs under Glucose/Oxygen Deprived Conditions In Vitro To study the underlying mechanism, we then determined mitogen‐activated protein kinase (MAPK) family proteins and cleaved‐caspase 3 protein expression in glucose/oxygen deprived induced apoptosis of NRCMs. It is known that p38 is involved in cell apoptosis and here we observed that silicon enriched 1/64CS extract affected p38 expression of NRCMs by inhibiting the phosphorylation of p38 ( Figure 3 a, b), while the expression of nonphosphorylated p38 was not affected. The time and duration of p38 phosphorylation in different conditions varied. At 90 min, we found that p38 phosphorylation was clearly inhibited by 1/64CS extract, while the inhibition disappeared after culturing for 120 min. This is in consist with the in vivo result that the expression of phosphorylated P38 protein was also increased in the border area of infarcted myocardial (Figure S8, Supporting Information). We further found that ERK1/2, another important protein which is known to inhibit cell apoptosis, was also affected by 1/64CS extract. The expression of the phosphorylated ERK1/2 of NRCMs was significantly enhanced both after 90 and 120 min culturing. In contrast, the expression of p‐JNK, AKT, and p‐AKT was not affect by 1/64CS extract (Figure 3 a, b). More interestingly, we found that the 1/64CS extract can also regulate the expression of another major protein of downstream apoptotic effector molecule, the cleaved caspase‐3 protein. As shown in Figure 3 c, d, the expression of cleaved caspase‐3 expression of NRCMs was significantly downregulated as compared to that in PBS groups both after 90 and 120 min culturing. Generally, these findings provided evidences that silicon‐enriched ion extract may execute post‐traumatic cardiac protection through attenuating cardiomyocyte apoptosis by regulating apoptosis‐associated proteins expression. Figure 3 Effect of silicon‐enriched ion extract on the expression of apoptotic‐associated MAPK family proteins and cleaved‐caspase 3 in NRCMs under glucose/oxygen deprived conditions in vitro. a) MAPK family protein expression measured by western blot and GAPDH was served as the loading control. b) Quantification of bands by densitometry. c) Cleaved‐caspase 3 protein expression measured by western blot and GAPDH was served as the loading control. d) Quantification of bands by densitometry. Data are obtained from three independent experiments. ** p < 0. 01 versus PBS‐90 min, †† p < 0. 01 versus PBS‐120 min; mean ± SD. 2. 3 Effect of Silicon‐Enriched Ion Extract on the Expression of Gap Junction Associated Cx43 in NRCMs under Glucose/Oxygen Deprived Conditions In Vitro Gap junction associated Cx43 is known as an important protein in cardiomyocyte metabolism, and in particular it may play a protective role in myocardial infarction. Glucose/oxygen deprived resulted in a significant decrease of Cx43 expression of the NRCMs cultured for 120 min in PBS as compared to that of PBS‐90 min group. Silicon‐enriched ion extract significantly reduced the suppression effect of glucose/oxygen deprived on Cx43 expression in both 90 and 120 min groups as the fluorescence intensity of Cx43 in silicon‐enriched ion extract treatment group was remarkably enhanced ( Figure 4 a). Moreover, both Cx43 protein and gene expression evaluated by Western blot and real‐time PCR analysis also confirmed the positive function of Si ions. We observed that silicon‐enriched ion extract significantly upregulated Cx43 protein and gene expression (Figure 4 b, c). As Cx43 is widely expressed in heart gap junctions and is purported to play a crucial role in the synchronized contraction of the heart, these results indicate the potential roles of Si ions in promoting cell–cell communications thus enhancing cardiac function post‐AMI. Figure 4 Effect of silicon‐enriched ion extract on the expression of gap junction associated Cx43 in NRCMs under glucose/oxygen deprived conditions in vitro. a) Representative images of Cx43 immunofluorescence staining of NRCMs cultured in silicon‐enriched ion extracts under glucose/oxygen deprived conditions for 90 and 120 min, respectively. Scale bars represent 25 µm. b) Protein expression of Cx43 measured by western blot and GAPDH was served as the loading control. c) Gene expression of Cx43 in NRCMs under glucose/oxygen deprived conditions measured by RT‐qPCR. All data are obtained from three independent experiments. * p < 0. 01 versus PBS‐90 min; mean ± SD. 2. 4 Effect of Silicon‐Enriched Ion Extract on Vascular Endothelial Growth Factor (VEGF)‐Mediated Angiogenesis of NRCMs and HUVECs Cocultures under Glucose/Oxygen Deprived Conditions In Vitro The proangiogenesis effect of silicon‐enriched ion extract was observed by using NRCMs and HUVECs cocultures under glucose/oxygen deprived environment in vitro. Figure 5 a, b shows that silicon‐enriched ion extracts of CS stimulated angiogenic responses in the coculture system. The vWF‐staining revealed that typical capillary‐like networks were formed in the cocultures both with and without bioactive ions, but it is much more evident in 1/64CS extract group as tube numbers formed in 1/64 CS extract group were markedly higher than that in control group. Next, in order to clarify whether the angiogenesis effect is mainly mediated by VEGF and whether this phenomenon was mainly functioned by paracrine effect, we explored the VEGF and kinase insert domain receptor (KDR) gene expression, the receptor of VEGF, in the monocultures and cocultures of NRCMs and HUVECs under glucose/oxygen deprived conditions. As shown in Figure 5 c, d, in monocultured NRCMs no significant difference of VEGFA gene expression was seen between PBS and 1/64CS groups while in monocultured HUVECs, VEGF 165 gene expression was increased in 1/64CS extract. In coculture system, VEGFA expression by co‐NRCMs and VEGF 165 expression by co‐HUVECs were projected to increase in 1/64CS extract. Consist with VEGF expression results, KDR expression in co‐HUVECs was also increased correspondingly (Figure 5 e). These findings also verify the confirmed view that paracrine effect plays a pivotal role in the recovery of myocardial post‐AMI. Figure 5 Effect of silicon‐enriched ion extract on VEGF‐mediated angiogenesis of NRCMs and HUVECs co‐cultures under glucose/oxygen deprived conditions in vitro. a) Representative fluorescence images of vWF‐stained tube formation of co‐HUVECs under glucose/oxygen deprived conditions in vitro. Scale bars represent 75 µm. b) Quantification of tune numbers in coculture systems (10 pictures for each group). * p < 0. 01 versus PBS group; mean ± SD. c) Gene expression of VEGFA in mono and co‐NRCMs under glucose/oxygen deprived conditions. d) Gene expression of VEGF 165 in mono and co‐HUVECs under glucose/oxygen deprived conditions. e) Gene expression of KDR in mono‐ and co‐HUVECs under glucose/oxygen deprived conditions for 120 min; * p < 0. 05 versus PBS group. ** p < 0. 01 versus mono‐cultured PBS group; † p < 0. 05 versus cocultured PBS. Data are obtained from three independent experiments; mean ± SD. 2. 5 Effect of “Ion Therapy” on Cardiac Function and Heart Remodeling as well as Hypertrophy Post‐AMI In Vivo In order to further verify our in vitro findings, investigating the protective effect of silicon ion on cardiomyocytes in vivo, and exploring the application possibility of the “ion therapy, ” a mouse model of AMI was established and left ventricular (LV) function was evaluated by echocardiography right after the surgery and 4 weeks later. For grouping, the detailed information and method were described in the experimental section. In addition, there was also no significant difference in the level of myocardial enzyme spectrum between the AMI group and the AMI+CS group. The result of cardiac enzymatic expression was shown in Figure S5 (Supporting Information). The changes in echocardiography before and post “ion therapy” were investigated. To study the real therapeutic efficacy of “ion therapy, ” different concentrations (1/2CS and CS extracts) of ion extracts were first test in a pre‐experiment. The results indicate that although both 1/2CS and CS ion extracts can improve cardiac function (Figure S6, Supporting Information) and alleviate fibrosis post‐AMI (Figure S7, Supporting Information), the CS group showed remarkably better therapeutic effect than 1/2CS group. Therefore, the CS extracts were chosen for further detailed in vivo experiments with increased animal numbers. As shown in Figure 6 a, b, right after LAD ligation, ejection fraction (EF) values of AMI mouse were decreased sharply to ≈40%. Compared to sham group, AMI mouse displayed remarkable LV enlargement indicating that the AMI mouse model were successfully established. After 28 d, EF values in AMI group were reduced by 20% when compared with that right after LAD ligation. However, the EF values in CS group did not show significant change after same time period. The similar results were also observed in fractional shortening (FS) values. Additionally, EF and FS values in CS group were significantly higher than those of AMI group after injection. These results indicate that cardiac function deteriorates slowly after myocardial infarction and CS can alleviate this decrease. Indeed the results revealed that the “ion therapy” significantly improved cardiac systolic and diastolic functions. Compared to AMI group, intravenous injection obviously alleviate the deterioration of contractile function of hearts post‐AMI defined by the changes of FS, EF, left ventricular end‐diastolic diameter (LVEDD), and LV end‐systolic diameter (LVESD) values. These results were also verified by histological Masson's trichrome staining of collagen deposition. Compared to AMI group, the “ion therapy” showed remarkably decreased content of collagen (Figure 6 c, e). In accordance, heart samples size, serum NT‐proBNP level, heart to body weight ratios “ion therapy” group were also significantly decreased than that of AMI group (Figure 6 d, f, g). These data indicate that the “ion therapy” clearly have therapeutic effect on AMI mice by attenuating left ventricular remodeling and hypertrophy. Figure 6 Effect of “ion therapy” on cardiac function and heart remodeling as well as hypertrophy post‐MI in vivo. a) Representative echocardiograms (left) and measurements of different groups obtained from the mid‐papillary muscle region of the left ventricle (right) of each groups before “ion therapy”(right after LAD ligation) and post “ion therapy” (4 weeks after LAD ligation); sham‐1, AMI‐1 and AMI+CS‐1: right after LAD ligation; sham‐2, AMI‐2 and AMI+CS‐2: 4 weeks after LAD ligation. b) Cardiac function measured by left ventricular end‐diastolic diameter (LVEDD), LV end‐systolic diameter (LVESD), the percentage of LV ejection fraction (EF) and fractional shortening (FS) values and changes before and post “ion therapy”. * p < 0. 05 versus AMI group, † p < 0. 01 versus AMI group. c) Representative Masson's trichrome staining of heart sections evaluating collagen deposition 4 weeks after surgery. d) Representative pictures of heart samples indicating hypertrophy degrees. e) Quantitative analysis of the area of fibrosis. Sham group was set as 0% and † p < 0. 001 versus AMI group (10 pictures for each group). f) Heart weight/body weight (HW/BW) ratio in each groups. g) Serum expression of NT‐proBNP level in in each groups. Sham group n = 4; AMI group n = 7 and AMI+CS group n = 13. * p < 0. 05 versus sham group, † p < 0. 05 versus AMI group; mean ± SD. 2. 6 Effect of “Ion Therapy” on Cardiac Cell Apoptosis In Vivo Post‐AMI Considering the in vitro antiapoptosis effect of Si ions, we also investigated the antiapoptosis effect of “ion therapy” on AMI by using TUNEL assay on histological sections of AMI animals in vivo. As shown in Figure 7 a, the number of TUNEL‐positive nuclei (stained in brown‐yellow ) in AMI mice was clearly more than that of sham group suggesting that ischemic condition induced apoptosis of cardiomyocytes in heart tissue. In contrast, the CS ions treated group showed a clear decrease of the TUNEL‐positive staining. The quantitative analysis of the assay further confirmed the observation that the percentages of TUNEL‐positive stained cells in CS extracts treatment groups, namely the “ion therapy” group, declined sharply as compared to that of AMI group (Figure 7 b). Considering our in vitro findings, it is clear that the silicon ion extract of the “ion therapy” has the function to protect cardiac cells and reduce myocardial apoptosis in AMI. Figure 7 Effect of “ion therapy” on cardiac apoptosis in vivo post‐AMI. a) Representative immunohistochemical pictures of TUNEL staining in border‐zone of infarction. Total nuclei (DAPI staining, blue) and TUNEL positive nuclei (brown‐yellow). Red arrows show TUNEL‐positive cardiomyocytes. Scale bar represents 50 µm. b) Quantitative analysis of TUNEL‐positive cardiomyocytes (10 pictures for each group). Sham group n = 4; AMI group n = 7 and AMI+CS group n = 13. ** p < 0. 01 versus AMI group; mean ± SD. 2. 7 Effect of “Ion Therapy” on the Expression of Gap Junction Associated Cx43 in Cardiomyocytes In Vivo Post AMI To study the effect of “ion therapy” on promoting cell–cell communications in AMI mice hearts, protein level of Cx43 were investigated. Myh6 protein is expressed specifically and predominantly in cardiomyocytes and is the major protein comprising the cardiac muscle thick filament, and functions in cardiac muscle contraction. In this study, Myh6 staining was used to distinguish cardiomyocytes to other cells. It is clear to see that myh6 protein structure and arrangement in AMI mice were disorganized and some even disappeared both in the infarcted area and border area as compared to sham group (Figure S3, Supporting Information). As expected, Cx43 proteins were abundantly expressed in the intercalated disc (Figure S3, Supporting Information and Figure 8 merge pictures). In normal heart (sham group), Cx43 was widely expressed while the expression was downregulated in AMI mice (both in infarcted area and border area). In the border area, Cx43 was slightly decreased while in infarct area the decrease was much more evident suggesting that ischemic condition indeed has great influence on Cx43 expression in AMI. Interestingly, the immunofluorescence staining showed that the “ion therapy” treatment (CS group) clearly increased Cx43 expression both in infarct area and border area as compared to that of AMI group (Figure 8 ). The in vivo results confirmed our in vitro findings and indicates that the “ion therapy” has the activity to promote cell–cell communications post AMI. Figure 8 Effect of “ion therapy” on the expression of gap junction associated Cx43 in cardiomyocytes in vivo post‐AMI. Representative immunofluorescence images of Myh6 (green) and gap junction associated Cx43 (red) staining in the infarcted myocardium 4 weeks after surgery. Scale bars represent 50 µm. 2. 8 Effect of “Ion Therapy” on VEGF‐Mediated Angiogenesis in Border Area of Ischemic Tissues In Vivo Post AMI One of the most critical issue in AMI treatment is to rebuild blood supply, which was damaged under ischemic condition. Therefore, the in vivo effect of the “ion therapy” on angiogenesis was investigated by using vWF (large blood vessels) and isolectin‐IB4 (for capillaries) immunofluorescence staining of heart tissues after different treatments. Figure 9 a, d illustrates that “ion therapy” stimulated angiogenic responses in the border area of infarction. More blood vessels and capillaries were observed in the “ion therapy” group as compared to AMI group. Quantitative analysis further confirmed that the blood vessel and capillary numbers of “ion therapy” group were significantly increased as compared to AMI group (Figure 9 b, e). To further confirm the angiogenic effect of bioactive ions, we analyzed serum VEGFA level by using a ELISA assay. The results showed that the serum level of VEGFA in AMI group was slightly higher than that of sham group due to ischemic stimulation of the AMI condition, which induced VEGF‐mediated neovascularization in the border area of infarction myocardial (Figure 9 c). Interestingly, it is observed that serum VEGFA level in “ion therapy” group was significantly increased as compared to AMI group (Figure 9 c). Taken together, our in vivo findings demonstrated that the “ion therapy” indeed activated angiogenesis and promoted blood vessel formation in AMI animals. Figure 9 Effect of “ion therapy” on VEGF‐mediated angiogenesis in border area of ischemic in vivo post‐ AMI. a) Representative immunofluorescence images of vWF‐stained blood vessels in the border area of infarcted myocardium 4 weeks after surgery. Scale bars represent 75 or 50 µm, respectively. b) Quantification of tube number/HPF in the border regions of infarctions (10 pictures for each group). ** p < 0. 01 versus AMI group; mean ± SD. c) Serum VEGFA expression post‐AMI. d) Representative immunofluorescence images of isolectin IB4‐stained capillaries in the border area of infarcted myocardium 4 weeks after surgery. Scale bars represent 50 µm. e) Quantification of capillaries numbers/HPF in the border regions of infarctions (five pictures for each group). ** p < 0. 01 versus sham group, †† p < 0. 01 versus AMI group. Sham group n = 4; AMI group n = 7 and AMI+CS group n = 13; mean ± SD. 2. 9 Biodistribution and Metabolism of Si and Acute Toxicity of “Ion Therapy” Si concentrations in blood and important organs including heart, lung, liver and kidney before and postinjection were measured, and the results are shown in Figure 10. The metabolism time of Si in blood was fast. The Si concentration was less than 1 ppm in serum before injection (at 0 min), increased after ion injection, reached maximum after 10 min, and then decreased quickly with time. After 120 min, the Si ion concentration was almost back to the level before the injection (Figure 10 a, b). In all organs such as heart, liver, lung, and kidney, the Si ion concentrations changed in a similar pattern with time after injection, which increased after injection and reached maximum at day 14, and then decreased back to the normal level at day 28, indicating no significant accumulation of Si in all important organs (Figure 10 c–f). It is worth to indicate that after injection Si ion concentration increased in heart indicating that Si ions indeed were infiltrated in heart tissue through intravenous injection. Then, to investigate whether these increased trace element content in organs would harm the functions of the organs, the acute toxicity of “ion therapy” was investigated. As shown in Figure 10 g, h, the results revealed that no acute inflammation reaction was observed as no significant differences in serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), and creatine kinase (CK) level between AMI and AMI+CS groups at day 7 and day 14. Figure 10 Bio‐distribution and metabolism of Si and acute toxicity of “ion therapy. ” a) Serum Si concentrations after initiation injection in AMI mouse ( n = 4). b) Serum Ca ion concentrations after initiation injection in AMI mouse ( n = 4). c) Si concentrations in hearts before and after “ion therapy” for 1, 7, 14, 21, and 28 d ( n = 3 for each group). d) Si concentrations in lungs before and after “ion therapy” for 1, 7, 14, 21, and 28 d ( n = 3 for each group). e) Si concentrations in kidneys before and after “ion therapy” for 1, 7, 14, 21, and 28 d ( n = 3 for each group). f) Si concentrations in liver before and after “ion therapy” for 1, 7, 14, 21, and 28 d ( n = 3 for each group). g) Serum expression of ALT, AST, and Cr level in each groups after injection for 7 d. h) Serum expression of ALT, AST, and Cr level in each groups after injection for 14 d. Sham group n = 4; AMI group n = 7 and AMI+CS group n = 13. * p < 0. 05 versus sham group or day 0, ** p < 0. 01 versus Day 0, † p < 0. 05 versus AMI group; mean ± SD. 3 Discussion AMI is the acute necrosis of myocardial tissue, which threats the public health seriously. 16 Previous tissue engineering therapies for AMI treatment has acquired certain effect but 17, 18, 19, 20, 21, 22 few observed robust myocardial regeneration and focused on the paracrine effects of angiogenesis and antiapoptosis effects stimulated by stem cells. 20 Here in this study, we proposed, for the first time, a new treatment concept for AMI therapy utilizing bioactive ions, which are bioactive components of bioceramics. By intravenous injection, the bioactive ions from biomaterials were delivered into the infarct area, which not only makes up for the deficiency of traditional medicine for AMI treatment in endogenous recovery, but also avoids drawbacks of in situ cell injection or local surgical cell/biomaterial implantation. Our results demonstrated that the “ion therapy” can improve cardiac function by promoting cell–cell communication, enhance gap junction associated Cx43 expression thus stimulate VEGF mediated angiogenesis, reconstruct blood flow and inhibit the MAPK family protein‐associated apoptosis ( Figure 11 ). Furthermore, it is for the first time revealed that bioactive ions such as Si are able to regulate cardiomyocyte behavior and stimulate angiogenesis during AMI recovery. This “ion therapy” not only provides a novel strategy for the treatment of AMI, but also give new insight for the application of biomaterials in systemic treatment of human diseases. Figure 11 Overall effects and mechanism of “ion therapy” on AMI treatment. “Ion therapy” can significantly improve cardiac function in mice post‐AMI by stimulating Cx43 mediated gap junction thus promoting VEGF‐mediated angiogenesis and by inhibiting caspase 3‐associated apoptosis. Cardiac myocytes apoptosis is a key factor in the pathogenesis of myocardial infarction and acute coronary artery occlusion is often accompanied by myocardial apoptosis. Inhibition of apoptosis has been considered as an effective treatment for AMI as hypoxia induced cardiomyocyte apoptosis plays an important role in the development of heart failure post‐AMI. 23, 24 This study shows that bioactive silicon ions not only promoted angiogenesis, but also played an active role in inhibiting cardiomyocyte apoptosis and enhancing myocardial viability. In vitro, the results of cell viability and gene expression indicated that silicon ions enhanced cardiomyocyte viability and stimulated expression of cardiomyocyte specific marker genes under normal and glucose/oxygen deprived conditions. In addition, the results of TUNEL staining and c‐caspase 3 protein expression indicated that Si ions may play a protective role on NRCMs by inhibiting apoptosis. To further explore the underlying mechanisms, we investigated the expression of apoptosis‐related MAPK family proteins in NRCMs under glucose/oxygen deprived conditions. Our findings showed that the expression of MAPK p38 and ERK 1/2 was regulated by Si ions, in which the phosphorylated p38 was downregulated while phosphorylated ERK1/2 was upregulated. Previous studies have found that selective inhibition of MAPK p38 or activation of MAPK ERK 1/2 pathway can reduce apoptosis induced by ischemic injury. Our previous studies have also revealed that silicate‐based biomaterials activated the AMPK/ERK1/2 and PI3K/AKT signaling pathways 25 in the process of osteogenesis. The extract of β‐CS/PDLGA significantly increased the expression of p‐ERK 1/2 in rBMSCs and the expression of p‐AKT in HUVECs. In this study, we demonstrated that the bioactive Si ions simultaneously inhibited expression of phosphorylated p38 and activated expression of ERK1/2 of NCRMs, which suggests a highly effective regulation of Si ions on NCRMs apoptosis. However, in this study, no significant difference was seen in the expression of phosphorylated AKT in silicon ions treated group. This may because of different treatment conditions and cell types. In previous study, the expression of p‐AKT in HUVECs was investigated under normal or hypoxia condition whereas in this study the expression of p‐AKT in NRCMs under glucose/hypoxia condition were estimated. Overall, it can be speculated that silicon ions may play a role in antiapoptosis by affecting the expression of MAPK proteins in NRCMs under glucose/oxygen deprivation conditions. By comparing ion concentrations between the culture medium and the control medium (PBS), it is confirmed that Si ions (the most effective concentration is 4. 53–9. 08 µg mL −1 ) play a major role in the antiapoptotic effect for NRCMs under glucose/oxygen deprived conditions. Most importantly, the in vivo results demonstrated that the “ion therapy” treatment lead to significantly decreased TUMEL‐positive cell staining in infarct tissue, and confirmed the inhibitory effect of bioactive Si ions on NRCMs apoptosis in vivo. However, the in vivo model could not distinguish whether the protection effect of “ion therapy” on cardiomyocyte survival results from the stimulation of angiogenesis or the direct prosurvival effect of the ions or both. It is important to clarify this issue in the future research. Additionally, in this study we only focused on the effect of Si and Ca ions, so we selected calcium silicate, one of the simplest silicates, which only contains Si, Ca, and O elements. However, previous reports have shown that some other elements such as potassium, 26 zinc, 27 lithium, 28 and magnesium 29 may affect the function of heart. Possible synergistic effect of these trace elements with Si ions in “ion therapy” should be considered in further studies. Recovery of blood perfusion can rescue the dying cardiomyocytes, reduce ventricular remodeling infarction, protect cardiac structure, reduce myocardial fibrosis and cardiomyocyte hypertrophy, and improve cardiac function. Re‐establishment of blood flow in infarct tissue is similar as irrigating the arid farmland, which can save the dying cardiac cells affected by AMI condition. 30 The formation of neovascularization is essential for revascularization post‐AMI. The evidence that the bioactive ions released from silicate biomaterials can promote endothelial cells to form blood vessels has been clearly identified in our previous studies. 11, 12, 31 Li et al. 13 demonstrated that the CS extract (mainly containing Si and Ca ions) promoted the interaction between fibroblasts and endothelial cells through paracrine effect and stimulated the formation of blood vessels. However, whether bioactive silicon ions can regulate intercellular communication between cardiomyocytes and further affect endothelia cells behavior under glucose/oxygen deprived conditions and promote angiogenesis remain unclear. In order to test our hypothesis that the bioactive Si ions may enhance angiogenesis in infarct tissue through activating cell–cell communication, we first investigate the expression of gap junction related Cx43 in cultured NRCMs under glucose/oxygen deprived conditions. Cx43 is the most abundant and widely expressed connexin family protein in atrial and ventricular myocytes. 32 It has been reported that mitogen activated protein kinases (MAPK) such as p42/p44 and p38 might be involved in the regulation of Cx43 expression and phosphorylation. 33, 34, 35 Increased Cx43 phosphorylation is mediated by p38 in neonatal rat ventricular cardiomyocytes. 34 In this study, we found that Si ion upregulated Cx43 expression in NRCMs with the increased phosphorylation level of p38 and p44/p42. Therefore, it is with high possibility that Si ion may affect gap junction and upregulate Cx43 expression by regulating the phosphorylation level of p38 and p44/p42. This is in consistence with our previous finding that the CS extract exerts protective effects via stimulating the expression of Cx43 between endothelial cells under hypoxia conditions, 10 but further study is needed to elucidate the detailed mechanisms of Si ion activation of NRCMs. Next, we tried to clarify whether silicon ions can regulate endothelial cells behavior and promote VEGF mediated angiogenesis via paracrine effect by affecting closely connected cardiomyocyte under glucose/oxygen deprived conditions. We investigated the expression of VEGF in mono‐ and cocultures of NRCMs and HUVECs system under glucose/oxygen deprived conditions. Interestingly, we found that Si ion can increase the expression of VEG 165 in mono‐HUVECs while no significant change of VEGF expression in mono‐NRCMs was observed. However, in cocultured NRCMs, Si ions significantly stimulated the VEGF expression. In addition, higher expression of KDR in co‐HUVECs was also observed, suggesting that the co‐NRCMs might activate the expression of KDR in cocultured HUVECs via paracrine effect. In our previous study, we have identified similar paracrine activation of Cx43 expression in hypoxia HUVECs and enhancement of the communication between endothelial cells by Si ions released from dressing materials in wound healing. 10 Being the important “logistics support” for working heart, endothelial cells are critical for angiogenesis and blood vessel formation. The vitality of the endothelial cells directly affects the function of the blood vessels thus influences the vitality of working cardiomyocytes. Here in this study, the similar paracrine effect has also been observed between cardiomyocytes and endothelial cells for angiogenesis. In myocardial infarction, the protection of cardiac myocytes and endothelial cells (in vitro that is glucose/oxygen deprived conditions) is essential and crucial for recovery of cardiac tissues from AMI condition. Here we demonstrated that the bioactive Si ions also activated interactions between cardiomyocytes and endothelia cells, and promote angiogenesis. And most importantly, not only in vitro, but also in vivo, we found that intravenously injected silicon ions were functioning as a “protector” and “activator” for both cardiomyocytes and endothelial cells in the damaged heart by promoting cell‐cell communications, and finally resulted in the re‐establishment of blood supply in AMI. In the past decade, silicate materials have been developed as implants for bone regeneration and wound healing. 10, 36, 37 However, it is unknown whether the bioactive function of these local applied biomaterials can be utilized for systemic application such as AMI treatment through intravenous injection of material derived bioactive ions. The fundamental scientific question is whether active silicon ions can be transported to AMI site through venous blood vessels and play bioactive roles in protecting cardiomyocytes and promoting angiogenesis. This investigation confirmed our hypothesis that the intravenous injection based “ion therapy” does have a significant effect on myocardial infarction. Using a mouse model of myocardial infarction, our results demonstrated that silicon ions derived from calcium silicate bioceramics indeed can function in myocardial infarction area by intravenous injection, promote neovascularization, restore blood flow, reduce the degree of myocardial injury, prevent heart failure, and thus significantly improve cardiac function. The therapeutic effect are clearly supported by the changes of myocardial systolic and diastolic function (EF, FS, LVEDD, and LVEDS), and the results of histology, enzymology and morphology. The findings of Masson staining showed that silicon ion can significantly reduce the myocardial fibrosis, inhibit scar formation and reduce fibrosis. Immunohistochemical staining demonstrated that silicon ions enhanced angiogenesis in vivo, provided nutrition and oxygen for damaged heart tissues and promoted endogenous recovery. Additionally, by immunofluorescence staining for specific marker of myocardial cells in vivo, we revealed that “ion therapy” enhanced cell–cell interaction in AMI area as Cx43 expression remarkably increased in Si ion treated group. These results demonstrated that intravenous injected bioactive silicon ions was able to penetrate into the infarcted site and affect cells in the border area of infarcted site, confirming that “ion therapy” indeed has therapeutic effects. Silicon, as an important micronutrient trace element in human and animal tissues, is a constituent of some glycosaminoglycans and polyuronides, which can bind to the polysaccharide matrix tightly. 38 It is mainly distributed in connective tissue, bones, tendons, muscles, hair, feathers, and skin. 39, 40 Lacking of silicon may lead to abnormal bone and cartilage formation as well as other diseases. Based on our previous findings on activity of Si ions in stimulating cell viability and angiogenesis, we propose “ion therapy” as a novel strategy for AMI. One of the key issues is whether these bioactive ions can be delivered to heart tissue through intravenous injection. By the measurement of ion concentrations in organs, we found that Si concentration in heart increased over time which indicating that the injected Si ions were indeed delivered to the heart tissue indicating that “ion therapy” is achievable by intravenous injection. Another important issue is the safety of the ion injection. It is known that, as compared to other medication ways such as enteral and topical administration, intravenous injection can avoid the first pass elimination effect and increase the bioavailability with fast metabolism time. Our results on serum ion concentration analysis revealed that the metabolism time of Si was very fast by intravenous injection. No Si in blood was detected before injection, which was in consistent with previous reports showing that the Si concentrations is less than 1 ppm in serum. 39, 41, 42 After injection, although the Si concentration in blood increased sharply first, but then falls back to normal level in 120 min and the peak concentration was 3 ppm, at which no acute toxicity was observed. Analysis of Si concentrations in other important organs such as heart, lung, kidney, and liver was also similar as the result in blood. After injection, Si concentrations in organs first increased with time, reached a maximal concentration at day 14, and fell back to normal level in 3–4 weeks. This result indicates no Si accumulation in the major organs, and the possibility of induced long‐term toxicity and side effects is low. Furthermore, our evaluation of the acute toxicity of “ion therapy” revealed no acute inflammation reaction 7 and 14 d after injection, and no significant differences of Si concentrations in heart, liver, lung, and kidney were found as compared to control group without injection at day 28 indicating that the “ion therapy” is safe. These results in some degree are consisted with our previous findings on Si concentration in kidney, liver, lung, and spleen after CS scaffold implantation in bone defect for 4, 8, and 12 weeks, and Si ions in the body are nontoxic, are maintained within physiologically safe range and the excess Si was finally excreted through the urine. 39, 41 Considering all of these results, we demonstrated that “ion therapy” is an effective and safety strategy for cardiac tissue engineering. However, although we have investigated the acute toxicity by estimating liver enzymatic expression and kidney creatinine level, the long‐term toxicity has not been investigated, in particular the long‐term organ function such as renal function after follow‐up, which should also be considered in the future. 4 Experimental Section Synthesis of CS Powders : CS powder was prepared by using a chemical coprecipitation method as previously described. 43, 44 At room temperature, by continuously mixing aqueous solution of Na 2 SiO 3 (1 mol L −1 ) and Ca (NO3) 2 (1 mol L −1 ) for 24 h, the reaction mixture with a molar mass of Na 2 SiO 3 :Ca (NO 3 ) 2 = 1:1 was obtained. The mixture was then filtered and washed thoroughly with deionized water and ethanol to give a CS suspension. Then, after drying at 80 °C overnight and calcining at 800 °C for 2 h, the resulting CS powder was sieved to obtain 100–150 µm particles for future use. Ion Extract Preparation and Ion Concentration Determination : According to previously reported procedures, 44, 45, 46 the ion extracts of CS bioceramics were prepared by soaking 1 g of CS powder in 5 mL of PBS or serum‐free DMEM (GIBICO) and incubated at 37 °C for 24 h. The suspension of CS powder was then centrifuged to collect the supernatant. A filter (Millipore, 0. 22 µm) was used to sterilize the supernatant and serial dilutions of extracts (1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, and 1/256) were prepared by using PBS or cardiomyocyte growth medium [DMEM+10% fetal bovine serum(FBS)+1% P/S (penicillin/streptomycin)] respectively and stored at 4 °C for future use. ICP‐AES was used to determine the concentrations of the Ca, P, and Si ions in the solution. Cells Isolation and Culture : Neonatal Sprague‐Dawley (SD) rats (1–3 d old, weighing 5–7 g, means 6. 1 ± 0. 7 g) were obtained from the Experimental Animal Centre of Southern Medical University. Primary neonatal rat cardiomyocytes (NRCMs) were isolated from hearts of neonatal SD rats as previously reported 47 and the purity of the isolated cardiomyocytes was determined by immunofluorescence staining of cTnT (cardiomyocytes specific marker, Abcam, ab6994) and vimentin(fibroblasts specific marker, Abcam, ab11370). The isolated NRCMs with purity as high as 93% were used for further studies (Figure S1, Supporting Information). After separation from fibroblasts, enriched cardiomyocytes were seeded at density of 1. 5 × 10 4 cell per well in 96‐well plates or 5 × 10 5 cell per well in six‐well plates in cardiomyocyte growth medium for 96 h before use. HUVECs were isolated according to previously described methods 48 and the obtained HUVECs were cultured in endothelial cell medium (ECM) for 12 h before use. HUVECs at passages from 3 to 5 were used in this study. Cells were seeded at a density of 40 000 cells cm −2 in six‐well plate for future experiments. Cells at ≈80% confluence were prepared for future experiments. To investigate the effect of Si ions on the proliferation of cells, NRCMs were cultured under normal condition for 5 d and under glucose/hypoxia deprivation condition for 120 min, respectively. The expression of NRCMs specific marker genes such as were investigated. For NRCMs and HUVECs coculture experiments, the transwell permeable supports (Corning, 12 mm Diameter Insert, 12 Well) with a 0. 4 µm polycarbonate membrane were used in the coculture system to separate HUVECs and NRCMs into different compartments. NRCMs were plated at a density of 5 × 10 4 cell per well in top chamber of a transwell insert and monocultured in cardiomyocyte growth medium for 96 h before use. HUVECs were seeded at a density of 40 000 cells cm −2 in the lower chamber and monocultured in ECM for 12 h before use. After that, NRCMs were precultured in top chamber of a transwell insert and HUVECs precultured in the lower chamber were combined and further cocultured under hypoxia conditions (94% N 2, 5% CO 2, and 1% O 2 ) with PBS or CS extracts diluted in PBS in an anaerobic system (Thermo Forma, Marietta, OH, USA) at 37 °C for 90 or 120 min. Cell Viability Detection : To determine the effect of CS extracts on cardiomyocytes viability in vitro under normoxia and glucose/oxygen deprived conditions, serial dilutions of CS extracts (1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, and 1/256) diluted in cardiomyocyte growth medium or PBS were prepared. At day 1, day 3, and day 5, the effect of CS extracts (diluted in NRCM growth medium) on NRCMs cell viability was detected under normoxia condition. For glucose/oxygen deprivation, NRCMs were cultured in CS extracts (diluted in PBS) under hypoxia condition for 90 or 120 min and cell viability were detected. The NRCM viability was measured by CCK8 assays. The medium was removed and replaced with 110 µL of DMEM medium or PBS containing 10 µL of CCK8 solution (Cell counting kit‐8, Dojindo, Kumamoto, Japan) according to the manufacturer's instructions. The absorbance was measured spectrophotometrically using a microplate reader (Bio‐Rad Benchmark Plus) at wavelengths of 450 nm. Determination of NRCMs Apoptosis In Vitro : For glucose/oxygen deprivation, NRCMs were cultured in PBS or 1/64 CS extract (diluted in PBS) under hypoxia conditions for 90 or 120 min. The cells were then fixed in 4% paraformaldehyde and apoptotic cells were detected by TUNEL staining using one‐step TUNEL Apoptosis Assay Kit (Beyotime, Jiangsu, China) according to the manufacturer's protocol. Cells were counterstained with DAPI (Sigma–Aldrich, St. Louis, MO, USA) and TUNEL positive cells were observed by a fluorescence microscope (Leica DMI8, Germany). Total nuclei (DAPI staining, blue) and TUNEL positive nuclei (green) in each field were counted in ten randomly chosen field, and the index of apoptosis (number of TUNEL‐positive nuclei/total number of nuclei × 100%) was calculated. DCFH‐DA Staining for Analysis of Intracellular ROS Activity Level : NRCMs (1 × 10 4 per well) were seeded in black bottomed 96‐well culture plate in PBS or 1/64 CS extract (diluted in PBS) under hypoxia conditions for 120 min. After treatment, cells were incubated with 10 × 10 −3 m DCFH‐DA for 30 min at 37 °C. After washing with PBS for three times, fluorescence intensity was measured with a multiwell microplate reader at an emission wavelength of 528 nm and at an excitation wavelength of 488 nm. All the values were expressed as percentage fluorescence intensity relative to the control. Protein Isolation and Western Blot Analysis : In vitro, the protein level of Cx43, AKT, p ‐AKT, MAPK p38, MAPK p ‐p38, Erk1/2, p ‐Erk1/2, and p ‐JNK were determined by Western blot analysis. Following treatment, cells were washed with cold PBS (pH 7. 4) and lysed for 15 min with RIPA lysis buffer (Beyotime, Nantong, China) supplemented with a cocktail of protease and phosphatase inhibitors (Sigma Chemical Co, St Louis, MO) in ice bath and isolated from the following standard protocol. Protein concentrations were determined by the bicinchoninic acid (BCA) method. Each sample (50 µg of protein per lane) was separated by sodium dodecyl sulfate–polyacrylamide gel (10%) electrophoresis followed by electrophoretic transfer of protein from the gel to a nitrocellulose membrane. The membranes were then treated with the blocking buffer for 1 h at room temperature, followed by incubation with primary antibodies at 4 °C overnight. Primary antibodies used here included anti‐Cx43(Abcam, ab11370), anti‐p44/42 MAPK(Erk1/2) (1:1000 dilution, CST, 4695P), anti‐Phospho‐p44/42 MAPK (Thr202/Tyr204) (1:1000 dilution, CST, 9101S), anti‐Phospho‐p38 MAPK (Thr180/Tyr182) (1:1000 dilution, CST, 4511S), anti‐p38 MAPK(1:1000 dilution, CST, 8690S), anti‐Phospho‐AKT (Ser473) (1:1000 dilution, CST, 4060), anti‐AKT (1:1000 dilution, CST, 9272), anti‐Phospho‐SAPK/JNK (Thr183/Tyr185)(1:1000 dilution, CST, 4668), and anti‐cleaved caspase‐3(1:1000 dilution, CST, 14220S). Normalization of results was conducted by running parallel Western blots for detecting glyceraldehyde 3‐phosphate dehydrogenase protein (GAPDH). The optical density was quantified using an image processing analysis program. Quantitative Real‐Time Polymerase Chain Reaction (Q‐RT‐PCR): For RNA extraction, cells were prepared as described before. Total RNA was extracted from cultured cells or total left ventricular tissue using the TRIzol (Invitrogen) reagent and following the manufacturer's protocol. The concentration of RNA was measured though a nanodrop 1000 reader (Thermo Scientific). cDNA was synthesized using a ReverTra Ace‐a kit (Takara, Japan) according to the manufacturer's instructions. Primer (all from Sangon Biotech Co. Ltd. ) were used as the final concentration of 400 × 10 −9 m. Glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) was used as a housekeeping gene. The sequences are shown in Table S1 (Supporting Information). Reactions were performed in triplicate. Data were analyzed with the SDS 2. 3 software and compared by the ΔΔC t method, and each Q‐RT‐PCR was performed in triplicate for yield validation. Data were normalized to GAPDH mRNA expression of each condition and were quantified relative to the corresponding gene expression from control sample (cells cultured with PBS) at 90 or 120 min, which were standardized to 1. In vitro vWF, Cx43, cTnT and vimentin immunofluorescence staining. In vitro, immunofluorescence staining of von Willebrand factor (vWF) was applied on HUVECs and NRCMs cocultures under glucose/oxygen deprivation conditions. Meanwhile, Cx 43 immunofluorescence staining was also applied in NRCM monoculture under glucose/oxygen deprived conditions. The cells were incubated either with PBS or PBS with 1/64 CS extracts under hypoxia conditions in an anaerobic system (Thermo Forma, Marietta, OH, USA) at 37 °C for 90 or 120 min. After that, the co‐HUVECs or mono‐NRCMs were fixed for 15 min in 4% paraformaldehyde at room temperature and permeabilized with 0. 5% Triton X‐100 (PBS) for 30 min and blocked with PBS containing 1% bovine serum albumin (BSA) for 1 h at 37 °C. The immunofluorescence staining was carried out using rabbit anti‐vWF (Abcam, ab6994) or rabbit anti‐connexin 43(Abcam, ab11370) diluted in 0. 5% BSA at 1/500 according to the manufacturer's instruction. To reveal the vWF or Cx43, cells were incubated with Alexa Fluor 488 goat anti‐rabbit IgG (Santa Cruz) or cy3 goat anti‐rabbit IgG (Santa Cruz), and then counterstained with DAPI (Sigma–Aldrich, St. Louis, MO, USA) and pictures were taken by using a fluorescence microscope (Leica DMI8, Germany). Ten randomly chosen images were taken per well, and tubules were manually counted and averaged over the ten images. Mouse AMI Modeling and Grouping : Adult male C57 BL/6 mice (8–10 week old) were purchased from laboratory animal center of Southern Medical University. Protocols were approved by the Southern Medical University animal care and use committee guidelines, which conform to the guide for the care and use of laboratory animals published by the US National Institutes of Health (8th edition, 2011). For the AMI model, mice were subjected to permanent left anterior descending (LAD) ligation as described previously. 49 In brief, the animals were anesthetized using tribromoethanol (1. 2%, 0. 02 mL g −1 ) and ventilated using a rodent ventilator (DW‐3000B; Xinsida, Beijing, China) with 100 breaths per min and a stroke volume of 100 mL. A left thoracotomy was performed, and the LAD was ligated using an 8‐0‐prolene suture with a mortality rate of 80%. Sham group mice underwent the same surgical procedure with the exception that the LAD was not ligated. Echocardiography was performed right after LAD ligation and the animals were mainly randomized based on echocardiography. The mouse with LAD ligation treatment were randomly divided into AMI ( n = 7) and AMI+CS ( n = 13) group by matching the values of cardiac function index. There were no differences in EF, FS, LVEDD, and LVESD values between AMI and AMI+CS groups before “ion therapy. ” In addition, no differences were seen in serum cTnT and CK‐MB level between AMI and CS groups. The results of cardiac enzymatic level were shown in Figure S5 (Supporting Information). After that, the AMI+CS group received 200 µL bioactive ion administration (CS in PBS) by intravenous injection through tail vein every 2 d for 2 weeks while sham and AMI groups received equal volume of PBS. The optimal injection concentrations of “ion therapy” were predetermined by a pre‐experiment, in which two concentrations of CS ion extracts (the CS extracts and 1/2CS dilution) were tested. Determination of Myocardial Apoptosis In Vivo : Mouse myocardial apoptosis was detected by TUNEL by using a commercial kit (catalog 11 684 817 910; Roche Diagnostics). The heart tissues were harvested 28 d after surgery and fixed in 4% paraformaldehyde at room temperature before embedded in paraffin. Heart samples were sectioned to 5 µm along the short axis and transversely across the infarct zone. Sections were then stained according to the manufacturer's protocol. Total nuclei (DAPI staining, blue) and TUNEL positive nuclei (brown‐yellow) in each field were counted in ten randomly chosen fields, and the index of apoptosis (number of TUNEL‐positive nuclei/total number of nuclei × 100%) was calculated. Echocardiography : 28 d after surgery, mice were anesthetized through inhalation of isoflurane (1–1. 5%) in O 2 and echocardiographic examination was performed using a Vevo 2100 System equipped with a 30 MHz transducer (FUJIFILM Visual Sonics, Inc. Toronto, Canada) as previously described 50 by an observer blinded to the experiment. The LVEDD, LVESD, and the percentage of LV ejection fraction (EF), and FS were detected. All measurements were repeated for at least three consecutive pulsation cycles and the data were averaged for statistical analysis. Histological Analysis : Mice were sacrificed 4 weeks after the induction of myocardial infarction, and hearts samples were collected and fixed in 4% paraformaldehyde overnight and were embedded in paraffin. The weight of hearts samples was measured and the heart weight/body weight (HW/BW) ratio was determined. Samples were sectioned to 5 µm along the short axis, transversely across the infarct zone. Following deparaffinage and dehydration, samples were stained with Masson's trichrome staining was used to detect the collagen deposition in infarct area (collagen was stained blue). The collagen area and LV area was measured by Image‐Pro Plus software (version 6. 0; Media Cybernetics, Silver Spring, MD, USA). Fibrosis (%) was calculated as the ratio of collagen area to LV area. In Vivo vWF, myh6, Isolectin‐IB4, and Cx43 Immunofluorescence Staining : In vivo, paraffin sections were prepared as described before. For angiogenesis investigation, paraffin sections were incubated with primary antibodies against von Willebrand factor (vWF, 1:500, Abcam) and then incubated with Alexa Fluor 488 goat antirabbit IgG (Santa Cruz), cell nuclei were stained by DAPI, images of ten randomly selected fields in infarct and border zone were captured by fluorescence microscope, and the blood vessel density is determined as vessels/HPF (high power field) (400×). For capillaries detection, isolectin‐IB4 conjugated with Alexa Fluor 488 (Molecular Probes, Eugene, OR, USA) was used to stain the sections. For gap junction investigation, paraffin sections were incubated with primary antibodies against rabbit anti‐connexin 43 (Abcam, ab11370) and mouse anti‐heavy chain cardiac myosin(Abcam, ab50967) for special labeling of cardiomyocyte. Sections were then incubated with Alexa Fluor 488 goat anti‐rabbit IgG (Santa Cruz) and cy3 goat anti‐mouse IgG (Santa Cruz). Plasma NT‐proBNP, VEGFA, AST, ALT and Cr Elisa Assays : A volume of 600 µL venous blood was collected into tubes containing disodium EDTA, which was centrifuged for 10 min at 3500 rpm. Then the supernatant was collected in EP tube and kept at −80 °C. Plasma NT‐proBNP, VEGFA, AST, ALT, and Cr level were determined by sandwich enzyme‐linked immunosorbent assay (ELISA) with commercially available kits (MSK, KT21109 and KT21108; Mmbio, MM‐44384M2, MM44625M2, and MM‐44455M2, China, respectively) according to the manufacture's instruction. Optical density of each well was determined by using a microplate reader set to 450 nm. The sample values were then read off the standard curve. Si Metabolism and Distribution Analysis : The metabolism of Si was investigated by determining Si concentrations in blood samples collected before and 10, 30, 60, 90, and 120 min after ion injection. The accumulation and distribution of Si in important organs such as heart, lung, liver, and kidney after seven times intravenous injection was also analyzed. The blood and tissue samples were chemically digested in concentrated nitric acid (spectral purity) using a microwave digester and the Si concentrations of the samples were determined by ICP‐MS (PerkinElmer NexION 350X, USA). Statistical Analysis : Data were expressed as mean ± standard deviation. Comparison between two groups was performed with two‐tailed Student's t‐test. Comparisons among more than two groups were performed using one‐way ANOVA followed by post‐hoc Bonferroni test. Significant differences were considered when p < 0. 05 and p < 0. 001. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201801290
| 2,018
|
Advanced Science
|
Matrix Rigidity‐Dependent Regulation of Ca
|
Abstract The dynamic regulation of signal transduction at plasma membrane microdomains remains poorly understood due to limitations in current experimental approaches. Genetically encoded biosensors based on fluorescent resonance energy transfer (FRET) can provide high spatiotemporal resolution for imaging cell signaling networks. Here, distinctive regulation of focal adhesion kinase (FAK) and Ca 2+ signals are visualized at different membrane microdomains by FRET using membrane‐targeting biosensors. It is shown that rigidity‐dependent FAK and Ca 2+ signals in human mesenchymal stem cells (hMSCs) are selectively activated at detergent‐resistant membrane (DRM or rafts) microdomains during the cell–matrix adhesion process, with minimal activities at non‐DRM domains. The rigidity‐dependent Ca 2+ signal at the DRM microdomains is downregulated by either FAK inhibition or lipid raft disruption, suggesting that FAK and lipid raft integrity mediate the in situ Ca 2+ activation. It is further revealed that transient receptor potential subfamily M7 (TRPM7) participates in the mobilization of Ca 2+ signals within DRM regions. Thus, the findings provide insights into the underlying mechanisms that regulate Ca 2+ and FAK signals in hMSCs under different mechanical microenvironments.
|
1 Introduction Cell‐based therapeutics are revolutionizing the medicine field. 1 One promising branch is stem cell‐based therapy, which has developed from preclinical to early clinical studies for treatment of various diseases. 2 Human mesenchymal stem cells (hMSCs) are a type of adult stem cells (ASCs) that are multipotent, easily accessible, and can be expanded ex vivo, providing great potential for clinical applications. 3 However, insufficient stem cell adhesion and survival in vivo remains a problem, even though it can be partly addressed by tissue engineering and ex vivo genetic modifications. 4 Indeed, the mechanical environment of cells, including factors such as substrate stiffness, has been shown to influence adhesion, 5 which is essential for hMSC survival, proliferation, and differentiation. 6 As such, the cell adhesion process not only links the extracellular matrix (ECM) and cytoskeleton chemically, but also establishes mechanical coupling between the ECM and the cell. 7 Therefore, understanding the molecular mechanism of hMSC adhesion, especially in the context of its mechanical environment, is necessary for the development of scaffold and active biological materials to enhance cell adhesion for hMSC‐based therapy. Cell adhesion to the surrounding matrix starts with the binding of integrin receptors in the plasma membrane to ECM proteins. Binding to matrix proteins such as fibronectin and collagen leads to integrin clustering and subsequent downstream assembly of both mechanical structures and chemical signaling complexes, including adaptor proteins, cytoskeletal components, catalytic signaling proteins, and secondary messengers. 8 Focal adhesion kinase (FAK) is a key component of integrin‐mediated signal transduction at focal adhesion complexes, which consequently mediates cell adhesion and migration. 9 FAK has been extensively studied during cell–matrix adhesion process on glass, but relatively less investigated when cells are adhered on softer substrates than glass. 10 Integrins also induce intracellular Ca 2+ increase in various cell types. 11 This Ca 2+ signal then directly and/or indirectly regulates adhesion through Ca 2+ ‐dependent proteins such as myosin II and calpain. 12, 13 Most studies designed to evaluate cell adhesion utilize cell types other than hMSCs. Our previous studies revealed that hMSCs are highly sensitive to the mechanical microenvironment, displaying spontaneous Ca 2+ oscillations which is dependent on the cell adhesion on mechanical environment. 14 However, there is little understanding on the regulation of FAK and Ca 2+ during the cell–matrix adhesion process of hMSCs under different mechanical microenvironment. Even less is known about the spatial organization of activation patterns of FAK and Ca 2+ at the plasma membrane, which is structurally organized into two functional microdomains called detergent‐resistant membrane (DRM) and non‐DRM regions. 15, 16 This spatial organization of FAK and Ca 2+ is crucial for the initiation of FAK and Ca 2+ signals. The DRM, known as lipid rafts membrane, is enriched with sphingolipids and cholesterol, whereas non‐DRM domains lack these lipid compositions. 17 Lipid rafts are not only enriched with resident integral membrane proteins such as caveolin and flotillin, they are also connected to extracellular proteins through glycophosphosphatidylinositol (GPI) anchors containing long chain fatty acids. 16, 18 Cytoplasmic proteins can be integrated into lipid rafts by the modification of a dual myristoylation/palmitoylation motif or palmitoylation on cysteine residues. 19, 20 Such distinct characteristics of membrane microdomains can contribute to the specific dynamics of cellular signaling and their physiochemical regulation owing to the differential distribution of membrane‐associated proteins at plasma membrane microdomains. However, it remains unclear how signaling events are compartmentalized by specialized microdomains due to limitations in available methodologies. In this study, we take advantage of membrane‐targeting fluorescent resonance energy transfer (FRET)‐based FAK and Ca 2+ biosensors to investigate the regulation of these two signals during cell–matrix adhesion process in hMSCs seeded on substrates with different stiffness. Through live single cell imaging, we first demonstrate that rigidity‐dependent FAK and Ca 2+ signals are selectively enriched at detergent‐resistant membrane (DRM) microdomains in a concerted manner, but not at non‐DRM during this adhesion process. Furthermore, we report that transient receptor potential subfamily M7 (TRPM7) is involved in the regulation of matrix rigidity‐dependent Ca 2+ signals at DRM microdomains, which is mediated by the functional FAK and the integrity of lipid rafts. 2 Results and Discussion 2. 1 FAK Activation at DRM Microdomains Is Mediated by Matrix Rigidity It is known that matrix rigidity has a significant impact on cell adhesion and spreading via integrin‐cytoskeleton linkages. 21 FAK plays a crucial role in the regulation of integrin‐cytoskeleton networks. 22 However, it remains a challenge to elucidate whether/how FAK activity is spatiotemporally controlled at the plasma membrane microdomains in the early stages of cell adhesion and spreading processes. To unravel this, we developed and utilized two kinds of membrane‐targeting FAK biosensors based on FRET technology. These biosensors were designed to detect changes in FAK activity within specific membrane microdomains. As illustrated in Figure 1 a, interaction between the SH2 domain of the biosensor and the substrate phosphorylated by FAK triggers a conformational change in the biosensor. This conformational change leads to alteration of the distance/orientation between the enhanced cyan fluorescent protein (ECFP) and the yellow fluorescent protein variant YPet (yellow fluorescent protein for energy transfer) with a FRET ratio change. Quantification of this change provides an index to measure levels of FAK‐induced phosphorylation (Figure 1 a). A DRM‐targeting FAK biosensor, called Lyn‐FAK, was then engineered containing a lipid raft‐targeting motif (MGCIKSKRKDNLNDDE) originated from Lyn kinase to the N‐terminus of cytosolic FAK, which enables the tethering of this sensor to the DRM microdomain. In contrast, a non‐DRM‐targeting FAK biosensor, called Kras‐FAK, holds a non‐raft‐targeting motif, a prenylation substrate sequence from Kras (KKKKKKSKTKCVIM) to the C‐terminus of the sensor. 23 These biosensors were separately transfected into hMSCs, and the cells were then seeded on polyacrylamide (PA) hydrogels with a Young's modulus of 0. 6 or 40 kPa prior to imaging experiments and analysis (Figure 1 b). We first showed that hMSCs seeded on 40 kPa gel displayed larger surface areas indicating a better spreading compared to cells on 0. 6 kPa gel ( n = 15, *** P < 0. 001) (Figure 1 c). Applying FRET biosensors, we further observed that the FRET ratio of Lyn‐FAK at DRM microdomains increased by 15% from the basal level within 1 h upon adhesion on 40 kPa gel surface ( n = 7), but not on 0. 6 kPa gel ( n = 10) (Figure 1 e–g). These results suggest that the FAK activity within DRM microdomains can be induced upon cell adhesion, which is dependent on matrix rigidity. FAK has been shown to participate in mechanosensing. 24 However, it was not known whether FAK is exclusively accumulated or recruited to adhesive partners at DRM microdomains during cell–matrix adhesion process. While previous studies have shown chemically mediated FAK recruitment to DRM microdomains, our findings provide the first evidence that such recruitment can also be mediated by mechanical factors such as matrix rigidity. Figure 1 Matrix rigidity‐dependent focal adhesion kinase (FAK) activation at DRM microdomains during the adhesion process. a) Schematic drawings of the underlying mechanism illustrating the conformational changes in FAK biosensors enabling detection of FAK signals at two different plasma membrane microdomains, DRM (Lyn‐FAK) and non‐DRM (Kras‐FAK). FAK phosphorylation of the substrate enables its interaction with the SH2 domain, leading to a conformational change in the FAK biosensor. Lyn‐FAK and Kras‐FAK biosensors can be tethered at DRM and non‐DRM microdomains, respectively. b) Schematic drawings of adhesive hMSCs exposed to either hard (HG, young's modulus, 40 kPa) or soft gels (SG, 0. 6 kPa). c) Analysis of cell area 40 min after seeding on substrates with different rigidities ( n = 15, *** P < 0. 001). d, e) Time‐lapse FRET images of Lyn‐FAK in hMSC grown on 40 and 0. 6 kPa during the cell–matrix adhesion process. Hot and cold colors indicate high and low FAK activities, respectively. Scale bar = 20 µm. f, g) Dynamic and average changes in the FRET ratio of Lyn‐FAK biosensor in hMSCs cultured on 40 and 0. 6 kPa gels. All error bars are s. e. m ( n = 7–10, ** P < 0. 01). 2. 2 Distinct FAK Activation at Membrane Microdomains To confirm whether the obtained FRET signals originated from FAK activity, we treated hMSCs with a specific FAK inhibitor, PF228, as a control to suppress the upregulated FAK activity. When the cells were in suspension, the FRET ratio of Lyn‐FAK was relatively lower than those in adhesion ( n = 3, * P < 0. 05, and ** P < 0. 01) ( Figure 2 a–c). Immediately after PF228 treatment, rigidity‐dependent Lyn‐FAK signals were attenuated, suggesting that the observed FAK activation was specific (Figure 2 b, c) and occurred predominantly (Figure 2 e) at DRM microdomains. Since myosin light chain kinase (MLCK) is known to regulate actomyosin contractility and mechanical support of the cells, which are crucial for the structural integrity of DRM microdomains, 13 we further examined whether ML‐7, an inhibitor of MLCK, regulates FAK activity at DRM microdomains. Our results clearly show that ML‐7 pretreatment markedly inhibited the FAK activity at DRM microdomains even when hMSCs were seeded on rigid matrix ( n = 5–7) (Figure 2 d), suggesting the involvement of MLCK and its associated actomyosin contractility and structural support in regulating DRM FAK activities. Our previous results suggest that FAK activities are mostly concentrated in DRM regions in HT1080 cancer cells. 23 Consistently, no significant increase in Kras‐FAK signal was detected at non‐DRM microdomains during the cell–matrix adhesion process in hMSCs ( n = 9) (Figure 2 e). Figure 2 Distinct FAK activations at different plasma membrane microdomains in response to matrix rigidity. a–c) Time‐lapse analysis of Lyn‐FAK signal during the cell–matrix adhesion process and treatment with PF228, a specific inhibitor of FAK. Bar graphs represent the FRET ratio in suspension, adhesion/spreading in the presence and absence of PF228 ( n = 3, * P < 0. 05, and ** P < 0. 01), Scale bar = 20 µm. d) The FRET ratio of Lyn‐FAK in response to 40 or 0. 6 kPa in the presence of ML‐7, an inhibitor of MLCK ( n = 5–7). e) The FRET ratio of Kras‐FAK biosensor reflecting the FAK signal at non‐DRM in hMSC cultured on 40 or 0. 6 kPa gels ( n = 9) in the adhesion process. 2. 3 Ca 2+ Mobilization at the Plasma Membrane Microdomains Similar to FAK, Ca 2+ signals also play an important role in cell adhesion and spreading. 13 Accordingly, we examined how Ca 2+ signals could be activated at plasma membrane microdomains. To perform this study, two distinct types of FRET‐based Ca 2+ biosensor were engineered to tether at DRM and non‐DRM microdomains using Lyn and Kras peptide sequences, respectively ( Figure 3 a, b). These membrane‐targeting Ca 2+ biosensors, Lyn‐D3cpv and Kras‐D3cpv, can detect the FRET changes caused by the change of Ca 2+ concentration at DRM and non‐DRM microdomains. These Ca 2+ sensors showed similar levels of FRET increase in response to treatment with ionomycin, an ionophore that increases the plasma membrane permeability and raises the intracellular level of Ca 2+ (Figure S1, Supporting Information), suggesting that similar Ca 2+ influx occurs at these two different membrane microdomains DRM versus non‐DRM regions with ionomycin treatment. Interestingly, the Ca 2+ increase during cell–matrix adhesion process selectively occurred only at the DRM microdomains in a rigidity‐dependent manner, but with undetectable signals at non‐DRM regions (Figure 3 c, d, n = 5–8, ** P < 0. 01, *** P < 0. 001, Figure 3 e–h, n = 7–8, *** P < 0. 001 and Figure S2, Supporting Information). This pattern of calcium regulation during adhesion process is similar to that of FAK activation. Figure 3 Matrix rigidity‐dependent Ca 2+ activation at membrane microdomains during the adhesion process. a, b) Schematic drawings of the underlying mechanism by which Ca 2+ biosensors target DRM (Lyn‐D3cpv) and non‐DRM regions (Kras‐D3cpv). Ca 2+ binding to mutated calmodulin (mCaM), which consequently interacts with the intramolecular m‐smMLCKp (smooth muscle myosin light chain kinase peptide), results in a conformational change in the Ca 2+ biosensors. c, d) The FRET ratio indicates a Ca 2+ signaling activity at DRM (Lyn‐D3cpv) in hMSCs cultured on 40 and 0. 6 kPa gels ( n = 5–8, ** P < 0. 01, *** P < 0. 001). Bar graphs indicate the quantitative analysis of Ca 2+ signaling activity 40 min after cell seeding. e, f) The FRET ratio showing the Ca 2+ signaling activity at DRM (Lyn‐D3cpv) and non‐DRM (Kras‐D3cpv) in hMSC cultured on 40 kPa. Sus; Suspension, Adh; Adhesion. ( n = 7–8, *** P < 0. 001). g, h) Time‐lapse FRET images of Lyn‐D3cpv and Kras‐D3cpv in hMSC during the cell–matrix adhesion process. The hot and cold colors represent high and low FRET ratios, indicating high and low Ca 2+ signaling activities, respectively. Scale bar = 20 µm. We have previously reported that the membrane targeting motifs should not affect the function of fused FRET biosensors. 23, 25 Our findings hence offer the first evidence that Ca 2+ signals are differentially mobilized at different plasma membrane microdomains during the process of cell–matrix adhesion in hMSCs. It is possible that there is a high level of buffer proteins such as calmodulin specifically localized at the non‐raft regions, which can bind to and neutralize free Ca 2+ diffused from other subcellular domains, for example, DRM and cytosolic organelles. This is consistent with earlier reports that there are different compositions of membrane proteins at different microdomains during cell adhesion processes, including Ca 2+ channels, integrins, and FAK. 26, 27 2. 4 The Regulation of Ca 2+ Mobilization at DRM Microdomain by FAK In this study, we found that within DRM microdomains, hMSCs displayed higher Ca 2+ mobilization and FAK activation during the adhesion process, both of which are dependent on the magnitude of matrix rigidity. In contrast, these two signals were poorly activated at non‐DRM regions regardless of matrix rigidity. To further examine whether Ca 2+ signals are correlated with FAK signals, we took advantage of two FAK mutants, FAK NT (a negative mutant, N‐terminal tail of FAK) and FAK KD (a kinase dead mutant). These mutants were cotransfected into cells expressing the calcium biosensor Lyn‐D3cpv. As shown in Figure 4 a, b and Figure S3 (Supporting Information), FAK NT and FAK KD caused reduced Ca 2+ signaling activity at DRM microdomains (Ctl, n = 3; FAK NT, n = 7; FAK KD, n = 8, *** P < 0. 001). Pharmacological inhibitor PF228 or ML‐7 also decreased Ca 2+ mobilization within the DRM region (Figure S4, Supporting Information). These results indicate that Ca 2+ mobilization at the DRM region is regulated by FAK during the cell–matrix adhesion process. Figure 4 Regulation of Ca 2+ signals and its associated molecules at DRM by FAK during the cell–matrix adhesion process. a, b) Ca 2+ mobilization and its quantitative analysis at DRM in hMSCs in response to 40 kPa in the absence or presence of FAK mutants (FAK NT and KD) (Ctl, n = 3; FAK NT, n = 7; FAK KD, n = 8, *** P < 0. 001). c, d) Effect of Caveolin‐1 mutation (Cav1S80E) and cholesterol depletion on Ca 2+ signaling activity. Cholesterol depletion and hence lipid raft disruption by MβCD, but not Cav1S80E, at DRM inhibits Ca 2+ mobilization during adhesion process in hMSCs cultured on 40 kPa ( n = 5–7, *** P < 0. 001). e, f) Calcium mobilization during the adhesion process on 40 kPa in hMSCs cotransfected with Lyn‐D3cpv and TRPM7 siRNA or control siRNA ( n = 5–9, *** P < 0. 001). The physiological interpretation of these phenomena requires further investigation to understand detailed mechanisms with regard to how FAK regulates Ca 2+ mobilization. One possibility is that FAK directly affects Ca 2+ channels at DRM regions. Even though FAK is mostly interacting with integrins at the plasma membrane, it is possible that FAK can physically interact with mechanosensitive membrane channels. 28 Another possibility could be that FAK affects Ca 2+ channels indirectly in an integrin‐dependent manner. In fact, FAK modulates different types of voltage‐gated Ca 2+ channels at the plasma membrane via integrins. 29 Ca 2+ influx through the plasma membrane has also been shown to closely correlate with the activation of high affinity β2 integrin and subsequent adhesive signals due to their dynamically coupled events. 30 2. 5 Depletion of Cholesterol by MβCD Inhibits Ca 2+ Mobilization at DRM Region, but Not by Caveolin‐1 It is reported that cholesterol, along with caveolin, plays an important role in signal transduction at DRM microdomains. 31 In fact, cholesterol depletion by MβCD appears to cause the downregulation of FAK. 32 It is possible that caveolin‐1, a key protein of caveolae, interacts with integrin β1 and promotes its localization at the DRM region, thereby affecting FAK activity. 33 Our previous study showed that the platelet derived growth factor (PDGF) induced FAK activation was inhibited by MβCD treatment. 23 Thus, we examined the role of DRM integrity in regulating the rigidity‐dependent Ca 2+ mobilization at these local regions during the cell–matrix adhesion process, by depleting cholesterol with MβCD. As shown in Figure 4 c, d ( n = 5–7, *** P < 0. 001), we found that MβCD treatment caused significant inhibition of FRET changes in Lyn‐D3cpv calcium biosensor, while a mutant of caveolin‐1 (Cav1S80E) did not affect the Ca 2+ mobilization at DRM microdomains (Figure S5, Supporting Information). These results suggest that cholesterol at DRM microdomains is crucial in Ca 2+ mobilization during cell–matrix adhesion process, which is not dependent on the caveolin‐1 function. Although it is known that caveolae disruption prevents Ca 2+ influx under mechanical stretch, 34 it may not have a significant impact on the adhesion process. Alternatively, caveolin may target other TRP channels, for example, TRPC1, instead of TRPM7, 35 in affecting the stretch‐induced Ca 2+ influx. While our previous study has shown that cholesterol disruption by MβCD treatment inhibited FAK activation at DRM region upon PDGF stimulation, 23 caveolin‐1 only specifically links the integrin α subunit, but not β subunit where FAK is coupled, and thus has possibly less effect on FAK activity during the adhesion process. 36 2. 6 TRPM7 Contributes to Ca 2+ Mobilization at DRM Region The distribution of TRP Ca 2+ channels at DRM regions is different from that of non‐DRM regions. For example, TRPC3 and TRPC6 are dominantly expressed at non‐DRM regions, but other TRPC channels are expressed throughout the plasma membrane. 27 TRPM7 is a mechanosensitive Ca 2+ permeable channel, which is known to dominantly accumulate at DRM microdomains, but not at non‐DRM regions. 37 Consistently, it was reported that TRPM7 serves as an adhesion‐associated channel that regulates actomyosin contractility. 38 This suggests that the integrin‐FAK complex might be connected with TRPM7 at DRM microdomain via cytoskeleton (CSK)‐actomyosin network to regulate Ca 2+ signals. Indeed, previous studies support the note that TRPM7 and integrin‐FAK complex may be closely interrelated. For example, TRPM7 regulates focal adhesions by controlling m‐calpain, 39 and the inhibition of TRPM7 disrupts the actin cytoskeleton, and the focal assembly of myosin IIA and vinculin, a focal adhesion protein. 40 We hence examined whether TRPM7 could be involved in Ca 2+ mobilization at DRM microdomains during the cell–matrix adhesion process. In our previous reports, we have shown that the delivery of TRPM7 siRNA clearly suppressed the expression of TRPM7 in hMSCs. 41 Using this siRNA method, we observed that TRPM7 knockdown significantly inhibited the rigidity‐dependent Ca 2+ mobilization at the DRM region, with nontargeting (NT) siRNA (control group) having minimal effects ( n = 5–9, *** P < 0. 001) (Figure 4 e, f and Figure S6, Supporting Information). Interestingly, TRPM7 activation by small molecule naltriben, a specific activator of TRPM7 rescued the lack of Ca 2+ response in Lyn‐D3cpv cells on 0. 6 kPa gel, but not in Kras‐D3cpv, suggesting that the local DRM‐specific Ca 2+ enrichment is mainly via TRPM7 during cell–matrix adhesion (Figure S7, Supporting Information). Our finding hence suggests that TRPM7 is a crucial regulator of matrix rigidity‐dependent Ca 2+ mobilization at DRM microdomains during the adhesion process. The colocalization of TRPM7 with Lyn, a DRM marker, as well as with paxillin and p‐FAK (Tyr397), provides additional supporting evidence that TRPM7 locally associated with FAK and integrin/focal adhesion complex (Figure S8, Supporting Information). As such, a putative connection between FAK and TRPM7 mediated by CSK‐actomyosin may contribute to the overall control of Ca 2+ mobilization. 42 3 Conclusions Previous evidence has shown that substrate stiffness directs hMSCs differentiation but with limited understanding on the underlying molecular mechanisms. 5 Our results reveal differential Ca 2+ and FAK signaling specifically occurring within DRM microdomains in hMSCs on hard and soft matrix at the initial stage of cell–matrix adhesion ( Figure 5 a). As the adhesion process establishes the link between ECM and cytoskeleton, these initial differences can have profound effects on subsequent cellular fate. Such a study may shed light on the impact of the segregation of membrane microdomains on cellular responses to mechanical environment and on consequent functional outcomes. Figure 5 A proposed model of matrix rigidity‐dependent Ca 2+ signals regulated by FAK during adhesion process. a) Ca 2+ and FAK signaling is currently activated at DRM regions during the cell–matrix adhesion process, while they are less activated at non‐DRM. b) Matrix rigidity‐dependent Ca 2+ mobilization at DRM is regulated by the functional support of the TRPM7 channel, but not Caveolin‐1. Integrin‐FAK complex could form interaction with TRPM7 via cytoskeleton (CSK)‐actomyosin network to regulate Ca 2+ mobilization at DRM during cell–matrix adhesion process. In conclusion, we have shown that Ca 2+ signals at DRM regions are regulated by FAK signaling during the adhesion process, and that this phenomenon is dependent on extracellular matrix rigidity in hMSCs. Additionally, our results demonstrate that FAK regulates Ca 2+ mobilization and that this can be controlled by the proper localization of cholesterol and the functional support of TRPM7 (Figure 5 b). These data provide new insights into the underlying mechanisms that regulate Ca 2+ and FAK signals, and the relationship between biochemical/mechanical factors and cellular differentiation in hMSCs. 4 Experimental Section Construction of DNA Plasmids : The Lyn‐FAK biosensor was generated by insertion of a raft‐targeting motif (MGCIKSKRKDNLNDDE) originated from Lyn kinase to the N‐terminus of the cytosolic‐FAK biosensor. Kras‐FAK biosensor was also constructed by insertion of a non‐raft‐targeting motif: a prenylation substrate sequence from Kras (KKKKKKSKTKCVIM) to the C‐terminus of the cytosolic‐FAK biosensor. 20, 23 The DNA encoding the FAK biosensors that contain ECFP‐YPet pair were subcloned with the BamHI/EcoRI sites in pRSetB for the protein purification from Escherichia coli, and in pcDNA3. 1 plasmid for the expression in mammalian cells. As FAK mutants, the kinase‐dead FAK with its kinase domain mutated (FAK KD) and the N‐terminal tail (containing 1–400 amino acids) of FAK (FAK NT) were used in this study. 23 The membrane‐targeting Ca 2+ biosensors based on FRET were generated in the same manner. The plasmids Lyn‐D3cpv and Kras‐D3cpv were constructed by fusion of a raft‐targeting motif: the myristoylation and palmitoylation sequence from Lyn kinase (MGCIKSKRKDNLNDDGVDMKT) to the N‐terminus of the D3cpv and a non‐raft‐targeting motif (KKKKKKSKTKCVIM) to the C‐terminus of the D3cpv. 43 A dominant negative Caveolin‐1 mutant (Cav1 S80E) was used to disrupt caveolar organization. 44 The caveolin‐1 was amplified by PCR and inserted into pcDNA3. 1 by BamHI and EcoRI sites. The primers are forward 5′‐GCGCGGATCCGCCACCATGTCTGGGGGCAAATACGTAG‐3′ and reverse 5′‐TCCGGAATTCTTATATTTCTTTCTGCAAGTTGATG‐3′. The Cav1 S80E mutant was generated by using QuikChange Site‐Directed Mutagenesis Kit (Stratagene, La Jolla, CA). Cell Culture and Chemicals : hMSCs (Lonza Walkersvile, Inc. , Walkersvile, MD) were purchased from Lonza and maintained in mesenchymal stem cell growth medium (MSCGM, PT‐3001, Lonza) containing 10% fetal bovine serum (FBS), 2 × 10 −3 m l ‐glutamine, 100 U mL −1 penicillin, and 100 µg mL −1 streptomycin in a humidified incubator of 95% O 2 and 5% CO 2 at 37 °C. The DNA plasmids were transfected into the cells (transfection efficiency, 29. 8%) by using Lipofectamine 2000 or LTX (Invitrogen, Carlsbad, CA) reagent according to the product instructions. PF228, ML‐7, methyl‐beta‐cyclodextrin (MβCD), and naltriben methanesulfonate hydrate were purchased from Sigma Aldrich (St. Louis, MO). RNA Interference Assays : Small interfering RNA (siRNA) sequences targeting human TRPM7 (ON‐TARGETplus SMARTpool siRNA) and nontargeting control sequences were designed by Dharmacon RNAi Technology (Dharmacon Inc. , Lafayette, CO). Cotransfection of 1–2 µg siRNA specific for TRPM7 (L‐005393‐00) or a nontargeting pool (D‐001810‐10‐05) along with the plasmid DNA was conducted according to the product instructions. Bis‐acrylamide‐PA Gel Fabrication : The fabrication of PA hydrogel with a defined modulus of elasticity ( E or stiffness), a characteristic of the ECM can be a useful technique to study the interactions of cells with their mechanical microenvironment. Such matrix substrate from PA gels can be created by simply changing relative concentration of acrylamide and bis‐acrylamide. 45 PA gels were cast on amino‐silanized glass coverslips. 40% w/v acrylamide and 2% w/v bis‐acrylamide stock solutions (Bio‐Rad) were mixed to prepare PA solution and then the gel's stiffness was achieved by varying the final concentrations of PA solution (3 and 7. 5%) and bis‐acrylamide cross‐linker (0. 06 and 0. 4%) for the corresponding stiffness of 0. 6 (soft gel) and 40 kPa (hard gel). To polymerize the solutions, 2. 5 µL of 10% w/v ammonium persulfate (APS; Bio‐Rad) and 0. 25 µL of N, N, N9, N9‐tetramethylethylenediamine (TEMED; Bio‐Rad) were added to yield a final volume of 500 µL PA solution. To crosslink extracellular matrix molecules onto the gel surface, a photoactive cross‐linker, sulfo‐SANPAH (0. 5 mg mL −1, sulfosuccinimidyl 6(4′‐azide‐2′‐nitrophenyl‐amino) hexanoate, Pierce) was used. For adhesion via integrins, 200 µL of a 0. 1 mg mL −1 fibronectin solution (from bovine plasma, Sigma) was incubated overnight with the PA gel at 37 °C. Image Acquisition and Microscopy : FRET sensor‐transfected cells were incubated for 1 h on 1% agarose dishes in a humidified incubator of 95% O 2 and 5% CO 2 at 37 °C to maintain in a state of suspension after detachment with 4 × 10 −3 m EDTA in phosphate buffered saline (PBS). During imaging process, the cells were maintained with MSCGM in a chamber, which was designed to provide the constant humidified air containing 5% CO 2, 10% O 2, and 85% N 2. The 37 °C degree of temperature throughout the samples in the chamber was maintained by a controlled heater (Nevtek ASI 400). Nikon Eclipse Ti inverted microscope with a cooled charge‐coupled device (CCD) camera was used for the image acquisition under perfect focus system (PFS) which allows for minimizing any focal change during cell–matrix adhesion. The pixel‐by‐pixel ratio images of FRET/ECFP were quantified after background subtraction in fluorescence intensity images of FRET and ECFP. The emission ratio images were computed and quantified by the MetaFluor software, and presented in the intensity modified display (IMD) mode. The following filter sets were utilized in the imaging experiments: dichroic mirror (450 nm), excitation filter for ECFP (420/20 nm), emission filter for ECFP (480/40), and FRET emission filter (535/25 nm). Immunofluorescence (IF) Staining and Confocal Microscopy : IF staining was carried out on hMSCs seeded on 40 kPa substrate. Fixation was performed with 4% paraformaldehyde in PBS for 10 min at room temperature, followed by washing three times with PBS. Fixed cells were permeabilized with 0. 1% Triton‐X100 (Sigma, Cat. No. STBG3972V) in PBS for 15 min and blocked with 3% BSA (bovine serum albumin) in PBS for 1 h at room temperature. The cells were then stained with primary antibodies, mouse anti‐TRPM7 (1:100, GeneTex, GTX41997), and rabbit anti‐phospho‐FAK (Tyr397; 1:200, ThermoFisher, Cat. No. 700255) overnight at 4 °C and secondary antibodies, FITC‐conjugated anti‐mouse IgG (1:100, Santa Cruz, sc‐516140) and CFL 555 conjugated anti‐rabbit IgG (1:200, Santa Cruz, sc‐516249) at room temperature for 1 h. After washing three times with PBS, the samples were mounted in VECTASHIELD medium containing DAPI (VECTOR Laboratories, H‐1200) and stored at 4 °C in the dark. Confocal images were acquired using a Leica laser scanning confocal microscope (TCS‐SP8) with a 40× Plan‐Apo objective / numerical aperture (NA) 1. 40. Statistical Analysis : All statistical data were expressed as the mean ± standard error of the mean (s. e. m. ). Statistical evaluation was performed by unpaired t ‐test and one‐way ANOVA using Graphpad Prism 6. 0 software to determine the statistical differences between groups. A significant difference was determined by the P ‐value (<0. 05). Conflict of Interest Y. W. is a scientific co‐founder of Cell E&G Inc. D. ‐H. K is a co‐founder and scientific board member of NanoSurface Biomedical Inc. However, these financial interests do not affect the design, conduct, or reporting of this research. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201801452
| 2,018
|
Advanced Science
|
In Situ Polyphosphate Nanoparticle Formation in Hybrid Poly(vinyl alcohol)/Karaya Gum Hydrogels: A Porous Scaffold Inducing Infiltration of Mesenchymal Stem Cells
|
Abstract The preparation and characterization of a porous hybrid cryogel based on the two organic polymers, poly(vinyl alcohol) (PVA) and karaya gum (KG), into which polyphosphate (polyP) nanoparticles have been incorporated, are described. The PVA/KG cryogel is prepared by intermolecular cross‐linking of PVA via freeze‐thawing and Ca 2+ ‐mediated ionic gelation of KG to form stable salt bridges. The incorporation of polyP as amorphous nanoparticles with Ca 2+ ions (Ca‐polyP‐NP) is achieved using an in situ approach. The polyP constituent does not significantly affect the viscoelastic properties of the PVA/KG cryogel that are comparable to natural soft tissue. The exposure of the Ca‐polyP‐NP within the cryogel to medium/serum allows the formation of a biologically active polyP coacervate/protein matrix that stimulates the growth of human mesenchymal stem cells in vitro and provides the cells a suitable matrix for infiltration superior to the polyP‐free cryogel. In vivo biocompatibility studies in rats reveal that already two to four weeks after implantation into muscle, the implant regions containing the polyP‐KG/PVA material become replaced by initial granulation tissue, whereas the controls are free of any cells. It is proposed that the polyP‐KG/PVA cryogel has the potential to become a promising implant material for soft tissue engineering/repair.
|
1 Introduction Advanced tissue engineering, as a branch of regenerative medicine, focuses on the replacement/regeneration of impaired/damaged tissues and organs to restore their biophysiological function within an organism. In the beginning of the last century, Alexis Carrel and Charles Lindbergh succeeded to keep entire organs alive outside of the body, using a perfusion machine (reviewed in: ref. 1 ). During this period, regenerative medicine was “cytocentric”, meaning that mainly the cells were assessed as the relevant players in biological complex systems (see: ref. 2 ). This concept changed with the conceptualization of the “OncoMouse or Harvard mouse” which introduced the technology of genetically modified laboratory animals to implant “relatively simple” body parts like vessels, bones or skin (ref. 3 ; reviewed in: ref. 4 ). The next phase started with the discovery that stem cells can initiate regeneration in complex tissue regions, which have lost this potency, through appropriate manipulation of those body parts. Those stem cells might re‐confer regeneration ability to quiescent cells (reviewed in ref. 5 ). It has been hoped that, e. g. , mesenchymal stem cells (MSCs) can repair damaged tissue not only by recruiting other somatic cells but also by initiating the secretion of functional, bioactive molecules, like growth factors and matrix proteins. Finally, in a biomimetic approach scaffolds have been fabricated from natural polymers that are biodegradable, less immunogenic as compared to synthetic polymers, and provided with biomechanical properties matching those of the native components of the tissues with their extracellular matrix (ECM). 6 At present, one major challenge in tissue engineering is the fabrication of a biomaterial as a scaffold that is not only biocompatible but also regeneratively active 7, 8 through eliciting regeneration‐inducing stimuli by providing the metabolic energy that is required for the organization and reconstitution of the fibrillar and nonfibrillar ECM. 9, 10 Such a scaffold, like thermoresponsive hydrogels, 11, 12 should also meet the physical properties of the replaced tissue, besides of the (bio)chemical prerequisites. Our group emphasized that it is inorganic polyphosphate (polyP) that acts as an extracellular storage for metabolic energy needed in the extracellular space. 9, 13 Within the cells polyP is stored in vesicles, termed acidocalcisomes. 14 These organelles contain polyP as a salt with inorganic divalent cations. In human platelets the chain lengths of polyP varies between 50 and 100 phosphate units. 15 Applying a new technology, we succeeded to entrap polyP, in a biomimetic way, into amorphous nanoparticles (Ca‐polyP‐NP) formed by the Ca 2+ salt of this polymer, 16 like those present in acidocalcisomes; similar particles have also been synthesized with other divalent counterions of the polyP. 17 These polyP nanoparticles have been shown to be morphogenetically active in vitro, both as nanoparticles and microspheres, 18 as well as in vivo. 19 Recently we showed that Ca‐polyP‐NP undergo coacervation after contact with proteins/peptides that reduce the zeta potential of the particles. 20 The polyP is prone to enzymatic hydrolysis with the alkaline phosphatase (ALP), the dominant intra‐ and extracellular enzyme that hydrolyzes the phosphoanhydride bonds which link the phosphate groups. 21 As consequence, Gibbs free energy is released, which is at least partially stored in the energy‐rich phosphoanhydride bonds of adenosine 5′‐diphosphate (ADP)/adenosine 5′‐triphosphate (ATP). 22 In turn, the extracellularly formed ADP undergoes enzymatic disproportionation to adenosine 5′‐monophosphate (AMP) and ATP, catalyzed by the adenylate kinase. 9, 13 Most interesting is the fact that the amorphous polyP nanoparticles meet the prerequisites of being fully biocompatible and provided with comprehensive functionalities allowing tissue incorporation and initiating tissue regeneration (ref. 17 ; reviewed in: refs. 23, 24 ). They are suitable for fabrication of bioinspired scaffold materials in tissue engineering bone and cartilage. 7 They are even suitable for application as an implant material that can be fabricated in a personalized way by using a molding procedure 25 or 3D bioprinting. 26 The polyP nanoparticles can also be fabricated in hybrid forms, e. g. , with N, O‐carboxymethyl chitosan 27 or with hyaluronic acid 8 ; the nanoparticles retain their biological function. First steps toward the formulation of a morphogenetically active bioink, applicable for 3D‐bioprinting at lower cell concentrations, have been successfully passed. 28 It is the aim of the present study to integrate the Ca‐polyP‐NP into a scaffold suitable for bone and cartilage regeneration. It is well established that hydrogels, being highly hydrated polymers, are promising replacement materials for impaired bone as well as damaged cartilage tissues, since they match the required mechanical performance (strength and toughness) of those tissues (reviewed in refs. 29, 30 ). Hence those nanoparticles were formed within a karaya gum (KG)/poly(vinyl alcohol) (PVA) hydrogel. PVA is a water‐soluble synthetic polymer with the cross‐formula [CH 2 CH(OH)] n, 31 which has been used for hydrogels in wound care, or for textiles or coatings. Since PVA readily forms hydrogen bonds in a symmetrical and regular arrangement, the polymer shows excellent filmability, water solubility, emulsification, and adhesion properties. 32 The properties of PVA can be further improved by addition of different polymers toward higher biocompatibility, porosity, or biodegradation. 33 A further promising feature of PVA, besides of its biomechanical properties, is the possibility of including a repeated freeze−thaw method, 34, 35 allowing physical cross‐linking of the PVA cryogel with other polymers and components. Among them is the natural polysaccharide KG 36, 37 which can be grafted with PVA to a hydrogel suitable for adsorption of compounds out of the aqueous milieu or of fluids in general. 38 KG is stabilized via ionic cross‐linking with Ca 2+. In this study, we show that Ca‐polyP‐NP can be formed in situ within a KG:PVA copolymer cryogel where they retain their biological and morphogenetic properties. Like in the free state, the nanoparticles within the cryogel undergo coacervation after transfer of the material into a protein/peptide‐containing medium and by that become biochemically active. 20, 39 The material is a dynamic hydrogel with pronounced viscoelastic properties. Incubation of the KG:PVA material with MSCs revealed that the cells respond with an increased growth potential and even display a higher propensity to infiltrate into the cryogel if the KG‐PVA gel, containing polyP in the form of Ca‐polyP nanoparticles, is used. Biocompatibility studies in vivo, performed by implanting the material into the muscle of rats, revealed that the polyP‐containing KG‐PVA material is superior to the polyP‐lacking cryogel. The data show that polyP nanoparticle formation can be performed in situ in a cryogel, qualifying this material as a potential new implant material offering new possibilities for soft tissue repair. 2 Results 2. 1 Preparation and Characterization of the Karaya Gum/Poly(vinyl alcohol) Hydrogel The hydrogel was prepared from KG and PVA after separate dissolution of these components in distilled water, as described under in the Experimental Section. After mixing and degassing, the viscous solution was poured into a Petri dish ( Figure 1 A) and subsequently subjected to a freezing‐thawing cycle to obtain a 3D structured KG/PVA hydrogel and allowing physical cross‐linking of PVA. During this process the hydrogel layer shrunk by ≈25% (Figure 1 B). From those 0. 2 to 3 mm thick layers smaller units were excised/punched out with a scalpel or a trephine, which were submersed into a 5% CaCl 2 solution (Figure 1 C) to induce ionic cross‐linking of the KG (ionic gelation). Three different weight ratios for KG and PVA have been chosen; 1:2, 1:1, and 2:1 (KG:PVA). Accordingly, the cryogels were termed “KG1/PVA‐cryogel”, “KG2/PVA‐cryogel”, and “KG3/PVA‐cryogel”. The polyP was added to the material after cryogel formation and prior to ionic gelation initiated with 5% CaCl 2 (Figure 1 D). The samples with polyP are termed “KG1/PVA/polyP:NP‐cryogel”, “KG2/PVA/polyP:NP‐cryogel”, and “KG3/PVA/polyP:NP‐cryogel”. Figure 1 Fabrication of KG/PVA‐cryogel samples. A) After mixing of KG with PVA the viscous solution was poured into a Petri dish. B) During the subsequent three freezing‐thawing cycles the layer shrunk to ≈75%. From those layers samples (sa) were drilled out which C) were submersed into a 5% CaCl 2 solution, allowing ionic gelation of the KG. The polyP was added as Na‐polyP prior to the ionic gelation. D) A cryogel slice without (left) and with polyP (right) is shown. Incubation of the lyophilized KG/PVA cryogel, either in the absence of polyP (“KG2/PVA‐cryogel”) or containing this polymer (“KG2/PVA/polyP:NP‐cryogel”), in phosphate buffered saline (PBS) did not result in a significant change ( p < 0. 05) of the weight of the dried scaffold after a 10 days' period. Mass Swelling Ratio : The cryogel shows an extensive property to take up aqueous solution. The uptake is dependent on the concentration ratio between KG and PVA. Increasing the KG portion from 1:2 (“KG1/PVA‐cryogel”) to 1:1 (“KG2/PVA‐cryogel”) and to 2:1 (“KG3/PVA‐cryogel”) results in an intensified PBS uptake from 3. 40 ± 0. 08 to 4. 29 ± 0. 13 and finally to 5. 26 ± 0. 19 during a 1 h incubation period ( Table 1 ). A prolongation to 24 h does not change the swelling ratio significantly. Supplementation of the cryogel with polyP causes a significant increase in the swelling ratio for “KG1/PVA/polyP:NP‐cryogel” to 3. 82 ± 0. 15, and for “KG2/PVA/polyP:NP‐cryogel” and “KG3/PVA/polyP:NP‐cryogel” to 4. 72 ± 0. 17 and 5. 69 ± 0. 29, respectively. We attribute this increase to the superior hydrophilic properties of KG and to an increased Ca‐polyP nanoparticle formation, 20, 39 paralleled with a decrease in hydrogen‐bonding between the polymer chains allowing a stronger uptake of water within the hydrogel network. Table 1 Swelling ratios of the different cryogel compositions after swelling in PBS. The means ± SD have been calculated from 10 independent experiments Sample designation Mass swelling ratio (after 1 h) Mass swelling ratio (after 24 h) “KG1/PVA‐cryogel” 3. 40 ± 0. 08 3. 65 ± 0. 60 “KG2/PVA‐cryogel” 4. 29 ± 0. 13 4. 36 ± 0. 19 “KG3/PVA‐cryogel” 5. 26 ± 0. 19 5. 30 ± 0. 30 “KG1/PVA/polyP:NP‐cryogel” 3. 82 ± 0. 15 4. 02 ± 0. 90 “KG2/PVA/polyP:NP‐cryogel” 4. 72 ± 0. 17 4. 98 ± 0. 18 “KG3/PVA/polyP:NP‐cryogel” 5. 69 ± 0. 29 5. 81 ± 0. 23 John Wiley & Sons, Ltd. Fourier Transform Infrared Spectroscopy (FTIR) Analysis : All the samples show a broad band in the region 3550 and 3200 cm −1 reflecting the stretching vibration of ν(O‐H) that originates from the intra/intermolecular hydrogen bonds of the PVA and KG chains and two bands around the region 3000 and 2800 cm −1 which refer to the stretching of ν(C—H) from the alkyl groups of the PVA and KG backbones (these data are not shown). A comparative FTIR spectral analysis of the KG/PVA‐cryogels and the KG/PVA/polyP:NP‐cryogels with their individual components KG and PVA in the region between 2000 and 600 cm −1 is given in Figure 2. For PVA, the weak bands between 1670 and 1550 cm −1 represent the stretching vibrations of the remaining acetyl groups. The peaks located at 1417, 1323, 1235 cm −1 most likely reflect the bending vibration (δ) of (CH 2 ), wagging vibration (ψ)‐(CH 2 ) and ψ for C—H, respectively. The stretching vibrations for ν(C—O), ν(C—O—C), ν(C—C), and ψ(CH 2 ) within the backbone of PVA polymer chains are observed at 1135, 1083, 912, and 835 cm −1, respectively. 40 The spectrum for KG shows the vibration bands at 1720 and 1242 cm −1 which are due to the stretching ν(C=O) and ν(C—O) from the acetate group, vibrations of the carboxylic group ν(COO − ) at 1603 and 1414 cm −1, the ν(C—H) vibration of methyl coming from the acetyl group at 1373 cm −1 and the ν(C—O) stretching and group vibration of the sugar rings at 1180–940 cm −1 (Figure 2 A). 41 Figure 2 Characterization of the cryogels by FTIR. The signals between the wavenumbers 2000 and 600 cm −1 have been recorded; those that are marked are discussed in the text. A) Cryogels with the components PVA and KG. In addition, the spectra for “KG1/PVA‐cryogel”, “KG2/PVA‐cryogel”, and “KG3/PVA‐cryogel” are shown. B) The spectra of the polyP‐containing cryogels “KG1/PVA/polyP:NP‐cryogel”, “KG2/PVA/polyP:NP‐cryogel”, and “KG3/PVA/polyP:NP‐cryogel”. The shift of those bands which are caused by the interaction between the KG‐carboxylic groups and Ca 2+ are boxed and shaded (COO‐Ca‐OOC) in (A). The FTIR spectra of the different “KG/PVA‐cryogels” show the major absorption peaks that are found also in KG and PVA, with some structural modifications. 42 In all cryogels new absorption bands within the regions 1700–1600 and 1400–1300 cm −1 are observed, which are due to the stretching vibration of the carbonyl groups (C=O), the acetyl as well as the carboxylic groups. These shifts can be attributed to the cross‐linking of the KG with its carboxylic groups to the Ca 2+ ions during ionic gelation and additionally also the formation of hydrogen bonds between the hydroxyl groups of KG and the PVA. It is notable that the intensities of the absorption peaks originating from the acetyl group decreased, reflecting a reduction of the free acetate groups within the KG occurring during the preparation (Figure 2 A). With respect to the polyP‐containing cryogels, the absorption signal for the carboxylic group was also shifted, i. e. , the carboxylic band at 1645 of “KG1/PVA‐cryogel” was shifted to 1640 cm −1 for “KG1/PVA/polyP:NP‐cryogel” (Figure 2 B). Furthermore, the “KG1/PVA‐cryogel” bands close to 1325 cm −1 were decreased with addition of polyP (“KG1/PVA/polyP:NP‐cryogel”), as well as the 1080 and 1027 cm −1 were shifted to 1133 and 1082 cm −1 for “KG1/PVA/polyP:NP‐cryogel”. These modifications are also seen in the spectra for “KG2/PVA/polyP:NP‐cryogel” and “KG3/PVA/polyP:NP‐cryogel” which could be attributed to the incorporation of polyP within the KG/PVA‐cryogel and reflect the ionic interactions between the Ca 2+ ions, the carboxylic group of the KG and polyP. 16 Taken together, the above results indicate the in situ formation of Ca‐polyP nanoparticles within the KG/PVA cryogel (Figure 2 B). Scanning Electron Microscope (SEM) Morphology : The samples were analyzed by SEM ( Figure 3 ). Applying the low‐resolution SEM apparatus for lower magnifications the dry samples appear as an extensive and widespread porous/channel‐traversed material (Figure 3 A–D). The dimensions of the channels reach up to 50 µm, irrespectively if the polyP lacking “KG2/PVA‐cryogel” (Figure 3 A, C) or the polyP supplemented “KG2/PVA/polyP:NP‐cryogel” is analyzed (Figure 3 B, D). Applying the ImageJ analysis software, the pore size of the freeze‐dried “KG2/PVA‐cryogel” was determined with 28 ± 8 µm and for the “KG2/PVA/polyP:NP‐cryogel” with 34 ± 9 µm. Figure 3 Morphology of the KG/PVA cryogel. A, C, F) The polyP‐free “KG2/PVA‐cryogel” sample and B, D, F) the polyP containing “KG2/PVA/polyP:NP‐cryogel” sample were analyzed; SEM. A–D) At lower magnification the cryogels comprises a meshwork, interspersed with extensive porous/channel‐traversed lacunas and openings. Using high resolution SEM microscopy the plain surfaces of the polyP‐lacking “KG2/PVA‐cryogel” samples are devoid of any nanoparticles (E), while those formed in the “KG2/PVA/polyP:NP‐cryogel” are abundantly interspersed with nanoparticles (NP) (F). The corresponding partial images are in the same magnification. At a high‐resolution SEM inspection it becomes overt that the plain surfaces within the cryogel in the samples lacking polyP do not show any nanoparticles (Figure 3 E), while they are present in a high density in the polyP‐containing “KG2/PVA/polyP:NP‐cryogel” sample (Figure 3 F). The nanoparticles are spherical and comprise a diameter of 100–150 nm. Energy Dispersive X‐Ray ( EDX) Analysis/Phosphate Content : The polyP‐free cryogel “KG2/PVA‐cryogel” ( Figure 4 B) shows no signals for phosphorus or calcium, if analyzed by EDX spectroscopy (Figure 4 A). Only two peaks for carbon and oxygen are recorded. In contrast, inspection of the “KG2/PVA/polyP:NP‐cryogel” (Figure 4 D) revealed besides of the peaks for C and O, within the electromagnetic emission spectrum, strong signals for Ca and P, comprising 4. 0 wt% for Ca and 3. 6 wt% for P (semi‐quantitative estimation) (Figure 4 C). If these signals would originate from the complex between Ca 2+ and phosphorus, a ratio of 0. 66 wt% would be expected. However, at present the distribution of Ca 2+ and PO 3 − within the nanoparticles is not known; we expect a higher ratio than 0. 66 wt%, because the particles are only formed in the presence of a surplus of Ca 2+. 16 In addition, KG also comprises some Ca 2+ as a counterion to galacturonic acid. 43 Figure 4 A, C, E) EDX spectra of the cryogels, prepared A, B) in the absence [“KG2/PVA‐cryogel”] or C, D) in the presence of polyP [“KG2/PVA/polyP:NP‐cryogel”], together with the B, D, F) corresponding SEM images. In (E) and (F), the cryogel was incubated for 1 h at 37 °C in DMEM medium with 10% FCS, prior to analysis. It is evident that the signals for P and Ca are present in the “KG2/PVA/polyP:NP‐cryogel”, while they are not visible in the “KG2/PVA‐cryogel”. The presence of Cl in the EDX spectrum cannot be attributed to a potential NaCl contamination, since the signal for Na is lacking in the spectrum. If the KG sample “KG2/PVA/polyP:NP‐cryogel”, after incubation (1 h at 37 °C) in Dulbecco's modified Eagle medium (DMEM) with 10% FCS (Figure 4 F), is analyzed, almost the same content for P and Ca is found with 3. 8 wt% for Ca and 3. 3 wt% for P (Figure 4 C). The content of polyP within the cryogel is dependent on its composition with respect to KG and PVA ( Figure 5 ). At all concentration ratios tested, 1:2 (KG:PVA) via 1:1 to 2:1, and using those samples as a matrix for in situ nanoparticle formation starting with Na‐polyP, the particles can be visualized by SEM (Figure 5 A–C). Semi‐quantitative analysis of the cryogel specimens by EDX revealed that “KG1/PVA/polyP:NP‐cryogel” contains 2. 1 wt% for Ca and 1. 3 wt% for P (Figure 5 D), reflecting a phosphate content of ≈3. 3 wt%. For “KG2/PVA/polyP:NP‐cryogel” (Figure 5 B), Ca–P wt% of 4. 1 and 3. 6 (phosphate content of ≈9. 1 wt%) (Figure 5 E) are determined, and for the “KG3/PVA/polyP:NP‐cryogel” (Figure 5 C) 4. 5 wt% for Ca and 3. 7 wt% for P (phosphate of ≈9. 5 wt%) (Figure 5 F). Figure 5 Morphology and composition of the KG/PVA cryogel, containing in situ formed Ca‐polyP nanoparticles, in dependence on the content of KG. Samples of KG:PVA were prepared in a ratio from 1:2 to 1:1 and finally to 2:1. The gels were supplemented with polyP and exposed to Ca 2+ in order to initiate in situ nanoparticle formation. The gels are termed A) “KG1/PVA/polyP:NP‐cryogel”, B) “KG2/PVA/polyP:NP‐cryogel” [sub‐surface region showing the porous morphology], and C) “KG3/PVA/polyP:NP‐cryogel”. D–F) In a separate series of experiments, the polyP‐containing samples were analyzed by EDX. The corresponding spectra are given. X‐Ray Diffraction ( XRD) Patterns : The XRD patterns of the pure PVA and KG powders as well as the fabricated “KG2/PVA‐cryogel” and the “KG2/PVA/polyP:NP‐cryogel” were recorded ( Figure 6 ). As shown, the pure PVA material exhibits a sharp intense peak at 2θ of 19. 7° and adjacent weak intense peaks at 2θ of 11. 6°, 22. 7°, and 40. 8° (Figure 6 A). This XRD pattern reflects the semi‐crystalline nature of PVA, originating from the strong intermolecular interactions within the PVA chains through hydrogen bonds. 44 In contrast, KG, as well as the “KG2/PVA‐cryogel” and the “KG2/PVA/polyP:NP‐cryogel” patterns do not show any sharp diffraction peak, indicating a highly disordered structure and amorphous nature of the KG used (Figure 6 B). It is notable that within the XRD patterns of “KG2/PVA/polyP:NP‐cryogel” around 2θ of 15°, 23°, and 30° signals appeared that can be attributed to the presence of the Ca‐polyP NP which were formed in situ within the cryogel matrix. This result again supports the amorphous phase of the in situ synthesized Ca‐polyP NP. Taken together, the data show that KG changes the ordered structure and the intense packing of PVA. Figure 6 Characterization of A) PVA and B) KG starting materials, as well as of the polyP‐lacking cryogel, “KG2/PVA‐cryogel”, and of the polyP‐containing material “KG2/PVA/polyP:NP‐cryogel” by X‐ray diffraction (XRD). While the PVA pattern reflects crystallinity, the KG‐containing cryogels are amorphous. 2. 2 Transformation of the Nanoparticles in the Cryogel to Coacervate In a recent study it was found that the polyP nanoparticles undergo coacervation after exposure to peptides/proteins. 20 In the absence of peptides the nanoparticles in the dried “KG2/PVA/polyP:NP‐cryogel” remain permanently present ( Figure 7 A–F). The sizes of the particles, formed in situ after exposure of the KG/PVA cryogel to 50 mg mL −1 Na‐polyP varies around 173 ± 70 nm (30 determinations) (Figure 7 A–C (surface view of the hydrogel)). The porous structure of the cryogel is obvious. In contrast, if the hydrogel is exposed to 100 mg mL −1 Na‐polyP the formed particles are larger with 210 ± 85 nm (Figure 7 D–F (the inner porous morphology is visible)). Figure 7 Transformation of the nanoparticles within the polyP‐containing “KG2/PVA/polyP:NP‐cryogel”; SEM analysis. KG/PVA was exposed either to A–C) 50 mg mL −1 Na‐polyP (surface view) or to D–F) 100 mg mL −1 Na‐polyP (showing the internal morphology) after the addition of CaCl 2 solution (as described under “Experimental Section”) to induce in situ nanoparticle formation; SEM analysis. The morphology of the cryogel containing the polyP nanoparticles (> <NP) was likewise inspected by SEM in the dried state, G–I) 30 min or J–L) 60 min after incubation of the “KG2/PVA/polyP:NP‐cryogel” at 37 °C in medium/FCS. After the 60 min incubation period no polyP particles can be visualized. Also, after immersing in PBS the particles are persistently existent for at least 2 weeks at room temperature (not shown here). In contrast, if the polyP‐containing cryogel is incubated for 30 min (Figure 7 G–I) or 60 min at 37 °C in medium/FCS (Figure 7 J–L) only a few nanoparticles are scattered within the hydrogel, especially at the shorter incubation period. However, this cryogel contains still phosphorus and calcium at the same proportion like in the nonincubated sample (previous paragraph). This finding strongly suggests that the polyP in the cryogel exists in the coacervate state. 2. 3 Mechanical Properties The mechanical properties of the specimens without (“KG1‐3/PVA‐cryogel”) or with the polyP (“KG1‐3/PVA/polyP:NP‐cryogel”) were assessed by force‐displacement‐time measurements under longitudinal oriented compressive or tensile loading conditions. The resulting stress–strain curves clearly show a soft‐tissue like nonlinear viscoelastic behavior of the material 45 ; a representative graph is given ( Figure 8 A). The graph shows a typical viscoelastic stress strain behavior with a distinct hysteresis loop. The ascending branch is nonlinear and followed by a small plateau which reflects the time‐dependent creep phase. The descending branch shows a short, delayed but complete recovery of the material. A comparative analysis between the different polyP‐free and polyP‐containing KG/PVA materials shows that in both series the materials with the higher KG content, like in “KG3/PVA‐cryogel” and “KG3/PVA/polyP:NP‐cryogel”, show a decrease in the mechanical resistance of the respective specimens as seen from the data as follows. “KG1/PVA‐cryogel” shows a compressive modulus/stiffness ( E C ) of 42. 44 ± 0. 01 kPa a value that decreases significantly to 11. 25 ± 01 kPa for “KG2/PVA‐cryogel”; the corresponding value for “KG3/PVA‐cryogel” is with 13. 39 ± 01 kPa again significantly lower than the one for “KG2/PVA‐cryogel”. The tensile tests, which were independently performed (Figure 8 B, C; Figure 9 ), confirmed these changes. The values, reflecting the material properties, and correlating with the ultimate tensile strength (UTS) were found to be for the “KG1/PVA‐cryogel” 399. 37 ± 14. 74 kPa, while for the “KG2/PVA‐cryogel” a sharp decrease is determined with 81. 32 ± 10. 15 kPa and for “KG3/PVA‐cryogel” a further decrease to 45. 74 ± 7. 21 kPa is recorded (Figure 8 B). Finally, the determinations revealed that “KG1/PVA‐cryogel”, which has the lowest KG to PVA ratio (1:2), shows the highest tensile stiffness (143. 67 ± 5. 51 kPa), while for “KG2/PVA‐cryogel” (32. 35 ± 5. 29 kPa) and “KG3/PVA‐cryogel” (19. 57 ± 3. 48 kPa) lower values are measured. Figure 8 Mechanical properties of the fabricated material. A) One exemplary stress/strain curve, the relationship for the “KG2/PVA‐cryogel”, is shown. The nonlinear ascending branch is followed by a distinct plateau which reflects the creep phase. The descending branch shows a delayed recovery. The prominent hysteresis loop is indicative for a viscoelastic material. B) Mechanical properties of the materials (absence or presence of polyP) with respect to the tensile modulus and the ultimate tensile strength. C) Maximum elongation characteristics. The significance was calculated for both the E T values (absence and presence of polyP) and for the UTS values (minus and plus polyP). The values for the polyP‐free “KG2/PVA‐cryogel” and “KG3/PVA‐cryogel” have been correlated either to the “KG1/PVA‐cryogel” (significance levels are indicated by asterisks [*]) or among each other (“KG2/PVA‐cryogel” and “KG3/PVA‐cryogel”; indicated by hashes [#]). The correlation of the polyP‐containing cryogels has been performed accordingly. The mean values ± SD with their significance levels are given (*[#], p < 0. 05; **[##], p < 0. 01; ***, p < 0. 001) versus control, n = 10. Figure 9 Tensile strength and elongation properties, using the “KG2/PVA/polyP:NP‐cryogel” material as an example. A) At the start of the experiment, the material is entirely homogeneous. B, C) With the progression of the experiment and the increase of the tensile force the material shows local ruptures (> <). D) After a 3. 4‐fold stretching the material breaks. E) The corresponding tensile test, curve of the stress/strain relationship, with the “KG2/PVA/polyP:NP‐cryogel” as the example. Stress and strain increased during the continuous stretching of the specimen. The discontinuous transitions within the curve are indicative for proceeding defects in the integrity of the material. At first, small ruptures appear at 300% strain (2 mm mm −1 ) (one arrowhead) (referring to B); then, after increasing the strain to 410% larger fissures of the material appear (two arrowheads; corresponding to C). Finally at ≈440% strain the material fails and the stress level returns to zero (three arrowheads; corresponding to D). The assessment of the maximum elongation (ME) of the three materials shows only slight variations (Figure 8 C). In general, the elongation behavior of “KG1/PVA‐cryogel” (3. 94 ± 0. 36 mm mm −1 ), “KG2/PVA‐cryogel” (3. 26 ± 0. 26 mm mm −1 ), and “KG3/PVA‐cryogel” (2. 86 ± 0. 21 mm mm −1 ) is almost identical. The values of “KG3/PVA‐cryogel” show significant differences, only if compared to “KG1/PVA‐cryogel”. The addition of Na‐polyP to the materials did not result in significant changes of the mechanical properties. The values for the UTS for “KG1/PVA/polyP:NP‐cryogel” (564. 75 ± 109. 15 kPa), “KG2/PVA/polyP:NP‐cryogel” (118. 25 ± 11. 76 kPa), and “KG3/PVA/polyP:NP‐cryogel” (59. 75± 23. 06 kPa) are in a similar range like the corresponding materials without polyP. However, a substantial increase of the standard deviation for “KG1/PVA/polyP:NP‐cryogel” and “KG3/PVA/polyP:NP‐cryogel” is calculated (Figure 8 B). This fact suggests to us a less homogenous behavior of the polyP enriched materials. Also for the E T the determined values are in the same order like those determined in the material without polyP (“KG1/PVA/polyP:NP‐cryogel”: 178. 53 ± 42. 11 kPa; “KG2/PVA/polyP:NP‐cryogel”: 41. 53 ± 2. 65 kPa; “KG3/PVA/polyP:NP‐cryogel”: 28. 25 ± 8. 51 kPa). Again, the standard deviation for “KG1/PVA/polyP:NP‐cryogel” and “KG3/PVA/polyP:NP‐cryogel” increased. The ME of the materials with polyP addition shows no significant difference to the materials without polyP (“KG1/PVA/polyP:NP‐cryogel”: 4. 10 ± 0. 24 mm mm −1 ; “KG2/PVA/polyP:NP‐cryogel”: 3. 65 ± 0. 25 mm mm −1 ; “KG3/PVA/polyP:NP‐cryogel”: 2. 49 ± 0. 36 mm mm −1 ); the extent of the standard deviations remains unchanged. The tensile properties of the KG‐PVA material are pictured in one image using the “KG2/PVA/polyP:NP‐cryogel” (Figure 9 ). At the beginning of the tensile tests the surface of the material is dense and closed; it appears to be homogeneous (Figure 9 A). After increasing the strain to ≈300% small ruptures appear that increase in size (Figure 9 B, C). After a 4. 4‐fold total increase the samples tears (Figure 9 D). The corresponding tensile stress‐strain curve is shown (Figure 9 E) with the marked abrupt discontinuities of these phases. In order to allow a correlation between those values for the fabricated cryogels with native tissue, one series of tensile experiments was performed with bovine muscle tissue. We oriented the muscle fibers perpendicular to the loading direction. The animal tissue gave a UTS value of 73. 96 ± 11. 93 kPa, an E T of 84. 74 ± 59. 39 kPa and an ME value of 1. 56 ± 0. 46 mm mm −1. These values are close to the ones measured for the polyP‐free KG/PVA‐cryogel samples. 2. 4 Cell Viability Studies The MSCs were seeded into 6‐well plates which were either free of any cryogel (control) or contained the cryogel, either as polyP‐free cushion “KG2/PVA‐cryogel” or as polyP‐containing “KG2/PVA/polyP:NP‐cryogel”. The cell growth/viability was determined with the XTT assay system, as described under “Experimental Section” ( Figure 10 ). After an incubation period of 36 h the cell number increased to the same extent in all three series, as can be deduced from the increase in the absorbance from 0. 48 ± 0. 06 (at the seeding) to ≈0. 8 absorbance units at 450 nm. After an incubation period of 72 h the cell concentration increased further to 1. 68 ± 0. 19 units (control) and 1. 84 ± 0. 21 (“KG2/PVA‐cryogel”), or to 2. 81 ± 0. 26 (“KG2/PVA/polyP:NP‐cryogel”), respectively. The significant increase in the growth rate, reflecting also the remarkable biocompatibility of the polyP‐containing hydrogel “KG2/PVA/polyP:NP‐cryogel” in comparison to the control or the “KG2/PVA‐cryogel” is even more pronounced after a total incubation period of 168 h. Figure 10 Viability studies using MSC. The cells were plated onto either hydrogel‐free wells (controls), or onto cryogel cushions lacking polyP (“KG2/PVA‐cryogel”) or the polyP‐containing material (“KG2/PVA/polyP:NP‐cryogel”). The numbers of viable cells were determined using the XTT assay (A 450 values are given). Three incubation periods (36, 72, and 168 h) were selected. Ten parallel assays were performed and the mean values (±SD) are given. Significant correlations are determined within a given incubation period; they are marked (*, p < 0. 001). 2. 5 Infiltration of the Stem Cells into the Cryogel As outlined under “Experimental Section” the polyP‐lacking and polyP‐containing cryogels were prepared onto the bottom of 6‐well plates. After 8 or 24 h the cells were fixed and inspected by SEM. In the series of cultivation onto “KG2/PVA‐cryogel” the cells adhere during the first 8 h to the cryogel matrix ( Figure 11 A, B) and remain there also after 24 h (Figure 11 C, D). In contrast, if the MSCs were plated onto the “KG2/PVA/polyP:NP‐cryogel” they appear to become coated with the hydrogel during the first 8 h (Figure 11 E, F) or even infiltrate into the channel‐like system of the material (Figure 11 G, H). Figure 11 Infiltration of MSC into KG/PVA matrices. The cells were layered onto either A–D) the “KG2/PVA‐cryogel” or E–H) the “KG2/PVA/polyP:NP‐cryogel”. Then both series were incubated for 8 or 24 h, as indicated; subsequently, the samples were fixed, coated, and subjected to low‐resolution SEM. Primarily the MSCs attached onto the polyP‐containing cryogel (cry) were found to migrate into the channels of the gel, as in (H). A–C, E–G) Surface aspects and D, H) view of cross fractures. In order to verify that the cells are infiltrating into the cryogel the samples were inspected by confocal laser scanning microscopy (cLSM) ( Figure 12 ). The cells (20 × 10 3 cells mL −1 ) were plated onto either the “KG2/PVA‐cryogel” or the “KG2/PVA/polyP:NP‐cryogel” and incubated for 8 h. Then the samples were stained and inspected by cLSM. The individual images were collected and the resulting 3D images were compiled. The computed 3D files showing the distribution of the cells distinctly show that only comparably few cells infiltrated into the “KG2/PVA‐cryogel” (Figure 12 A), while the accumulation of the cells within the “KG2/PVA/polyP:NP‐cryogel” is more dense (Figure 12 B). In order to document also the individual layers of the distribution of the cells within the polyP‐free (Figure 12 C‐a–C‐c) and polyP‐containing gel (Figure 12 C‐d–C‐f) individual images from the top to the bottom of the stack are given. Optical cell counts of defined areas within the cLSM stacks (100 µm × 100 µm × 100 µm) revealed an approximate number of 68 ± 28 cells, infiltrated within the “KG2/PVA/polyP:NP‐cryogel”, while only 18 ± 12 cells were counted within “KG2/PVA‐cryogel” samples. Figure 12 Analysis of the distribution of MSC within A) the “KG2/PVA‐cryogel” and B) the “KG2/PVA/polyP:NP‐cryogel” by cLSM. The cells were incubated for 8 h, then stained with Calcein AM and inspected. The compressed stacks are given in (A) and (B). In (C), individual slices from the poly‐free and the polyP‐containing cryogel (from the top to the bottom) are shown: C‐a–C‐c) aspects within the “KG2/PVA‐cryogel” and C‐d–C‐f) within “KG2/PVA/polyP:NP‐cryogel”. This series shows the increasing density of the cells within the polyP containing cryogel. 2. 6 In Vivo Biocompatibility Studies in Rats The KG/PVA microspheres were prepared as described in the Experimental Section. An emulsion procedure was selected to obtain round microspheres of a diameter of 820 ± 70 µm ( Figure 13 ). The inner structure of the spheres is porous, interspersed with channels with dimensions ranging between 15 and 30 µm (not shown). No pronounced morphological differences between the polyP‐free “KG/PVA:MS” and the polyP‐supplemented spheres, “KG/PVA/polyP:MS”, were seen (Figure 13 A, C and B, D). The polyP content of the “KG/PVA/polyP:MS” was determined to be 8. 5 ± 0. 7 wt%. Figure 13 KG/PVA microspheres, used for the in vivo biocompatibility studies. The spheres contained as a scaffold KG and PVA and were prepared by an emulsion technology. A, C) The control, polyP‐free spheres (“KG/PVA:MS”); B, D) polyP‐containing microspheres “KG/PVA/polyP:MS”; A, B) light microscopic and C, D) SEM images. Samples of microspheres (20 mg) were implanted into the back of the rats, as described in the Experimental Section. After a healing period of 2 weeks or 4 weeks tissue samples were taken, sliced and stained with the hematoxylin. Eye inspection of the samples revealed that in none of the animals histopathological alterations evolved. Histological analyses of all sample sections of the two groups were performed. Typical images for the stained sample sections are shown ( Figure 14 ). It is striking that after 2 weeks the regions where the polyP‐free implanted microspheres, “KG/PVA:MS”, were placed contained “hollow” spaces which did not contain infiltrated cells (Figure 14 A, B). In contrast, the regions into which the polyP‐supplemented spheres, “KG/PVA/polyP:MS”, were inserted were already partially replaced by cells forming an initial granulation tissue‐assembly (Figure 14 C). An extension of the implant regeneration period to 4 weeks intensified the extent of tissue regeneration with “KG/PVA/polyP:MS”; granulated regions which contained initially the spheres were surrounding the muscle fasciae (Figure 14 D). Those regions were not found in places which contained the polyP‐free samples (Figure 14 B). Figure 14 Implantation of the KG/PVA microspheres into back muscle regions of rats. A, B) Insertion of polyP‐free spheres (“KG/PVA:MS”); C, D) placing of polyP‐containing microspheres, “KG/PVA/polyP:MS”. Samples were taken after A, C) 2 weeks and B, D) 4 weeks, sliced, stained with hematoxylin and inspected. MS, microspheres; gt, granulation tissue‐assembly; m, muscle fasciae. 3 Discussion Any kind of animal tissue is structured, based on the underlying fibrous network primarily due to the complex and organized collagen meshes. This characteristics is required for providing the tissue form, strength, and stability. Surely, the degree of complexity between different organs varies; among the highly hierarchically structured, fibrillar collagen‐based tissues are the cornea 46 and the muscle. 47 Focusing on muscle, this tissue is macroscopically composed of muscular fibers, built of myofibrils constructed from actin, myosin, titin, and other proteins that hold these fibrils together. 48, 49 The muscles have viscoelastic properties since they comprise both viscous and elastic characteristics during the process of deformation. Especially if connected with the tendon, the muscles comprise a high degree of stress relaxation, 50 optimizing this tissue for considerable stiffness and pronounced damping to the loading response. 51 Since the characteristics of most tissues is that they do not exhibit linear elasticity because of the presence of the locally and spatially distinct arrangement of the cells and their surrounding extracellular matrix, an artificial scaffold suitable for soft tissue repair is preferentially a hydrogel showing viscoelastic properties. 52 Even more advanced would be hydrogels that are poro‐viscoelastic in order to facilitate vascularization. 53 In the present study a hybrid viscoelastic hydrogel is introduced that is constructed from two polymer systems that are cross‐linked through two different interacting forces; first by PVA whose molecules are linked together via hydrogen bonds and second by KG which allows cross‐linking through their anionic sugar resides via Ca 2+ (see Section 1 ). In a first step, the PVA/KG solution was subjected to freezing‐thawing cycles to introduce a porous channel system and to allow physical cross‐linking of the PVA chains to occur, and subsequently to Ca 2+ that causes ionic gelation (Scheme in Figure 15 A). Both bonding types are noncovalent, allowing a considerable degree of flexibility and elasticity of the scaffold, which are prerequisites for cells to invade into the scaffold or to migrate onto this material. By alteration of the pH conditions in the vicinity of the cells or by peptides released by them and interacting with the ionic and hydrogen bonds the strength of the network is substantially affected. 20 This property of the scaffold substantially controls the relaxation and retardation times of the MSCs, and by that also the viability and the differentiation capacity of those cells. 54 Figure 15 Formation of porous KG‐PVA hydrogel based on physically cross‐linked PVA (during the freezing‐thawing process) and on ionic gelation of KG (scheme flanked with SEM images on the right panel). A) Formation of the cryogel in the absence of polyP. The resulting material is porous (SEM insert). After further processing with Ca 2+ a porous “KG/PVA‐cryogel” is formed that allows the MSC to attach. B) If Na‐polyP is added together with KG and PVA and subsequently processed by freeze‐thawing, the physical cross‐linkage of the PVA matrix occurs. The subsequent exposure to Ca 2+ causes cross‐linkage of the KG under formation of the hybrid KG/PVA cryogel by ionic gelation (upper SEM image in panel B). In parallel 100–150 nm large Ca‐polyP nanoparticles (NP) are formed in situ that appear abundantly on the KG‐PVA matrices (lower SEM image in panel B). After lyophilization the material is stable and can be stored. If transferred to medium/serum the polyP‐containing cryogel, “KG/PVA/polyP:NP‐cryogel”, allows the nanoparticles to undergo coacervation; this matrix provides the MSC a suitable matrix to adhere and also to infiltrate into the cryogel. The PVA/KG‐based cryogel comprises a considerable potential to take up water/PBS. For all cryogel species synthesized here the swelling ratio was larger than 3, qualifying the material as a suitable hydrogel for tissue engineering. 55 Increasing the KG content increased the PBS uptake capacity by > 50%. The increase in the KG content also decreases the tear strength, without considerable effect on the elasticity of the cryogel. Likewise, the tensile modulus decreases allowing a lower elastic deformation of the material. The mechanical properties of the PVA/KG‐based cryogel were compared with those of bovine muscle. Those studies revealed that the UTS value and, in parallel, also the E T value are close to those measured for the KG/PVA‐cryogel samples, especially for “KG2/PVA‐cryogel” and “KG3/PVA‐cryogel”. 45, 56 It is important to note that, in the XRD pattern, PVA shows sharp crystalline reflections, with a strong maxima at 2θ of 19. 7° and adjacent weak intense peaks at 2θ of 11. 6°, 22. 7°, and 40. 8°. This characteristic has been described earlier 57 and is confirmed here. If PVA is added to KG, even under the different weight ratios selected, the amorphous phase of KG remains unchanged. This observation underscores the usefulness of the cryogel as a potential implant material in tissue engineering. This assessment bases on the observation disclosing that implants having an amorphous phase are, in general, more suitable than crystalline materials. 58 The distinguished feature of the PVA/KG cryogel is its characteristics to allow an in situ synthesis/formation of nanoparticles from polyP and Ca 2+ ions (Figure 15 B). Usually hydrogels provide ideal environment to support cell adhesion and tissue formation, based on their bioinert nature. 59 Additional functionalization of the materials can be introduced, e. g. , by cell‐binding domains that serve as ligands for integrins and by that provide anchorage and triggering signals that direct cell function and the expression of the differentiated phenotypes. 60 In the present study, a procedure is described that allows the in situ synthesis of nanoparticles of morphogenetically active polyP into the PVA/KG cryogel. Those Ca‐polyP particles have a size similar to the ones that have been previously synthesized in aqueous solution. 16 The maximum content of polyP within the scaffold was around 9. 5 wt%. This level is wanted since a much higher amount would open the risk of a local pH drop during the process of enzymatic hydrolysis by ALP. Taken together, the fabricated PVA/KG cryogels show a tissue‐like nonlinear viscoelastic behavior with comparable maximum elongation values, as seen for muscle and cartilage. 45, 52 The shift of the KG/PVA ratio to higher PVA contents leads to an increased tensile strength as well as a higher stiffness of the materials and vice versa. The addition of polyP does not result in a substantial alteration of the mechanical properties of the hybrid material. It might be highlighted that both components of the scaffold described here, PVA and KG, are biocompatible and biodegradable. 38, 61 PVA is more stiff and tough in comparison to KG. 62, 63 This finding is corroborated by our studies revealing that an increased KG content of the cryogel decreased the Young's modulus. In a recent study, we described that biomimetically fabricated Ca‐polyP nanoparticles are not degraded/metabolized unless they are transferred into the coacervate state. 20 This transformation from the amorphous state as nanoparticles to the coacervate has been demonstrated to occur after transfer to a proteinaceous environment. Now, we show that this transformation to the biologically active polyP coacervate also proceeds with in situ formed Ca‐polyP nanoparticles. After exposure of the KG/PVA‐cryogel to a protein‐containing medium/serum the nanoparticles within the cryogel disappeared without changing the polyP content. These data are strong indications that the transition of the nanoparticles from the amorphous nanoparticulate state to the coacervate state also occurred in situ, implying that the bioavailability of the nanoparticles in the fabricated cryogel is not impaired. As demonstrated 20 the polyP‐containing cryogel, most of the studies have been performed with “KG2/PVA/polyP:NP‐cryogel”, provides an excellent matrix for MSCs to proliferate. The growth rate onto the polyP‐containing cryogel is higher than the one onto the polyP‐lacking cryogel. Intriguing is the observation that the cells cultivated onto the polyP‐containing hybrid cryogel show a high propensity to infiltrate in the matrix. This SEM observation was corroborated by cLSM analysis. In addition, previous studies have shown that in response to polyP the expression of the chemokine SDF‐1α is upregulated. 64 In order to test the biocompatibility of the new hybrid material in vivo, the material was fabricated as microspheres using the emulsion technology, 65 which were implanted into the muscle tissue. After a 4 week's regeneration phase the regions into which the polyP‐free microspheres were implanted were free of any cells. In contrast, the regions which were implanted with the polyP‐containing (better: in situ formed Ca‐polyP nanoparticles enriched) KG/PVA hydrogel were replaced by cells under formation of granulation tissue‐like assemblies. This finding highlights the beneficial component, polyP, in the microspheres as a morphogenetically active constituent. 4 Conclusion The cryogel presented here is not only equipped with suitable, functionally adaptable tissue‐matching properties and space‐filling characteristics, but also with the potential to stimulate the growth of MSCs in vitro and—most importantly—accelerate the cell‐based repair/regeneration processes in vivo. The technology presented here, in which morphogenetically active nanoparticles are formed in situ, is a versatile method that might be applicable also for other hydrogel scaffold materials. Even though this study describes a one‐stage in situ synthesis procedure, an adaptation of the method to multistep processes appears to be straightforward. In ongoing studies the application of the in situ technology for the inclusion/supplementation of Ca‐polyP nanoparticles into i) microspheres, suitable to embed stem cells, ii) artificial, biomaterial‐based blood vessels, and iii) 3D printed implants is elaborated. 5 Experimental Section Materials : Na‐polyphosphate (Na‐polyP) with an average chain length of 40 phosphate units was from Chemische Fabrik Budenheim (Budenheim; Germany). Preparation of the Porous Karaya Gum/Poly(vinyl alcohol) Hydrogel : Ten grams of PVA ( M w 146 000–186 000; #363065 Sigma‐Aldrich, Taufkirchen; Germany) were dissolved in 100 mL of distilled water at 90 °C, while stirring for 3 h. In parallel, 5 g of KG (from sterculia tree; #G0503 Sigma) were dissolved in 100 mL of water at 90 °C for 3 h. Then, different volumes of the KG solution were added to a fixed amount of PVA and stirred for 1 h. In this study, three different weight ratios of KG:PVA were chosen, namely 1:2, 1:1, and 2:1. After mixing, the resulting KG/PVA viscous solutions were allowed to stand for 3 h to remove air bubbles. Subsequently, 20 mL of the respective KG/PVA solution were poured into a 12 cm Petri dish (plastic) and frozen down to −80 °C for 6 h. Afterward, the frozen dishes were thawed at room temperature for a period of 3 h. This freezing‐thawing cycle was applied three times to obtain a 3D structured KG/PVA hydrogel (based on physical cross‐linking of PVA chains). The obtained KG/PVA hydrogel was cut out by using a sharp biopsy punch of 10 mm inner diameter. One batch was directly immersed into a calcium chloride solution (5%; CaCl 2 ·2H 2 O) for 6 h under agitation to allow ionic cross‐linking of the KG (ionic gelation). The obtained KG/PVA hydrogel was washed three times with water to remove the nonreacted ions and freeze dried at −80 °C to obtain the KG/PVA hydrogel; the three samples were termed “KG1/PVA‐cryogel” (ratio between KG and PVA 1:2), “KG2/PVA‐cryogel” (ratio KG and PVA 1:1), and “KG3/PVA‐cryogel” (ratio KG and PVA 2:1). For the in situ synthesis of Ca‐polyP nanoparticles (NPs) within the KG/PVA matrix, the obtained hydrogels were immersed into Na‐polyP solution (1 g/20 mL distilled water) for 2 h, prior to the addition of 5% CaCl 2 solution to allow the formation of Ca‐polyP NP within the KG/PVA hydrogel. In one series of experiments the concentration of Na‐polyP solution was increased to 2 g/20 mL water. The resulting samples were termed “KG1/PVA/polyP:NP‐cryogel”, “KG2/PVA/polyP:NP‐cryogel”, and “KG3/PVA/polyP:NP‐cryogel”. All the outlined samples were lyophilized prior to a further characterization. The swelling ratio ( Q M ) was defined as the fractional increase of the cryogel in the weight of the hydrogel due to PBS (phosphate buffered saline) uptake and was calculated after incubation of the cryogel for different time periods at 37 °C according to the following equation; Q M = mass of fully swollen gel/mass of dried hydrogel. 66 Stability of In Situ Formed Nanoparticles in the Cryogel : Samples of 50 mg of “KG2/PVA‐cryogel” were exposed to 1 mL of PBS or to 1 mL of Dulbecco's modified Eagle medium (DMEM) with 10% FCS for a period of 60 min at 37 °C. Then the cryogel samples were washed and processed. In one additional series the samples, exposed to FCS+medium, were already inspected after 30 min by SEM. Fourier Transform Infrared Spectroscopy : The analysis by FTIR was performed with an attenuated total reflectance‐FTIR spectroscope/Varian IR spectrometer (Agilent, Santa Clara; CA). The ground composite powder was analyzed. X‐Ray Diffraction : The XRD patterns of dried powder samples were recorded with a D8 Advance A25 diffractometer (Bruker, Billerica, MA; USA) using monochromatic Cu‐Kα radiation. 67 EDX Analysis : The experiments were performed with an EDAX Genesis EDX System attached to a scanning electron microscope (Nova 600 Nanolab, FEI, Eindhoven; The Netherlands). The analyses were operated at 10 kV with a collection time of 30–45 s. Areas of 10 µm 2 were analyzed. In a first approximation, the signals corresponding to the different elements were used for a semi‐quantitative assessment. 68 Determination of the Mechanical Properties: Compression Testing : In order to determine the bulk mechanical properties of the respective KG/PVA materials the samples were analyzed with the MultiTest 2. 5‐xt force testing system (Mecmesin Ltd. , Slinfold; UK), connected with the 100 N Load Cell unit. The experiments were performed with standardized cylindrical scaffold samples which were punched out from the sample material using a stainless steel biopsy trepan. They measured 10 mm in diameter and ≈1. 5 mm in height, and were loaded in the longitudinal direction with an acceleration of 1 mm min −1. The data for the calculation of the compressive modulus were continuously recorded at a frequency of 50 Hz using the Emperor XT Force software. To determine the compressive modulus, forces of 0. 25, 0. 5, 1, and 2. 5 N, respectively, were applied to the samples and kept for 60 s. Subsequently, an unloading period (0 N) followed for 300 s. The resulting force‐displacement‐time data were used to calculate the compressive modulus ( E C ) of the material and the tensile modulus ( E T ). 69 Determination of the Mechanical Properties: Tensile Testing : For the tensile experiments standardized 3–4 mm thick stripes, measuring 30 mm in length and 12 mm in width, were cut out from the sample material. These stripes were loaded in the longitudinal direction with an acceleration of 5 mm min −1 until tearing off. The continuously recorded force, displacement and time data (sampling rate: 50 Hz) were used to calculate the UTS, the ME, as well as the tensile modulus ( E T ). 69 The same procedure was applied also for bovine muscle samples from the lower leg; the muscle was obtained from a local butcher. The samples were cut to the same dimensions as the KG/PVA materials and analyzed as described above. Cultivation of Human Mesenchymal Stem Cells : Human MSCs, obtained from bone marrow of healthy nondiabetic adult volunteers, were purchased from Lonza Cologne (Cologne; Germany). They were cultivated as described before. 17 Shortly, the cells were maintained in 75 cm 2 flasks and cultivated in DMEM medium, supplemented with 10% FCS (Biochrom, Berlin; Germany) and 0. 5 mg mL −1 of gentamycin, 100 units mL −1 of penicillin, 100 µg mL −1 of streptomycin, and 1 × 10 −3 M pyruvate (#P2256 Sigma‐Aldrich). The assays were incubated in a humidified incubator at 37 °C. For plating of the MSCs onto the KG/PVA cryogel, the material was prepared first in 6‐well plates (#M8562; Sigma). A final volume of 1 mL of “KG2/PVA‐cryogel” was prepared in each well following the described procedure by freezing‐thawing, followed by incubation with 5% CaCl 2. For the preparation of the “KG2/PVA/polyP:NP‐cryogel” in the 6‐well plates the samples were treated with Na‐polyP and finally with 5% CaCl 2. After extensive washing with medium, 3 mL cell suspension (20 × 10 3 cells mL −1 ) were placed onto the cryogel prepared onto the bottom of the well plates. For the determination of the attachment properties of the MSC the incubation was performed for 8 or 24 h. In one series of experiments, the samples were inspected by low‐resolution SEM and after fixation with paraformaldehyde and glutaraldehyde. 9 The samples were coated with gold; in a second one the cryogels were exposed to Calcein AM (#17783; Sigma) to analyze the cells 70 within the cryogel by cLSM. 71 Cell Proliferation/Viability Assay : MSCs were seeded into the 6‐well plates at a density of 10 4 cells per well and were cultured in DMEM medium/10% FCS for the indicated period of time (total volume of 3 mL). The parallel assays were set. First, in the control wells no further material/cryogel was added. In the second and third series, the polyP‐free cryogels (height of the cushion: 1 mm) were prepared using the “KG2/PVA‐cryogel” composition and for the polyP‐containing cushions the “KG2/PVA/polyP:NP‐cryogel” formula was applied. The cells were layered onto the gels. The proliferation/viability assays were performed using the XTT Cell Proliferation Kit II [2, 3‐bis‐(2‐methoxy‐4‐nitro‐5‐sulfophenyl)‐2 H ‐tetrazolium‐5‐carboxanilide], purchased from Roche (Mannheim; Germany), as outlined. 72 The cell growth/metabolic activity of the living cells was determined on the basis of the extent of oxidation of the tetrazolium salt. The absorbance was determined at 450 nm and the values were subtracted by the background values (500 nm). The cells were incubated for up to 7 days. Ten parallel experiments were performed. Preparation of Karaya Gum/Polyvinyl Alcohol Microspheres : The microspheres were prepared as follows. The aqueous phase was composed with 10 g of PVA in 100 mL of distilled water and heated to 90 °C (while stirring for 3 h) to dissolve the polymer. In parallel, 10 g of KG were dissolved in 100 mL of water at 90 °C for 3 h. Then, 50 mL of both solutions were mixed together on a magnetic stirrer. The oil phase contained in 200 mL of paraffin oil (#18512, Sigma) 10% (wt/wt) Tween‐80 (#P4780, Sigma). To prepare the microspheres 20 mL of the aqueous phase were added to the oil phase under stirring (1500 rpm; room temperature) during a 30 min period. After that, 10 mL of a CaCl 2 solution (from the stock solution of 1 g CaCl 2 /10 mL water) were filled into a syringe, and in separate but at the same time, 10 mL of a Na‐polyP solution (stock solution of 0. 2 g/10 mL water) were likewise filled into a syringe. Then, in parallel, these two solutions (both 10 mL) were added under stirring, drop‐by‐drop, to the oil phase. After stirring for 3 h at 500 rpm the suspension was taken. The particles were sedimented during 10 min. The oil phase was removed by decantation. The sediment was washed three times in distilled water. Finally the material was centrifuged at 1000×g and the spheres were freeze‐dried at −80 °C; they are termed “KG/PVA/polyP:MS”. The polyP‐free microspheres were prepared in the same way without adding polyP; “KG/PVA:MS”. The “KG/PVA/polyP:MS” samples contained ≈8. 5 wt% polyP, as determined by EDX. Animal Studies : The biocompatibility studies were performed with Wistar rats (male gender; age of two months) weighting between 250 and 300 g. 73 All experimental procedures were approved by the ethics committee at the Dongzhimen Hospital at the Beijing University of Chinese Medicine (No. 5 Haiyuncang Road, Dongcheng District, Beijing 100700; Beijing Committee of Science and Technology). The certificate number for the approval is 2012‐0001a; the experimental studies were performed by Dr. Xing YU. Four animals per each group were used. The animals were kept at constant room temperature (22 ± 2 °C), controlled 12 h light/dark cycle and humidity (≈50%); diet and water were provided ad libitum. As described, 73 preoperatively the animals were treated with Ciprofloxacin (Bayer, Leverkusen; Germany) at a dose of 10 mL kg −1 of body weight for antibiotic prophylaxis. Subsequently, the animals were narcotized with chlorpromazine (Smith, Kline & French, Philadelphia; PA)/Ketamine (Ketanest; Pfizer, Groton; CT) using the intramuscular route. After routine disinfection, incisions of ≈1 cm were made in the right and left half and oriented perpendicularly to the vertebral axis at the upper limb level. After skin incision, the muscle was incised and dissected to allow the insertion of the microspheres. The implanted material (≈20 mg in a volume of 100 mL) was introduced into the muscle and stabilized there in the deeper layer. 74, 75, 76 After a period of 2 or 4 weeks the animals were sacrificed and the specimens with the surrounding tissue were dissected and sliced. The samples were inspected macroscopically for inflammation, infection and discoloration. Two groups of studies were performed; one group received the microspheres without polyP “KG/PVA:MS”, and one group with polyP “KG/PVA/polyP:MS”. For histological inspection the samples were fixed in formalin, sliced, stained with Mayer's hematoxylin (#MHS1; Sigma) and then analyzed by optical microscopy (using an Olympus AHBT3 microscope). 77 Microscopic Analysis–Pore Size Determinations : The high‐resolution SEM images were obtained with a Zeiss Gemini 1530 (Zeiss Oberkochem; Germany). For low‐resolution SEM studies an ESEM XL‐30 machine (Philips, Eindhoven; Netherlands) was applied. 78 cLSM was performed with a Leica apparatus (Microsystems, Bensheim; Germany) of Calcein AM‐stained cells. Serial confocal images (64 z‐slices) were captured and the 3D overlaid image stacks were computed using an average projection algorithm provided with the TCS‐SP software. The settings of the microscope have been given recently. 79 Pore‐size determinations were performed with SEM/ESEM XL images obtained from gold sputtered cryogel samples, as described. 80 Then ImageJ analysis software was applied to determine the averaged mean pore sizes; randomly selected samples have been analyzed. Statistical Analysis : The values reported are the average ± standard deviations. Statistical analyses were performed with the one‐way ANOVA test, by using SigmaStat 3. 5 software (Dundas Software Ltd, Toronto; Canada). Values of p < 0. 05 were considered statistically significant. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201801555
| 2,018
|
Advanced Science
|
Self‐Healing and Injectable Hydrogel for Matching Skin Flap Regeneration
|
Abstract The fabrication of highly biocompatible hydrogels with multiple unique healing abilities for the whole healing process, for example, multifunctional hydrogels with injectable, degradation, antibacterial, antihypoxic, and wound healing–promoting properties that match the dynamic healing process of skin flap regeneration, is currently a research challenge. Here, a multifunctional and dynamic coordinative polyethylene glycol (PEG) hydrogel with mangiferin liposomes (MF‐Lip@PEG) is developed for clinical applications through Ag–S coordination of four‐arm‐PEG‐SH and Ag +. Compared to MF‐PEG, MF‐Lip@PEG exhibits self‐healing properties, lower swelling percentages, and a longer endurance period. Moreover, the hydrogel exhibits excellent drug dispersibility and release characteristics for slow and persistent drug delivery. In vitro studies show that the hydrogel is biocompatible and nontoxic to cells, and exerts an outstanding neovascularization‐promoting effect. The MF‐Lip@PEG also exhibits a strong cytoprotective effect against hypoxia‐induced apoptosis through regulation of the Bax/Bcl‐2/caspase‐3 pathway. In a random skin flap animal model, the MF‐Lip@PEG is injectable and convenient to deliver into the skin flap, providing excellent anti‐inflammation, anti‐infection, and proneovascularization effects and significantly reducing the skin flap necrosis rate. In general, the MF‐Lip@PEG possesses outstanding multifunctionality for the dynamic healing process of skin flap regeneration.
|
Hydrogels are biomaterials widely used for tissue repair in tissue engineering. With their biocompatibility and tunable architecture, hydrogels can simulate the physical structure of the extracellular matrix to promote cell proliferation and tissue regeneration. 1, 2, 3 Hydrogels are able to encapsulate and deliver different drugs for various kinds of tissue repair processes, for example, flap regeneration, skin healing, and bone reconstruction. 4, 5, 6 Injured tissue repair is a dynamic and complex process. During the early stage of skin flap recovery, tissue suffers from damaged homeostasis and cell apoptosis caused by inflammation, ischemia, and hypoxia; then, in the proliferation period, the cells proliferate rapidly after endothelial cells exhibit adequate vascularization, while fibroblasts and mesenchymal stem cells help with healing and regeneration. Finally, during the remodeling period, the tissue undergoes structure and vascular network remodeling and maturing, which take months or years to for the tissue to eventually regain its normal function. 7, 8 However, little research has been done to optimize hydrogels' drug releasing and degradation property to match with each dynamic healing process. Therefore, multifunctionalized hydrogels that could synchronize with tissue physiological healing process would have significant potential in bioengineering applications. For example, in the early stage of random skin flap reconstruction, deprivation of adequate blood supply causes flap hypoxia, ischemia/reperfusion, and inflammation, especially in the distal part. The unfavorable hypoxic environment hinders flap rapid regeneration during the early stages. To address this problem, previous studies mainly focus on using hydrogels as various drugs and cytokine delivery systems to reduce distal end necrosis in the skin flap. However, using cytokines to stimulate angiogenesis does not protect cells from hypoxia, high concentrations of cytokines and drugs can cause side effects and even impair healing. 9, 10, 11 Moreover, in these studies, the hydrogels only act as physical supports to deliver cargos, and thus, the healing effect of these hydrogels is not an area of concern. 12 Hydrogels can be modified to improve their performance in tissue engineering, including cell proliferation, new microvascular network formation, and tissue extracellular matrix structure construction. 13, 14 Although many studies have focused on the chemical and physical internal structure of hydrogels, the healing improvement capacity of hydrogels themselves has received little attention. 15 Hydrogels delivered to the wound area during the tissue healing process can directly contact and interact with the wounded cells. The hydrogel's drug releasing and degradation property could be modified to match with the dynamic healing process. 16 Herein, we investigate the biological repair and healing effect of hydrogels with the goal to uncover its role and mechanisms in tissue repair and regeneration. In this work, we crosslinked four‐armed polyethylene glycol (PEG) with Ag + by Ag–S coordination to form a dynamic coordinative and injectable PEG‐Ag hydrogel that was loaded with mangiferin liposomes (MF‐Lip) to promote skin flap survival rate through the combination of the cytoprotective effect of mangiferin and sustained anti‐infection and accelerated wound healing activity of Ag + under hypoxic conditions. 17 The dynamic coordinative PEG hydrogel constructed by Ag–S coordination exhibited outstanding injectable, self‐healing, degradable, and anti‐infection properties. In addition, mangiferin (MF), a naturally occurring glucosyl xanthone commonly found in mango and papaya, has been suggested to possess various pharmacological activities, such as antioxidant, anti‐inflammatory, antiapoptotic, and immunomodulation effects. 18, 19, 20, 21 Herein, we fabricated a hydrogel that is able to locally release Ag + and MF and that exhibit anti‐infection, cytoprotective, and angiogenesis activity under hypoxic conditions. The hydrogel delivery system slowly degrades with the skin flap healing process while guaranteeing enough mangiferin concentration in situ to promote skin flap regeneration. With the combination of cytoprotective and angiogenesis of mangiferin and the anti‐infection and healing promotion ability of Ag +, the hydrogel delivery system is expected to match the renewal of random skin flap regeneration progress and promote skin flap survival. ( Scheme 1 ). Scheme 1 Schematic illustrations of MF‐Lip@PEG matching the renewal of random skin flap regeneration. a) Four‐arm‐PEG‐SH, MF and Ag + constructed by Ag–S coordination form the dynamic coordinative hydrogel. b) MF‐Lip@PEG was injected into the derma, and the MF‐Lip was encapsulated in situ during the Ag–thiol coordination. The hydrogel continued to release MF‐Lip and Ag + while slowly degrading with the flap healing process. c) The MF‐Lip@PEG promote skin flap survival rate through the combined cytoprotective, neovascularization‐promoting, anti‐inflammation, and antibacterial effects and the reduction in skin flap necrosis rate under hypoxic conditions, matching the renewal of flap regeneration. MF was loaded in the lipid bilayer of liposomes, which consists of phosphatidylcholine and cholesterol (Figure S1, Supporting Information). The morphology of the MF‐Lip was investigated by transmission electron microscopy (TEM) (Figure S2, Supporting Information) and scanning electron microscopy (SEM) (Figure S3, Supporting Information). According to these images, the nanoparticles showed a bilayer structure; specifically, the darker parts were the inner water phase of the liposomes, and the lighter parts indicated the lipid bilayer. For the surface properties, lyophilized liposome powder with a round smooth surface was observed, as shown in Figure S3 in the Supporting Information. To further characterize the MF liposomes, their size distribution and polydispersity index (PDI) were investigated (Figure S4, Supporting Information): the average size of the MF‐Lip was 142. 7 ± 2. 4 nm, and the PDI was 0. 140 ± 0. 023. Moreover, the zeta potential distribution is shown in Figure S5 in the Supporting Information, and the mean zeta potential was −8. 57 ± 1. 11 mV. The entrapment efficiency of these liposomes was 66. 7 ± 3. 2% and the drug release in vitro lasted for almost 24 h, with the cumulative percent release reaching over 75% (Figure S6, Supporting Information). To explore the process of gelation ( Figure 1 a, b), it was observed that before mixing the solution of four‐arm PEG‐thiol with AgNO 3, the solution of the materials had excellent flowability. After mixing these two solutions, a composite gel containing MF‐Lip was formed. Furthermore, the injection capability of the gel was confirmed, as shown in Figure 1 c. The rhodamine‐dyed MF‐Lip@PEG was continuously injected though the syringe with a 0. 5 mm needle. The hydrogels were also injected into a centrifuge tube filled with deionized (DI) water to prove that they can not only be injected but also maintain their shape in solution, as shown in Figure 1 d−f. The surface structure of the hydrogel was observed by SEM at 1 h, as shown in Figure S7 in the Supporting Information. At low magnification, there were no significant differences among these diverse groups from PEG to MF‐Lip40@PEG (Figure S7, Supporting Information). In contrast, some nanoparticles were observed on the surface of the MF‐Lip@PEG with different concentrations of MF‐Lip under high magnification; particularly, increasing amounts of MF‐Lip led to an increase in the number of nanoparticles on the surface of the composite hydrogels. Furthermore, the magnified SEM images also showed a successful combination between the MF‐Lip and the PEG hydrogel. Figure 1 Related properties of MF‐Lip, morphological examination of the hydrogels, and characterization of the hydrogels. a) Solution of four‐arm PEG‐thiol with MF‐Lip. b) The MF‐Lip@PEG hydrogel. c) Photograph of the injectable hydrogel. d–f) Images of a hydrogel injected into DI water. h) Magnified SEM results (liposomes are marked by yellow circles). i) Strain sweep measurements of the storage moduli ( G ' denotes the elastic modulus, and G ” denotes the loss modulus), measured in kPa, and their effect on the strain of the hydrogels (%), which was used to determine their tensile strength. j) Measure of viscosity parameters in relation to time in seconds. The strain shearing rate alternated between 0. 05% strain for 100 s and 500% strain for 50 s. k) The process of self‐healing. l) The mechanism of the self‐healing ability and injectability. m) Degradation. n) Swelling percentage. o) In vitro drug release. Before considering its drug‐carrying capabilities, the hydrogel has to be able in exhibit self‐healing properties and endure external strain. As shown in Figure 1 i, both the elastic modulus ( G ') and loss modulus ( G ”) could withstand an increase in strain of ≈50%. Furthermore, upon examination of the recoverability by a rheometer, it was observed that before applying high shear rate, the hydrogel was able to maintain a colloidal state while keeping the viscosity at ≈7. 5 × 10 5 Pas; then, when a high shear rate was applied, the hydrogel's structure was destroyed, and the viscosity of the hydrogel dropped to almost 0. 7 × 10 5 Pas. However, a few seconds after the high shear rate was removed, the hydrogel recovered to a state similar to the one prior to being subjected to the external shear rate (Figure 1 j). From a macro view, the same process was also observed in photographs (Figure 1 k), where a hole was created in the center of the hydrogel after ≈10 min, followed by recovery from the damage. The mechanism of the process is described in Figure l. During the mixing of four‐arm PEG‐thiol with AgNO 3, the covalent bonds formed between S and Ag, as well as the simultaneous formation of interactions between Ag and Ag, transformed the solution to the gel. 22, 23 When an intense external shear rate was applied, these bonds were destroyed, and when these external forces were removed, the network regenerated. Therefore, we predicted at this point that the hydrogels were capable of recovery after injection. Regarding implantation, a hydrogel should have the capability to not only maintain its integral construction during the treatment process, but also undergo degradation. As shown in Figure 1 m, MF‐Lip40@PEG showed the longest endurance period for 12 d compared to 9 d for both MF‐Lip20@PEG and MF‐Lip10@PEG, while the shortest time was 7 d for PEG. Regarding the swelling percentages, PEG exhibited the highest value (almost 1200%). In contrast, the swelling percentages in all MF‐Lip@PEG groups were noticeably lower, ranging from ≈900% for both MF‐Lip20@PEG and MF‐Lip10@PEG to just 800% for MF‐Lip40@PEG. Increasing amounts of MF‐Lip were added to the hydrogel system, leading to a denser network structure in the composite hydrogels, and resulting in lower swelling percentages and longer endurance periods of the MF‐Lip40@PEG, because the dense network could serve as a barrier to the movement of water molecules. 24, 25 MF is a hydrophobic drug and tends to aggregate upon direct blending with a solution of four‐arm PEG‐thiol and AgNO 3. Thus, MF‐PEG produced by this method showed poor drug dispersion, and the release percentage of MF from PEG‐MF was the lowest (≈50%). Almost half of the loaded MF was not released from the MF‐PEG because of severe aggregation of MF in the hydrogel matrix. In contrast, the liposomes were able to carry MF in their bilayer, leading to excellent dispersibility of the drug in the composite hydrogel networks. As a result, the cumulative release percentages of MF were significantly improved from 50% in MF‐PEG to ≈95% in the three MF‐Lip@PEG groups. Specifically, these drugs would be released in several methods in vitro. First, these drugs would be released from the bilayers with the leakage of liposome, and second, these drug‐loaded liposomes would be released from the composite hydrogel, and then these drugs could be detected in the release medium. In particular, the MF‐Lip40@PEG exhibited more favorable release characteristics than those of the other two MF‐Lip@PEG groups, because the dense network is capable of making contributions to sustain the in vitro drug release. 26, 27 The beneficial effects of MF‐Lip@PEG were investigated in human endothelial vein cells (HUVECs), and the angiogenic activity of the liposome was studied in vitro. HUVECs were treated with different concentrations of MF‐Lip@PEG leaching liquor for 7 d, and the protein expression levels of vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) were identified by western blot analysis. As shown in Figure 2, at concentrations below 40 × 10 −6 m, the MF‐Lip40@PEG could significantly increase the expression level of VEGF and bFGF in a dose‐dependent manner. Moreover, the expression level of VEGF and bFGF in the groups treated with MF‐Lip10@PEG, MF‐Lip20@PEG, and MF‐Lip40@PEG was significantly higher than those in the control group. However, the promotion effect of MF‐Lip60@PEG was dramatically decreased, similar to the results observed for MF's effect on cell viability (Figures S8 and S9, Supporting Information). Therefore, the concentrations of MF‐Lip10@PEG, MF‐Lip20@PEG, and MF‐Lip40@PEG were chosen for further studies. Figure 2 The effects of MF‐Lip@PEG on HUVEC angiogenesis ability and the protective effect of MF‐Lip@PEG on 400 × 10 −6 m CoCl 2 ‐damaged HUVECs. a) Western blotting data showing the levels of angiogenesis‐related growth factors with different MF treatments after 7 d. Western blotting data of the levels of b) VEGF and c) bFGF with different MF‐Lip@PEG treatments after 7 d. d) Western blotting data showing the levels of apoptosis pathway‐related protein expression with different treatments after 24 h. Western blotting data of the levels of e) Bax, f) Bcl‐2, and g) cleaved‐caspase 3 with different MF‐Lip@PEG treatments 24 h. h–n) Flow cytometric analysis of cell apoptosis after 24 h of treatment with 400 × 10 −6 m CoCl 2 and PEG or different concentrations of MF‐Lip@PEG. Hypoxia: cells were treated with 400 × 10 −6 m CoCl 2 to stimulate hypoxia‐induced damage and were not treated with MF‐Lip@PEG; PEG: PEG leaching liquor; 10: MF‐Lip10@PEG leaching liquor; 20: MF‐Lip20@PEG leaching liquor; 40: MF‐Lip40@PEG leaching liquor; 60: MF‐Lip60@PEG leaching liquor. * p < 0. 05. Endothelial cells are essential cells for blood vessels, and they play an important role in blood vessel formation. 28 The vascular endothelial cells can secrete a variety of vasoactive substances via autocrine and paracrine mechanisms. 29 A study reported that MF could increase endothelial cell migration in a dose‐dependent manner, indicating that MF has beneficial effects on the formation of new blood vessels. 30 Angiogenesis is very important in the early stage of healing. VEGF and bFGF are growth factors that play a central role in angiogenesis by promoting neovascularization and endothelial cell proliferation. A lack of these factors can result in abnormal vessel formation. 31, 32 In the present study, we found that following treatment with MF‐Lip@PEG leaching liquor for 7 d, the expression levels of VEGF and bFGF in HUVECs increased significantly in a dose‐dependent manner at concentrations below 40 × 10 −6 m (Figure 2 b, c). The present results indicate that the MF‐Lip@PEG leaching liquor could promote the cells' angiogenesis ability. As a result of the MF‐Lip's antiapoptotic effect, the MF‐Lip@PEG was expected to have cell protection ability under hypoxic conditions. To investigate this effect, we stimulated a hypoxic environment in vitro. A hypoxic microenvironment for HUVECs was induced with 400 × 10 −6 m CoCl 2 (Figures S10 and S11, Supporting Information). The HUVECs were further treated with PEG, MF‐Lip10@PEG, MF‐Lip20@PEG, or MF‐Lip40@PEG leaching liquor, and cell apoptosis was assessed by flow cytometry. In the hypoxia group, the cell apoptosis rate was 39. 8 ± 4. 5%. Significantly decreased cell apoptosis rates were observed in the MF‐Lip@PEG leaching liquor‐treated groups. Additionally, the apoptosis rate in the MF‐Lip10@PEG group was 17. 5 ± 2. 1%, whereas that in the MF‐Lip20@PEG group was 12. 2 ± 2. 0% and that in the MF‐Lip40@PEG group was 10. 0 ± 3. 0%. In the MF‐Lip@PEG leaching liquor groups, CoCl 2 ‐induced cell apoptosis was significantly inhibited in a dose‐dependent manner. Significantly decreased cell apoptosis was observed in the MF‐Lip@PEG‐treated groups compared to the PEG‐only group and the hypoxia group (Figure 2 h−n), indicating that MF‐Lip@PEG leaching liquor could ameliorate apoptosis in the hypoxia‐induced cells. It is worth noting that the apoptosis rate in the PEG group was 27. 3 ± 3. 2%, which was also significantly lower than that in the hypoxia group, therefore suggesting that the PEG leaching liquor can also significantly inhibit 400 × 10 −6 m CoCl 2 ‐induced cell apoptosis. It was observed that the MF‐Lip@PEG was capable of attenuating CoCl 2 ‐induced cell apoptosis in HUVECs. We next sought to determine the underlying mechanisms by examining protein expression of the apoptosis‐related proteins cleaved‐caspase‐3 and Bax/Bcl‐2. Treatment with MF‐Lip@PEG leaching liquor significantly reduced the cells' cleaved‐caspase‐3 expression levels in a dose‐dependent manner (Figures 2 d, g). Similarly, the MF‐Lip@PEG leaching liquor significantly reversed the CoCl 2‐ induced expression of Bax/Bcl‐2 (Figure 2 d−f). The Bax/Bcl‐2/caspase‐3 pathway is very important in the process of apoptosis. 33 Bax is a proapoptotic protein that is polymerized after apoptotic provocation and generates pores in the mitochondrial membrane, subsequently inducing cell apoptosis. 34 Bcl‐2 is a major antiapoptotic protein and can regulate apoptosis by binding to Bax and inhibit its apoptotic effect. 35 Therefore, the Bcl‐2/Bax ratio is a crucial factor in cell survival. Members of the caspase family are essential proteins in cell apoptosis, among which caspase‐3 is crucial effector during the apoptotic process. 36 The current study indicates that the MF‐Lip@PEG can protect hypoxia‐damaged HUVECs through regulation of the Bax/Bcl‐2/caspase‐3 pathway. Moreover, it was observed that PEG leaching liquor also protected the HUVECs from apoptosis and upregulated the Bcl‐2/Bax ratio, suggesting that the PEG leaching liquor alone exerted a protective effect on the hypoxia‐damaged HUVECs cells. To further demonstrate the persistent skin flap regeneration‐promoting capability of the hydrogel in vivo, the effect of the hydrogel on a random‐pattern skin flap rat model was investigated. An ≈1. 5 × 5 cm random skin flap was elevated on the dorsal side of the rats ( Figure 3 a), and the hydrogel was intradermally injected into the skin flap before the flap was put back in its original position (Figure 3 b, c). Different concentrations of MF directly blended in the hydrogel for the skin flap treatment were also investigated (Figure S12, Supporting Information). At 7 d after treatment, the rats were anesthetized, and the skin flap survival rates were assessed using a moorFLPI (Moor instruments, UK) to detect the real‐time flap blood flow. The contrast images obtained were color‐coded to correlate with blood flow of the flap (Figure 3 e). The average necrosis rate in the control groups was ≈38. 6 ± 1. 25% ( n = 6), while the average necrosis rate in the 40 × 10 −6 m MF‐Lip group was 23 ± 0. 1% ( n = 6) and the average necrosis rate in the PEG group was 28. 4 ± 1. 4% ( n = 6). Lower skin flap necrosis rates were observed in flaps treated with the MF‐Lip@PEG. For the MF‐Lip10@PEG group, the average necrosis rate was 25. 4 ± 2. 3% ( n = 6); for the MF‐Lip20@PEG group, the average necrosis rate was 20. 8 ± 0. 8% ( n = 6); and for the MF‐Lip40@PEG group, the average necrosis rate was 12. 8 ± 1. 2% ( n = 6). It was observed that, without loading of the MF‐Lip into the hydrogel, PEG alone could significantly increase the skin flap survival rate. Additionally, the skin flap survival rates further increased with the MF‐Lip@PEG in a dose‐dependent manner. Figure 3 The skin flap survival rates after treatment. a) A random skin flap animal model, in which the flap was 1. 5 cm (width) × 5 cm (length) on the rats' dorsal side with the pedicle at the tail end. b, c) After flap elevation, 0. 3 mL of hydrogel solution was intradermally injected into the flap skin, and the hydrogel was injected evenly into the flap at 20 dots for each flap (yellow arrow). d) Flap necrosis area percentages of different groups; e, first row): the laser speckle contrast imaging captured real‐time blood flow images of different groups. e, second row): photos of skin flaps in different groups; e, third row), H&E staining of the necrosis and survival junction area of the skin flaps in different groups. e, last row), Magnified view of the survival area in different groups. * p < 0. 05. Skin flap angiogenesis was studied to further investigate the neovascularization effect of the hydrogel on the skin flap. CD31 immunohistochemical staining was performed in order to study the skin flap neovascularization. Cross‐sections of the flap revealed that PEG and the MF‐Lip@PEG significantly increased the densities of the flap microvessels. After 7 d, massive microvessels were seen in the PEG‐treated groups ( Figure 4 a, the red dots marked the microvessels). The average microvessel densities were 20. 2 ± 12. 0 microvessels/spot in the control group; 45. 0 ± 5. 0 microvessels/spot in the 40 × 10 −6 m MF‐Lip group; 70. 6 ± 5. 0 microvessels/spot in the PEG group; 42. 0 ± 3. 2 microvessels/spot in the MF‐Lip10@PEG group; 50. 0 ± 5. 5 microvessels/spot in the MF‐Lip20@PEG group; and 55. 7 ± 4. 2 microvessels/spot in the MF‐Lip40@PEG group. The CD31 staining was significantly increased in the treatment groups, and there was no difference between the 40 × 10 −6 m MF‐Lip group and the MF‐Lip40@PEG group (Figure 4 c). Immunohistochemical staining of CD31 proved the dose‐dependent pro‐angiogenic effect of the MF‐Lip@PEG. The microvessel density in the PEG group was the highest compared to the other groups (Figure 4 ). This was probably caused by the mild local inflammation or immune response caused by PEG hydrogel delivery. Figure 4 Effect of MF‐Lip@PEG on random‐pattern skin flap neovascularization, inflammation, and infection. Immunohistochemical images of flaps highlighting a) blood vessel CD31‐positive endothelial cells (red dots mark the microvessels) and CD68‐positive macrophage/monocytes (red arrows mark the cells). b) Immunofluorescence images of flaps highlighting S. aureus (green). The c) average microvessel densities, the d) average macrophage densities and b) the average bacterial densities in different groups. * p < 0. 05. In addition to angiogenesis, local inflammation after hydrogel treatment is also vital to skin flap regeneration. Immunohistochemical staining for CD68 was performed in order to assess the effect of MF‐Lip@PEG on skin flap inflammation during the regeneration process. As shown in Figure 4, the CD68 staining in the PEG group was significantly higher than that in the MF‐Lip, MF‐Lip@PEG and control groups. The average macrophage density was 50 ± 19 cells per spot in the control group and 244 ± 10 cells per spot in the PEG group. The average macrophage density was 120 ± 9 cells per spot in the 40 × 10 −6 m MF‐Lip group. Additionally, CD68 staining decreased as the concentration of MF‐Lip@PEG increased. The average macrophage densities were 199 ± 15 cells per spot in the MF‐Lip10@PEG group; 145 ± 20 cells per spot in the MF‐Lip20@PEG group; and 107 ± 12 cells per spot in the MF‐Lip40@PEG group. The results showed that the MF‐Lip@PEG could alleviate the mild local inflammatory reaction. Hydrogels, as foreign bodies, usually give rise to the mild local inflammatory reaction. This response could cause reactive neovascularization, which is beneficial for skin flap regeneration (Figure 4 ). However, excessive and persistent inflammation is not beneficial to flap recovery. However, the above results showed that the skin flap distal end survival rate was significantly increased in the PEG group compared to the control group. This finding indicates that PEG may cause mild inflammation in the flap tissue but does not affect the skin flap survival rate. The MF‐Lip@PEG is dynamic coordinative hydrogel that is formed by four‐arm‐PEG‐SH and Ag + Ag–S coordination. During hydrogel degradation, the MF‐Lip@PEG can slowly release Ag + into the skin flap. To explore the antibacterial ability of the hydrogel, Staphylococcus aureus immunofluorescence staining was performed. As shown in Figure 4 b, compared to the control (39 ± 4 bacteria per spot), 40 × 10 −6 m MF‐Lip (20 ± 3 bacteria per spot) showed an outstanding antibacterial property. The S. aureus bacteria numbers could be further decreased by treatment with the MF‐Lip10@PEG (15 ± 2 bacteria per spot), MF‐Lip20@PEG (13 ± 3 bacteria per spot), and MF‐Lip40@PEG (8 ± 2 bacteria per spot). This result significantly demonstrates the effective antibacterial property of the MF‐Lip@PEG. Bacterial infection is another leading cause of random skin flap necrosis, as flap tissues are especially vulnerable in an early stage hypoxic environment. Ag + is a well‐known antibacterial agent and has long been used for targeted therapy of wound bacterial infections to promote wound healing. 37, 38 However, excessively high concentrations of Ag + irons in the early stage of wound healing can damage the tissue cells and induce innate cell apoptosis. The MF‐Lip@PEG in vitro release and degradation characteristics (Figure 1 ) indicate that dense network of the MF‐Lip@PEG could retard the drug release rate and ensure prolonged degradation during the endurance period. This behavior leads to constant but more favorable release concentrations of Ag + for accelerating skin flap healing. The number of S. aureus bacteria in the PEG group was 55 ± 5 bacteria per spot, higher than that in the control (39 ± 4 bacteria per spot) group and the rest of the hydrogel treatment groups. This result suggests that PEG and the other hydrogels used for the skin flap are potential infection sources, which may induce local tissue inflammation (as observed in Figure 4 a) and infections. However, the MF‐Lip@PEG could significantly reverse the flap bacterial number, which was in accordance with the reverse of the flap inflammation results observed upon treatment with the MF‐Lip@PEG (Figure 4 a). From these results, it was observed that MF‐Lip@PEG possessed an outstanding antibacterial ability. In summary, we have developed a hydrogel consisting of multifunctional four‐arm‐PEG crosslinked with Ag + and loaded with MF‐Lip. The MF‐Lip@PEG was injectable and had self‐healing properties, low swelling percentages, a long endurance period, and excellent drug dispersibility and release characteristics for prolonged drug delivery. In addition, the hydrogel possessed the distinctive functions of the combined effects of PEG, Ag + and MF‐Lip. The in vitro study results demonstrated that the MF‐Lip@PEG exhibited an outstanding proangiogenic effect and a strong protective effect on cells against hypoxia‐induced apoptosis through regulation of the Bax/Bcl‐2/caspase‐3 pathway. The in vivo study demonstrated that the MF‐Lip@PEG promoted skin flap angiogenesis, decreased skin flap inflammation, reduced skin flap infection, and significantly enhanced the skin flap survival rate. Overall, the injectable multifunctional MF‐Lip@PEG holds great promise for effectively matching the renewal of tissue regeneration. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201801664
| 2,019
|
Advanced Science
|
Self‐Healing Hydrogels: The Next Paradigm Shift in Tissue Engineering?
|
Abstract Given their durability and long‐term stability, self‐healable hydrogels have, in the past few years, emerged as promising replacements for the many brittle hydrogels currently being used in preclinical or clinical trials. To this end, the incompatibility between hydrogel toughness and rapid self‐healing remains unaddressed, and therefore most of the self‐healable hydrogels still face serious challenges within the dynamic and mechanically demanding environment of human organs/tissues. Furthermore, depending on the target tissue, the self‐healing hydrogels must comply with a wide range of properties including electrical, biological, and mechanical. Notably, the incorporation of nanomaterials into double‐network hydrogels is showing great promise as a feasible way to generate self‐healable hydrogels with the above‐mentioned attributes. Here, the recent progress in the development of multifunctional and self‐healable hydrogels for various tissue engineering applications is discussed in detail. Their potential applications within the rapidly expanding areas of bioelectronic hydrogels, cyborganics, and soft robotics are further highlighted.
|
1 Introduction In recent years, tissue engineering has emerged as a promising technology to grow organs from scratch, 1, 2, 3 replicate biological mechanisms of various diseases, 4, 5, 6, 7 address tissue‐related ailments 8, 9, 10, 11, 12 and enable life extension in the growing aging population. 13, 14 So far, most of the tissue engineering approaches has relied on the encapsulation of stem cells within native‐like and highly porous biomaterials; 15, 16, 17, 18, 19, 20, 21, 22 or scaffolds as the tissue engineers prefer to say. The scaffold‐based biomaterials enable encapsulated cells to spread and reorganize into tissue‐like architectures, while permitting sufficient nutrient and waste material exchange with the surrounding environment. Of the many scaffolding biomaterials currently utilized for tissue engineering applications, hydrogels are among the most promising ones. Hydrogels are composed of polymeric networks that are capable of absorbing and retaining high amount of water. 19, 23 Hydrogels are also tunable (both physically and chemically), are injectable, and have been used over the years for tissue engineering and various drug delivery applications. 24, 25, 26, 27, 28 However, as one of the fascinating properties of natural tissues is their ability to self‐heal after minor injuries, to truly recapitulate the physical properties of native tissues, such human‐made biomaterials also need to spontaneously heal and regenerate injuries inflicted on them. This inherent ability of native tissues to regenerate on demand has initiated enormous motivation to develop intelligent hydrogels with similar self‐repair mechanisms. In spite of their many similarities to the extracellular matrix (ECM) of the native tissues, self‐healable hydrogels still face several shortcomings, which limits their specific application for replacement of electrically active and elastic tissues ( Figure 1 ). 29, 30, 31, 32 For example, current self‐healable hydrogels are typically nonconductive and exhibit significantly lower fracture energies (<10 J m −2 ) 33 than that of cartilage, 33 skin, 34 tendon, 35 and muscle tissues 36 (kJ m −2 regime). Conventional hydrogels implanted within the load‐bearing and dynamic environments of the human body are thus inclined to acquire some minor defects. These microcracks gradually propagate and grow in size and will ultimately lead to failure of the material if they are not repaired in due time. Moreover, in case of cell‐encapsulated hydrogels, cells are prone to rapid migration and interface pulling, and will eventually disrupt the structural integrity of the hydrogel matrix due to traction forces. Therefore, to achieve optimal implant lifetime, it will be necessary to engineer mechanically tough hydrogels with the ability to quickly remedy material defects. 37, 38, 39, 40, 41, 42, 43 Figure 1 Human organs are made of elastic, tough, and electroactive tissues, which can spontaneously heal. The microenvironment within most tissues is also highly dynamic and load‐bearing. Tissue engineering hydrogels, therefore, need to heal on command and incorporate the same mechanical and electrical properties as those found in natural tissues. Injectability is also a sought‐after property, as injectable hydrogels can be used to deliver stem cells to the target tissue in a minimally invasive manner. Although the literature on self‐healing hydrogels is growing fast, only a few practical applications for these biomaterials exist in tissue engineering; this is because most self‐healable hydrogels do not match with the above‐described electromechanical milieu of the body (Figure 1 ). Moreover, the long‐standing incompatibility between hydrogel toughness and rapid self‐repair has not yet been fully addressed. To address this unmet need, nanomaterials are rapidly emerging as an exciting approach to develop self‐healable and multifunctional hydrogels through one‐step strategies that are based on simple mixing procedures ( Figure 2 ). Figure 2 Nanoreinforcement can be used to generate multifunctional hydrogels that blend in with electrically and mechanically active tissues. With the right combination of nanoreinforcer and hydrogel polymer, it is possible to manufacture mechanically tough, electrically conductive, and bioactive tissue engineering systems for organ regeneration. In this review article, we discuss the unexplored and enormous possibilities of self‐healable hydrogels in tissue/organ regeneration and repair. We will mainly focus on the synergy between hydrogel toughness and dynamic self‐repair, as we believe this unique combination will ultimately induce a paradigm shift in the field of hydrogel‐based tissue engineering. Additionally, we will also highlight the emerging area of self‐healable hydrogels made through nanoreinforcement and review the many recent impressive applications of these systems in tissue engineering. Finally, we will review the application of self‐healable hydrogels in cyborganics, soft robotics, and bioelectronics, since these fields will rise in the coming decades and define an entire new frontier in health sciences. 2 Self‐Healing mechanisms Self‐healable hydrogels rely on one common principle involving a so‐called mobile phase, which enables crack closure through a combination of mass transfer and reconnection of broken links within the hydrogel matrix. The reconnection within this matrix is typically mediated by either noncovalent or covalent bonds ( Figure 3 ). The noncovalent interactions are based on weak sacrificial links such as ionic, 44 hydrogen, 45 or hydrophobic bonds, 46 while the chemical bonds are based on dynamic covalent bonds and metal coordination bonds. 47, 48, 49, 50, 51, 52 The combination of the above‐mentioned covalent and noncovalent bonds has also recently been used to generate mechanically tough and elastic double‐network hydrogels with rapid self‐repair properties. 53, 54, 55 In this section, we will highlight the key mechanisms behind the repairing properties of such self‐healable hydrogels (Figure 3 ). Specifically, we have divided this section into three independent subsections, the first two dealing with self‐healing mechanisms based on either noncovalent or covalent bonds, and the last section on the working principle behind double‐network hydrogels. Figure 3 The various chemical and noncovalent interactions behind self‐healable hydrogels are highlighted here. 2. 1 Noncovalent Interactions This section focuses on self‐healable hydrogels made from noncovalent crosslinks including hydrogen bonds, 53, 56, 57, 58, 59, 60 host–guest interactions, 61, 62, 63 ionic bonds, 64 and hydrophobic interactions. 65, 66, 67 Although the listed interactions can generate hydrogels that display rapid healing time and good self‐healing efficiency, these attributes come at the cost of inelasticity and mechanical weakness. In the following subsections, we describe the pros and cons of each of these interaction schemes from a mechanical‐point‐of‐view and provide possible solutions that could remedy their current shortcomings. 2. 2 Hydrogen Bonds Hydrogen bonds are among the most common noncovalent interactions found in nature. For instance, it is hydrogen bonds that keep complementary strands of DNA together in their unique helix structure, 68 and they are also a key player in water's many solvent properties (Figure 3 ). 69 They also play a determining role in the secondary and tertiary structure of folded proteins 70 and the properties of various solid polymers such as cellulose and wool. 71, 72 Hydrogen bonds occur when the positive hydrogen atom establishes an electrostatic link with electronegative acceptor atoms such as oxygen, nitrogen, or fluoride. 73 The strength of the hydrogen bonds depends mostly on the negative charge of the acceptor atom and can thus vary greatly depending on the atom in question and the pH value of the solvent in which the bonding occurs. Examples of hydrogen bonds are those formed between two hydroxyl groups (OH···OH), between carboxyl and amide groups (NH···O=C), and between hydrogen and fluorine (H···F). The bonding strength of hydrogen bonds varies greatly, from 0. 25 to 15 kcal mol −1, with the weakest being those formed with fluorine and the strongest those that incorporate hydroxyl or amide groups. 73 Even though a hydrogen bond is several times weaker than ionic (100–1000 kcal mol −1 ) and covalent bonds (35–240 kcal mol −1 ), 74 they still contribute significantly to the mechanical properties of hydrogels, when a lot of them are present within the hydrogel matrix. Moreover, the association and dissociation of hydrogen bonds occur rapidly; typically on sub‐picosecond to picosecond time scales. 75, 76, 77 This property is a governing factor behind the rapid healing time seen in repairable hydrogels based on hydrogen bonds. For these reasons, hydrogen bonds have been exploited in the manufacture of numerous self‐healable hydrogels, the most prominent of these being those using poly(vinyl alcohol), 53, 56 ureidopyrimidinone (UPy), 57, 58, 59, 60 and polyacrylamide (PAM); 53, 78 with the hydrogen bonds being mediated by either hydroxyl or amide groups. However, due to their weakness, hydrogen bonds need to be combined with other stronger bonds to yield self‐healable hydrogels with the desired combination of mechanical strength and fast healing time. We will return to cover this important point in Section 2. 4. , when discussing how double‐network hydrogels work and what they do. 2. 2. 1 Hydrophobic Interactions Hydrophobic interactions are perhaps as important as hydrogen bonds when it comes down to protein folding, the properties of solid polymers, and the interaction between molecules in different solvents. 79 However, hydrophobic interactions are slightly stronger than hydrogen bonds and easier to control; as they can be fine‐tuned by varying the shape of the hydrophobes and the number of hydrophobic moieties on them. Hydrophobic interactions occur because of the formation of a clathrate cage around the hydrophobe, which is an ice‐like cage structure of water molecules formed through hydrogen bonds among the water molecules (Figure 3 ). 80 This highly ordered arrangement leads to entropy decrease and constitutes an unstable configuration, which easily breaks when two hydrophobes come close enough to one another to enable the formation of hydrophobe–hydrophobe assemblies, and the subsequent release of trapped water between the two. The highly adhesive force between two hydrophobes is thus caused by a physically driven increase in entropy, which is also a governing factor behind the highly reversible nature of hydrophobic interactions. 79 Some different hydrophobic schemes based on 1) host–guest interactions, 61, 62, 63 2) micelles, 81, 82 and 3) hydrophobic moieties 65, 66, 67 have been used over the years to generate self‐healable hydrogels. Self‐healable micelle‐based hydrogels are generated by incorporating amphiphilic polymers and surfactants into the hydrogel. The self‐healing mechanisms of such hydrogels are attributed to the cyclic dissociation and reassociation of the micelles. The self‐repair mechanisms underlying hydrogels that incorporate hydrophobic moieties, on the other hand, arise from reversible interactions between such moieties. Host–guest interactions are more complicated and, in most cases, are ruled by the conjugation of cyclodextrin—a molecule that consists of a lipophilic inner cavity and a hydrophilic outer surface—onto the hydrogel backbone. 61 To this end, cyclodextrin can enable a so‐called host–guest interaction with hydrophobic guest molecules, as they can become restrained within its lipophilic inner cavity due to hydrophobic interactions. One important concern of the above‐mentioned self‐healable hydrogels is the possible low water‐uptake caused by the presence of hydrophobic regions within the hydrogel matrix. However, numerous studies have shown that this does not need to be the case when hydrophilic regions outbalance the number of hydrophobic moieties in the hydrogel matrix. 81, 83, 84 2. 2. 2 Ionic Bonds As an alternative strategy, ionic bonds can also be used to develop self‐healable hydrogels via reversible electrostatic interactions between oppositely charged moieties. Such interactions can happen between oppositely charged polymers or through ionic bridges between same charged polymers mediated by oppositely charged ions. 85, 86, 87 A common example of the later is alginate hydrogels made from negatively charged alginate pre‐polymers that crosslink into a hydrogel through divalent ions; typically calcium. 88, 89 A less common but yet highly promising approach is to use electrostatic interactions between charged nanomaterials and the hydrogel backbone, as this approach can add a range of additional properties to the system in question because of the multifunctional attributes of nanomaterials. 8, 90, 91, 92 For instance, nanomaterials are known for their ability to dissipate energy within hydrogels, and can, therefore, increase the toughness and durability of hydrogels significantly. 93 Although ionically bonded hydrogels are simple, as they are manufactured by one‐step‐mixing procedures, their inelasticity and brittleness are major drawbacks. One avenue to overcome the shortcomings of ionically crosslinked hydrogels is blending them with a covalently cross‐linkable polymer to yield a double‐network hydrogel consisting of interpenetrating and adaptable polymeric networks. 94, 95 2. 3 Covalent Interactions In addition to the noncovalent self‐healing mechanisms discussed in the previous section, covalent binding schemes in the form of chelation and dynamic covalent bonds can also be used to generate self‐healable hydrogels. In this section, we will focus on the chemistry of these with special focus on dynamic covalent bonds made from imine, disulfide, boronate ester, acylhydrazone, and Diels–Alder reactions (Figure 3 ). 2. 3. 1 Chelation Chelation can be described as a number of coordinate bonds between ligands (organic molecules) and one positively charged transition‐metal ion. 96 Notably, the transition metal ion is surrounded by the ligands to yield highly complex lattice structures. Each ligand can donate electrons to the metal ion; two or more electrons from each ligand are typically donated to the metal. The bonding is thus essentially a covalent bond involving two electrons from the same atom instead of one electron from each atom, which is the typical case in standard covalent bonds. However, due to the lattice structure of these metal complexes and the many donor atoms involved, the binding energy of the complexes is typically stronger than covalent bonds. 97, 98 But what makes chelation unique in comparison to covalent bonds is the fact that they can display high adhesivity, elasticity, and reversibility at the same time. For this reason, chelation can be used to yield highly adhesive, elastic, and self‐healable materials. A prime example of this can be found in nature in the form of the sticky feet of mussels, which in recent years have been linked to chelation between Fe +3 and catechol ligands. 99 Indeed, the binding strength of catechol‐Fe +3 can reach 33 kcal mol −1 and catechol can form reversible bonds with titanium interfaces with a bond strength of 800 pN, which is almost 40% of the bonding strength between silicon and carbon. 100 This gives an idea of some of the incredible features of chelation complexes, and why they have constituted an integral component in many self‐healable hydrogels. 48, 101, 102 2. 3. 2 Dynamic Covalent Bonds Dynamic covalent bonds are unique because of their ability to reconnect without physical stimuli, which is in contrast to standard covalent bonds. Dynamic covalent bonds, therefore, combine the stability of covalent bonds and reversibility of noncovalent interactions into a potent healing force that works entirely on its own. The general principle behind such dynamic bonds is the presence of an equilibrium phase between various fates from the same reaction process. Under certain conditions, one of the fates is more stable and dominates over the others. However, a return to the original compounds followed by a reversion to another outcome is still possible, which makes such reactions highly dynamic. 103, 104 As such, these are highly sought out in the field of chemistry, but their numbers are relatively few compared to conventional covalent reactions. Nevertheless, various self‐healable hydrogels have been produced over the years through dynamic covalent crosslinks, with imine bonds being the most widely used. Thus, we will first focus on the chemistry of these bonds and their usefulness for self‐healing. The famous German chemist, Hugo Schiff, discovered imine bonds in 1864 and imine‐based compounds are therefore also commonly referred to as “Schiff's bases. ” 105 An imine bond essentially involves a reaction between an aldehyde and a primary amine with the generation of a water molecule; and is considered a strong covalent bond (150 kcal mol −1 ) 106 that can occur both at neutral and acidic pH values. In this reaction, the amine nitrogen (a nucleophile) attacks the electrophile carbonyl atom in the aldehyde to yield a double nitrogen–carbon bond. If the water molecule is not removed, this reaction can still go back through hydrolysis and, therefore, under certain conditions, a dynamic equilibrium is possible. Since imine bonds involve an amine group, they are frequently used to turn amino‐rich polymers, such as chitosan and polyacrylamide, into self‐healable hydrogels by combining them with other aldehyde‐functionalized polymers, such as oxidized alginate and hyaluronic acid (HA), as described in Section 3. 50, 107 Acylhydrazone bonds are very close relatives of imine bonds, as they are synthesized by reacting a hydrazine with an aldehyde group; 108, 109 typically through a condensation reaction. 50, 110 Acylhydrazone bonds can be spontaneously formed under physiological conditions; albeit at a significantly slower rate than under acidic environments. 111, 112 However, recent studies have shown that this bonding scheme can be used to yield self‐healable hydrogels with a crosslinking time that makes them amenable as injectable hydrogel carriers for stem cells. 50, 113 In one of these studies, a self‐healable polyethylene glycol (PEG) hydrogel was developed through a condensation reaction between two PEG macromers; one functionalized with benzaldehyde and the other with an aldehyde. This system could self‐assemble rapidly under physiological conditions and provided a viable environment for encapsulated muscle cells. 113 Another dynamic covalent bond type used in self‐healable hydrogels, although a bit weaker than imine bonds (50 kcal mol −1 ), 106 is disulfide bonds. They are essentially based on thiol/disulfide dynamic exchange reactions, in which the thiol groups needs to be oxidized. 114, 115 The reaction is, therefore, highly sensitive to the pH value and needs the involvement of an oxidation agent, which can make some of the manufacturing protocols for such hydrogels cytotoxic for cells. For these reasons, thiol‐based hydrogels with self‐healing capacity are not stable in physiological tissues due to the presence of reducing agents such as glutathione, which is found in most tissues in the body. 116, 117, 118 These are, therefore, not the preferred choice for the design and development of self‐healable hydrogels. The combination of diols and boronic acid can also yield reversible covalent links in the form of boronate esters. 52, 119 However, the stability of this reaction is highly pH sensitive and the resultant self‐healing efficiency and mechanical properties of such systems are therefore sensitive to pH changes. Indeed, the formation of diol‐boronic acid links only happens at pH values greater than or equal to the pKa value of boronic acid; which is typically greater than 8 pKa. 120, 121 From a tissue engineer's point‐of‐view, this can be a disadvantage, as most tissues in the human body operate at neutral pH and the fact that cells perish at pH values above 8. However, in a recent study, it was shown that the combination of PEG–phenylboronic acid macromonomers with PEG–diol macromonomers could yield injectable hydrogels with sufficient self‐healing efficiency and mechanical properties at neutral pH. 122 In another recent study, a self‐healing hydrogel was formed using intramolecular interactions between 2‐acrylamidophenylboronic acid (2APBA) moieties. This hydrogel could self‐heal at both neutral and acidic pH values and is thus suitable for tissue engineering applications. 51 These studies, taken together, have demonstrated that, despite the highly pH‐sensitive nature of boronate esters, they can be used to develop self‐healable hydrogels that are compatible with tissues and cells. In addition to the above‐mentioned reactions, Diels–Alder reactions have also been rapidly adopted by scientists in the field to yield self‐healable hydrogels. Diels–Alder reactions are considered click reactions, which are widely recognized for their ability to yield outstanding reaction specificity through simple synthesis procedures typically done in water and with no offensive byproducts. 123 In simple terms, a Diels–Alder reaction is a reaction between a conjugated diene and a dienophile, typically an alkene or alkyne. 124 Such reactions are essentially electrocyclic reactions that involve π electrons from the HOMO and LUMO molecular orbitals of diene and the dienophile, respectively. To this end, to enable the reaction to proceed optimally, it is important that there is an energy‐band overlap between the two molecular orbitals. 124 Diels–Alder reactions are also thermoreversible, meaning that they break at elevated temperatures—typically above 100 °C—and can reform again once the temperature is lowered. This process is cyclic and, therefore, enables the manufacture of materials that can break and heal indefinitely by applying the appropriate thermal healing procedure. 125, 126, 127 This is interesting but, unfortunately, also implies that the self‐healing potential of conventional Diels–Alder bonds is not fit for use in tissue engineering. However, some recent studies have shown that it is possible to generate Diels–Alder‐based self‐healable hydrogels that can solidify and self‐heal under physiological conditions. 128, 129 The family of Diels–Alder reactions, therefore, has promise for delivering injectable and self‐healable hydrogels that can crosslink autonomously within the target tissue, if prepared correctly. 2. 4 Double‐Network Hydrogels An on‐going challenge in the design and development of hydrogels is the conflict between strength, toughness, and high water content (>90 wt%). Given that strong materials tend to be brittle, softer materials usually tougher, and highly hydrated materials weak, this challenge is difficult to overcome. 130, 131 Another challenge in the field is to combine fast self‐healing kinetics with hydrogel strength, as mechanically strong hydrogels are typically achieved by increasing the polymer concentration and the number of crosslinks within the hydrogel matrix. The latter, unfortunately, reduces the mass transfer into the crack site and, therefore, significantly increases the healing time. Double‐network hydrogels address all of these challenges by combining a strong and rigid network with a much weaker network that is typically made from reversible crosslinks ( Figure 4 ). 94, 95, 132, 133, 134 Hydrogels made in this manner have demonstrated some amazing properties, including high elastic modulus (0. 1–1. 0 MPa), strength (1–10 MPa), stretchability (1000–2000%), and toughness (100–10 000 J m −2 ). 94, 132 Notably, double‐network hydrogels can be used to develop mechanically strong hydrogels with rapid self‐healing properties using much lower polymer concentrations than their single‐network counterparts. 63, 102 Indeed, the mechanical properties that double‐network hydrogels display are highly sought‐out attributes in the field, as they resemble those of conventional rubbers (1000 J m −2 ) and load‐bearing cartilage tissues (100–9000 J m −2 ); 94, 132 while being tougher than commercially available hydrogels, which are notoriously weak and brittle (0. 1–10 J m −2 ) and, therefore, unable to resist the cyclic forces that tissues such as muscle, bone, heart, and cartilage endure during daily routine activities. 94, 132 Figure 4 A brief depiction of the concept behind double‐network hydrogels and what they can do. The working principle behind double‐network hydrogels is essentially based on a combination of weak reversible bonds and strong irreversible bonds. Together these can yield tough and highly extendible hydrogels due to the dissipation mechanism embedded in the reversible bonds. a) A prime example of systems like this was given in a recent publication in nature letter that was based on b) ionically bonded alginate (reversible) and covalently bonded polyacrylamide (irreversible) polymers. c) This hydrogel could stretch up to 21‐times its original length and d) exhibited a toughness value around 10 000 J m −2. Reproduced with permission. 94 Copyright 2012, Macmillan Publishers Ltd. The toughening mechanism of double‐network hydrogels stems from the presence of strong irreversible and weak reversible bonds within the hydrogel matrix. The weak bonds can reform again and thus limits the amount of stress accumulation within the matrix, as these bonds can undergo many destruction–reconnection cycles. In simple terms, one can therefore attribute the amazing mechanical properties of double‐network hydrogels to a rigid skeleton that keeps the system intact, and sacrificial bonds that enable energy dissipation, delaying the onset of critical stress accumulation. 94, 132, 133, 134 This mechanism is similar to the mechanism used by the body to toughen bone, 135 and by engineers to generate structural materials that are both strong and tough at the same time. 130 In a recent study, a double‐network hydrogel made from a covalently crosslinked polyacrylamide and ionically crosslinked alginate network displayed some interesting properties, as this system could stretch up to 2000% of its original length with a toughness value that could reach 9000 J m −2 (Figure 4 ). 94 The authors speculated that the reversible ionic bonds enabled an energy dissipation scheme that could keep going as the hydrogel stretched, leading to the reported combination of high toughness, strength, and elasticity. The working principle behind double‐network hydrogels therefore presents an exciting avenue to remedy the inherent mechanical weakness of hydrogen‐bonded self‐healable hydrogels; 53, 136, 137 and as we will see in Sections 3 and 4, this combination can generate self‐healable hydrogels that could transform the world of tissue engineering if they can be translated into the clinic. 2. 5 Outlook and Future Opportunities In conclusion, a wide selection of reversible bonds is currently available for the design and development of self‐healable hydrogels. However, neither the physical nor the chemical bonds can, on their own, address the many requirements for self‐healable tissue engineering hydrogels. These requirements include high toughness, mechanical strength, high water content, injectability, elasticity, and fast self‐healing kinetics. Nevertheless, the combination of physical and chemical bonds in the form of double‐network hydrogels can address this current lack‐of‐methodology in the field, as evident from recent publications in Nature Letters 94 and Nature Materials. 134 In the authors' opinion, this area of research is still ripe for investigation, and a successful outcome to this end could yield the next gold standard in implantable materials for tissue engineering applications. One avenue that might open the field even further is the development of a double‐network hydrogel with well‐defined matrix architecture. Indeed, some recent studies have shown that by aligning polymers 138, 139 or nanomaterials 140, 141 within hydrogels, it is possible to significantly improve mechanical integrity without increasing the polymer concentration or the crosslinking density. We, therefore, anticipate that this unique combination can result in self‐healable hydrogels with mechanical properties that might match that of skeletal tissues, without compromising the self‐healing or water retention properties. On a more fundamental level, we envision that other dynamic covalent bonds could be used to further improve the state‐of‐the‐art in the emerging area of self‐healable tissue engineering hydrogels. For instance, dynamic covalent bonds, such as those based on amide and ester exchange reactions, 142 could—together with weak hydrogen bonds—yield some exciting double‐network hydrogels for the field. 3 Polymers for Self‐Healing Hydrogels One of the most important components in hydrogels, are polymers, that can be united into water friendly 3D environments for cells to attach, grow, and differentiate within. 143 The logical way to obtain self‐healing hydrogels is, therefore, by altering these essential building blocks. This is typically accomplished by incorporating the aforementioned self‐healing mechanisms into the polymeric hydrogel backbone through various nontoxic and nonhazardous chemical modifications. The healing capacity of polymeric hydrogels is typically accessed by monitoring the rejoining of two broken pieces by either optical or scanning electron microscopy (SEM). Another more precise quantification of the healing capacity includes mechanical characterization of the ultimate tensile strength and Young's modulus of the hydrogels before and after the healing process. The healing efficiency is then defined as the ratio of the modulus or tensile strength of the healed and unbroken hydrogels. Finally, it is also possible to use rheology to gain a detailed picture of both the healing efficiency and healing time by monitoring the real‐time changes in storage and loss modulus during the healing process. Self‐healing hydrogels have so far been made from either natural or synthetic polymers ( Figure 5 ). The natural polymers are for the most part polysaccharide‐based polymers, such as alginate, chitosan, and HA. The synthetic polymer systems, on the other hand, are based on polymers such as polyethylene glycol, poly(acrylic acid), poly(vinyl alcohol), and polyacrylamide. These polymers provide certain advantages and disadvantages and abandoning one in favor of another typically involves some trade‐offs. For example, the natural polymers are typically more biocompatible, as polymers such as gelatin and chitosan are famed for their cell attachment properties, 144, 145, 146, 147, 148 while poly(vinyl alcohol) and polyacrylamide typically yield stronger and more elastic hydrogels at the cost of reduced biofriendliness. In the following subsections, we will review these polymers as potential candidates for self‐healing hydrogels in the field of tissue engineering. Figure 5 The various synthetic and natural polymers used to generate self‐healable hydrogels are highlighted here. 3. 1 Natural Polymers Over the centuries, natural materials have been a great source of inspiration for materials scientists and physicians; noteworthy examples in this regard are silk‐based sutures 149 and cellulose‐based building materials. 150, 151 In recent years, biomedical engineers have also started to tap into natural sources to develop even better biomaterials. Because one of the prime‐ingredients in native tissues is HA—a polysaccharide—the focus has for the most part been directed toward polysaccharides, such as alginate, pectin, and chitosan, which can be procured from natural sources. These naturally derived polymers are cheaper than HA and less immunogenic than, for instance, gelatin. The focus of this section is, therefore, directed toward alginate and chitosan‐based self‐healable hydrogels, however, we will also briefly highlight some of the emerging trends in self‐healable hydrogels made from gelatin and HA. 3. 1. 1 Alginate Alginates are naturally occurring polysaccharides typically retrieved from marine algae and brown seaweed. 152 Because of their good biocompatibility, exceptional water retention, and tunable gelation properties, alginates have been extensively studied as soft scaffolds in various tissue engineering applications, and as microcapsules for delivery of drugs. 153, 154, 155, 156, 157 However, alginate‐based hydrogels are brittle and mechanically unstable, which significantly limits their use in many biomaterial applications. 89 One approach to remedy this is by making self‐healing alginate hydrogels with the capacity to spontaneously self‐repair in the event of mechanical damages. Recently, a variety of methods have been applied to generate self‐healing alginate hydrogels; something that has enticed much attention in the field of biomaterial science. 50, 158, 159 These methods include various chemical methods to incorporate dynamical covalent bonds into the polymeric backbone of alginate 50, 159 and through the concept of host–guest interactions. 158 For instance, in a recent study, the noncovalent interaction between a host [β‐cyclodextrin grafted alginate (alg‐ g ‐CD)] and a guest (Pluronic F108) was shown to facilitate a fast healing performance alongside desirable thermoresponsive gelling properties and negligible cytotoxicity. 158 As alginates contain functional hydroxyl groups, which are transformable into aldehyde groups, dynamical imine bonds arising from the interaction between aldehyde groups and amines have also become a central theme in the development of self‐healable alginate hydrogels. 50, 159 As an example of this feasibility, Wei et al. used chitosan (amine rich) and oxidized alginate (aldehyde rich) to yield a mechanically robust hydrogel with excellent self‐healing ability (up to 95% healing efficiency) and good cytocompatibility, as tested through encapsulation with 3T3 fibroblast cells. 50 The hydrogel developed in this study was an engineering masterpiece as it achieved the rare union between high self‐healing efficiency and high mechanical strength, while still keeping the hydrogel biocompatibility intact. Such hydrogels will assist scientists to carve new avenues in the field of tissue engineering, which we anticipate will enable the manufacture of scaffolds with the ability to integrate with native tissue over sustainable periods. Along the same line, a hydrogel composed of dopamine‐grafted oxidized sodium alginate (OSA‐DA) and PAM, showed efficient self‐healing ability (80% mechanical recovery in 6 h), high tensile strength (0. 109 MPa), and ultrastretchability (2550%). Remarkably, due to plenty of catechol groups on the OSA‐DA chains, the hydrogel possessed unique cell affinity (NIH‐3T3 fibroblasts) and tissue adhesiveness. Furthermore, the in vivo rat experiment showed that this hydrogel could promote tissue regeneration and accelerate the process of wound healing. 160 Quite often multicomponents are utilized to construct hydrogels with self‐healing ability and the resulting hydrogel systems often exhibit batch‐to‐batch dependent inconsistent results, problems in multicomponent mixings, and involve unpredictable cross‐talk between the added components. To this end, Hong et al. , developed a “single polymeric component, ” based on alginate‐boronic acid (alginate‐BA) to overcome the aforementioned problems. 161 This stretchable hydrogel showed efficient self‐healing performance (up to 98% healing efficiency), owing to reversible inter‐ and intramolecular interactions. In addition, subcutaneous implantation of this hydrogel under mice skin showed inflammatory response at day 3 which mostly disappeared on day 7, indicating the low toxicity of the hydrogel. 3. 1. 2 Chitosan Chitosan is a positively charged and amino‐rich polysaccharide typically derived from the exoskeleton of shellfish and possesses similar structural characteristics as glycosaminoglycan (GAG); one of the major components of the ECM. 162, 163 Due to its similarities with the native ECM, chitosan is also biocompatible, biodegradable, and hydrophilic, which makes it an ideal 3D microenvironment for cell encapsulation. For these reasons, chitosan has frequently been used as a scaffolding material for tissue engineering 144, 164, 165 and various cell delivery therapies. 166, 167, 168 As chitosan is rich in amino groups, the ideal pathway for generating self‐healing hydrogels with chitosan is through dynamic covalent imine/enamine bonds. 43, 50, 107, 169, 170 To this end, a recent study by Huang et al. used an enamine bonding between carboxymethyl chitosan (CMC) and an aldehyde functionalized polyethylene glycol polymer (PEG‐BA) to develop a self‐healing hydrogel ( Figure 6 ). 43 The two polymers formed a stable hydrogel film after 5 min with a good storage modulus (3. 2 kPa) and a healing efficiency ranging from 80% to 94% at physiological temperature (37 °C), depending on the healing time (6–12 h). Encapsulation studies with fibroblast cells demonstrated excellent cell viability for up to 7 d, indicative of chitosan's potential as a regenerative milieu for tissue engineering applications. In a similar vein, chitosan and cellulose acetoacetate were mixed to endow a hydrogel with good self‐healing ability on account of dynamic enamine bonding between aldehyde groups of the functionalized cellulose and amino groups of chitosan. 169 Specifically, the broken gels were able to heal after 40 min at 37 °C in physiological conditions. Despite increasing usage of chitosan in self‐healing hydrogels, low water solubility of this polymer at physiological pH can be problematic for its application in tissue engineering. To this end, Khan et al. synthesized a water‐soluble derivative of chitosan by grafting l ‐glutamic acid onto its backbone (chit‐glu). Subsequent mixing with benzaldehyde‐terminated 4‐arm poly(ethylene glycol) (PEG‐BA) led to fast formation (<60 s) of a gel, based on imine bonds. The prepared hydrogels were shown to be injectable and self‐healing, and the healed hydrogels fully recovered their initial elastic modulus in a compression test. Most remarkably, human fibroblast cell lines (WI‐38) cultured on top of the hydrogel were viable and capable of proliferating. 171 Figure 6 A mechanically strong and self‐healable carboxymethyl–chitosan (CMC) hydrogel based on Shiff‐base bonds. a) The gel formation is shown here along with b) the chemical mechanism behind its self‐healing properties. c, d) The self‐healing properties of the CMC hydrogel. e) The mechanical properties of the as‐prepared hydrogel and those that were allowed to heal for 6 and 12 h. f) The healing efficiencies (the compressive load ratio of as‐prepared and healed samples) at 25 °C and 37 °C are shown here; there were no significant differences between bars with the same letters. Adapted with permission. 43 Copyright 2016, Wiley‐VCH. Noncovalent interactions based on hydrogen bonds 172, 173 and hydrophobic forces 174 have also been utilized to fabricate self‐healable chitosan hydrogels. For instance, a pH‐sensitive hydrogel made from hydrogen‐bonded chitosan (CS) and polyvinyl alcohol (PVA) was capable of completely healing itself after 1 h and showed good cytocompatibility with HeLa cells. 175 In a similar study, a multilayered polyelectrolyte composite film comprised of CS and polyacrylic acid (PAA) was demonstrated to pose excellent self‐healing ability at low PH values. 172 In another study, hydrophobic interactions were used to develop a chitosan‐based self‐healable hydrogel. 176 Specifically, in this study, chitosan was modified with the hydrophobic compound—ferrocene—to enable self‐healing through hydrophobic interactions between adjacent ferrocene sites. This hydrogel was also highly stimuli‐responsive and could be triggered to deliver small molecules and biologics by changing the pH value. In the authors' opinion, stimuli‐responsive and self‐healable hydrogels are ideal for on‐demand delivery of tissue growth factors and stem cells to injury sites in the body and could potentially change the course of the field of tissue engineering. 3. 1. 3 Hyaluronic Acid HA is a nonsulfated glycosaminoglycan that is found in ECM of vertebrate tissues, 177 and is a suitable tissue engineering polymer due to its ability to bind cell surface receptors such as CD44, as well as its involvement in regulating cell differentiation and proliferation. 178 In addition, the presence of reactive carboxylic groups on HA enables various chemical functionalization schemes for further downstream applications. 179 Nonetheless, similar to other hydrogels made from naturally derived polymers, HA‐based hydrogels are susceptible to mechanical disruption, and accordingly, a variety of methods have been applied to develop self‐healable and durable HA hydrogels. These methods include the use of chemical bonds such as Diels–Alder and acylhydrazone bonds 180 or noncovalent interactions such as guest–host interactions 38, 181, 182, 183, 184 and hydrogen bonds. 45 For instance, a recent study has used a combination of Diels–Alder and acylhydrazone bonds to make a self‐healing hydrogel (healing after 3 h) with good mechanical property (storage modulus of 18 kPa) from a mixture of oxidized furan–HA, furan–adipic dihydrazide‐functionalized HA, and dimaleimide–PEG. 180 Specifically, Diels–Alder click chemistry between furan and maleimide was utilized to give the hydrogel matrix its mechanical integrity, while the self‐healing mechanism was built into the system via reversible acylhydrazone bonds between acylhydrazine and the aldehyde groups present on the oxidized furan–HA backbone. Most remarkably, this hydrogel showed great adhesion to native cartilage tissue (adhesive strength of ≈10 kPa), owing to the aldehyde‐amine Schiff‐base reaction. Based on a similar principle, a hydrogel composed of hydrazide‐modified HA (HA‐HYD) and aldehyde‐modified HA (HA‐ALD), exhibited fast self‐healing (healing after 10 min) with 100% healing efficiency. 185 Additionally, 3T3 fibroblast cells encapsulated within the hydrogel demonstrated high cell viability (after 14 d of culture) in accordance with a high biocompatibility. With regard to self‐healable hydrogels generated through physical crosslinking, a series of milestone studies from Jason Burdick's laboratory are worthy of mentioning. 38, 181, 182, 183, 184 In most of these studies, guest–host interactions between adamantine(guest)–HA and β‐cyclodextrin(host)–HA were used to develop self‐healable hydrogels, which at the same time also were suitable for printing 3D scaffolds due to their shear‐thinning properties. 38, 182, 186, 187, 188 In some of these studies, a methacrylated HA precursor polymer was added to the system for stabilizing the finalized scaffolds through UV‐crosslinking of the methacrylate groups (covalent bonding). 38, 182, 188 Specifically, the authors used this method to generate highly complex 3D scaffolds into which 3T3 fibroblast cells were encapsulated in a viable state for up to 5 d of culture ( Figure 7 ). 188 In the authors' opinion, the combination of 3D printing and self‐healable hydrogel inks can pave the way for 3D‐printed tissue engineering constructs with the capacity to heal in the same way that natural tissues do within the human body. To this end, the authors also envision that the addition of shape memory properties to such constructs could enable the delivery of scaffolds with complex architectures using a noninvasive syringe injection methodology. We will revisit injectable scaffolds with shape memory and self‐healing capacity in Section 6. 3. Figure 7 A self‐healable and printable HA‐hydrogel based on host‐guest interactions. a) The chemistry behind the self‐healing mechanism. b) A depiction of the cell‐laden hydrogel patterns with helical features, which were generated through extrusion printing. c) The team also used the extrusion printing technique to generate a grid‐like 3D construct consisting of 3T3 cells, which displayed high viability. Adapted with permission. 188 Copyright 2016, American Chemical Society. 3. 1. 4 Gelatin Gelatin is a natural protein that is derived from the denaturation of collagen via hydrolysis. 189, 190 Consequently, gelatin retains the bioactive sequences of collagen (such as an arginine–glycine–aspartic acid peptide), while exhibiting limited antigenicity. 191 Gelatin has numerous advantages making it an exciting biopolymer for tissue engineering and regenerative medicine. 148, 192, 193, 194 This includes biocompatibility, biodegradability, cost effectiveness, and ease of modification. 190, 191, 192 Yet, gelatin undergoes gel–sol transition at body temperature, which emphasizes the importance of chemical cross‐linking of gelatin‐based hydrogels for tissue engineering applications. 195 Because gelatin contains aromatic residues in its backbone (e. g. , tryptophan, phenylalanine, and tyrosine), the most logical method for developing self‐healable gelatin is through guest–host interactions. 196 In a noteworthy example, Feng et al. used the guest‐host interactions between aromatic residues (guest) of gelatin and photo‐crosslinkable acrylate β‐cyclodextrin (Ac‐β‐CD) (host) to institute a highly stretchable (failure strain above 400%) hydrogel with quick healing ability (5 min) and good biocompatibility, as tested through encapsulation with human mesenchymal stem cells (hMSCs). 197 The reversible nature of guest–host crosslinks made this hydrogel injectable and allowed infiltration and migration of hMSCs into the hydrogels without comprising the structural integrity of the hydrogel matrix. Most importantly, in vivo implantation of the hydrogel into rat calvarial bone defect led to enhanced tissue deposition in the defect, which was correlated with the ability of the hydrogel to recruit endogenous osteoblastic cells as the reversible crosslinks within the gel enabled cell migration without disrupting its mechanical integrity. As it was described in Section 2. 4, double‐network hydrogels are another approach to produce self‐healing hydrogels. Along these line, gelatin methacrylate (GelMA) was used in conjugation with tannic acid (TA) to institute a self‐healing hydrogel with adhesive properties. 198 The double‐network hydrogel in this study was comprised of precrosslinked GelMA hydrogel accompanied with TA as a multifunctional hydrogen bond provider. The resulting hydrogel showed significant increase in ultimate stress (4. 3‐fold), compressive modulus (2. 5‐fold), and elongation (sixfold), when compared to pristine GelMA hydrogel. Furthermore, the GelMA‐TA hydrogel generated sufficient adhesiveness to various surfaces (rubber, plastic, metal, glass, and porcine skin) corresponding with the TA content in the hydrogel. 3. 2 Synthetic Polymers Synthetic polymers have played an integral part in the technological revolution, which have been witnessed in modern times. This is in part due their chemical inertness, strong elastic modulus, flexibility, and ease of custom modification. 199, 200 However, some synthetic polymers are highly toxic to humans, not biodegradable, and biologically inert, and thus not suitable for tissue engineering applications. 201 Furthermore, many synthetic polymers are oil based, which is a great disadvantage, as the global oil supplies are expected to become greatly exhausted in the near future. Nevertheless, synthetic polymers offer a range of exciting properties that one has to consider when engineering self‐healing hydrogels. As was noted in the previous section, the natural‐based polymeric hydrogels are typically not stretchable, do not offer instant self‐healing, and are mechanically weak. Their synthetic counterparts can address all of these shortcomings, and we will, therefore, highlight their recent uses in the field of self‐healable tissue engineering hydrogels. In particular, we will mainly focus on self‐healable hydrogels made from polyethylene glycol, poly(acrylamide), and poly(vinyl alcohol) under conditions that make them amenable to tissue engineering applications (Figure 5 ). 3. 2. 1 Polyethylene Glycol PEG has been cemented as one of the most imperative hydrophilic polymers for biomedical applications thanks to desirable characteristics such as good biocompatibility, nonimmunogenicity, and versatile physical properties. 23, 202, 203 The widespread recognition that PEG has garnered over the years has made it the first choice for most engineers attempting to develop self‐healing hydrogels. This is evident from the many studies published on the preparation of self‐healable PEG hydrogels. 57, 101, 122, 128, 204, 205, 206 Examples include hydrogels with repair mechanisms based on borate esters, 122, 206 Diels–Alder reactions, 128 imine bonds, 205, 207 hydrogen bonds, 57, 204 and hydrophobic forces. 84 Especially, a series of landmark achievements, wherein borate esters were used to impart good self‐healing efficiency and high mechanical stability are worthwhile mentioning. 122, 206 In these milestone studies, the authors develop self‐healable hydrogels that were elastic and could reach high storage modulus values (up to 10 kPa) without rupturing. For instance, a team led by Robert Langer and Daniel G. Anderson managed by using boronate ester bonds between a four‐arm PEG–phenylboronic acid (three different PEG–phenylboronic acid derivatives: PEG–FPBA, PEG–PBA, and PEG–APBA) and four‐armed PEG–diol macromonomers, to develop a stretchable hydrogel with instant healing properties ( Figure 8 ). 122 Moreover, this hydrogel displayed shear‐thinning properties and could thus be used as an injectable and self‐healable hydrogel carrier for stem cell therapy. However, to this end, it is pivotal that the hydrogel carrier can sustain viable cells over longer periods. The team validated this by encapsulating fibroblast cells within the hydrogel and demonstrated that the cells remained viable for up to 72 h with viability of around 80%. This team also examined the potential of this system for the targeted delivery of both biologics and cells to possible target sites in the body with promising results. Figure 8 A stimuli‐responsive and self‐healable PEG‐hydrogel based on phenyl–boronic acid–cis–diol binding. a) The chemistry behind the self‐healing mechanism. b) Photographic images showing the self‐healing properties of the hydrogel. c) In addition to its self‐healing properties, this hydrogel was also shear‐thinning. d) Strain amplitude sweep and the associated e) step‐strain measurements of the PEG‐hydrogel to elucidate its self‐recovery properties. f) The viability of cells within the self‐healable hydrogels was ≈80% and remained at this level for up to 72 h. g) The release profiles of insulin, BSA, and IgG at different time‐points and h) cell viability at different time points are shown here. Adapted with permission. 122 Copyright 2016, Wiley‐VCH. In another series of experiments, self‐healable PEG‐based hydrogels were synthesized by functionalizing telechelic difunctional PEG (DF‐PEG) with two aldehyde end groups through esterification of the hydroxyl groups on PEG and mixing the DF‐PEG with an aminated polymer such as chitosan 205 or polyethylenimine (PEI). 207 The combination of DF‐PEG and PEI resulted in an exciting hydrogel that could stretch up to 400% and recover fully to its original state. Moreover, once broken, it could instantly self‐heal and withstand extensive stretching without rupturing. However, one drawback of this study is the use of PEI, a molecule that can be toxic to cells at high doses. 208 Along the same lines, a hydrogel comprised of dialdehyde‐functionalized polyethylene glycol (DF‐PEG) and agarose–ethylenediamine conjugate (AG–NH2) showed fast self‐healing (after 5 min) and exhibited higher yield stress (≈ 2000 Pa) when compared to pure agarose (126 Pa) or agarose–PEG hydrogels (315 Pa). 209 This hydrogel demonstrated excellent adhesion to porcine skin and after being applied on the incision of the skin, it was able to withstand a bursting pressure that was even bigger than the arterial blood pressure (120 mmHg), indicating the capability of the hydrogels to assist in wound healing. Furthermore, human umbilical vein endothelial cells (HUVEC) in contact with the hydrogel extracts demonstrated high cellular viability (>80%), indicating low cytotoxicity of the hydrogels. Most importantly, hydrogels showed remarkable hemostatic capability after being applied to a rabbit liver incision immediately, in contrast with conventional sterile gauze. Another way of preparing self‐healing PEG hydrogels is through thiol‐ene interactions, where the PEG alkene (or alkyne) reacts with a thiolated polymer to establish a self‐healing polymeric network. 210, 211, 212 To this end, Macdougall et al. developed a self‐healing hydrogel from a mixture of four‐arm PEG alkyne and difunctional linear PEG thiol, yet this hydrogel suffered from low compressive strength and stretchability. 212 In an attempt to address these shortcomings the same group developed interpenetrating networks (IPN) by incorporating a range of unfunctionalized natural polymers (alginate, chitosan, gelatin, heparin, or hyaluronic acid) into the hydrogel. Accordingly, PEG/alginate and PEG/gelatin systems showed higher compressive strength, tensile strength, and stretch‐ability when compared to control PEG hydrogel. The observed properties were attributed to existence of a secondary electrostatic loose network that complemented the efficient nucleophilic thiol‐ene cross‐linking chemistry. Most remarkably, human mesenchymal stem cells encapsulated in PEG/alginate hydrogels showed higher cell viability after 72 h (95%) when compared to that of PEG‐alone systems (77%), which implied PEG/Alginate provided a cytocompatible matrix that supports cell growth. Recently, UPy has also been shown to be a promising tool for making self‐healable PEG hydrogels owing to hydrogen‐bond‐assisted dimerization of the UPy moieties. 57 Notably, PEG oligomers were functionalized with UPy and immersed in a buffered solution, in which they began to solidify into stable hydrogels through hydrogen bonding mediated by UPy. As this system was injectable, it was used for further down‐stream studies to examine its ability to deliver bone morphogenetic protein 7 (BMP‐7) into kidney tissue via a minimally invasive injection method. Therefore, this UPy‐based approach could potentially lead to significant clinical break‐through discoveries in the foreseeable future. 3. 2. 2 Poly(vinyl alcohol) Another synthetic, biocompatible, non‐toxic and water‐soluble polymer—PVA—has also been employed in numerous biomedical applications ranging from contact lenses to scaffolds and various drug delivery platforms. 152, 213, 214, 215 PVA is unique because of its many hydroxyl groups, which, together with its high water retention properties, makes it an ideal candidate for self‐healing hydrogels. Accordingly, over the years, several PVA‐based self‐healing systems have emerged, in which the healing mechanism was based on reversible hydrogen bonds. 63, 216, 217, 218 In one such example, a double‐network hydrogel consisting of PVA and PEG was combined to yield an exceptionally tough hydrogel that could reach an ultimate tensile strength of almost 1. 3 MPa while exhibiting shape memory and self‐healing properties. 219 The secret behind the many amazing properties of this PVA–PEG system stemmed mainly from the presence of weak reversible hydrogen bonds between PVA polymers that yielded deformability, and strong chemical crosslinks between PEG polymers that could keep the system intact in high strain regimes and, thus, enable a return to its original shape. Even though a plethora of seal‐healable hydrogels have been presented to the field of biomaterial science, this combination of high material toughness and self‐healing ability is rather unusual, as high stiffness typically goes hand‐in‐hand with strong and stable intermolecular bonds, which conflict with the unstable and highly dynamic links behind self‐healable systems. Similarly, a hydrogel blend made from PVA, borax, and nanofibrillated cellulose resulted in a reinforced self‐healable hydrogel with improved mechanical properties as a result of reversible interchain hydrogen bonding between PVA and the nanocellulose, and strong boronate ester links mediated among the PVA polymers by borax. 220 In a similar vein, a mixture of PVA and 2APBA yielded a hydrogel that exhibited self‐healing ability not only in PBS but also in a culture media (with or without the serum). 221 Moreover, encapsulated fibroblast cells (CCL‐151) or breast cancer cells (MDA‐MB‐231) exhibited good viability over the course of 7 d, indicating cytocompatibility of the materials. More remarkably, the self‐healing ability of this hydrogel was successfully used to create a dynamic coculture system by separately encapsulating fibroblast cells and breast cancer cells in the hydrogel pieces and subsequently connecting them. In our opinion, this study holds much value as it can inspire researchers to create more complex culture environments for probing dynamic cell−cell and cell−matrix interactions. Another challenge in the field of biomaterial science is the design and development of materials with high stiffness but fast molecular dynamics to yield a hydrogel that can quickly heal itself. To address this challenge, a recent study combined highly crystalline nanocellulose with PVA to generate a promising hydrogel for tissue engineering applications. Specifically, this self‐healable hydrogel was composed of hard, modified cellulose nanocrystals and soft, polymeric PVA domains crosslinked by dynamic host–guest interactions brought about by cucurbituril (CB). 63 This combination resulted in a mechanically strong hydrogel that could heal rapidly (within few seconds). From our viewpoint, such systems that address the conflict between strength and fast molecular dynamics could lead to new resilient biomaterials with the capacity to mend defects within load‐bearing tissues, such as cartilage and bone, as they can readily heal possible small fractures arising from the cyclic stretches and compressions imposed on natural bone and cartilage tissues. 3. 2. 3 Poly(acrylic acid)/Poly(acryl amide) Another class of synthetic self‐healing hydrogels—albeit less used—are those made from acrylate‐based polymers, especially acrylic acid and acrylamide. These hydrogels are special because of their good water absorption capacity and high density of functional chemical groups, such as carboxylic acid and amine groups, which in combination with their nontoxicity, makes them ideal to use for many biomedical applications. 222, 223, 224, 225 Over the years, acrylate‐based self‐healable hydrogels have, for the most part been generated through hydrophobic interactions derived from either host‐gust interactions or micelles entrapped within the hydrogels. 81, 82, 226, 227 In particular, the incorporation of micelles into poly(acrylic acid)/poly (acrylamide) hydrogels has proven a versatile pathway to produce self‐healable hydrogels with desirable attributes, such as high extensibility, elastic modulus, and toughness. A common approach to this end is the spontaneous polymerization of stearyl methacrylate (C18) into micelles within hydrophobically modified acrylic acid/acrylamide hydrogels in the presence of either of the surfactants; cetyltrimethylammonium bromide (CTAB) or sodium dodecyl sulfate (SDS). 81, 82, 226 A key example of this methodology can be found in a recent study by Gulyuz et al. , 81 wherein a poly(acrylic acid) hydrogel encapsulated with C18‐based micelles was developed. This self‐healable hydrogel could stretch up to 800% of its original length before it broke and displayed a high tensile strength in the range of 0. 7–1. 7 MPa. Similarly, the copolymerization of N, N ‐dimethylacrylamide (DMA) and stearyl methacrylate (C18) in the company of SDS led to a poly(acrylamide)‐based hydrogel with self‐healing properties arising from the incorporation of C18/SDS‐based micelles within it. 82 This hydrogel was extremely stretchable (broke at 4200% strain), displayed shape memory behavior, and could heal itself. However, one disadvantage of using the above‐mentioned methodology is the long hydrogel self‐healing duration; typically ranging from 20 to 60 min. Another way of preparing self‐healing poly(acrylamide) hydrogels is through host‐guest interactions, in which the amine groups on the polymer backbone are functionalized with a host—cyclodextrin—and a hydrophobic and aliphatic guest molecule ( n ‐butyl acrylate, adamantane, and ferrocene). 61, 62 Kakuta et al. embraced this approach to develop a self‐healable hydrogel based on a adamantine–ferrocene host molecule, which could immediately mend itself after rupturing, however, the self‐healing efficiency in terms of the adhesion strength between the rejoined pieces was relatively low, and it took up to 24 h to reach the prerupture adhesion strength. 61 In another study, researchers attempted to develop a double‐network hydrogel by using host–guest interactions, between isocyanatoethyl acrylate modified with β‐cyclodextrin (host; β‐CD‐AOI 2 ) and 2‐(2‐(2‐(2‐(adamantyl‐1‐oxy)ethoxy)ethoxy)ethoxy)ethanol acrylate (Guest; A‐TEG‐Ad), accompanied by a second covalent bonding between acrylate groups (achieved by UV‐initiated polymerization). 228 The resulting hydrogel showed fatigue resistance and resistance to slicing because of the combination of host–guest interactions and covalent bonds in the hydrogel networks. Furthermore, in a strain sweep test (1% strain–1000% strain–1% strain), these hydrogels fully recovered their initial storage modulus even after the hydrogel had undergone several cycles. Most significantly, mouse bone marrow stromal cells (mBMSCs) or myeloid‐derived suppressor cells (MDSCs) cultured in the presence of the hydrogel extracts had a proliferation rate consistent with that of the positive control group, indicating their good cell compatibility. Most recently, the same group used a similar composition except they added GelMA to the recipe to benefit from its excellent biocompatibility. 229 The resulting hydrogel exhibited superior mechanical properties compared to pristine GelMA and was able to completely heal itself after 1 h. Owing to shear‐thinning properties of this hydrogel, it was successfully 3D printed into a multilayer scaffold and mouse bone marrow stem cells (mBMSCs) cultured on top of these scaffolds were shown to be viable and proliferating after 7 d of culture. Furthermore, subcutaneous implantation of this hydrogel on the backs of nude mice (for 40 d) revealed that the scaffolds were completely integrated with the autogenous tissue of the nude mice, and new subcutaneous muscle tissues and blood vessels were formed in their pores with no immunological rejection occurred. These results suggested that these self‐healing scaffolds had favorable bioactivity and histocompatibility, making them useful in biomedical applications. Hydrogen bonds are another alternative interaction scheme, which has been explored extensively over the years to develop self‐healable acrylic acid/acrylamide hydrogels. 53, 60, 230 In one such study, acrylic acid and acrylamide were mixed in the company of glycogen to polymerize into a hydrogel through hydrogen bonds between carboxyl groups and hydroxyl groups in glycogen. This hydrogel could completely heal itself at neutral pH after 12 h while exhibiting shear modulus values as high as 1000 kPa and a swelling ratio that could reach 3500% after ≈ 500 min of swelling in water. 230 The combination of self‐healing capacity at neutral pH together with high mechanical strength makes this system interesting, as many of the self‐healing hydrogels reconnect in nonphysiologically conditions and display low elastic modulus values. Based on the same concept, a double‐network hydrogel composed of poly(acrylamide‐ co ‐acrylic acid) (PAM‐ co ‐PAA) and PVA was generated via copolymerization and hydrogen bonds between the carboxyl groups of PVA and acrylamide groups of acrylamide. Moreover, the PVA polymeric backbone was made highly crystalline through a well‐established freeze/thawing procedure. Overall, this double‐network hydrogel could stretch up to almost 600% and reach 1230 kPa in tensile strength before breakage. The authors speculated that these formidable mechanical properties arose from the combination of weak reversible hydrogen bonds and the rigid crystalline PVA domains. 53 3. 3 Others Other self‐healable hydrogels based on peptides, 231, 232, 233, 234, 235 mussel‐inspired proteins, 47, 48, 101, 236, 237 conductive polymers, 85, 233, 238, 239, 240, 241, 242, 243, 244, 245 and zwitterionic polymers, 246, 247, 248, 249 have also gained acceptance in the field. Peptide‐based hydrogels with self‐healing properties have been described elsewhere in detail, and the interested reader is referred to these recent reviews on this topic. 231, 250 In this section, we will instead focus on conductive, mussel‐inspired and zwitterionic self‐healable hydrogels. The working mechanism behind most mussel‐inspired hydrogels stems from metal chelating cross‐links between positively charged Fe +3 ions and negatively charged catechol molecules. 48, 101, 102, 237 In simple terms, the polymer in question is functionalized with negatively charged catechol groups—typically in the form of 3, 4‐dihydroxyphenylalanine (DOPA)—which then crosslinks into a hydrogel through sacrificial bonds mediated by metal‐catechol coordination. These bonds are stronger than hydrogen bonds yet sufficiently weaker than covalent bonds to enable dynamic binding schemes and, ultimately, good self‐healing properties. A recent forerunner toward such Fe +3 –DOPA‐mediated hydrogels was based on a DOPA‐chitosan (DOPA‐CHT) derivative, which could spontaneously crosslink into a double‐network hydrogel in the presence of Cl 3 Fe. 6H 2 O (metal–ligand coordinate) and genipin (covalent) ( Figure 9 ). 102 This bioinspired hydrogel displayed a range of interesting properties including: i) a high compressive strength in the MPa range, ii) good cytocompatibility and iii) fast self‐healing time ranging from 8 to 15 min. In addition, the mussel‐inspired hydrogel was injectable and could easily recover from repeated cyclic loadings. For these reasons, this is an ideal hydrogel carrier for stem cell therapies targeted against load‐bearing tissues, such as bone and cartilage. Figure 9 A tough, durable, and self‐healable chitosan–dopamine–hydrogel based on chelation. a) The chemistry behind the self‐healable hydrogel. DN indicates a double‐network hydrogel made from dopamine–chitosan (DOPA–CHT) and medium molecular weight chitosan (M M w ‐CHT), while DC indicates a double crosslinked DOPA–CHT hydrogel made through genipin (crosslinks the amine group in CHT) and Fe 3 (mediates coordination bonds between catechol groups). b) Compressive stress–strain curves of DOPA–CHT hydrogels that are solely crosslinked through Fe 3+ (SFe) or through the addition of 0. 5% genipin (SC 0. 5%). SN (0. 5%) and DN (0. 5%) indicate single‐network DOPA‐CHT hydrogel with 0. 5% genipin and double‐network hydrogel with 0. 5% genipin, respectively. c) The self‐healing properties of DN 0. 5% and SN 0. 5% hydrogels are shown. d) Cyclic stress–strain curves and e) their associated hysteresis recovery. Adapted with permission. 102 Copyright 2017, Wiley‐VCH. Mussel‐inspired hydrogels with self‐healing capacity have also been generated under metal‐free conditions by using catecholamine‐3, 4‐dihydroxyphenethylamine (dopamine) instead of DOPA. These hydrogels typically obtain their self‐healing properties from noncovalent interactions between aromatic rings, hydrogen bonds, and via imine bonds mediated by the NH2‐groups in polydopamine. An elegant solution toward such hydrogels was recently reported, wherein gelatin was functionalized with aldehyde groups and mixed with polydopamine under metal‐free condition (via sodium periodate) to generate an injectable, moldable, self‐healable hydrogel with fast recovery time, good self‐healing efficiency (95%), and a good adhesiveness (10–40 kPa). 237 Such properties were an outcome of a complicated cross‐linking scheme comprised of hydrogen bonds in combination with imine and π–π interactions. Another remarkable concept in the field is the development of electrically conductive and self‐healable hydrogels—as many tissues in the body—such as heart, muscle, and brain tissues are electroactive, and therefore need to be matched with similar electroactive biomaterials to yield good biointegration. 93 Despite extensive studies on the design and development of self‐healable hydrogels, only a few studies have attempted to capture other properties, such as electrical conductivity. An elegant forerunner toward this goal has been based on polypyrrole‐based hydrogels; as polypyrrole is a highly conductive yet biocompatible polymer. 85, 233, 238, 239, 240, 241 To this end, Darabi et al. 85 generated a conductive, injectable, and self‐healing hydrogel by decorating chitosan with pyrrole and mixing it with acrylic acid monomers in the presence of Fe +3 and N, N ˝‐methylenebis‐acrylamide (MBA) crosslinkers ( Figure 10 ). These components altogether generated a double‐network hydrogel compromised of reversible ionic interactions between carboxylic groups on poly(acrylic acid) and amine groups on polypyrrole mediated by Fe +3, and irreversible covalent bonds between neighboring poly(acrylic acid) chains mediated by MBA. The combination of reversible and irreversible bonds resulted in a highly conductive hydrogel that could stretch up to 1500% and which could self‐heal its electrical and mechanical properties after just 1–2 min. Specifically, the mechanical healing efficiency was complete (100%), whereas the electrical healing efficiency saturated at 96% after 1 min. Figure 10 A conductive and self‐healable hydrogel made from chitosan (CSH), poly (acrylic acid) (PAA), and polypyrrole (PPY) based on electrostatic interactions. a) The self‐healing properties of the resulting hydrogel are displayed. b) The mechanical and electrical self‐healing efficiency of the hydrogel. c) Schematic depicting the preparation of the wearable sensor. d) The 3D printed wearable sensor could detect human‐body motions by measuring the associated resistance variation. Adapted with permission. 85 Copyright 2017, Wiley‐VCH. Another series of milestone concepts recently pursued is the development of electrically conductive, thermoplastic, moldable, and self‐healing hydrogels by mixing polypyrrole with agarose. 240, 251 The polypyrrole makes the system conductive, while the agarose gel imposes self‐healing properties onto the system as the gelation of agarose is thermally reversible. This system could, therefore, heal itself through both external heat and near‐infrared light because of reversible liquidation and gelation in response to thermal stimuli. Moreover, the authors managed to adhere this conductive gel directly on the human skin for human motion detection and demonstrated that the system could yield an electrical circuit that could self‐heal via external heat or near‐infrared light stimuli. 240, 251 Most recently, a new class of self‐healing hydrogels based on zwitterionic polymers have emerged— taking the field by storm owing to their unique biocompatible attributes. These polymers contain a balanced pairs of cationic and anionic groups, and mimic the phospholipids comprising the membranes of native cells or the mixed‐charge surfaces of many proteins. 246 In fact, the positive and negative charges of the overall neutral zwitterionic molecules make a high dipole moment and such strong dipolarity endows excellent adhesion of zwitterionic hydrogels to many surfaces through ion–dipole or dipole–dipole interactions. 252 Furthermore, the association of zwitterionic polymers can provide physical cross‐linking to enhance the mechanical properties of hydrogels. Most remarkably, zwitterions can assist the ion transportation along the highly dipolarized skeleton to promote ion conduction, which endows zwitterion hydrogels with good conductivity. 253 Despite such great properties, most zwitterionic hydrogels are mechanically weak, hence researchers have incorporated other functionalities into them to make them self‐healable. In a noteworthy study, researchers utilized two different types of zwitterionic monomers, carboxybetaine acrylamide monomers with either one‐carbon (PCB‐1) or two‐carbon (PCB‐2) spacing between the charged groups, and separately cross‐linked them using carboxybetaine diacrylamide. 246 Both hydrogels exhibited self‐healing behaviors in a strain sweep test (1% strain–300% strain‐1% strain), owing to ionic and hydrogen bonds between polymeric chains. Yet, PCB‐2 hydrogels demonstrated higher storage modulus compared to PCB‐1 hydrogels, which was attributed to stronger hydrogen bonds in PCB‐2 hydrogels. Additionally, both hydrogels were injectably and human embryonic kidney cells (HEK‐293T cells) encapsulated in these hydrogels retained a high cell viability even after being injected through a 28‐gauge needle. Additionally, hMSCs encapsulated in these hydrogels had greater population expansion over 14 d than that of cells in standard flask culture. Notably, hMSCs encapsulated in the hydrogels maintained their multipotency after 28 d of culture, while half the population of flask‐cultured hMSCs lost their multipotency. This phenomenon was shown to be a result of ROS‐scavenging capacity of zwitterionic hydrogels, which in turn causes stem cells favoring self‐renewal and mitigating nonspecific differentiation. In summary, this hydrogel presented a promising new platform for a wide variety of clinical applications requiring biocompatible injectable materials. Another approach to make zwitterionic‐based self‐healing hydrogels is by modification of these polymers with boronic acid to allow formation of boron ester bonds. Along these lines, a group of researchers used a zwitterionic monomer (2‐methacryloyloxyethyl phosphorylcholine; MPC) and copolymerized it with a benzoxaborole‐containing monomer (5‐methacrylamido‐1, 2‐benzoxaborole) to yield the PMB hydrogel. Separately, they have also copolymerized the MPC monomer with a glucose‐containing monomer (2‐gluconamidoethyl methacrylamide) to make the PMG hydrogel. Accordingly, mixture of PMB and PMG resulted in the formation of a hydrogel (PMBG) that showed fast self‐healing (after 20 s) owing to boron–ester bonds between the boronic acid (on PMB backbone) and hydroxyl groups (on PMG). The resulting PMBG hydrogels was injectable and showed pH‐responsive behavior due to the nature of boron–ester bonds. Additionally, both normal skin fibroblast cells (NSFB) and cancerous HeLa cells treated with the gel extracts maintained a high level of cell viability (>80%). 247 Similarly, the same group copolymerized the MPC monomer with catechol‐containing monomer (dopamine methacrylamide; DMA) to yield poly (MPC‐ co ‐DMA) and, mixed it with poly (MPC‐ co ‐DMA) to facilitate a rapidly self‐healing hydrogel (after 1 min) with pH‐responsive behavior. 248 3. 4 Outlook and Future Opportunities In summary, the synthetic polymers give rise to stronger self‐healable hydrogels with faster self‐healing kinetics than their natural counterparts. However, this is accompanied by much lower biocompatibility and potential adverse reactions within the body. The incorporation of cell adhesive motifs, such as RGD peptides, into the polymeric backbone of synthetic hydrogels, could potentially address some of these issues by enhancing the spreading, proliferation, and differentiation of encapsulated cells into mature tissues. The combination of a synthetic and natural polymer also offers the interesting possibility of generating a hydrogel that taps into the positive properties of both realms. As the field advances, we anticipate that other sophisticated polymers grown inside bacteria—through recombinant technology—will grab the attention of scientists in the field, as this methodology can enable highly customized routes toward the generation of polymers with even better self‐healing efficiency and mechanical properties. Although attempts have already been made to introduce this technology into the field, most of these hydrogels have been weak (typically below 5 kPa) and, for the most part, intended to be used as injectable stem cell carriers that could mechanically shield the cells during the injection phase and retain them within the target site in a postinjection scenario. 231 We, therefore, conjecture that further research into this area could yield some exciting self‐healable hydrogels for the field of tissue engineering. 4 Nanomaterial‐Based Self‐Healing Hydrogels Nanomaterial‐based hydrogels are defined as hydrated polymeric networks held together by noncovalent or covalent bonding with each other and nanomaterial reinforcers. 93, 254 These nanomaterials are incorporated into the hydrogels to either reinforce their physical properties (e. g. , mechanical properties, thermal and electrical conductivity, and swelling degree) or to bestow superior biological properties to them. 255, 256, 257, 258, 259, 260, 261, 262, 263 Accordingly, such nanocomposite hydrogels are highly desirable from a biomedical point‐of‐view, as many‐at‐the‐same‐time modifications are no longer needed to yield the combinatorial biomaterials required for optimal tissue regeneration. 90, 254, 259, 264, 265, 266, 267 Recently, scientist have been working toward a new scenario, in which the nanomaterials are used to add autonomous repairing mechanisms to hydrogels. 268, 269 Indeed, the combination of the multifunctional nature of nanomaterial‐based hydrogels with self‐healing characteristics would offer a promising prospect for generating even better tissue engineering hydrogels. 270, 271, 272, 273, 274, 275, 276, 277 Overall, the nanomaterials can be categorized into three groups: a) carbon‐based, b) mineral‐based, and c) magnetic ones. The carbon‐based group typically refers to carbon nanotubes (CNT) and graphene, 278 whereas the mineral based refer to clay‐based platelets and ceramic nanoparticles 254 while the magnetic nanomaterials mostly include iron oxides. 279 In this section, we will highlight some of the most important advances in generating such self‐healable hydrogels. 4. 1 Carbon Based Carbon‐based nanomaterials, such as CNT and graphene, represent a unique class of materials with fascinating properties, including excellent mechanical, 280 electrical, 281, 282, 283 and optical properties. 284 They are currently widely used in the field for the design and development of tissue engineering scaffolds with the capacity to generate electroactive and load‐bearing tissues. 93 Because carbon‐based nanomaterials are easy to modify they have also steered much attention among researchers as a new pathway toward self‐healable hydrogels; their popularity stems primarily from the many unique physical and chemical nanomaterial interactions, which they can establish with the polymer backbone of hydrogels. Through manipulation of these interactions, a range of self‐healable systems for tissue engineering applications has been developed in recent years, including hydrogels with hard‐to‐get properties, such as high self‐healing efficiency, good electrical conductivity, flexibility, and mechanical toughness. 274, 285, 286, 287, 288 The following subsections will first describe the many chemical and physical properties of CNTs and graphene, and then the potential of these attributes in the engineering of self‐healable hydrogels with multifunctional capacities. 4. 1. 1 Graphene Graphene can be described as a sheet of 2D monolayers of sp 2 ‐hybridized carbon atoms that are arranged into a high‐aspect ratio hexagonal pattern. 289, 290 Since its discovery in 2004, some extraordinary properties have been associated with graphene. For instance, graphene is a potent conductor of electricity (2. 50 × 10 5 cm 2 V −1 s −1 ) and heat (3000 W m −1 K −1 ) 291 and has, in some cases, been reported to have a specific tensile strength 100 times that of steel. 292 Another exciting property of graphene is its large surface area (2600 m 2 g −1 ), which is an important contributing factor to the many strong interactions that it can establish with polymers. Graphene can be modified into an even more functional nanomaterial, graphene oxide (GO), which offers numerous modifiable oxygen‐based groups, such as OH‐groups, useful for further downstream applications. 293 The electrical conductivity of graphene can also be further improved by reducing GO into reduced GO (rGO) through chemical and thermal treatment, 294 however, this pathway can sometimes lower the quality of graphene and thus presents a tradeoff between chemical functionality and conductivity. 295 The negatively charged OH‐groups on GO enable many interaction schemes with polymeric chains that can lead to self‐healable hydrogels. To date, most of these schemes have been based on either hydrogen bonds 286, 296, 297 or chelating‐crosslinks mediated by metal ions. 54, 298 The polymers used have for the most part been derived from acrylamide 47, 286, 296, 297, 299 or acryl‐based polymers. 54, 298 In general, the GO‐acrylamide hydrogels displayed much better properties than the acryl‐based ones, as they could elongate up to 4900%, displayed higher fracture strengths ranging from 0. 18 to 1. 0 MPa, and could heal more efficiently with a self‐healing efficiency that in some instances could reach up to 97. 9% after 4 min of healing time. 297 On the other hand, the GO‐acryl‐based hydrogels broke at strain values in the range 1000%–3000% with fracture strengths ranging from 0. 07 to 0. 8 MPa. GO‐acryl‐based hydrogels also displayed low healing efficiency, which at best could reach 50% after 48 h of healing time. 298 Thus, it is clear that the GO‐acrylamide hydrogels are most worthy of attention and we will accordingly focus on them in this section. An elegant forerunner to this end was a self‐healing GO‐acrylamide hydrogel that exhibited high elongations (4900%) with an ultimate tensile strength value of 0. 35 MPa. 286 Although this hydrogel displayed some amazing mechanical properties, its nonoptimal self‐healing efficiency (88% after 24 h) significantly limited its use in tissue engineering applications. This low self‐healing efficiency most likely stemmed from the incorporation of GO into the system as the healing efficiency dropped concomitantly with GO concentration. The authors speculated that this effect was caused by the large GO sheets acting as diffusion barriers, thus preventing the polymers from reconnecting broken cross‐links in the broken hydrogels. This lack of methodology was also evident in another publication by Han et al. , 47 wherein the authors incorporated PDA into the GO‐acrylamide system. In brief, these authors used PDA to reduce GO to rGO, after which acrylamide monomers were added to the system in the presence of an initiator to commence the formation of a fully crosslinked hydrogel. The inclusion of PDA into the system had two functions: one that resulted in higher conductivity by turning GO into rGO and the other enabled the system to heal itself after damage. The healing capacity was caused by some reversible bonds between neighboring PDA chains, such as hydrogen bonding and cation–π interactions, as well as hydrogen bonds between PDA and polyacrylamide. In addition to all of these exciting properties, this hydrogel could also stretch up to 40 times its original length before breaking and exhibited an ultimate tensile strength that could reach almost 0. 2 MPa under certain conditions. These additional mechanical properties were achieved through the combination of the above‐mentioned reversible bonds and strong covalent bonds between the polyacrylamide chains. Although the self‐healing efficiency was less than optimal (80% after 24 h), the system provided some exciting opportunities, such as reasonable adhesive strength to skin tissue (30 kPa), long‐term biocompatibility in vivo, high conductivity, and a good electrical healing efficiency (95% after 24 h). Overall, these exciting properties made the system able to perform various tasks, such as measuring the cyclic movement of articular joints without failing, detect electromyographic signals from the skin, and regenerate broken cartilage tissue in a rabbit animal model. Some studies have also focused on the incorporation of graphene into hydrogels that were made from much smaller polymers such as amino acids, 300 amine‐terminated branched oligomers, 136 and DNA strings. 301 The most noteworthy of these are, in the author's opinion the one based on branched oligomers. This hydrogel could self‐assembly spontaneously after mixing carbonyl‐functionalized GO sheets with the branched and amino‐rich oligomers through imine bonds mediated between the GO sheets and the NH2‐groups, as well as hydrogen bonds between neighboring NH2‐groups. In its essence, the developed system was, therefore, a double‐network hydrogel, and the end‐product was a strong mechanical hydrogel with fast self‐healing time. Specifically, the GO‐based hydrogel could reach a tensile modulus close to 1 MPa while offering self‐healing properties. These healing properties included a self‐healing efficiency close to 100% after 1 h of healing at room temperature. However, just as in some of the other studies reviewed herein, 47, 286 the self‐healing efficiency dropped significantly as function of graphene concentration. The authors speculated that this was caused by the many strong imine bonds established between GO and the polymer due to the large GO surface area, which in turn resulted in a higher density of bonds that were dynamic, but yet, much stronger and stable than hydrogen bonds. The authors propose that these issues severely restricted the movement of polymer chains; this was also reflected in their differential scanning calorimetry (DSC) results, as the glass transition temperature ( T g ) increased significantly from −5 °C to 9 °C with increasing GO concentration (1% → 4%). In conclusion, GO‐based self‐healable hydrogels have added exciting attributes such as good conductivity, sensing‐abilities, and increased mechanical strength to self‐healable hydrogels. For these reasons, there is no doubt that the inclusion of GO in the field has led to some interesting scientific possibilities. However, these additional opportunities have come at the cost of a much lower self‐healing efficiency caused by restricted polymer mobility mediated by the large GO sheets. Therefore, further investigations are needed to develop new systems that can address this lack‐of‐methodology and unleash the full potential of GO‐based self‐healable hydrogels. 4. 1. 2 Carbon Nanotubes Since their discovery in 1991, 302 CNTs have been applied widely in the field of materials science, largely due to their electrical and mechanical properties. 303 The unique electromechanical properties of CNTs have made them useful in a great variety of applications, such as in electronics, optics, biomedical engineering, and nanotechnology. 304, 305, 306 The huge interest in CNTs is evident from the annual production of CNTs, which is currently in the megaton range and rapidly increasing. Indeed, CNTs are already being used in some commercial products, including water filters, rechargeable batteries, automotive parts, and sporting hulls. 307 CNTs can be described as sheets of carbon atoms that are rolled into high‐aspect‐ratio (>1000) nanotubes. These carbon atoms are arranged in well‐ordered hexagonal structures mediated by sp 2 bonds between neighboring carbon atoms. The CNTs themselves can have diameters ranging from 0. 8 to 20 nm—depending on whether they are single or multiwalled—with lengths ranging from a couple of nanometers to several centimeters. 307 Additionally, CNTs can be functionalized with highly reactive groups, such as hydroxyl, carbonyl, carboxyl, and amines and are, therefore, easy to incorporate into various hydrogel systems. They also pose unique properties such as high tensile strength (11–63 GPa), 308 electric conductivity (10 9 A cm −2 ), 309 Young's modulus (1–1. 8 TPa), 295 thermal conductivity (2000–6000 W m −1 K −1 at room temperature), 310 and a specific tensile strength which outcompetes that of steel by up to 100 times. 311 CNTs are, therefore, perfect reinforcing agents for self‐healable hydrogels as they are mechanically strong, easy to functionalize, thermally conductive, and electrically active. Compared to the many exciting reports on GO‐based self‐healable hydrogels, the literature on the CNT‐based hydrogels is rather sparse. 287, 312, 313, 314 Indeed, in our opinion, these systems need much future attention to fully open up the field of multifunctional and self‐healable hydrogels. Nevertheless, two recent studies from 2014 287 and 2017 314 have led to a series of milestone contributions in this regard. In one of these studies, a supramolecular hydrogel based on both weak and strong hydrogen bonds was made from carboxylic functionalized CNTs and polyethylene polyamine (PPA). In brief, this system consisted of weak intermolecular hydrogen bonds (N—H···N) between individual PPA molecules and strong bonds (N—H···O) between PPA and the carboxylic groups on CNTs. This hierarchical bonding scheme gave the system self‐healing, adhesive, and sol–gel properties. Specifically, the system could restore 90% of its mechanical properties within 90 s and adhered to a Teflon surface with an adhesion force of 10 kPa. The hydrogel also spontaneously solidified into a gel at 55 °C after 30 s and could return to its original liquid state when the temperature was shifted back to room temperature. Since CNTs are famed for their incredible thermal conductivity, this reversible sol‐gel transition could easily be controlled through external NIR light stimuli, and therefore stimuli responsiveness can be added to these systems already impressive list of properties. In another study, a self‐healing, conductive hydrogel was created through the incorporation of CNTs and borax into a PVA hydrogel ( Figure 11 ). 314 The self‐healing mechanism in this hydrogel was also primarily based on reversible hydrogen bonds, which in this case was generated between PVA and borate ions. Interestingly, the device was embedded into a Scotch permanent clear mounting tape to enable it to stretch up to 1000%. Specifically, the device could heal after 3 s with a self‐healing efficiency of 98%; and because of the CNT reinforcement, this hydrogel system was also tough, highly conductive, and piezoelectric. The piezoelectric properties of the system were used to turn the device into a high‐fidelity sensing device. To this end, the authors demonstrated a remarkable human‐motion sensing capacity of the device by utilizing its piezoelectric properties to sense resistivity changes during human‐joint motions. Figure 11 A conductive and self‐healable hydrogel made from CNTs and PVA, in which the self‐healing mechanism is governed by hydrogen bonds. a) The chemistry behind the self‐healable hydrogel. b) Photographic images of the self‐healing properties of the hydrogel and its electrical self‐healing efficiency. c) The hydrogel was embedded within a VHS scotch tape to increase its d) twistability and stretchability (up to 1000%). e) The hydrogel was utilized in a sensing device for the detection of various human motions. Adapted with permission. 314 Copyright 2016, Wiley‐VCH. In conclusion, despite their great promise as self‐healing and multifunctional reinforcing agents for tissue engineering hydrogels, CNTs have not yet been fully used to generate such advanced hydrogels. However, a few exciting studies have emerged, which have demonstrated the huge hidden potential of CNT‐based self‐healable hydrogels. Indeed, we expect interesting developments in CNT‐based hydrogels in the coming years. 4. 2 Mineral Based Minerals such as calcium, silicate, lithium, magnesium, phosphate, and zinc are important components in bone and cartilage. In fact, it is the mineral phase of bone and cartilage that gives them their incredible load‐bearing properties, which in turn enables the human skeletal system as a whole to withstand repeated mechanical stimuli directed at the body during daily routine activities. 315, 316, 317 For these reasons, mineral‐based nanomaterials made from clay materials (e. g. , montmorillonite and laponite), bioactive glasses, calcium phosphates, and calcium carbonates have been used extensively in recent years to make hydrogels more appealing for skeletal tissue engineering, providing a manifold increase in load‐bearing and osteogenic properties compared to pristine hydrogels. 93 Nevertheless, most of these hydrogels cannot resist the cyclic in vivo biological forces in skeletal tissues for long periods; instead, they quickly rupture and disperse into the body. Therefore, the design and development of mineral‐based hydrogels that can self‐heal within the load‐bearing microenvironment of skeletal tissues are needed to transform these systems into truly “master‐healers” of damaged skeletal tissues. In this section, we will review various self‐healable systems with a specific focus directed towards clay‐based nanoreinforcments, as they, in our opinion, constitute the key to overcoming the current challenges in bone tissue engineering. 4. 2. 1 Nanoclays Silicate is the principal component of nanoclays, but nanoclays are much more than just nanosized silicate, as they can contain traces of important skeletal minerals such as magnesium, calcium, zinc, and lithium. 93 Nanoclays are also potent mechanical nanoreinforcers by virtue of their ultrathin (≈1 nm) and high‐aspect ratio geometry (up to ≈1000), and like their carbon‐based counterparts, they also display a high‐surface area consisting of many easy‐to‐modify OH‐groups, which enables a multitude of possible self‐healable crosslinks with the polymeric backbone of hydrogels. These properties make nanoclays ideal for use as multifunctional and self‐healable agents for skeletal tissue engineering. In recent years, several studies have demonstrated that the above‐mentioned feats of nanoclays are realistic in real‐world applications. 93, 254, 318 Most of these studies have been centered on either laponite or montmorillonite; two of the most well‐renowned clays in this field. Therefore, in this section, we will focus on laponite‐ and montmorillonite‐based self‐healable hydrogel systems. Laponite is perhaps one of the most widely used nanoclays in skeletal tissue engineering since several studies have demonstrated its ability to turn bone‐marrow‐derived stem cells into bone cells. 8, 93, 259, 319 In brief, laponite is a silicate‐based nanoplatelet with a diameter of 25 nm and a thickness of 1 nm. 93 It contains traces of lithium, magnesium and natrium and displays a negative charge, which makes it easy to disperse in water at low concentrations. In general, the working principle behind laponite‐based self‐healable hydrogel is a result of ionic interactions between negatively (at the face) or positively (at the rim) charged sections of laponite with charged sections present within the backbones of various polymeric systems. 64, 320, 321, 322, 323, 324, 325, 326, 327, 328 In two groundbreaking studies, dendritic molecules consisting of amine‐terminated end groups (G3‐binder) were mixed with laponite to yield a moldable, freestanding, and self‐healable hydrogel; possibly because of the reversible ionic interactions between the negatively charged laponite and the positively charge amine groups on the dendritic molecules ( Figure 12 ). 64, 325 This hydrogel was injectable, could heal in less than 1 h with a mechanical recovery close to 100%, retain almost 98% water, and was able to memorize its original shape after a drying and rehydration process. This combination of injectability and shape memory properties makes it well‐suited for various injectable tissue engineering strategies. In another study, laponite was mixed with a carboxybetaine methacrylamide (CBMAA‐3)–2, hydroxyethyl methacrylate (HEMA) polymer to yield a highly hydrated hydrogel, which could extend to 1800% of its original length before breaking, and rapidly heal fractures (within 5 min). 321 A major advantage of these laponite‐based self‐healable systems is their simple working principle combined with their facile manufacturing process. This important attribute does not only make them readily injectable into the body, but also makes it much easier to bring them beyond the FDA approval phase and into the clinic. Figure 12 An injectable and nanoreinforced hydrogel with self‐healing properties based on electrostatic interactions between laponite and a G3 binder. a) The chemistry behind the self‐healable hydrogel. b) Photographic images of the hydrogel crosslinking process. c) The shear‐thinning and self‐healing properties of the resulting hydrogel were examined through rheology. d) Photographic images showing the self‐healing and shape memory properties of the laponite‐based hydrogel. Reproduced with permission. 64 Copyright 2012, Macmillan Publishers Ltd. Compared to laponite, montmorillonite is a less‐studied nanoclay in the field of skeletal tissue engineering, which makes the incorporation of montmorillonite into existing skeletal regenerative strategies an approach that is ripe for investigations. In the following paragraphs, we will highlight some of these promising systems with a special emphasis directed towards their self‐healing capacity. In simple terms, montmorillonite can be described as a negatively charge silicate clay with a thickness of 1 nm and a length that can range from 100–1000 nm. 93 This makes montmorillonite a much higher‐aspect ratio nanomaterial than laponite and thus a better mechanical reinforcer. Similar to laponite, montmorillonite also contains traces of magnesium and natrium; however, unlike lithium, montmorillonite also contains significant amounts of aluminum. The self‐healing mechanism in most of the reported montmorillonite‐based systems (like the laponite‐based systems) is governed by interactions between positively charged NH2‐groups on the polymer backbone within the hydrogel matrix and the many OH‐groups present on montmorillonite. 78, 329, 330, 331 Gao et al. recently published an elegant forerunner to montmorillonite‐based systems by incorporating montmorillonite into a covalently crosslinked polyacrylamide hydrogel. In this study, the NH2‐groups present on polyacrylamide were able to physically interact with montmorillonite because of a combination of hydrogen bonds and possible interactions between the oppositely charged NH2 and OH‐groups, to yield a double‐network hydrogel. Indeed, because of its double‐bonded crosslinking nature, this hydrogel could stretch to 11800% its original length before breaking, and could easily twist into various complex shapes. Nevertheless, the self‐healing efficiency of this hydrogel was relatively poor; it took almost 3 d for the system to self‐heal. In another recent study, 331 a similar approach was used to develop a system with a much better self‐healing property. In brief, the authors in this study choose to use a much smaller aminated polymer than polyacrylamide, namely a poly(amidoamine) dendrimer, which, because of its low molecular weight and higher mobility, was able to reconnect broken links much faster. Specifically, it was found that it only took 400 s for a broken hydrogel to reach its original shear modulus. In conclusion, despite their great promise as self‐healing nanomaterials for tissue engineering hydrogels, the nanoclay‐based hydrogels have not yet reached a mature stage. Especially, the montmorillonite‐based systems, despite their amazing stretching properties, require further improvement to yield a self‐healing efficiency that matches their carbon‐based counterparts. In the author's opinion, this is for the most part linked to the large size of montmorillonite, which significantly increases the diffusion barrier within the hydrogel matrix and, therefore, puts severe restraints on the polymer mobility. However, the studies highlighted in this section have shown sufficient promise, which once fully addressed could result in exiting self‐healable hydrogels for skeletal tissue engineering. 4. 2. 2 Others Other self‐healable hydrogels embedded with mineral‐based nanomaterials, such as silicate, 332, 333, 334, 335 hydroxyapatite, 273 magnesium silicate, 271 and calcium carbonate 86 nanoparticles, have also been developed, with promising results. Of these, self‐healable systems made from silicate and calcium carbonate nanoparticles have shown the most promise. The skeletal system of the human body utilizes some minerals to give it the load‐bearing support it needs during various day‐to‐day activities. For instance, minerals such as calcium carbonate and calcium phosphate have a role in the mechanical properties of bone, adding strength to and hardening the soft phase. In a recent study by Sun et al. 86 calcium carbonate nanoparticles were implemented within a polyacrylic acid hydrogel to yield a biomineral‐like material with self‐healing capacity ( Figure 13 ). This system was made through a simple mixing of CaCl 2, Na 2 CO 3, and polyacrylic acid in water followed by vigorous stirring. Scanning electron microscopy of the freeze‐dried hydrogel clearly showed the presence of nanosized calcium carbonate nanoparticles that were crosslinked with the hydrogel backbone. The resulting hydrogel exhibited shape memory properties, was moldable, injectable stretchable and could self‐heal within 5 s. The authors speculated that the driving mechanism behind the listed properties was reversible ionic interactions between Ca 2+ and the negatively charged polyacrylic acid backbone. Especially, the combination of injectability, shape memory property, and rapid self‐healing makes this biomineral‐like hydrogel promising for bone tissue engineering applications, as it can readily be injected into the target size to form a stable and predetermined mineral structure. Figure 13 A self‐healable polyacrylic acid (PAA) hydrogel based on electrostatic interactions between PAA and amorphous calcium carbonate (ACC) nanoparticles. a) The chemistry behind the self‐healable hydrogel. Photographic images demonstrating the b) moldability, c) stretchability, and d) self‐healing properties of the hydrogel. e) The shear‐thinning and moldability of the hydrogel were measured through rheology and are displayed here. Adapted with permission. 86 Copyright 2016, Wiley‐VCH. Behind nanoclays, silicate nanoparticles are the second most frequently incorporated mineral‐based nanomaterial within self‐healable hydrogels. 332, 333, 335, 336 Like the nanoclays, silicate nanoparticles also offer many modifiable OH‐groups, and therefore enable a multitude of self‐healing schemes. For example, in a recent study a polymer consisting of polyacrylamide and small traces of stearyl methacrylate was covalently grafted to silica nanoparticles to form a self‐healable hydrogel that could retain up to 90% water. 333 In addition to strong covalent bonds, this system also consisted of hydrogen bonds mediated by the NH2‐groups of polyacrylamide and hydrophobic interactions among the stearyl methacrylate groups. The system was thus essentially a triple‐bonded hydrogel, which was a contributing factor to its mechanical properties. This hydrogel could stretch to 2830% of its original length before breaking and displayed an ultimate tensile strength of 256 kPa. Moreover, the hydrogel was able to heal itself with an efficiency of almost 70% after it was left broken at 60 °C for 24 h. The inclusion of silicate nanoparticles was an essential factor in reaching the above‐mentioned mechanical and self‐healing properties, and the authors speculated that this was caused by the ability of the silicate nanoparticles to delay the onset of crack propagations within its matrix by redistributing the applied mechanical stress. In another study, positively charged poly(2‐dimethylaminoethyl methacrylate) was grafted onto silicate nanoparticles and then mixed with the highly anionic poly(acrylic acid) to yield a hydrogel that was crosslinked through electrostatic interactions between its oppositely charged polymers. 334 Notably, the silicate nanoparticles enabled this hydrogel to effectively dissipate accumulated energy during the extension phase of the hydrogel, enabling it to reach 2000% strains before breaking. Moreover, due to the many reversible electrostatic interactions, this system could self‐heal with an efficiency ranging from 80% to 100% after 12 h of healing time. Besides its excellent stretchability and reasonable healing‐efficiency, the as‐prepared hydrogels also displayed shape memory behavior and were therefore highly suited as a hydrogel carrier for stem cells targeted against skeletal tissue disorders. 4. 3 Magnetic Nanomaterials In recent years, magnetic nanomaterials have been incorporated into tissue engineering hydrogels to make them responsive to magnetic fields and thus easier to control from outside the body using externally applied electromagnetic fields. 337, 338, 339 This marriage between magnetic nanomaterials and tissue engineering hydrogels can, in the long run, yield stimuli‐responsive hydrogels that can actuate and heal on command within the human body. The working principle behind these systems relies on reversible links between the magnetic nanomaterial and the hydrogel backbone. However, such intelligent tissue engineering systems are still in their infancy, with only a few published studies. Most of these studies have focused on the incorporation of iron nanoparticles into hydrogels and have utilized catechol–iron coordination bonds to endow self‐healing properties to these systems. For instance, a mixture of magnetic iron nanoparticles and a four‐arm terminated polyethylene glycol (4cPEG) polymer enabled spontaneous hydrogel cross‐linking through metal coordination interactions between the polymer chains and iron nanoparticles. 338 This hydrogel was magnetic, self‐healable, biocompatible, and stretchable (up to 1000% strain values). Despite the many interesting properties of the iron‐4cPEG system, its magnetic properties were, in the author's opinion, not fully utilized. It would be interesting to examine the self‐healing time and its efficiency as a function of an externally applied magnetic field. Another hydrogel with both self‐healing and magnetic properties was also recently manufactured from the combination of chitosan and negatively charged iron‐coated graphene oxide (FeGO) nanomaterials. 339 These films demonstrated self‐healing ability due to electrostatic interactions between positive amine groups on chitosan and the negatively charged FeGO nanomaterials. Specifically, the incorporation of FeGO into chitosan had a direct effect on the hydrophobicity of the resulting hydrogel, as well as its mechanical and magnetic properties. Interestingly, the FeGo–chitosan system also displayed some noteworthy antibacterial properties, which could be utilized to develop hydrogels that can regenerate chronic wounds, while at the same time reducing the probability of wound infection and possible foreign body responses. However, in our opinion, the authors of this study did not fully tap into the exciting magnetic properties of their system. Indeed, the area of magnetic and self‐healable tissue engineering hydrogels presents a host of yet unexplored scientific possibilities, which once fully harnessed might push the field of tissue engineering to new exciting highs. 4. 4 Outlook and Future Opportunities Attributes such as high mechanical strength, conductivity, and osteoconductivity have made nanomaterials suitable for various tissue engineering applications. In recent years, they have also been incorporated into soft matrixes to yield multifunctional hydrogels with self‐repair capacity. Even though some of these systems could deliver the promise of hydrogels that are both multifunctional and able to rapidly self‐heal, some of the nanomaterials, such as montmorillonite and graphene, significantly reduced the self‐healing efficiency to the point that made them almost useless in the clinic. We noted that this was most likely caused by their high‐aspect ratio, which in turn resulted in the formation of nanomaterial‐based barriers that prevented efficient polymer diffusion. One possibility to address this problem is to incorporate the nanomaterials into the hydrogel matrix in a more orderly fashion. For instance, the use of a honeycomb‐like nanomaterial network within the matrix—instead of a random network—could result in a significant mechanical enhancement, together with improved mass transfer through the many pores within the honeycomb network. In fact, some recent studies have shown that this is easy to achieve, as such networks can be generated within hydrogels by simply freeze‐drying the nanomaterials, 340, 341 and then immersing them into a precursor hydrogel solution that is subsequently crosslinked into a solid construct. We also notice that the combination of a magnetic field and carbon nanotubes could yield long carbon nanotube treads 342 that could easily be incorporated within various self‐healable hydrogels to exhibit truly outstanding tissue engineering materials both from a mechanical and electrical point‐of‐view. However, despite the great promise of nanoreinforced and self‐healable hydrogels, their performance inside the body is still not fully elucidated. The major concern here is related to possible cytotoxic and immune responses in a postimplantation scenario from the carbon‐based nanomaterials, as they are nondegradable (in contrast to the mineral‐based nanomaterials), and therefore might elicit unwanted in vivo responses. 93 Nevertheless, there are several roadmaps that could potentially reduce such outcomes by making the carbon‐based nanomaterials easier to degrade by the human body through various chemical functionalization strategies. 93 As the field advances, we anticipate that this current gap between the laboratory and the clinic will be further minimized to enable the patients to fully benefit from the combined healing powers of self‐healable hydrogels and nanomaterials. 5 Tissue Engineering Applications Besides, biological and structural factors, the native microenvironment of tissues within the body relies on mechanical (i. e. , bone, muscle) and electrical cues (i. e. , cardiac, nerve). Therefore, tissue engineering hydrogels, aiming to integrate with native organs also need to be durable from a mechanical point‐of‐view and (in some situations) also electroactive. In this regard, major efforts have been directed toward engineering of hydrogels that are simultaneously mechanically though, self‐healable and eletroactive. In this section, we will review some of the recent progress in these areas with special emphasis on musculoskeletal, cardiac, and neural tissue engineering. 5. 1 Bone Bone is the second most transplanted tissue in the world, which underscores the immense need for off‐the‐shelf bone grafts. 343 Although the transplantation of fresh autologous bone is the current gold standard in the clinic, this option is limited due to the significant donor site morbidity associated with removal and reinsertion of the patients own bone tissue inside the target defect site. 344 While allografts offer some exciting advantages, they are also associated with significant flaws related to possible foreign body responses and graft rejection. 154, 343, 345 Consequently, bone tissue engineering has emerged as a promising alternative to auto and allografts, wherein stem cells, scaffolds, and biological factors are harnessed to create a native‐like microenvironment capable of facilitating the formation of new bone tissue. 346, 347 One of the key requirements of bone tissue engineering is the design and development of tissue engineering scaffolds that recapitulate important characteristics of bone, including its mechanical strength, durability, and self‐healing capacity. 348, 349, 350 Correspondingly, recent efforts have been redirected toward engineering of self‐healable scaffolds that can cope with the load‐bearing conditions in the native bone, while facilitating new tissue ingrowth. Thus far, a number of studies have explored the enormous potential of self‐healable hydrogels in this field, utilizing prepolymers such as HA, 182, 351, 352 (PEG), 272, 353, 354 elastin‐like polypeptides (ELP), 355 chondroitin sulfate, 356 protein–DNA complexes, 357 and silk fibroin. 358 These hydrogels have mostly relied on crosslinks based on thiol‐ene click chemistry, 353 supramolecular interactions (involving DNA blocks), 357 electrical interactions, 351, 352, 358 guest–host interactions, 182, 197 imine bonds, 355 acylhydrazone bonds, and Diels–Alder reactions. 356 The most frequently used mechanism to form self‐healable hydrogels for bone tissue engineering is based on electrostatic attractions; wherein various ions are interacting with oppositely charged ligands on polymeric chains. 351, 352, 358 For instance, in a recent study, HA macromolecules modified with bisphosphonate (BP) groups were shown to be able to bind reversibly to Ca 2+ ions that were conjugated onto silk microfibers (mSF) through a calcium phosphate (CaP) coating ( Figure 14 ). 358 Due to the presence of reversible electrostatic interactions, this mixture resulted in an injectable and self‐healable hydrogel. However, the silk‐based hydrogel demonstrated poor mechanical properties and insufficient stability in physiological condition as it quickly degraded (5 h) in PBS. To circumvent this drawback, the authors transformed this system into a double‐bonded one by adding a UV cross‐linkable HA–BP–acrylamide (Am–HA–BP) prepolymer to the system. This double‐network hydrogel showed significantly higher stability and an increase of ≈1500% in storage modulus. Moreover, the in vitro viability of encapsulated hMSCs was high, while the hydrogel itself showed some exciting osteogenic properties. Most importantly, results from an in vivo study in mouse revealed a significant acceleration in bone regeneration rate in a cranial size defect model, as the bone formation rate was 220% faster than in untreated groups. 358 In another paramount study, electrostatic interactions were used to generate an injectable and self‐healable hydrogel by mixing HA‐BP and acrylated BP (Ac‐BP) with magnesium chloride (MgCl 2 ) ( Figure 15 ). 352 Interestingly, the shear thinning property, compressibility, and stress relaxation profile of this hydrogel allowed it to be injected and rapidly fit within custom‐made defect sites. Finally, encapsulation studies with hMSCs in these nanocomposite hydrogels demonstrated their capacity to induce cell spreading and osteogenesis, as evident from the expression of important osteogenic markers. Figure 14 A silk‐based self‐healable hydrogel for bone tissue engineering. a) The chemistry behind the self‐healable hydrogel. b) Photographic images showing the self‐healing properties of the hydrogel. c) The hMSCs spreading within the hydrogels was examined through phalloidin (red) and DAPI (Blue) staining of the cell cytoskeleton and nucleus, respectively. The relative gene‐expression of important bone markers from hMSCs encapsulated within hydrogels is also displayed in this panel. d) The self‐healable hydrogel could promote significant bone formation within rat cranial defect site. Adapted with permission. 358 Copyright 2017, Wiley‐VCH. Figure 15 An HA‐based self‐healable hydrogel for bone tissue engineering. a) The chemistry behind the self‐healable hydrogel and its b) self‐healing and c) modability properties. d) The relative gene‐expression of important bone markers from hMSCs encapsulated within hydrogels is also displayed in this panel. Adapted with permission. 352 Copyright 2017, Wiley‐VCH. Perhaps the most innovative self‐healable hydrogel for bone tissue engineering is a recently developed protein–DNA hydrogel that can crosslink via DNA hybridization. In this work, a polypeptide copolymer from chemically modified human serum albumin was combined with a rationally designed DNA prepolymer, which could undergo rapid gelation via the addition of a complimentary multiarm DNA cross‐linker. One of the unique features of this protein–DNA hydrogel is its capability to readily assemble bioactive molecules within its structure using properly functionalized DNA adaptors and subsequently release them on demand via the use of a DNA‐cleaving enzyme (DNase). The authors utilized this unique property to custom‐engineer a hydrogel that could remedy the bone resorption caused by osteoclasts in osteoporotic patients through the controlled release of an osteoclast inhibitor. Specifically, the authors reported a significant decrease (≈96%) in the expression of important osteoclast markers, as well as the resorption activity of osteoclast cells. Furthermore, the authors found that the protein–DNA loaded hydrogels did not interfere with the metabolic activity, proliferation or osteogenesis of osteoblast cells. 357 In our opinion, this protein–DNA hydrogel holds great promise as an agent with the capacity to locally enhance the bone quality in the osteoporotic regions of bone through targeted delivery of bone forming hMSCs and targeted inhibition of osteoclast cell activity. 5. 2 Cartilage Articular cartilage lesions are among the most prevalent injuries in the general population and typically caused by trauma, osteoarthritis, or sport‐related injuries. 359, 360, 361 Unfortunately, the self‐repair of damaged cartilage is problematic due to the lack of vascularization and slow turnover of its ECM. 362, 363 One avenue for remedying the current situation is through the development of tissue engineering hydrogels to create a biomimetic microenvironment that promotes cell migration, proliferation, and neovascularization. Other properties, such as mechanical strength, engraftment stability, and elasticity, can be further integrated into these hydrogel systems to make them stronger and enable them to better withstand the load‐bearing forces within native cartilage tissue. We envision that tissue engineering hydrogels with sufficient mechanical strength and self‐repair ability can add a much‐needed dimension to the field of cartilage tissue engineering through the manufacture of durable grafts that match the dynamic and load‐bearing microenvironment of native cartilage. This urgent need for self‐healable cartilage grafts was addressed by Yu et al. 180 through the development of a novel class of biodegradable, self‐healable, and biocompatible hydrogels based on the integration of dynamic acylhydrazone covalent bonds into an aldehyde‐rich HA‐based hydrogel; as described in detail in Section 3. 1. 3. Specifically, the aldehyde groups contained in this hydrogel enabled it to adhere to cartilage; as aldehyde groups can bind to the amine groups of local cartilage tissue via Schiff‐base reactions. The adhesive strength between the hydrogel and cartilage was further probed by retrieving native cartilage tissue from porcine, carving a defect into the tissue, pouring the hydrogel into the defect‐site, and performing a push‐out test on the entire composite system. The resistance experienced by the push‐out probe provides a qualitative measure of the adhesiveness between the cartilage and the hydrogel. From these tests, the authors reported an adhesive strength of approximately 10 kPA, which was almost 10‐fold larger than a nonadhesive control hydrogel. In another study, dextran was functionalized with Upy to enable the generation of a hydrogel made from reversible hydrogen bonds between adjacent UPy moieties ( Figure 16 ). 364 To this end, the strong quadruple hydrogen bonding capacity of the UPy moieties endowed the hydrogel with both self‐healing and shear‐thinning properties. The authors emphasized that these properties were strongly affected by the concentration of Upy in the backbone of dextran. The self‐healing capacity of this system was utilized by the authors to merge a pro‐osteogenic hydrogel with a prochondrogenic hydrogel to generate an artificial cartilage–bone interface. This interface was later injected subcutaneously into mice to demonstrate that the construct could mature into an osteochondral‐like graft for furtherer down‐stream tissue engineering applications. Indeed, this concept could open a new paradigm to regenerate arthritis‐infected knees, as one of the devastating consequences of this debilitating disease is the gradual breakdown of the bone‐cartilage (osteochondral) interface. Figure 16 A self‐healable hydrogel for osteochondral engineering made from dextran (DEX) and UPy. a, b) The chemistry behind the self‐healable hydrogel and governing principle behind the formation of osteochondral constructs. c, d) Photographic images showing the hydrogel crosslinking and its self‐healable properties. e) The self‐healing properties were additionally examined through rheology. f) The bone–cartilage interface was stained for bone (Alizarin Red S) and cartilage (Alcian Blue). g) The volume of bone and cartilage after subcutaneous implantation that lasted for 8 weeks. Adapted with permission. 364 Copyright 2017, Wiley‐VCH. As mentioned in Section 4, the combination of nanomaterials and self‐healable hydrogels has gained substantial interest in the field due to a host of highly sought‐after properties, such as elasticity, adhesiveness, and bioactivity. Nanoreinforced and self‐healable hydrogels are, therefore, an ideal choice for cartilage tissue engineering, as they can withstand and adapt to the mechanical resilience of native cartilage while also enabling tissue ingrowth. To this end, a recent study demonstrated that incorporation of reversible bonds between calcium phosphate nanoparticles and bisphosphonate‐functionalized hyaluronic acid could facilitate the formation of a highly robust hydrogel with injectable and self‐healing properties. Furthermore, the authors injected the hydrogel into the knee of rats to evaluate its in vivo performance. 365 The grafts could withstand the highly dynamic environment of the knee for up to 4 weeks and also exhibited some bone formation after 1 week. Although this system is opening new possibilities in cartilage and osteochondral tissue engineering, further efforts should be focused on exploring their stability and in vivo performance to bring this system to the clinic. 5. 3 Skin The organs of the human body are not all internal like bone, cartilage, muscle, brain, or the heart. There is one that we wear on the outside of our bodies to protect internal organs, namely skin, the largest organ in the human body. The skin is the body's first line of defense against invasion by pathogens, toxins, or injuries; however, due to its delicate nature, external forces experienced during daily activities easily damage it. Different approaches have been established so far to make the skin healing process faster for patients with acute or chronic wounds. Most of these approaches have focused on the transplantation of skin grafts, such as auto, allo, or xenografts. 366 Although these strategies have been used clinically and have resulted in promising outcomes for the patients, there are drawbacks associated with the limited number of donors and possible rejection of allo and xenografts. 367 To remedy this standstill, regenerative wound dressings have in recent years emerged to effectively treat damaged skin tissues. 368 To this end, different types of wound dressings have been proposed, such as membrane, 369 rubber, 370 foam, 371, 372 nanofiber, 373, 374 or hydrogel‐based dressings. 375 Among these different varieties, there is growing interest in using hydrogel‐based dressings owing to their unique material characteristics. These features include the ability to maintain a moist‐like environment around the wound site, absorb tissue exudates and a high oxygen permeability to facilitate optimal tissue regeneration. 376 Although hydrogels are ideal candidates for the regeneration of damaged skin, some major challenges still remains unsolved. These challenges mainly include an insufficient mechanical strength to cope with the cyclic movements of human skin, since most hydrogel‐based wound dressing are brittle and unable to heal like natural skin is able to. 377, 378 Therefore, researchers are currently moving toward a new strategy based on tough and self‐healable hydrogel dressings with the capacity to autonomously heal after damage. 377, 379, 380, 381, 382, 383, 384, 385, 386 Along these thoughts, Zhao et al. 379 recently developed an adhesive, conductive, and self‐healable wound dressing by mixing chitosan‐g‐polyaniline (QCSP) with benzaldehyde functionalized poly(ethylene glycol)‐ co ‐poly(glycerol sebacate) (PEGS‐FA). 379 Notably, this system could crosslink without any external stimuli at 37 °C and, thus, was also injectable. Its self‐healing properties stemmed primarily from dynamic covalent Schiff‐base links between amine groups from QCSP and benzaldehyde groups from PEGS‐FA. In addition to these properties, the positively charged amino groups from the QCSP backbone and the highly anionic nature of polyaniline made the system electrically conductivity. The authors of this study later used their wound dressing on a skin‐defect site in a mouse model to study the in vivo performance of their system (up to 15 d). From these studies, it was shown that use of an OCSP–PEGS–FA dressing leads to a higher expression of wound healing markers (EGF, TGF‐β, and VEGF) as compared to a commercial wound dressing (Tegaderm), which ultimately promoted an almost flawless wound healing process. The authors speculated that these properties were caused by the combination of electrical conductivity and free radical scavenging capacity of the hydrogel, as well as its self‐healing ability, which enabled it to remain intact in the body for 15 d without rupturing. Indeed, several recent studies have demonstrated an interesting link between skin tissue regeneration and electromagnetic fields. Therefore, the system developed by Zhao et al. fits within this exciting and emerging area in the field of skin tissue engineering. 387, 388, 389 In other studies, the investigators have relied on polydopamine to develop self‐healable hydrogel‐based wound dressings, as they display formidable adhesiveness to native tissues, excellent biocompatibility, and the ability to endow hydrogels with self‐healing properties due to their highly reactive catechol groups. Some recent studies have tapped into the above‐mentioned portfolio of properties that polydopamine brings by incorporating polydopamine nanoparticles into various hydrogels. 377, 380, 381 For instance, polydopamine nanoparticles were recently incorporated into a poly ( N ‐isopropylacrylamide) (PNIPAM) network to generate a promising hydrogel for wound healing applications. This hydrogel displayed exciting properties that included responsiveness to near‐infrared light, self‐healing ability, and tissue adhesiveness. 380 Specifically, the hydrogel could heal when it was exposed to near‐infrared irradiation and displayed a significantly higher elastic modulus and a greater adhesive strength to porcine skin as compared to pure PNIPAM. The hydrogel also promoted attachment and proliferation of fibroblasts cells, which might be useful for regenerating damaged skin tissues. The regenerative properties of the polydopamine–PNIPAM‐based hydrogel dressing were further tested in an in vivo skin‐defect model in mice, in which a complete wound closure was achieved after 15 d. 380 In other similar studies, the polydopamine chains were linked to a polyacrylamide network via interactions between free catechol groups and amino groups on polyacrylamide to yield a tough, adhesive, and self‐healable hydrogel‐based wound dressing. 377, 390 This hydrogel could self‐heal after 2 h in an ambient environment without the need of external stimuli, and was also conformable and could adhere to a human arm undergoing twisting and bending due to wrist movements. It could also promote fibroblast adhesion, proliferation, and migration into the wound area, and ultimately the generation of new skin tissue for the treatment of dermal wounds. In another recent study, it was shown that the combination of dopamine‐modified four‐armed PEG (4‐arm‐PEG‐DA) with phenylboronic acid functionalized four‐armed PEG (4‐arm‐PEG‐PBA) could yield an injectable hydrogel with a good adhesion to skin tissue and rapid self‐healing (30 s). 381 Especially, the presence of reversible and dynamic phenylborate ester bonds between 4‐arm‐PEG‐DA and 4‐arm‐PEG‐PBA endowed the hydrogels with sufficient self‐healing and mechanical properties and multiple stimuli‐responsive behaviors toward various external stimuli, such as pH, glucose, and dopamine. Most importantly, these hydrogels displayed good cytocompatibility and promoted fibroblast cell attachment. Owing to their rapid self‐healing and exciting adhesion properties, this type of wound dressing provides an exciting opportunity for various wound closure therapies. 5. 4 Cardiac Acute myocardial infarction is considering one of the leading causes of mortality worldwide. 391 Several biomaterial‐based strategies, including epicardial bioengineered patches, and injectable hydrogels, have been developed to prevent heart failure because of postmyocardial infarction. 392, 393, 394, 395, 396 Therapies based on injectable hydrogels are currently rated among the most attractive solutions to deliver therapeutic agents (e. g. , drugs or cells) to restore the damaged myocardium because these therapies tend to be highly regenerative while being minimally invasive. 397 Although hydrogels can retain the delivered cells within a confined region, the pulsatile movements of the heart inevitably lead to stresses that might cause damage and a premature loss of the delivered cellular materials. Self‐healable hydrogels might represent an advantage here; as such, carrier systems can rapidly heal even under high mechanical strains to enable the delivered cells to remain engrafted at the desired location for prolonged periods In this direction, Bastings and co‐workers have developed an injectable pH‐responsive hydrogel for myocardial drug delivery by using a PEG‐based hydrogel with pH‐sensitive UPy moieties. 398 The hydrogel is fluid at basic pH (allowing easy injection) but is reversibly transformed into a gel state at neutral pH (when it reaches the heart). Notably, when used as a carrier of growth factors (HGF and IGF‐1), this hydrogel was capable of significantly reducing the size of an infarct scar within a porcine myocardial infarction model as such growth factors enhanced the activation of resident regenerative cells to promote rapid cardiac tissue regeneration. 399 Another class of self‐healable hydrogels that has been investigated for myocardial repair are based on host–guest interactions between adamantane and beta‐cyclodextrin‐modified HA. 186, 187 The feasibility of using this hydrogel for cardiac therapy has been assessed using an ischemic rat model, in which the stiffness and retention capacity of the hydrogel was significantly enhanced relative to untreated hydrogel controls. 186 The myocardial function was also remarkably improved when the hydrogel was loaded with endothelial progenitor cells, as the cell‐laden hydrogels caused a significant increase in neovascularization compared to cells delivered alone. 187 Several recent studies have also shown that scaffold materials displaying electrical conductivity can positively influence the behavior of cardiac cells. 396, 400 Therefore, it could be desirable to use self‐healable hydrogels with electroconductive properties for myocardial repair. 401, 402 To this end, Dong and co‐workers have investigated a self‐healable and electrically active hydrogel that was fabricated by incorporation of aniline tetramer into a dibenzaldehyde‐terminated PEG polymer. 403 The material displays excellent biocompatibility, as revealed by the viable encapsulation of C2C12 myoblasts and cardiac myocytes, and in vivo cell retention of these cells after subcutaneous implantation. Also, the hydrogel exhibited a fast degradation profile, as subcutaneously injected grafts were completely resorbed over the course of 45 d without inducing a significant inflammatory reaction. Another electroconductive and self‐healable hydrogel for cardiac repair was recently made from the combination of chitosan and polydopamine coated GO nanomaterials. 404 This hydrogel could crosslink spontaneously by utilizing reversible interactions between catechol groups on adjacent polydopamine polymers and electrostatic interactions between GO and chitosan. Specifically, the GO nanomaterials enabled a simultaneous increase in both the electrical conductivity and mechanical stiffness of the composite due to its multifunctionality. The suitability of this material for cardiac tissue engineering was assessed in vitro using human embryonic stem cell‐derived cardiomyocytes, which exhibited a fast and spontaneous beating rate when combined with the GO‐based hydrogel. Although these proofs of concept studies demonstrate the huge potential of electrically active and self‐healable hydrogels for cardiac repair therapies, the in vivo behavior of this novel class of biomaterials remains to be explored. In particular, it would be relevant to investigate whether these materials could be used as vehicles for simultaneous delivery of active biomolecules and relevant progenitor cells that could efficiently relieve the consequences of acute myocardial infarction. 5. 5 Neural Physical insults on the central nervous system (CNS) from car accidents, falls or stab wounds can lead to serious traumatic tissue injuries, which can have devastating consequences for the patients, including lifelong disabilities or even death. 405 Recent progress in tissue engineering and regenerative medicine has uncovered new therapies for the treatment of such CNS injuries. These approaches involve the use of biomaterial scaffolds to provide a microenvironment that promotes the survival and differentiation of transplanted cells. 406 Hydrogel‐based scaffolds are ideal for such tissue engineering approaches to target the degenerated CNS as they first reduce the risks of cell damage from the excessive shear stress caused by the injection process, combined with the fact that they present a minimally invasive cell‐engraftment option for the patient. 407, 408 While conventional hydrogels require changes in pH, temperature, or ionic strength to induce a gel phase transition, self‐healable hydrogels are special in the sense that they enable in situ gelation without the need of environmental triggers. 231 Heilshorn and co‐workers have developed a two‐component hydrogel that exploits the hydrophobic association of peptide‐based domains consisting of hydrophobic amino acids. 409 This self‐healable biomaterial enabled efficient encapsulation and stable culture of neural‐like PC12 and murine neural stem cells. However, the weak mechanical properties of the hydrogel (shear moduli from 9 to 50 Pa) could hinder its application inside the body, as the shear stress values observed within the human CNS are many times larger. To obtain mechanical properties closer to that of neural tissues, hydrogels comprising amyloid nanofibrils have been suggested as a way to address the current biomaterial challenges in this field. 67 Notably, these protein‐based hydrogels could support attachment and growth of fibroblasts, neuroblastoma, and mesenchymal stem cells. Other injectable hydrogels that recapitulate the biomechanical properties of the CNS have been manufactured with polysaccharide‐based prepolymers. 40, 410, 411 In one of these approaches, Tseng and co‐workers generated a chitosan‐based self‐healable hydrogel with a stiffness of 1. 5 kPa—in the same range as the stiffness of neuronal tissues (0. 1–1 kPa)—and demonstrated that this system could lead to an enhanced differentiation of encapsulated murine neural progenitors ( Figure 17 ). 40 Notably, these authors used a zebrafish‐based neural injury model to confirm that only neural spheroids that are encapsulated into hydrogels of appropriate stiffness can significantly heal neural damages. 40, 410 In another approach, a self‐healable hydrogel made from dynamic imine bonds between chitosan and oxidized sodium alginate was developed to carry neural stem cells (NSCs) into the CNS (Figure 17 ). 411 In this study, it was shown that it is possible to finely adjust the stiffness of the hydrogel between 100 and 1000 Pa to precisely fine‐tune the growth and differentiation of rat NSCs. The feasibility of this system for CNS cell‐based therapy was also demonstrated by confirming a uniform cell distribution upon injection of the NSC‐loaded hydrogels into mice brains. Figure 17 An alginate‐based self‐healable hydrogel for neural tissue engineering. a) The chemistry behind the self‐healable hydrogel. b) Photographic images showing the shear‐thinning properties of the hydrogel. c) Photographic images showing the self‐healing properties of the hydrogel. d) The self‐healing properties of the hydrogel were furtherer quantified through rheology. e) Neural stem cells (NSCs) encapsulated within the hydrogels could from spheroids after only 3 d of culture. f) The expression of important neural markers from the encapsulated NSCs is shown here. Adapted with permission. 40 Copyright 2015, Wiley‐VCH. Self‐healable hydrogels with tunable electrical properties are also promising candidates for the engineering of neural tissues. As studies have shown, not only mechanical but also electrical cues from the microenvironment are crucial to support the development of neural cells. 412, 413 To this end, Hou and co‐workers have developed an electroconductive graphene‐based hydrogel with self‐healing capacity, which supported the growth of neural‐like PC12 cells. 414 Although this material appeared to be well‐suited as a substrate for neural growth, its gel‐phase transition occurred under nonphysiological conditions, and thus did not allow 3D encapsulation of cells. This therefore warrants further investigation on self‐healable and electroconductive hydrogels for neural tissue engineering purposes. In summary, the few seminal in vitro and in vivo studies have demonstrated that it is possible to adjust the structure and biophysical properties of self‐healable hydrogels to control survival, growth, and differentiation of neural cells. Such hydrogels could eventually offer a unique platform for developing injectable stem cell therapies to target the degenerated CNS. However, more work is needed to achieve the potential of these biomaterials as their clinical significance remains to be fully investigated. 5. 6 Others In addition to neural, cardiac, bone and cartilage, self‐healing hydrogels have also been utilized to regenerate vascularized, 211, 415 muscle, 113, 416 and gastric tissues. 58, 198, 417 For instance, a glucose‐sensitive self‐healing hydrogel based on PEGDA was used as a sacrificial material to generate vascularized constructs ( Figure 18 ). 211 In this concept, the hydrogel was generated by mixing PEGDA and dithiothreitol followed by the addition of a borax solution to form reversible boronate ester bonds between the individual polymer blocks. The authors speculated that the breaking and reforming of boronate ester bonds were responsible for the self‐healing and glucose‐sensitive properties of this hydrogel. Specifically, these authors fabricated vascularized constructs in a two‐step process, first, self‐healable and microfabricated hydrogels were embedded into a nonglucose‐sensitive fibrin hydrogel. Then, to generate branched tubular channels, the combined hydrogels were immersed in a glucose‐containing culture medium to remove the sacrificial hydrogel from the composite construct. Endothelial cells were then seeded within the channels to facilitate the formation of capillary‐like structures. After 2 weeks of cell culture, a successful generation of 450 µm capillary‐like structures was observed inside a self‐healable hydrogel‐based construct Figure 18 A PEG‐based self‐healable hydrogel for vascularized tissue engineering. a) The chemistry behind the self‐healable hydrogel. b) The self‐healing properties of the hydrogel as quantified through rheology. c) Schematics showing the preparation of the vascularized and self‐healable tissue constructs. d) Photographic images of the tissue engineered construct. e) Fluorescence imaging of the constructs demonstrating the formation of premature lumen‐like structures within neural‐like tissues. Adapted with permission. 421 Copyright 2017, Elsevier. Self‐healable hydrogel systems have also shown promise for the delivery of stem cells, proteins, drug, and signaling molecules to promote muscle tissue regeneration. 248, 416, 418, 419 As an example of this feasibility, Mulyasasmita et al. used a peptide/protein modified PEG‐based hydrogel to yield a shear‐thinning and self‐healable hydrogel that could deliver human induced pluripotent stem cell‐derived endothelial cells (hiPSC‐ECs) into ischemic muscle tissues in mice. 416 This approach resulted in much higher cell retention and better hydrogel graft stability, which ultimately facilitated muscle tissue regeneration within the mouse model. In another study, McKinnon et al. 113 developed a multi‐arm PEG‐based hydrogel that could crosslink via reversible covalent interactions between hydrazone and aldehyde groups. Notably, this hydrogel could be fine‐tuned to promote proliferation and spreading of C2C12 myoblasts into multinucleated myotubes, which is an important event towards the maturation of myoblast cells into native‐like muscle tissue. Finally, self‐healing hydrogels have been used as tissue adhesives to seal stomach wounds. 58, 417 In an enlightening study, Phadke et al. demonstrated that, by using a pH‐responsive self‐healable hydrogel [poly (acryloyl‐6‐aminocaproic acid) (PA6ACA)], it is possible to create a stable, elastic, and adhesive hydrogel that can thrive within the highly acidic environment of the stomach, and ultimately close stomach wounds. Specifically, the damaged PA6ACA hydrogel could heal within 2 s upon exposure to an acidic solution (pH ≤ 3) and was thus applicable as an injectable and durable hydrogel for regenerating stomach wounds. In another recent study, a self‐healable carboxymethyl cellulose‐based hydrogel was developed through dynamic ionic coordination interactions between Al 3+ ions and carboxylate groups (COO − ) of carboxymethyl cellulose for sealing of stomach wounds. 417 In vitro studies with fresh gastric tissue from a pig stomach have shown that this hydrogel could adhere well to the gastric mucosa without any external intervention to enable a potentially strong and long‐term sealing of stomach wounds in possible future down‐stream applications. 6 Emerging Directions and Future Trends With the recent advances in self‐healing systems, tissue engineers have the necessary tools to generate electrical conductive and self‐healable hydrogels. The field could therefore easily be expanded into the emerging fields of cyborganics, bioacutators, and injectable bioelectronic hydrogels with shape memory properties. In the following sections, we will give a brief description of these emerging trends and revisit self‐healable hydrogels with electronic properties since they are pivotal in the above‐mentioned applications. 6. 1 Electronic Hydrogels with Self‐Repair Properties The field of self‐healable and hydrogel‐based electronics is still in its infancy. However, the emergence of wearable devices, such as electronic skin, google glass, apple watch, and various healthcare monitors, has pushed the field to a new high. In brief, self‐healable electronics encompass electrical conductive materials or circuits that can rapidly self‐repair during wear and tear. When combined with hydrogels and tissues, these could potentially generate self‐healable cyborganics, bioactuators, and injectable soft‐robotic systems for various tissue engineering applications. As briefly mentioned in Section 4, the easiest way to develop hydrogel systems that are both self‐healable and electrically active at the same time is through nanoreinforcement with carbon‐based materials such as graphene or carbon nanotubes. Indeed, several recent studies have demonstrated that this pathway can lead to some exciting electronic devices. Examples include a range of hydrogels that were incorporated with carbon‐nanotubes, 314, 420 calcium carbonate nanoparticles, 421 clay nanoparticles, 252, 390 ferric ions, 87, 422, 423, 424 or graphene 47, 425 to yield self‐healable bioelectronics, bioactuators, and electronic skin (e‐skin) devices. For instance, in a recent study Lei et al. 421 embedded calcium carbonate nanoparticles into an alginate–polyacrylic acid hydrogel to generate an ionic conductor ( Figure 19 ). As briefly highlighted in Section 4. 2, such systems exhibit self‐repair properties due to the ionic interactions between Ca 2+ and the negatively charged chains of alginate and polyacrylic acid. The Ca 2+ ions also gave the system its ionic conductivity, and by integrating these ionic conductors with a dielectric layer it was possible to generate a self‐healable capacitor. As the capacitance is intimately linked with the area of the capacitor, any deformation experienced by such devices will immediately result in measurable capacitance changes. The device was therefore used for human motion detection and blood pressure measurements through area‐facilitated changes in the capacitance. Interestingly, because the ionic conductor itself was self‐healable this device also displayed self‐healing properties and mimicked many of the fascinating properties such as self‐repair, elasticity, and deformability of natural skin. In a similar study, poly(vinyl alcohol) was functionalized with N, N, N ‐trimethyl‐1‐(oxiran‐2‐yl)methanaminium chloride to yield a self‐healable hydrogel in the presence of borax via dynamic diol–borate ester bonds. 426 In addition, a KCL solution was incorporated into the hydrogel to generate a self‐healable electrolyte; likewise activated carbon and acetylene black was embedded within the hydrogel to yield self‐healable electrodes. The sandwiching of the self‐healable electrolyte between the electrodes therefore resulted in the formation of a self‐healable capacitor with the self‐healing mechanism being governed by diol–borate ester bonds. Amazingly, the device could retain up to 84. 8% of its capacitance after bending for up to 1000 times. The device was also able to completely heal after 5 min and sustain its current‐voltage profile after 15 cycles. It therefore has the potential to transform into a self‐healable and wearable e‐skin device by using the same method reported in the recent study by Lei et al. 421 Figure 19 A hydrogel‐based electronic skin with self‐healing properties. a) The chemistry behind the self‐healable hydrogel. b) The working principle behind the developed self‐healable capacitor. c) Photographic images demonstrating the moldability of the hydrogel. d) A photo of the fractured sensor, e) the healed hydrogel, f) and healed hydrogel sensor. g) The capacitance of the sensor before and after the healing process. h) Photographic images of the sensor attached to a finger that is undergoing bending, and the associated change in capacitance during finger movement. i) Real‐time capacitance measurements from a blood‐pressure sensor when the pressure is raised and lowered. Adapted with permission. 421 Copyright 2017, Wiley‐VCH. A series of milestone contributions in the field has also lead to polymer‐based lithium batteries 423, 427 and semiconductors 300, 428 with self‐repair capacity. Notably, in a recent study, carboxymethylcellulose and aligned CNT sheets were combined with one another to yield an electrically active self‐healable hydrogel. 427 The self‐healing mechanism was mediated by hydrogen bonds between the cellulose fibers and weak Van der Waal bonds among the CNTs. By loading this hydrogel with LiMn 2 O 4 and LiTi 2 (PO 4 ) 3, it was possible to generate a self‐healable and flexible lithium battery for wearable healthcare monitoring. Specifically, it was demonstrated that the device could retain almost 90% of its capacitance after 200 bending cycles, heal instantly, and maintain 92% of its mechanical properties after five cutting‐healing cycles. 427 Despite the significant progress in the field of self‐healable and hydrogel‐based electronics none of these systems really incorporated complicated electronic circuitries, which is a requirement for cyborganic, bioactuating, and soft‐robotic devices. One avenue around this bottleneck is through the microfabrication of healable circuits based on carbon nanotubes or liquid metals. These circuits can then become encapsulated within a hydrogel and conjugated to the hydrogel backbone through various healable links. In the following sections, we will look into such future applications and briefly discuss their implications for the field of tissue engineering. 6. 2 Cyborganics The term cyborg tissues/organs first emerged in 2012 in a series of news articles to define the possible merger between artificially grown tissues and inanimate nanomaterials. 429 Such cyborg organic constructs (cyborganics) utilize the recent technological advances in the field of materials science, chemistry, electronics, and tissue engineering to yield tissues that are half‐man and half‐synthetic. These can be divided into two levels based on their level of complexity; those made from simple reinforcement of artificially engineered tissues with nanomaterials; 93, 430, 431, 432, 433 while the “level twos” are mergers between artificial tissues and more complex matters such as electronic circuits, robotic systems, and intelligent materials. 434, 435, 436, 437, 438, 439 Being the most “easy‐ones‐to‐manufacture” the level ones have been rapidly picked up by the field to give rise to a series of milestone contributions. On the other hand, the “level twos” have not soared as high as the level ones and therefore needs much future attention to reach the same level of maturity. An important gateway into this almost “sci. fi‐like” future was opened with a number of earth shattering publications from Charles Liebers research group. 434, 437, 438, 439 Especially, one study has truly opened up the realm of real‐world applications for level two cyborganics. 439 In this study, 439 intricate silicon‐based nanocircuits were intertwined with living cardiac tissues to generate something “out‐of‐this‐world”; something that could potentially in situ monitor the irregular beating of dysfunctional cardiac cells, heart inflammation, and potentially dangerous microenvironmental pH changes. In the same spirit as the futuristic studies from Liebers group, we will in the following paragraphs proposed a feasible roadmap for the manufacture of self‐healable cyborganics based on carbon‐based nanomaterials. The key here is the possibility of reversible π–π interactions between neighboring CNT's or RGO sheets, which we anticipate will enable the formation of highly conductive and self‐healable links. For instance in a recent study it was shown that the freeze‐drying of reduced graphene could lead to complex conductive structures that were held together by reversible π–π interactions. 314, 341 Such high‐order structures could easily be embedded within self‐healable hydrogels for further down‐stream cell culturing to yield cyborganics with the capacity to spontaneously heal as natural tissues do. Another interesting development is liquid metal circuits as they are both easy to mold and self‐healable 440 Again, like the other ones we also anticipate that such circuits are readily embeddable within various self‐healable hydrogels. Self‐healable cyborganics will likely facilitate a breakthrough that could potentially reform society into a more versatile form, wherein individuals fully master their health through wireless monitoring. For instance, the expansion of the field into controllable and self‐healable cyborganics will enable mankind not only to transcend his own biology beyond its current limitations but also enable him to control when and how his body heals. As the field advances, we also anticipate a number of unanswered ethical questions, that needs to be carefully addressed before the gates into the realm of cyborganics can fully open as this field is per say a highly controversial one. 6. 3 Bioactuators Bioactuators are devices that use the combined powers of biology and materials science to generate life‐like systems that can move, crawl, swim, sense, and assist in important tissue functions. 441, 442 They have been explored in a wide range of applications such as various robotic systems in the industry, 442 tissue engineering 93 and smart drug delivery systems. 441 Bioactuators are typically made from a combination of locomotive cells derived from muscle and cardiac tissues, and conductive polymers that can contract and expand in an almost inexhaustible manner. As cells within locomotive tissues typically form anisotropic structures, one of the key principles behind cardiac and muscle‐based bioactuators is the alignment of cells into linear microstructures. Examples include, the usage of cardiac cells that were genetically engineered to respond to light and aligned within a stingray‐like elastomeric body to yield a stingray‐like creature capable of controlled locomotion within a light field; 443 and a bioactuator consisting of CNT's aligned within GelMA to generate a more native‐like electrical coupling between encapsulated cardiomyocytes to enable a more coherent bioactuation in physiological conditions. 444 Bioactuators possesses a great deal of potential and if used properly they can open up for new groundbreaking opportunities in the defense sector, industry, healthcare system, and maybe even in the private sphere; who knows the end of the road? One thing is certain, though: the opportunities as we speak seem limitless and the field has established itself as one with the capacity to create a series of landmark changes in society. Even still, in the author's opinion the best is yet to come in the form of self‐healable bioactuators with the ability to spontaneously heal in demanding scenarios such as in the battlefield, hazardous, and dangerous working environments and within the sometimes highly strenuous milieus of the body. Along these thoughts, a recent study from Stanford University showed that a self‐healable bioactutaor could be made from the combination of a Fe(III)‐2, 6‐pyridinedicarboxamide (PDCA) chelation complex and poly(dimethylsiloxane) (PDMS) polymer ( Figure 20 ). 423 The PDCA–PDMS system broke the elasticity record for elastomers by stretching to 10000% its original length without breaking. This system also displayed autonomous self‐healing ability, as it was able to almost completely self‐heal after 48 h at room temperature. In addition to the self‐repair ability and high stretchability, the PDCA–PDMS system displayed interesting electrical properties because of the iron ions, which enabled it to actuate with ease during electrical stimuli. The authors of this study speculated that the native‐like actuating properties of the PDCA–PDMS system could potentially be used to manufacture artificial muscles for various self‐healable soft robotic systems. In our opinion, the incorporation and alignment of muscle cells within the PDCA–elastomers could make its bioactuating properties more human‐like, making it more amenable to human–machine interface applications. Figure 20 A bioactuator with self‐healing properties. a, b) The self‐healing chemistry behind the generated device. c) Tensile measurements of the device demonstrated a stretchability that could reach 10 000%. d, e) Photographic images demonstrating the self‐healing properties of the device. f) The self‐healable device could actuate in response to an alternating electrical field. Adapted with permission. 423 Copyright 2012, Macmillan Publishers Ltd. Building on the exciting results from the Stanford study 423 it would be interesting to also incorporate complex electronic structures into the PDCA–elastomer to develop more sophisticated actuators with self‐repair properties. Indeed, several methods are already available for the incorporation of self‐healable liquid metal circuits into PDMS‐based elastomers. 440 The most frequently used one is based on the generation of microfluidic channels within PDMS followed by microfluidic injection of liquid metal. Due to the shear‐thinning properties of most liquid metals, the metal will readily flow into such microchannels and form well‐controlled electronic circuits for further downstream applications. 445 With regard to future research directions, special attention should be given to the merger of bioactuators and cyborganics ( Figure 21 ). In our opinion, such hybrid creatures can use the working principle behind cyborganics to become more powerful soft robots; as well as being healed from a distance through remote controlling. The prospect of such self‐healable android‐like robots is certainly of huge interest for governments worldwide, being potentially useful for militaries preparing for the challenges of the next century and potentially enabling humans to journey to hard to reach and inhabitable places on Earth. Figure 21 Emerging trends and future directions in the field. 6. 4 Electronic Tissue Engineering Hydrogels with Shape‐ Memory Properties In recent years, injectable hydrogels have emerged as a promising alternative to minimize the many risks and complications associated with surgical implantation of tissue engineering scaffolds. 446 Especially, hydrogels that can completely restore their geometric properties after needle injection through a small‐bore needle have garnered a huge interest in the field, as they enable delicate native‐like architectures to fully survive the injection phase. 447, 448 For electroactive tissues, it is also desirable to incorporate certain electronic circuits within the tissue engineering scaffold to both improve the performance of the scaffold and enable remote monitoring of what is going on in the target tissue. Of course, the many delicate features of such circuits will not survive the injection phase under normal circumstance and, therefore, need to encompass materials that can self‐restore within the body. The research group of Charles Lieber has pioneered the development of electronic circuits that can survive the injection phase into the body. 438, 449, 450 To this end, Charlies Lieber's group used flexible and mesh‐like electronics that can fold and unfold many times, while keeping their electrical properties intact. We believe that it would be interesting to incorporate such mesh‐like electronics into self‐healable hydrogels with shape memory properties to develop injectable cyborganics, cardiac patches, or artificial muscle actuators that have sufficient durability for proper performance within the load‐bearing and dynamic tissues of the body. 7 Conclusion Self‐healable hydrogels have been rapidly adopted by tissue engineers as a new class of biomaterials with the potential to push the field of tissue engineering to new heights. The union between hydrogel toughness and self‐healability, as well as the multifunctional properties of self‐healable and nanoreinforced hydrogels, have been thoroughly discussed in this review, as we believe these systems will reshape the field. Especially, the nanoreinforced hydrogels present a new avenue in the field that could potentially be exploited to yield self‐healable electronic hydrogels, cyborganics and soft biorobots. We believe that these systems encompass some exciting concepts that will push the field of biomedical engineering to a new high in the coming decades. Although the above‐mentioned hydrogels hold great promise, the design and development of such systems pose several challenges that need to be addressed to enable the regeneration of damaged tissues and the manufacture of cybernetic devices that are partly living and partly machine. To this end, it is important to control the biodegradability of the hydrogels to both enable tissue ingrowth and cell migration, but also, to enable the hydrogels to thrive in physiological environments. Otherwise, they might degrade too quickly and jeopardize the long‐term performance of the devices within the human body. The tissue engineering hydrogels also need to promote sufficient cell spreading, proliferation, and differentiation. Otherwise, their regenerative performance within the body would be significantly compromised. Also, many of the described systems in this review paper have not yet been tested in animals, and their biocompatibility in the human body remains an open question. We anticipate that the recently intensified synergy between the fields of medicine, physics, chemistry, nanoscience, biology, and mechanical engineering can address some of these issues. In our opinion, a combined effort from these cross‐disciplinary fields is needed to develop nontoxic (chemistry), mechanically strong (mechanical engineering), electronic (physics), and clinically relevant hydrogels. Such interdisciplinary collaborations will undoubtedly contribute to bridging the current gaps between the laboratory and the clinic. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201802033
| 2,019
|
Advanced Science
|
On‐Demand Coalescence and Splitting of Liquid Marbles and Their Bioapplications
|
Abstract Coalescence and splitting of liquid marbles (LMs) are critical for the mixture of precise amount precursors and removal of the wastes in the microliter range. Here, the coalescence and splitting of LMs are realized by a simple gravity‐driven impact method and the two processes are systematically investigated to obtain the optimal parameters. The formation, coalescence, and splitting of LMs can be realized on‐demand with a designed channel box. By selecting the functional channels on the device, gravity‐based fusion and splitting of LMs are performed to mix medium/drugs and remove spent culture medium in a precise manner, thus ensuring that the microenvironment of the cells is maintained under optimal conditions. The LM‐based 3D stem cell spheroids are demonstrated to possess an approximately threefold of cell viability compared with the conventional spheroid obtained from nonadhesive plates. Delivery of the cell spheroid to a hydrophilic surface results in the in situ respreading of cells and gradual formation of typical 2D cell morphology, which offers the possibility for such spheroid‐based stem cell delivery in regenerative medicine.
|
1 Introduction Liquid marbles (LMs) are mainly composed of liquid droplets that are encapsulated and stabilized by solid micro or nanoscale lyophobic particles at the spherical liquid–air interface. 1 They show extreme mobility, elasticity, and stability on both hydrophilic and hydrophobic surfaces. LM has attracted increasing attention in the past decade due to its potential applications in gas sensing, 2 adhesive materials, 3 microchemical reactors, 4, 5, 6, 7, 8, 9 water pollution indicator, 10 and so forth. 11, 12, 13, 14, 15 Many types of liquid, with reactants and precursors dissolved inside, can be encapsulated by the particle shell to obtain LMs with a wide range of volumes in microliter scale. 16, 17, 18, 19, 20 In contrast to the naked liquid droplets, the as‐obtained LMs can be deemed and manipulated as nonsticky materials, which dramatically decrease the surface free energy and boost the handleability of the tiny drops. Moreover, the LMs were demonstrated to possess a significantly longer lifetime than that of the naked droplets. 21 The lifetime is related to many aspects, like particles size, surface chemistry, and the external conditions. 21, 22, 23 The manipulation of single LMs, by the mechanical force, 24, 25 magnetic locomotion 26, 27, 28 and light‐driven motivation 29, 30 have been developed in the past decade. Benefiting from the single LM manipulation and liquid control, the coalescence of the well‐prepared LMs can mix the precursors in a precise amount and the outer particle shell, a protective screen, can avoid the inner liquid loss and protect the inner liquid from being contaminated by the external environment. The splitting process allows people to isolate the useless wastes after each stage of reactions, especially for the splitting of metabolites during the long‐time culture of cells. The coalescence and splitting of two or multiple LMs are the key factors to realize the extensive application in the precise and trace amount reactions in both stoichiometric chemistry and biomedical applications such as cell culture. To date, however, the interaction and coalescence/splitting of LMs are not yet be fully studied and remain in an early stage with few deeper diggings in both the methodology, theory, and applications. The coalescence of LMs was conducted mainly through a magnetic attraction for opening and closing of the particles shell of LM, or manual injection with a microsyringe to achieve the LM‐based reactions. 31, 32 In addition, acoustic levitation strategy has been applied to realize the manipulation, and the opening and closing of the levitated LMs could be achieved from variation of the sound intensity and frequency. 33, 34 However, such a method is irreversible, power consuming, and generally limited to light and small LMs. The sustained shaking of the LMs is inevitable, which may interfere the inner liquid reactions especially for the cell activities in the biomedical applications. Recently, researchers have applied the LM to culture cells in the 3D format by manually rolling the medium on a hydrophobic powder. 35, 36, 37, 38, 39 However, the proposed method applied a tedious manual culture process and the sustained supply of nutrients and medium during medium renewal was difficult to achieve, 40 which hampered the long‐time culture of the 3D cell spheroid. Here, we introduced a centrifugal force‐ and gravity‐based method to achieve rapid formation, coalescence, and splitting of the LMs for 3D culture of stem cells in vitro, drug/nutrient addition, and waste medium removal using a highly automated process to ensure that the cellular microenvironment of such 3D cell culture was maintained under optimal conditions for long‐term. An additional benefit of the method is the versatility for various kinds of LMs. The method was systematically investigated to obtain optimal parameters to realize both coalescence and splitting. The size of the cell spheroid generated from the system could be controlled within a wide range, from the micrometer to sub‐millimeter scales. The prepared cell spheroids showed highly compact cell‐to‐cell cross‐contacts and formed the in vitro tissue, which is more valuable for the simulation of a real in vivo tissue than the 2D cell monolayers. More importantly, the obtained stem cell spheroid (SCS) from the coalescence and splitting system (CSS) showed much higher cell viability than that of the traditional culture method, which is probably attributed to the high gas permeability of the LMs. The CSS shows potentials as a bioreactor for controlled long‐time in vitro tissue cultures used in regenerative medicine, drug screening, and tissue engineering. 2 Results and Discussion The SiO 2 microspheres were prepared by hydrolysis and condensation of TEOS, followed with grafting of a layer of trimethoxymethylsilane. The resultant powders are white and water‐repellent with a water contact angle (5 µL) of ≈152° ( Figure 1 a). When a water droplet was dropped onto the bed of the powder, the droplet would be wrapped with the particles while rolling around due to the tendency to minimize the surface free energy, forming the so‐called liquid marble. The resultant LM is white, semitransparent, and sphere‐like with the SiO 2 particles evenly distributed on the shell. The shell of millimeter‐sized LM is composed of spherical SiO 2 particles with a uniform size distribution of about 300 nm as illustrated by scanning electron microscopic (SEM) image and transmission electron microscopic (TEM) image in Figure 1 b, c. The chemistry of the particles was investigated by XRD diffractogram which suggests a broad peak located at 2θ = 22. 9° (Figure S1, Supporting Information). The characteristic amorphous peak is attributed to silica according to the JCPDS data with card No. 01‐086‐1561. The volume of the LM can be precisely controlled in a wide range, from 0. 5 to 100 µL. Figure S2 in the Supporting Information selectively gives the microscopic images of the LM with different liquid volumes. The diameter as well as the height of LMs increase with the increase of the inner liquid volume (Figure 1 j). When the LMs were put into a large volume culture dish with water, the LMs tended to self‐assemble to form specific patterns on water which may attribute to the capillary force, 41 as is shown in Figure 1 d–i. The process was spontaneous and the geometry of the LM patterns varied with the increase of the LM numbers. Owing to the semitransparency of the LMs, the LMs can be applied as the sensor for gas detection. As shown in Figure 1 k, LM with pH = 1 aqueous resazurin solution shows pink color whereas LM with pH = 14 aqueous resazurin solution shows a blue color. Figure 1 l gives the demo of the LMs with different liquid pH values forms “CUHK” logo. The LMs show high stability on various kinds of surfaces/interfaces, such as hydrophilic surfaces, hydrophobic surfaces, and even aqueous solution surfaces. The LMs could reside on the surfaces/interfaces for a long period until the inner liquid is totally evaporated to form a “deflated ball”. Figure 1 a) Optical images showing the silica particles and the contact angle of the sample. b) SEM image showing the surface morphology and particle size of the SiO 2 particles. c) TEM image showing the surface morphology and particle size of the SiO 2 particles. d–i) Self‐assembly of LMs (6 µL) on the water surface: d) one LM, e) two LMs, f) three LMs, g) four LMs, h) five LMs, and i) seven LMs. j) Relationship between the diameter, height, and the inner liquid volume of a LM. k) Optical image showing the colors of the LMs with resazurin aqueous solution at different pH values. l) Optical image showing several LMs with resazurin aqueous solution arrange to the word “CUHK”. The letter “C” and “H” were composed of pH = 14 ammonium hydroxide solution and the letter “U” and “K” were composed of pH = 1 hydrochloric acid solution. Taking the advantages of the gravity and centrifugal forces, we develop the automatic CSS ( Figure 2 ), via 3D printing of polylactic acid, to realize fabrication, coalescence, and splitting of LMs all‐in‐one by selecting the specific channels. The system includes, from bottom to up, the orbital shaker, core 3D printing device, detection sensor, spiral channel, and autosampler of the droplet. The top panel contains the central spiral groove required for the formation of LMs and the all‐round channels used for transporting and steering the LMs on the track. The side panels are functional channels required for the fusion process and the splitting process, respectively. Alternatively, they can be designed on‐demand. As illustrated in Figure 2, the CSS system could be used for in the cell growth related potential applications and the Function 1, 2, and 3 can be completed simply by selection of the channel from rotating the central spiral channel to match the transportation channels. The Function 1 is to realize the LMs autoformation via dropping of media with stem cells to hydrophobic powder at the central spiral channel and the centrifugal motion driven by the orbital shaker can make both the uniform coverage of powers on media and transportation of the as‐formed LMs to the destination. Compared with the conventional method to fabricate LMs by manual rotating on a dish, this method shows the high automation and uniformity of all the prepared LMs. The Function 2 is to coalesce the fresh‐formed LMs to make the bioreactions, such as nutrient addition, drug screening, coculture of different kinds of cells. Similar to the previous work, the gravity can be utilized to coalesce the LMs with the guidance of the channel. 42 The direct addition of nutrient/drugs with a needle injection can also realize the mixing of bioreactant. However, the manual method is difficult for the operation of the microliter level droplet. More important, the ultimate capacity of the LM is limited by the outer hydrophobic particles. The shell can only tolerate slight expansion originated from the external addition, otherwise, the surface defects will generate, resulting in break up. The amount of media by the needle injection is therefore extremely confined. As for the coalescence‐based mixture, a large amount of media/bioreactant can be mixed into the original LM since the hydrophobic powder on the two LMs are sufficient to enwrap the coalesced marble without any surface defects. The Function 3 is to separate the formed LMs to realize the waste media removal and further analysis of the waste media after a certain period of culture. As the 3D cell spheroid has been formed inside the LM, the gravity‐driven cutting of the LM will separate the LM into two counterparts and the 3D cell spheroid will stay inside one of the sub‐LMs. The fluorescence sensor can be installed for screening in the outcome channel to inspect which LM has the spheroid. The LM without SCS can be screened for further analysis and other disposals. The LM with spheroid in it can be used to do another coalescence process with a LM of fresh media via the system. Each functional panel has several channels which are used for the LMs with different sizes/volumes. Figure 2 Schematic of the CSS, an on‐demand LM manipulation system, toward its application in the 3D SCS culture. 3 Formation The 3D view and unfolded view of the device is shown in Figure 3 a. The central spiral groove is moveable so as to select the specific pathway for realizing on‐demand functions. LM formation was completed by simply dropping the liquid in the center of the spiral channel followed by turning on the orbital shaker, as shown in Figure 3 b. After the shaker was turned on with a final stable rotating speed of 180 rpm, the liquid droplet on the bed of the SiO 2 quickly formed an LM. The centrifugal force‐driven circular motion with a gradually increasing radius ensured both the uniform coverage of powders on the medium and transportation of the as‐formed LMs to the ultimate destination. Figure 3 c shows the photograph of a batch of as‐prepared LMs with uniform size and acceptable error range as indicated by the inset at the top right corner. Compared with the conventional method to fabricate LMs by manual rotating on a dish, this method shows the potential for mass production of LMs with high uniformity in an automatic manner. Figure 3 a) Schematic of the CSS with 3D view and unfolded view. b) Photograph showing the CSS with an orbital shaker underneath and the successive images of the LM formation process after a liquid droplet is added on the center of the spiral channel. c) Optical image of the as‐prepared LMs (10 µL). The inset shows the statistical average diameter of the as‐prepared LMs with high consistency. 4 Coalescence To coalesce the LMs, the liquid inside the LMs should overcome the energy barrier generated from the outer hydrophobic powder to get into contact with each other. The energy barrier between the LMs can hardly be overcome via normal compression because the shell of an LM is composed with multiple layers of silica nanoparticles. The hydrophobic layers of silica nanoparticles of two LMs prevent all of the formation of liquid bridges. (see Movie S1 in the Supporting Information). After liquid‐to‐liquid contact occurred, the coalescence of the LMs would be a spontaneous process since they always tried to reduce their total surface area so that the total surface free energy of the system would be reduced accordingly to realize a lowest energy state, i. e. , quasi‐spherical shape (Figure S3, Supporting Information). To investigate the coalescence caused by LMs impact, we first investigated the process of an LM impacting a solid substrate, which is also demonstrated by Marston and co‐workers. 43 The substrate we used is a hydrophilic glass slide. As shown in Figure S4a in the Supporting Information, the still frames of speedy movies depict the integrated process of LM (6 µL) impacting a glass slide from different heights (from 2 to 7 cm). The LMs show repetitive bouncing on the glass slide with gradually decreased jumping heights. With the increase of the falling height, the jumping heights after the first surface collision also increased. When the falling height is within the range of 5–6 cm, the LM would split into two sub‐LMs with the upper sub‐LM always larger than the lower one. While the falling height is no less than 7 cm, the LM would collapse as soon as the LM collide the surface (within several milliseconds). In addition to the integral view of the whole dynamic impact process, the first collision between the LM and the surface was taken into consideration by enlarged successive images as showed in Figure S4b, c in the Supporting Information, with both side and bottom views (falling height is 3 cm, LM is intact during the entire process). As shown in Figure S4b in the Supporting Information, the liquid droplet was evenly wrapped in white SiO 2 shell with multiple layers. After collision with the substrate, the LM showed a serious deformation and became pancake shaped within only 3 ms. The serious deformation and complicated flow inside the LM made a small number of the outer layer of hydrophobic particles without direct contact and trapping on the water–air interface be splashed out. The peeling off of the shell particles mainly occurred at the moments of either impact of LM with the substrate at the lowest pancake thickness at the border region of the LM (3 ms, Figure S4b, Supporting Information) or bounce process at the center of the LM (6. 5 ms, Figure S4b, Supporting Information). The major part of the hydrophobic particles on the shell would make themselves adapt the shape variation of the inner droplet and surface area increase of the droplet. As is known, a droplet shows its lowest surface area at the spherical shape. The deformation of the LM made the surface area of the LM increase, therefore producing the defects, especially when the LM deformed into a pancake shape, as showed by side view and bottom view in Figure S4b, c in the Supporting Information. The LM showed pancake shape at the maximum deformation of the impact process and the diameter of the pancake shape is measured to be 3. 8 mm, which is 1. 4 mm larger than that of the 6 µL LM at its spherical stage (60% expansion on diameter, the second image in Figure S4c in the Supporting Information). Figure S4c in the Supporting Information gives the bottom view of the impact process at its spreading stage, the pancake shape at the maximum deformation, and the retraction stage. The increase of the surface area at the pancake state causes the shortage of the hydrophobic particles to have a full coverage of the droplet, making rich surface defects and bare regions. 44 These surface defects may facilitate the coalescence process of the LMs during impact processes. The surface defects and bare regions are highly dependent on the inner flow field of the LMs. The simulation results of the impact between a droplet and hydrophilic glass were presented (Figure S4d, e, Supporting Information). This simulation is established using the phase field method in COMSOL Multiphysics to investigate the velocity field during the impact process, where a droplet and surrounding air are modeled as two separate phases. We use 2D axisymmetric computational domain due to the axisymmetry of the whole process. The droplet with a radius of 1 mm is initially positioned at a 5 cm distance above the substrate with zero initial velocity. The droplet freely falls downward under the influence of the gravity force and reaches the substrate at an impact velocity. Open boundary conditions are applied at both the top and side, simulating an infinite domain. A wetted wall boundary condition is used for the substrate with a contact angle of 152°. The successive images in Figure S4d, e in the Supporting Information show the inner flow field variation tendency during the process of drop‐solid contact to pancake shape. The inner velocity field near the spherical edge shows increased intensity and finally reaches its maximum value at the pancake shape as the time elapse (Figure S4e, Supporting Information). Apart from the sharply increased surface area, the nonuniform flow field inside the droplet may also contribute to the generation of the surface defects. To overcome the energy barrier caused by the outer layer of hydrophobic particles, we use the gravity‐induced collision to overcome the energy barrier and coalesce the LMs. As shown in Figure 4 a, one of the LM is resided on a glass slide and the other LM is released at a certain height right above the bottom one. When the height is too small (height = 1 cm, Figure S5, Supporting Information), the collision between the LMs could not produce sufficient surface defects to make the liquid‐to‐liquid contact occur. Then the LMs would be bounced away and the coalescence effect did not occur. When the height increased to a certain height, the coalescence behavior would occur. Figure 4 d gives the LM impact at a height of 5 cm, both the top and bottom LMs were squashed by the collision and became pancake shaped at a maximum deformation (5. 2 ms) as boxed by the blue line. The two LMs coalesced during the process and bounced off the substrate within 20. 8 ms. As is shown in Figure S4 in the Supporting Information, the impact of LM could make the surface particles splash and the rearrangement of the surface particles makes the distribution of the particles uneven. The circinate sparse areas of the particles should appear on the surface of LMs and these areas were breakthrough for the instantaneous coalescence. As the height of the upper LM increased, the coalescence effect could not sustain due to the overlarge impact force, forming high‐adhesion wetting (Figure S5, Supporting Information). During the process of LM collision, the maximum deformation is an important aspect to measure the sustaining tolerance ability of the LM under collision. Here we use the high‐speed camera to record the entire process of LM collision and especially the thickness of the pancake shape at a maximum deformation. As plotted in Figure 4 b, the curve shows that the thickness of the pancake at the maximum deformation decreases with the increase of the falling height. When the height was larger than 5 cm, the LMs would collapse and the thickness of the pancake is recorded as 0. When the height was in the range of about 3–5 cm (safety height without high‐adhesion wetting for the LM‐substrate impact), the coalescence of LMs would occur efficiently. LM with a height smaller than 3 cm was insufficient to make a coalescence. Consequently, the height of the falling should be within a certain range to obtain an efficient coalescence effect. The impact forces between the LMs at each falling height were estimated at each height as plotted in Figure 4 c. Figure 4 a) Schematic of the coalescence and reorganization process via collision. b) The relationship between the thickness of the pancake shape at the maximum deformation and the falling height of the upper LM. The inset shows the pancake shape at the maximum deformation. c) The relationship between the impact force between LMs and the falling height of the upper LM. d) Successive images showing the side view of the coalescence of two LMs with a falling height of 5 cm (also see Movie S3 in the Supporting Information). e) Successive images showing the side view of the coalescence of two LMs at a falling height of 8 cm with a tilt angle of 60° (also see Movie S4 in the Supporting Information). f) Photographs give the demo of the coalescence process of the LMs with the CSS. The coalescence process could also be realized by a sliding impact process as is shown in Figure 4 e. As proposed in Figure 4 b, the LMs collapsed while the falling height is 8 cm. Here, the 6 µL LM was released in the upper terminal of a fixed and tilted hollow pipe with a length of 8 cm and variable tilt angles to trigger coalescence. From the successive side‐view images in Figure S6 in the Supporting Information, when the tilt angle of the hollow pipe is smaller than 30°, the coalesce does not occur and the LMs slide away separately. When the tilt angle of the hollow pipe increased to 60° (Figure 4 e), the two LMs could coalesce to form a larger LM intact. The coalescence process completed within a short time of 20 ms. We conclude that the LMs bounced away while the tilt angle was too small, coalesced while the tilt angle was within a certain range, and collapsed while the tilt angle was too large. Fresh medium and drug additions, as well as coculture of different types of cells, could be achieved by the coalescence of LMs (Figure 4 f). The direct addition of medium/drugs through injections with a needle also mixes the bioreactant. However, the manual method is difficult for microliter droplets. More importantly, the ultimate capacity of the LM is limited by the outer hydrophobic particles. The shell only tolerates slight expansion from the addition of an external solution; otherwise, surface defects are generated and the shell easily collapses. The amount of medium injected with a needle is therefore extremely limited. Regarding the fusion‐based mixture, large amounts of medium/bioreactant can be mixed into the original LM since the hydrophobic powder on the two LMs is sufficient to encompass the fused marble without any surface defects. 5 Splitting Apart from the coalescence of two small LMs into a larger one, the inverse process ( Figure 5 a), which separates one LM into two or several smaller ones can also be realized. Previous work has demonstrated that a manual cutting process could be applied for the splitting of LMs. 45 Here the splitting of LM could be realized by the simple impact process as is shown in Figure 5. A copper wire with a diameter of 30 µm was tightly straightened and the 12 µL LM was made to fall down from a 2 cm height to impact the copper wire through the sphere center. As shown in Figure 5 b, the impact between LM and the copper wire made a severe deformation of the LM and the LM was cut into two counterparts within 10 ms. The surface particles self‐rearranged on the separated two droplet surface during the short impact process, to form evenly distributed sub‐LMs. The size of the sub‐LMs could be controlled by controlling the cutting position of the parent‐LM. As shown in Figure 5 c, d, the copper wire was straightened and fixed on the outlet of the channels (diameter is 5 mm) with different positions. The position ratio (PR) represents the value of the shorter part of channel diameter cut by the copper wire divided by the diameter (0 ≤ PR ≤ 0. 5). With the increase of the PR, both the diameter ratio and the volume ratio of the obtained sub‐LMs increased (Figure 5 e). The larger diameter ratio and volume ratio denote the closeness of the sizes of the sub‐LMs. When the copper wire is fixed across the center of the cross section of the channel, the dropped LM (60 µL) could be cut into two counterparts with 30 µL. Figure 5 a) Schematic illustration of the splitting processes of LMs. The inset shows the microscopic image of the copper wire with a diameter of 30 µm. b) Successive images showing the top view of a 12 µL LM cut through the center by a copper wire via falling down process at a 2 cm height. The LM was cut into two sub‐LMs (also see Movie S3 in the Supporting Information). c) Top view of the channels (inner diameter is 5 mm) with the copper wires fixed at a different position. d) Optical images showing the LMs separated by the channels assembled with copper wires. e) The relationships between the diameter ratio and the volume ratio of the separated LMs, and the copper wire PR. f) Photographs demonstrating the splitting process of the LM with the CSS. Waste medium removal was achieved by separating the formed LMs using a gravity‐driven cutting process after certain culture period (Figure 5 f). After the 3D cell spheroid is formed inside the LM, the gravity‐driven cutting of the LM will separate the LM into two counterparts and the 3D cell spheroid will remain in one of the sub‐LMs. Bright‐field (BF) and fluorescence microscopy are used to screen the LMs by inspecting which LM contained the spheroid. 6 3D Culture of Stem Cell Spheroids The application of LM as a chemical reactor has been reported previously and here we demonstrated this application in Figure S7 in the Supporting Information. The aqueous solution‐based reactions with mild reaction condition can be precisely performed by our coalescence process to form nanomaterials. For example, the synthesis of the Ag NPs was proposed in the LM via an impact‐induced coalescence process and could be easily extracted for further usage. The specific information was presented in Figure S7 in the Supporting Information. Another application we must emphasize is the LM‐based SCS culture by applying our technique. In a conventional 3D cell culture protocol, cell spheroids are commonly obtained with either nonadhesive plates or the hanging drop method. For nonadhesive plates, the flow of the process includes the addition of culture medium containing stem cells, centrifugation of the medium to form aggregates at bottom of the well, incubation of the stem cells to form spheroids, replenishment of the medium with a pipette, and collection of the spheroids by pipetting using wide bore pipette tips, as shown in Figure 6 and Figure S8 in the Supporting Information. The hanging drop method uses similar culture protocols, and compared with the nonadhesive plate method, the hanging droplet is less extendable and the medium of the hanging droplet is difficult to replenish without affecting the inner spheroid. Using our protocol, 3D SCSs were easily harvested with the CSS using the LM as the carrier. As shown in Figure 6, we compared the 3D stem cell culture process of our protocol and the standard nonadhesive plate in three stages divided by the dotted lines. The formation of the spheroid in our protocol applied the hydrophobic silica particles to wrap the medium drop to ensure that the medium was well separated from the outer environment and permitted gas exchange. We applied a splitting process to remove the spent medium and a fusion process to replenish fresh medium. The last stage is the extraction of the LM. Instead of removing the LM from the nonadhesive plate‐based culture with a pipette, the LM was directly poured into the collection container with medium/phosphate‐buffered saline (PBS) solution, followed with a slight shaking process to break the LM. The SCS would then sink to the bottom. Figure 6 Comparison between the protocols for culturing 3D SCSs using a nonadhesive plate and our strategy. We compared the cellular configurations of the standard 2D cell culture and our method at different cell concentrations ( Figure 7 a), as shown in Figure 7 b, c. Stem cells cultured within a standard 96‐well plate spread and formed 2D stem cell layers on the bottom of the plate at various stem cell densities after 48 h in culture. The enlarged fluorescent image (FI) shows the elongated laminar morphology of cells from 2D cell cultures and extensive contacts in the surrounding area. Meanwhile, when stem cells were cultured in 50 µL of LMs for 48 h, cells in the LMs tended to aggregate, forming a large entity known as a 3D spheroid. The enlarged view of the spheroid shows the sphere‐like surface of the spheroid and cells that were tightly bound to one another, as shown in Figure 7 c. The size of the SCS increased with the cell density (Figure 7 d). As illustrated in Figure 7 d, when the cell density was less than 10 4 cells mL −1, spheroids were rarely formed and the size of the cell aggregates was approximately the size of single stem cell. When the cell density was greater than 10 5 cells mL −1, the 3D SCS was well formed, as shown in Figure 7 c. Cells cultured in the LMs tended to form single spheroids and the average sizes of spheroids were 250 µm at a cell density of 10 5 cells mL −1 and 450 µm at a cell density of 10 6 cells mL −1 (Figure 7 d; Figure S9, Supporting Information). The 3D SCSs that self‐assembled at a density of 10 6 cells mL −1 was tightly packed with ≈5000 cells in each spheroid, creating an “in vivo‐like” micro‐bioenvironment for organoid development that better preserved the stem cell phenotype and its innate properties. In a recent publication, the researchers suggested that the liquid marble–based 3D stem cell culture showed its potential applications in embryonic body formation and differentiation. 46 Figure 7 a) Schematic of the strategies for culturing cells in a 2D format and using our method. b, c) FIs showing stem cells cultured on a 96‐well plate b) and in 50 µL of LM c) for 48 h with different initial concentrations of stem cells, i. e. , 10 4, 10 5, and 10 6 cells mL −1. d) The column chart showing the size of the spheroid at different initial concentrations of stem cells. e) MTS assay showing the O. D. values of stem cells cultured in LMs after short‐term (1 h) and long‐term incubations (48 h). The error bars in (d) and (e) were obtained from 3 to 5 groups of experiments. The viability of cells growing inside the LMs was determined using the 3‐(4, 5‐dimethylthiazol‐2‐yl)‐5‐(3‐carboxymethoxyphenyl)‐2‐(4‐sulfophenyl)‐2H‐tetrazolium (MTS) assay. As shown in Figure 7 e, the optical density (O. D. ) at 490 nm was measured for stem cells cultured in LMs for 1 and 48 h. The cell viability of the 1 and 48 h cultures were comparable when the cell density was less than 10 4 cells mL −1, which may be attributed to the lack of spheroid formation at low cell concentrations. Cells were uniformly dispersed inside the LMs and had a sufficient number of contacts with the culture medium. When the initial cell density was greater than 10 5 cells mL −1, the cell viability of the 48 h culture was slightly lower than the 1 h culture, since the stem cells in the 48 h culture formed tightly bonded spheroids. As proposed in a previous report, 47 cell spheroids exhibited three zones along the radius, i. e. , a proliferating zone, quiescent viable cell zone, and necrotic core. The spherical cell aggregates represent an avascular tissue with a limit of diffusion of ≈150–200 µm for many molecules, including O 2. 48 Since molecules are mainly digested in the proliferating zone and the cell viability in the 48 h culture was less than the 1 h culture, a well‐formed SCS developed when the cell density was greater than 10 5 cells mL −1. In addition to the apparent cells, the internal cell‐to‐cell communication in the spheroid was inspected by capturing images with a confocal laser scanning microscope (CLSM) layer‐by‐layer with a gap of 9 µm, as shown in Figure S10 in the Supporting Information. Cells inside the spheroid (derived from a cell density of 10 5 cells mL −1 ) also showed a compact configuration. The shape of the cells varied, enabling them to conform to the environment. The cells inside the spheroid were in different shapes. The shape diversity inside a spheroid can affect the differentiation direction of stem cells. For instance, the chondrogenesis requires the stem cell to be spherical shape whereas the osteogenesis requires the stem cell to be spindle shape. The in situ inspection of the spheroid in an LM with BF and fluorescence microscopy is shown in Figure S11 in the Supporting Information. After the cutting‐based splitting, the LM containing spheroid was then used in another coalescence process with an LM containing fresh medium using our system and the other LM was subjected to further analysis and disposals. We can quickly inspect whether the separated LMs contain an SCS with the BF and fluorescent microscopy without breaking the LM (Figure S11, Supporting Information). The SCS derived from our system at a cell density of 10 6 cells mL −1 was visible to the naked eye after the LM on medium/PBS solution was subjected to a simple shaking process, as indicated by the arrows in Figure 8 a, b; this step is important and quite easy for the handling of the spheroid for further applications. We compared the cell configurations in the nonadhesive plate‐based 3D cell culture with our method at the same cell concentration, as shown in Figure 8. The 2D control group cultured in a 96‐well plate is shown in Figure S12 in the Supporting Information. Cells emitted green fluorescence under blue light illumination due to the expression of the green fluorescent protein (GFP), and the cell nuclei were stained with 4′, 6‐diamidino‐2‐phenylindole (DAPI) to label the nuclei with blue fluorescence. Compared with the spheroids obtained from standard 3D cell culture with nonadhesive plates, the cell configurations resulting from the nonadhesive plate and LM indicated that the sizes of the spheroids were comparable, with some differences in shape, as shown in Figure 8 c–h. The viability of the cells in the spheroids cultured using these two methods was evaluated with the MTS assay by measuring the O. D. value at 490 nm, as shown in Figure 8 k. The viability of cells cultured in the spheroid using the LM was approximately threefold higher than cells cultured on the nonadhesive plate, denoting the better cell viability of the LM‐based spheroid. The high cell viability of the spheroid produced using our method may be attributed to the structural advantage of the LM compared with the standard nonadhesive plate. As CO 2 and O 2 are indispensable for the cell metabolism, the LM is gas‐permeable and contains numerous microscale pores on the LM surface (Figure S13, Supporting Information) that enable the medium inside the LM to participate in gas exchange with the surrounding environment, whereas the standard nonadhesive plate only participates in gas exchange on the upper surface of the medium, as schematically illustrated in Figure 8 i. We measured the weight variation of 100 µL of culture medium in a 96‐well plate and 100 µL of culture medium in an LM over time at room temperature in an open environment to confirm that gas exchange was more efficient in the LM. As shown in Figure 8 j, greater changes in weight were observed for the LM than the 96‐well plate (6. 5‐fold), indicating better gas exchange by the stem cells in the LM. Figure 8 a) Schematic of the batch extraction process used for the SCSs in LMs. b) The 3D SCSs cultured in LMs at a cell density of 10 6 cells mL −1 were visible to the naked eye. The 3D SCSs cultured in the LMs were released into the PBS solution by simply shaking the plate with the LMs on the surface of the PBS solution. The 3D SCSs visible to the naked eye sank and were easily extracted with a pipette/dropper. c–e) CLSM images of the 3D spheroids cultured in LMs at a cell density of 10 6 cells mL −1. FITC and DAPI channels were examined. f–h) CLSM images of a 3D spheroid cultured in a 96‐well nonadhesive plate at a cell density of 10 6 cells mL −1. i) Gas exchange in the two types of culture. j) The evaporation curve of 100 µL of the culture medium in 96‐well plate and LM at room temperature in an open environment. k) The results of the MTS assay showing the viability of stem cells cultured at a cell density of 10 6 cells mL −1 in 96‐well nonadhesive plates (control) and LMs. The error bars shown in (j) were obtained from 3 groups of experiments. Asterisks denote the level of significance: ** p < 0. 01. l, n) The CLSM images of the morphology of a SCS with two days' culture period in the LM and the morphology after delivery onto a hydrophilic surface for another two days' culture. m) Schematic of the change of the cell morphology after delivery and replanting process (in vitro). The obtained 3D spheroid with high‐viability may show their potential for regenerative medicine, such as tissue repairing. As demonstrated in Figure 8 m, after two days' culture in LM, the cells form a spheroid, the spheroid can be directly delivered and planted onto a hydrophilic surface with ease for the spreading and diffusing. Figure 8 l, n gives the CLSM images of the morphology of an SCS with two days' culture period in the LM and the morphology after delivery onto a hydrophilic surface for another two days' culture. The result also suggests that the morphology of the individual stem cells changed from the sphere to the fried egg shape with an increased cell‐substrate contact area. Meanwhile, the results further proved the high cell viability of the spheroid. The comparison of 3D cell culture tools is described in detail in Table 1, showing evaluations of the size and shape of the formed spheroid, the cell viability, the internal cell‐to‐cell communication of the as‐obtained spheroid, the complexity of the extraction of the as‐cultured spheroid, the manipulation process, and the toxicity of the materials. As illustrated in the table, a standard culture dish is commonly used for 2D spread cell cultures and cell spheroids rarely develop. The advantages of our CSS include the formation of a viable spheroid with a wide range of sizes using simple manipulation strategy compared with the other tools. Nonadhesive plates are limited by their ability to obtain spheroids at high cell viability. Hanging drop and spinner flask cultures are limited by the ability to culture spheroids with large sizes or controllable shapes. In addition, the hanging drop method, 3D scaffolds, and magnetic cell levitation‐based cultures introduce external materials that have close contact with the cells and may produce long‐term toxicity to the cells. Overall, our CSS offers researchers a choice for producing SCSs with a large range of size, high cell viability, strong internal cell‐to‐cell contact, a simple extraction process, and low toxicity of the materials. Table 1 Comparison of 3D cell culture methods Method Spheroid size Spheroid shape Cell viability Cell‐to‐cell contact Extraction of spheroid Process Materials toxicity Standard culture dish 49 n. a. n. a. +++ Only at edge (2D) n. a. Easy and standard None Nonadhesive plate 50 Large (≈100–1000 µm) Spherical + Sufficient in 3D Simple Easy and standard None Hanging drop 51 Generally <500 µm Spherical ++ Sufficient in 3D Simple Easy and standard None 3D scaffolds 52 Large and tunable (Up to macroscale depend on the size of scaffolds) Controllable ++ Mainly contact with scaffold Simple Sophisticated to fabricate the scaffold Potential toxicity of the scaffold Spinner flask 49 Large amount with small size (microscale) Irregular ++ Sufficient in 3D Simple Simple spinning process None Magnetic cell levitation 53 Large (<1 mm) Quasi‐spherical ++ Cell‐to‐cell and cell‐to‐particles contact both exist Simple Additional magnetic levitation step based on culture dish Potential toxicity of the nanomaterials Microfluid 54 Generally <500 µm Spherical + Sufficient in 3D Difficult due to the oil–water barrier Easy and automated for droplet generation Low‐toxicity CSS Large (≈100–1000 µm) Spherical ++ Sufficient in 3D Simple shaking of LM on PBS solution Relatively simple due to the manipulability of LMs Low‐toxicity John Wiley & Sons, Ltd. 7 Conclusion In summary, the coalescence and splitting of LMs by applying simple mechanical force were systematically investigated to obtain the optimal parameters for the processes. By selecting the functional channels on the device, gravity‐based fusion and splitting of LMs were performed to mix precise volume of medium/drugs and remove spent culture medium, thus ensuring that the microenvironment of the cells was maintained under optimal conditions. The size of the cell spheroid was controlled within a wide range, up to sub‐millimeter in size. The cultured SCS showed highly compact cell‐to‐cell cross‐contacts, which are more valuable for the simulation of a real in vivo tissue than the 2D cell monolayer. The high gas permeability and the liquid repellent properties of the formed LMs endow the “in vitro tissue” with good cell viability, which is crucial for stem cell differentiation and tissue engineering. After transferring and delivering the cell spheroid to a hydrophilic surface, the cells were respread and transformed into typical 2D cell morphology gradually, which may offer the possibility for spheroid‐based stem cell delivery. The system offers researchers a new choice for producing SCSs with a large size, high cell viability, and strong internal cell‐to‐cell contact, which is a useful in vitro model to study the formation and growth mechanism of tissues in vivo. The coalescence and splitting of LMs show the great prospect for ultrafast, tiny, and stoichiometric chemical reactions, synthesis of functional nanomaterials, and cell spheroid culture in biomedical engineering. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201802045
| 2,019
|
Advanced Science
|
Bioinspired Preservation of Natural Killer Cells for Cancer Immunotherapy
|
Abstract The ability to cryopreserve natural killer (NK) cells has a significant potential in modern cancer immunotherapy. Current cryopreservation protocols cause deterioration in NK cell viability and functionality. This work reports the preservation of human cytokine‐activated NK cell viability and function following cryopreservation using a cocktail of biocompatible bioinspired cryoprotectants (i. e. , dextran and carboxylated ε‐poly‐L‐lysine). Results demonstrate that the recovered NK cells after cryopreservation and rewarming maintain their viability immediately after thawing at a comparable level to control (dimethyl sulfoxide‐based cryopreservation). Although, their viability drops in the first day in culture compared to controls, the cells grow back to a comparable level to controls after 1 week in culture. In addition, the anti‐tumor functional activity of recovered NK cells demonstrates higher cytotoxic potency against leukemia cells compared to control. This approach presents a new direction for NK cell preservation, focusing on function and potentially enabling storage and distribution for cancer immunotherapy.
|
Natural killer (NK) cells represent our body's first line of defense against cancers. 1, 2, 3 These cells play a key role in innate immune system, and have a unique ability to fight cancer cells without prior sensitization (unlike T and B‐ lymphocytes). 4, 5 Recently, various cell lines (e. g. , NK‐92, NK‐YS, and KHYG‐1) have been established with immunotherapeutic capabilities and potential to develop third‐party good manufacturing practice‐compliant cell banks (e. g. , NantKwest, Culver City, CA, USA). 6, 7 Although these cell lines are in clinical trials for treatment of various cancers (e. g. , leukemia, lymphoma, and melanoma), 6, 8, 9, 10, 11, 12 long‐term storage that facilitates off‐the‐shelf logistic distribution and repeated clinical administrations presents a challenge. 13 Cryopreservation has emerged as a powerful tool to extend the shelf life of cells by maintaining them at sub‐zero temperatures and slowing down their metabolic activity. 14, 15 During cryopreservation, cryoprotectants (also known as cryoprotective agents (CPAs)) are used to protect the cells from injury, which might result from the cooling and rewarming processes. 16, 17 NK cells are conventionally cryopreserved using dimethyl sulfoxide (DMSO), as a cryoprotectant, in a slow freezing methodology. 18, 19, 20, 21, 22, 23, 24, 25 A detailed description of existing cryoprotectants and cryopreservation methods is shown in Supplementary Table S1 (Supporting Information). Although cryopreservation of NK cells is currently employed, deterioration in cell viability and functionality has been reported. 20, 21, 26, 27, 28 This could be as a result of using DMSO, a well‐known toxic cryoprotectant. 20, 29, 30 In addition, although such a toxic cryoprotectant is usually removed (during the unloading process) from cells after cryopreservation, the trace amounts left inside cells are associated with side‐effects in human (e. g. , neurotoxicity, cardiovascular failure, respiratory arrest, and fatal arrhythmias). 31, 32 The deterioration effects posed due to exposure of cells to conventional cryoprotectants can be avoided by replacing them with nontoxic bioinspired cryoprotectants. Recently, bioinspired biocompatible materials, such as ectoine and trehalose, have been utilized to cryopreserve human cells (e. g. , red blood cells, oocytes, and stem cells). 33, 34, 35, 36 This study describes the development of an innovative approach to cryopreserve human cytokine‐activated NK cells using a cocktail of nontoxic bioinspired cryoprotectants (i. e. , dextran and carboxylated ε‐poly‐L‐lysine (CPLL)). The cells were cryopreserved using the slow‐freezing approach. The recovered cells were evaluated using a viability assay and compared to the selected control (i. e. , DMSO‐based cryopreservation). Subsequently, the recovered cells were incubated with leukemia cells, and their cytotoxic functional potency was evaluated by flow cytometry. To cryopreserve NK cells, we utilized a cocktail of biocompatible naturally occurring cryoprotective solution containing dextran and CPLL. The cocktail solution was based on experiment performed on a pool of candidates shown in Figure S1 (Supporting Information). Dextran is a bioinspired polysaccharide first discovered by Pasteur, 37 and produced in nature by certain bacteria such as Leuconostoc and Streptococcus. 38 The cryoprotective capability of dextran has been demonstrated for various cell types including stem cells 39, 40 and red blood cells. 41, 42, 43 PLL is a biocompatible L ‐lysine homopolymer naturally occurring during the fermentation process of multiple bacteria such as Streptomyces 44, 45 ( Figure 1 A). It has been used as a food and cell preservative 45, 46, 47 as well as for other biomedical applications (e. g. , tissue engineering and drug delivery). 48, 49, 50 Recently, the cryoprotective property of CPLL has been reported for numerous cell types such as red blood cells and stem cells. 51, 52, 53, 54 Carboxylated PLL acts in a way similar to antifreeze proteins (AFPs); however, the exact mechanism of action is still unknown. 55, 56 The CPLL exhibits antifreeze activities by inhibiting ice crystal growth and recrystallization during cryopreservation. 57, 58 During the cryopreservation process, small ice crystals tend to convert to large ice crystals by a process called recrystallization. This phenomenon occurs because small ice crystals have lower melting point (compared to larger crystals), and once they melt they release liquid water, which tend to merge with adjacent larger crystals and recrystallize. 55 Therefore, ice recrystallization forms a matrix of large ice crystals, causing more damage to cells as a result of membrane rapture and cell dehydration. 55, 59, 60 In addition, the ice crystals in CPLL are more hexagonal or bipyramid in shape since the carboxyl group binds to crystals. 51, 53, 55 This hexagonal crystal shape reduces the damage to cell membrane, in contrast to round and flat ice crystals observed when cooling pure water. 51, 53, 61 Furthermore, the CPLL shows high adsorption to cell membrane during cooling, which also contributes to the protection of cell membrane from outside and maintaining membrane integrity. 51, 62 This might be related to the binding affinity of CPLL to cell membrane, in a similar way of AFPs. 51 In addition, combining both dextran and CPLL might also provide synergic cryoprotective effect 62 (Figure 1 B). The high affinity of side‐chain amino groups of polyamines of CPLL and dextran molecules to cell membrane might help in maintaining the integrity of cell membrane. These molecules show high affinity to water and thus could also help removing intracellular water to the surrounding of cells during freezing. In this process, dextran and CPLL might build up at the cell boundary in such a manner that the diffusion of solutes is restricted; therefore, reducing the mechanical damage caused by ice crystal formation and growth. The combination of dextran and CPLL could also interact with concentrated salts controlling the degree of dehydration to a level sufficient to avoid intracellular ice formation during freezing. To further elucidate the exact cryoprotective mechanism(s) of the developed solution, the interactions of dextran/CPLL molecules with water, salts, and lipid membranes during freezing can be investigated using spectroscopic methods, such as solid‐state nuclear magnetic resonance. 55, 62 Prior to cryopreservation, the NK cells were loaded with precooled (≈4 °C) cryoprotective cocktail solutions at different dextran (5 or 10% w/v) concentrations with CPLL (7. 5% w/v) at room temperature for 5 min. A low concentration of dextran (i. e. , 5%) was chosen since the results show no significant difference in cell viability between solutions containing 5% dextran compared to 10% dextran (Figure 1 C). Thus, based on these data a combination of 5% dextran with 7. 5% CPLL was chosen to be the basis of our subsequent experiments. In this study, we focused on replacing DMSO, a toxic cryoprotectant used for cryopreservation NK cells, with biocompatible bioinspired cryoprotectants (i. e. , dextran and CPLL). In addition, our solution does not include animal or human serum, which prevents potential transmission of zoonotic infections or causing allogeneic reaction, respectively. Figure 1 Bioinspired biocompatible cryoprotectants for cryopreservation of natural killer cells. A) Schematic showing the chemical structures of dextran and carboxylated poly‐L‐lysine (CPLL). B) Schematic demonstrating the potential mechanism of action of dextran and CPLL during cryopreservation of natural killer (NK) cells. The synergic effect of CPAs is related to their high affinity to cell membrane, water molecules, and solutes. This characteristic might provide cell protection while removing intracellular water, restricting solute diffusion, and controlling the degree of dehydration to a level sufficient to minimize intracellular ice formation during cooling. Carboxylated PLL also might limit cryoinjury to cells by binding to ice crystals and inhibiting their growth and recrystallization during rewarming. C) Determination of percentage (%) cell viability following CPA loading and unloading. Low level (i. e. , 5% w/v) of dextran/CPLL‐based cocktail solution was used for subsequent experiments since there is no significant difference in cell viability between 5 and 10% dextran concentrations. The data shown are averages with standard error of the mean (SEM) from various independent experiments. For 5% dextran/CPLL group N experiments = 3; n total cells = 315 and for 10% dextran/CPLL N experiments = 3; n total cells = 416. To achieve cryopreservation, we utilized a slow freezing method, which consists of four steps: (i) CPA loading; (ii) cooling down to cryogenic temperatures (i. e. , cryopreservation); (iii) warming up to ambient temperature (i. e. , rewarming or thawing); and (iv) CPA unloading ( Figure 2 A). DMSO has shown to be toxic to cells, and the damage is proportional to the exposure time 22, 63 ; therefore, we decided to rapidly cryopreserve cells, a process that usually takes around 5 min to load cells with CPA and transfer them to −80 °C freezer (for CPA loading step). In addition, to prevent the prolonged exposure of the cells to DMSO after thawing, we decided to immediately unload the CPA, a process that usually takes around 5 min to wash the cells by resuspending them in fresh cell media and centrifugate (for CPA unloading step). Therefore, in our protocol we decided to be consistent with loading and unloading the cells for 5 min for each step. This protocol is consistent with other studies in the literature. 25, 29, 64 Furthermore, we decided to unload the dextran/CPLL solution from cells to be consistent with the DMSO group, and because we are not sure about their effects in circulation after transfusion, which is beyond the scope of this study. However, investigation of the systemic effect of the dextran/CPLL solution is an interesting field of further research. To evaluate the effect on cells during CPA loading and unloading, cell viability was evaluated (Figure 2 B). The results demonstrated that there is no significant difference between cells loaded (for 5 min) and subsequently unloaded (for 5 min; without cryopreservation) with a dextran/CPLL‐based solution (82. 8 ± 3. 1%) compared to DMSO‐based solution (69 ± 16. 5%) and fresh cell medium (83. 3 ± 7. 3%) groups. Subsequently, the NK cells were cryopreserved and stored for 1 week using a slow freezing method. Following cryopreservation and rewarming, the viability of recovered cells was evaluated (Figure 2 C). The results demonstrated that there is no statistically significant difference between cells cryopreserved using dextran/CPLL‐based and DMSO‐based solutions right after thawing (73. 3 ± 5. 3% and 65. 5 ± 1. 2%, respectively). These viability results are consistent with other reports, which were based on using DMSO‐based solutions. 65, 66, 67, 68, 69 In addition, we cultured cells after cryopreservation and rewarming for up to 1 week. Although cells in the dextran/CPLL group showed a drop in viability after 1 d in culture compared to fresh cell medium (uncryopreserved) and DMSO‐based groups, the cryopreserved cells using dextran/CPLL catch up after 1 week (Figure 2 D). This observation is in agreement with other studies, which reported decline in cell viability during the first 24 h in culture. 27, 30 It is important to note that the cryopreserved NK cells are intended to be used in real‐life scenarios, in which patients at clinical centers receive these cells immediately after thawing without an extended culture post‐thaw. 6, 7, 8, 13, 70 However, their fate in circulation is an interesting field of further research. 1, 6, 8, 11 Figure 2 Assessment of NK cell viability following dextran/carboxylated poly‐L‐lysine (CPLL) based cryopreservation and rewarming. A) Schematic showing the cryopreservation protocol used for preservation of natural killer (NK)‐92 cells. The concentrated NK cells are loaded with bioinspired dextran/CPLL‐based cryoprotective agent (CPA) at room temperature (24 °C). The cells are subsequently placed into cryovials, cryopreserved using slow freezing method at −80 °C. The cells were then stored for 1 week. Following rapid rewarming at 37 °C, the CPAs washed out from the cells by re‐suspending the cells in NK media. B, C) Determination of percentage (%) cell viability following B) CPA loading and unloading; C) cryopreservation, rewarming, and washing the CPAs; and D) in culture for up to 1 week. The data shown are averages with standard error of the mean (SEM) from various independent experiments. For CPA loading and unloading experiments: (i) cell medium group ( N experiments = 7; n total cells = 1316), (ii) DMSO group ( N experiments = 6; n total cells = 877), and iii) dextran/CPLL group ( N experiments = 3; n total cells = 315). For cryopreservation experiments: (i) DMSO group ( N experiments = 3; n total cells = 654) and (ii) dextran/CPLL group ( N experiments = 3; n total cells = 281). For 1 d in culture experiments: (i) cell medium group ( N experiments = 3; n total cells = 1152), (ii) DMSO group ( N experiments = 3; n total cells = 772), and (iii) dextran/CPLL group ( N experiments = 3; n total cells = 416). For 1 week in culture experiments: (i) Cell medium group ( N experiments = 3; n total cells = 3353), (ii) DMSO group ( N experiments = 3; n total cells = 3657), and (iii) dextran/CPLL group ( N experiments = 3; n total cells = 2219). NK cells demonstrate cytotoxic capability to various cancer cells including leukemia cells (e. g. , K562 cells). 7, 8, 71 K562 cells are chronic myelogenous leukemia cells that have attained widespread use due to their highly sensitive in vitro target for NK cells. 72 To evaluate cell functionality after cryopreservation and rewarming, we measured the cytotoxic potency of NK cells against K562 cells. The measurement was performed on NK cells cryopreserved with dextran/CPLL‐based and DMSO‐based solutions. Flow cytometry analysis was carried out after 4 h incubation of NK cells (effector cells) with K562 cells (target cells; prefluorescently labeled). Two different effector‐to‐target ratios (E:T: 5:1 and 10:1) were tested ( Figure 3 A). We observed a significant ( p < 0. 05) increase of cytotoxic potency of NK cells recovered after dextran/CPLL‐based cryopreservation compared to DMSO‐based cryopreservation (Figure 3 B). At both E:T ratios (5:1 and 10:1), dextran/CPLL‐based cryopreserved NK cells showed a significant killing efficiency of 67 ± 3. 1% (5:1) and 71 ± 3. 7% (10:1) compared to DMSO‐based cryopreserved cells 32 ± 8. 2% (5:1) and 44. 8 ± 3. 3% (10:1). These results are in agreement with other studies, which reported a reduction of killing efficiency of NK cells cryopreserved with DMSO‐based solutions. 27, 28, 29, 65, 73, 74 It is important to notice that cells cryopreserved with dextran/CPLL‐based showed higher functionality compared to cells cryopreserved with DMSO‐based solutions. This important observation indicates that there might be hidden factors in cryopreservation with dextran/CPLL‐based solution that select most potent NK cells or trigger stronger phenotypic changes of NK cells. Although, other groups have observed this phenomenon using DMSO‐based solution, 25, 73 it has not been reported before using dextran/CPLL‐based solution. Therefore, such results give a hint that there is a lot of exploration needed beyond the conventional DMSO‐based solution. Although we see a higher killing efficiency of dextran/CPLL‐based cryopreserved NK cells compared to DMSO group, this was observed in a small set of experiments ( n = 3–4). Therefore, to be able to make a strong conclusive remark on overall cell functionality, a larger sample size is required to achieve a strong statistical power analysis. Further, to evaluate the effect of dextran/CPLL‐based and DMSO‐based solutions on K652 cells, cell viability was evaluated (Figure S2, Supporting Information). The results demonstrated no significant difference in viability of K652 cells exposed to dextran/CPLL‐based solution (94. 2 ± 0. 6%) compared to cells exposed to a DMSO‐based solution (95. 3 ± 0. 3%) and a fresh cell medium (95. 9 ± 0. 2%). In addition, we noticed a significant difference in membrane stability of effector cells cryopreserved with both dextran/CPLL‐based and DMSO‐based solutions compared to fresh (uncryopreserved) cells (Figure S3, Supporting Information). A representative sample of a complete set of samples for each experiment was performed and its internal controls are shown in Figure S4 (Supporting Information). In addition, fresh NK and K562 cells were used as baselines to detect auto‐fluorescence or background staining (Figure S5, Supporting Information). Figure 3 Assessment of NK cell functionality following dextran/carboxylated poly‐L‐lysine (CPLL) based cryopreservation and rewarming. Anti‐tumor functional activity of recovered NK cells after dextran/CPLL‐ and DMSO‐based solutions was evaluated against K562 leukemia cell line using cytotoxicity assay. Two different effector cells: target cells ratios were assessed (i. e. , 5:1 (50 000:10 000) and 10:1 (100 000:10 000)). A) Representative flow cytometry dot plots. B) Quantification of flow cytometry analysis. The data shown are averages with standard error of the mean (SEM) from various independent experiments ( n = 3–4). In this study, we report the preservation of human NK cell viability and function following cryopreservation using an innovative cocktail of biocompatible bioinspired solution based on dextran and CPLL. The NK cells were cryopreserved using a slow freezing method. Right after rewarming and CPA unloading, NK cells preserved with dextran/CPLL‐based solution maintained their viability at a comparable level to DMSO, i. e. , the standard cryoprotectant used in cryopreservation. However, the lower viability observed in the first day of culturing the cells cryopreserved with dextran/CPLL‐based solution indicates that this cocktail solution can be further improved. Further, we demonstrate the preservation of the anti‐tumor functional potency of recovered NK cells. These results represent an important exploration toward evaluating the phenotypic changes that occur to NK cells during cryopreservation, which might preserve their functional capabilities. The developed bioinspired cocktail solution has the potential to pave the way for the development of similar approaches, which look at the functional capability of cryopreserved NK cells using biocompatible materials available in nature with broad applications for other cell types. Experimental Section Cell Lines and Cultures—Human Cells : Human natural killer (NK‐92) cell line and chronic myelogenous leukemia (K562) cell line used in this study were acquired from ATCC (Manassas, VA, USA). Cell Lines and Cultures—Cell Culture : The NK‐92 cells were cultured in suspension in X‐VIVO (without gentamicin or phenol; Lonza, Basel, Switzerland) supplemented with human AB serum (5%; Corning, New York, USA), and the human interleukin‐2 (500 UI mL −1 ; Rehovot, Israel) was used as a cytokine for activation. The K562 cells were cultured in suspension in Iscove's Modified Dulbecco's Medium (ATCC) supplemented with fetal bovine serum (FBS; 10%; EMD Millipore, Hayward, CA) with penicillin/streptomycin (1%, Thermo Fisher Scientific, Waltham, MA, USA) and L‐glutamine (1%; Sigma‐Aldrich, St. Louis, MO, USA). All cells were maintained at concentration of 2 × 10 5 cells mL −1 in 5% CO 2 incubator at 37 °C, and the cell medium was changed every 3 d. Cryopreservation Procedure—Bioinspired Cryoprotective Materials : The cryoprotective cocktail solutions were prepared using CPLL solution and various concentrations of dextran‐40 in Dulbecco's modified Eagle's medium (DMEM). ε‐poly‐L‐lysine solution (25%, Akron Biotech, West Palm Beach, FL, USA) and succinic anhydride were combined and allowed to react for 1 h at 50 °C resulting in a CPLL solution. Succinic anhydride and DMEM were both obtained from Sigma‐Aldrich (St. Louis, MO). The CPLL solution was then added to a room temperature DMEM. The CPLL and DMEM were mixed until homogenous. This medium was used as a base for the preparing solutions at various concentrations of dextran‐40 (Akron Biotech, West Palm Beach, FL, USA). Dextran‐40 used in this study has a normative molecular weight of 40 000 Da, ranging from 35 000 to 45 000 Da. The polydispersity, or M w / M n, is in the range of 1. 4–1. 9, as provided by the supplier. PLL has a molecular weight of 4700, with a polydispersity of 1. 14. No purity is being reported for the material. Cryopreservation Procedure—CPA Loading : Human NK‐92 cells were loaded with the precooled (≈4 °C) cocktail of biocompatible CPA solution (Akron Biotech, West Palm Beach, FL, USA), which consists of dextran (5% w/v (equal to 50 mg mL −1 )) and carboxylated poly‐L‐lysine (7. 5% w/v (equal to 75 mg mL −1 )) in DMEM for 5 min at room temperature. For the control group, DMSO (10% (v/v); Sigma‐Aldrich) in FBS was loaded into cells for 5 min at room temperature. Fresh NK‐92 medium was used for the negative control group. Cryopreservation Procedure—Cryopreservation : For cryopreservation, the NK‐92 cells that were loaded with different CPA (cell density: 1 million cells in 1 mL CPA solution) transferred into cryovials (Corning). The cryovials were immediately transferred into a freezing container (Mr. Frosty; Thermo Fisher Scientific) and placed into a −80 °C freezer (Thermo Fisher Scientific) for overnight. The cryovials were then transferred to a liquid nitrogen tank for storage. Cryopreservation Procedure—Thawing and CPA Unloading : For thawing, the cryovials were placed into water bath (Thermo Fisher Scientific) for 1–2 min (until it was clear that the media in the cryovials were thawed). The cells were then pipetted into a 9 mL of prewarmed (≈37 °C) fresh NK‐92 cell medium, and allowed to settle for 5 min. Following that cells were centrifuged (at 1280 rpm for 5 min), media discarded, cells washed one more time, and suspended with fresh NK‐92 media. Viability Assay : To evaluate the viability of NK‐92 cells after cryopreservation and rewarming, the recovered cells were stained using trypan blue (Sigma‐Aldrich), as a dye exclusion method. The live cells were then counted using a hemacytometer (Thermo Fisher Scientific). Percentage of cell viability was then calculated by dividing the number of live cells over total number (live + dead) of cells. Cytotoxicity Functionality Assay : For the cytotoxicity functionality assay, the target (K562) cells were first counted and stained with calcein AM (0. 1 × 10 −6 m ; BD Biosciences, San Jose, CA, USA) in phosphate‐buffered saline (PBS; Thermo Fisher Scientific) for 5 min in 5% CO 2 incubator at 37 °C. After staining, the K562 cells were washed and suspended in Roswell Park Memorial Institute (RPMI) media (Thermo Fisher Scientific) supplemented with FBS (20%). Subsequently, the cells were plated in a 96‐well plate (Sigma‐Aldrich). Second, the activated effector (NK‐92) cells were added onto the top of target cells at different effector‐to‐target ratios (E:T: 5:1 and 10:1; Supplementary Table S2) co‐incubated for 4 h in RPMI media supplemented with FBS (20%) in the dark at 37 °C in 5% CO2 incubator. Subsequently, the dead cells in the cell mixture were stained with propidium iodide (PI; 10 µg mL −1 ; Sigma‐Aldrich) for 15 min at room temperature. The cells were then washed and suspended in flow cytometry buffer (PBS supplemented with 2% FBS). Flow cytometry assessment was immediately performed on flow cytometer (BD Biosciences, San Jose, CA, USA) with an excitation wavelength of 488 nm. The fluorescence signals were collected through a 530 nm band pass (filter for the fluorescein isothiocyanate (FITC) signals), and a 650 nm long pass filter for PI fluorescence. To determine the percent‐specific killing efficiency of NK cells, K562 cells were gated and quantified. K562 target cells that both stain for calcein‐AM and PI represented dead target cells. The effector cells, which are unstained (Q4) or stained with PI (Q1), were gated out. For 5:1 ratio, the number of gated out cells was 0. 5 × 10 5, while for 10:1 ratio, the number of gated out cells was 1 × 10 5. Internal controls of each group and the baseline (fresh) cells were used to detect the auto‐fluorescence or background staining as well as determine the gating (Figures S4 and S5, Supporting Information). Subsequent data analysis was performed using FlowJo software (FlowJo, LLC, Ashland, OR, USA). Statistical Analysis : Experiments were carried out multiple times (≥3) on different NK‐92 cultures. Means, standard deviations, and standard errors were calculated. Data were analyzed using t ‐test and one‐way and two‐way analysis of variance (ANOVA) with Tukey's and Bonferroni's honestly significant difference (HSD) tests. Statistical significance was set at p < 0. 05. All statistical analyses were performed with GraphPad Prism (GraphPad Software). Error bars in the figures represent the standard error of the mean (SEM). Conflict of Interest Dr. U. Demirci is a founder of and has an equity interest in (i) DxNow Inc. , a company that is developing microfluidic and imaging technologies; (ii) Koek Biotech, a company that is developing microfluidic IVF technologies for clinical solutions; and (iii) LEVITAS Inc. , a company that develops biotechnology tools for genomic analysis in cancer. U. D. 's interests were viewed and managed in accordance with the conflict of interest policies. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201802077
| 2,019
|
Advanced Science
|
Cytocompatible, Injectable, and Electroconductive Soft Adhesives with Hybrid Covalent/Noncovalent Dynamic Network
|
Abstract Synthetic conductive biopolymers have gained increasing interest in tissue engineering, as they can provide a chemically defined electroconductive and biomimetic microenvironment for cells. In addition to low cytotoxicity and high biocompatibility, injectability and adhesiveness are important for many biomedical applications but have proven to be very challenging. Recent results show that fascinating material properties can be realized with a bioinspired hybrid network, especially through the synergy between irreversible covalent crosslinking and reversible noncovalent self‐assembly. Herein, a polysaccharide‐based conductive hydrogel crosslinked through noncovalent and reversible covalent reactions is reported. The hybrid material exhibits rheological properties associated with dynamic networks such as self‐healing and stress relaxation. Moreover, through fine‐tuning the network dynamics by varying covalent/noncovalent crosslinking content and incorporating electroconductive polymers, the resulting materials exhibit electroconductivity and reliable adhesive strength, at a similar range to that of clinically used fibrin glue. The conductive soft adhesives exhibit high cytocompatibility in 2D/3D cell cultures and can promote myogenic differentiation of myoblast cells. The heparin‐containing electroconductive adhesive shows high biocompatibility in immunocompetent mice, both for topical application and as injectable materials. The materials could have utilities in many biomedical applications, especially in the area of cardiovascular diseases and wound dressing.
|
1 Introduction Hydrogels have drawn a lot of attention as promising extracellular matrix (ECM) mimicking biological materials. Various characteristics, such as swelling degree, water content, stiffness, porosity, viscoelastic properties, and biocompatibility can be tailored for different applications. 1, 2, 3 The incorporation of electroconductive elements to hydrogels is an attractive design approach that combines the viscoelastic and mechanical properties of hydrogels with the conductivity of organic electronics. 4, 5, 6, 7 Electroconductive hydrogels can be used in bioelectrical interfaces as bioresponsive electrodes, substrates that facilitate the electrical stimulation of cells or tissues (neuron or muscle), biosensors, implants, or for drug delivery under electrical stimulation. 8, 9, 10 Conductive polymers, such as polypyrrole (PPy), polyaniline (PAni), and poly (3, 4‐ethylenedioxythiophene) (PEDOT), have been used in the synthesis of electroconductive hydrogels due to their high conductivity and ease of processing. 11, 12, 13, 14 However, the state‐of‐the‐art technology suffers from particular properties of conductive polymers, such as high hydrophobicity and insolubility, resulting in low adherence to wet substrates and poor penetration into living tissues. 5, 15, 16 A highly hydrated electroconductive network, with biochemical and mechanophysical properties similar to the natural extracellular matrix, can provide a cell‐compatible scaffold for 3D cell culture. 17 Moreover, a soft conductive adhesive, which can be utilized as injectable material, as well as bio‐ink for in situ bioprinting, will close the gap between electroconductive materials and biomaterials, lead to numerous medical and clinical applications. 18, 19, 20 To incorporate new functions to a conductive polymer for biomedical utilities, there are several limits of “add‐on” approach because it does not alter the intrinsic properties of the electroconductive hydrogel network. For example, injectability, adhesion to biological tissues under wet conditions, and cell encapsulation are important for many biomedical applications but have proven to be very challenging. An attractive avenue is to develop self‐assembled physical electroconductive hydrogels. 13 The noncovalent network can be disintegrated and re‐established, resulting in self‐healing and shear‐thinning properties important for developing injectable and printable materials. 21, 22, 23 Moreover, noncovalently assembled electroconductive hydrogels are also particularly useful for cell encapsulation. Alternatively, recent work has shown that fascinating material properties can be realized with the bioinspired hybrid network, especially through the synergy between irreversible covalent crosslinking and reversible noncovalent interaction. For example, Guo and co‐workers designed a host–guest and covalent bonding‐based multifunctional conductive hydrogel with a desirable mechanical property and self‐healing ability and the resulting material can be used to build pressure‐dependent sensors. 12 While the materials have shown high cytocompatibility for 2D cell culture, the covalent crosslinking reaction does not permit cell encapsulation in 3D hydrogel and completely impair the network dynamics mediated only by host–guest interaction. Developing a cell‐compatible, injectable, and electroconductive soft adhesive is vital for the advancement of many biomedical emerging fields, such as artificial skins, adhesive wound‐healing patches, and injectable cardiac patches. However, a material with multiple functions cannot be easily realized with an add‐on approach and remains very challenging. In this work, we report a polysaccharide‐based electroconductive hydrogel system formed by a hybrid covalent/noncovalent dynamic network composed of aldehyde‐modified hyaluronic acid (HA‐ALD), glycol chitosan (GC), and PEDOT:Heparin (PEDOT:Hep) or PEDOT: poly(styrenesulfonate) (PEDOT:PSS). The dynamic network is mediated by noncovalent interactions (electrostatic, hydrogen bonding, π–π, and π–ion interactions), and dynamic covalent bonds (Schiff‐base formation). We aimed to develop a viscoelastic hydrogel that displays enhanced mechanical stability, as well as mechanical responsiveness, such as self‐healing, stress relaxation, shear‐thinning, and adhesion behavior. The biological functions of the electroconductive hydrogels were evaluated for their cytocompatibility for 2D/3D cell culture, effect on the induced myogenic differentiation of myoblast cells, and in vivo biocompatibility with immunocompetent mice both for topic application and as injectable materials. 2 Results and Discussion We designed and synthesized polysaccharide‐based electroconductive hydrogels, which were formed by a hybrid network of aldehyde‐modified hyaluronic acid (HA‐ALD), glycol chitosan (GC), and PEDOT:Hep (or PEDOT:PSS) ( Figure 1 A–C). Schiff base (imine) bond formation between HA‐ALD and GC resulted in reversible covalent crosslinking, while the noncovalent crosslinking was mainly mediated through the electrostatic interaction between negatively charged PEDOT:Hep or PEDOT:PSS and positively charged GC. Polysaccharides and PEDOT polymers were chosen not only because of their biocompatibility and electroconductivity, respectively, but also for introducing different types of noncovalent interaction to the dynamic matrix. The abundant hydroxyl groups of polysaccharides can provide a hydrophilic network as well as many hydrogen bond donors and acceptors. The PEDOT polymers were formed through π–π stacking between PEDOT nanoparticles and ionic interaction with the anionic polymer (Heparin or PSS). We speculated that the combination of covalent crosslinking and different noncovalent interactions can contribute to the adhesion to surfaces of various properties and cohesion within the network, which are both essential for an adhesive. Figure 1 Characteristics of the dual cross‐linked electroconductive PEDOT:Heparin/HA‐ALD/GC hydrogels. A) Oxidation of HA by NaIO 4, resulting in the formation of HA‐ALD with the characteristic presence of aldehyde groups. B) Structure of glycol chitosan. C) PEDOT:Hep and PEDOT:PSS synthesis by polymerizing EDOT to form PEDOT particles using heparin or PSS as dopant. The chemical structures of PSS and PEDOT:PSS are not illustrated. D) Representative hydrogel formation by homogeneous mixing of HA‐ALD + PEDOT:Hep with GC solutions in microtubes. E) Scheme of the 3D structure of the dual cross‐linked hydrogel network. F) Scheme of the interactions involved in the formation of the double network. G–I) Scanning electron microscopy (SEM) images of hydrogel networks. The arrows in the dual cross‐linked sample (I) indicate features, such as lamellar‐like structures (yellow arrow) or fibers (green arrow). Scale bar: 10 µm. J) Bioprinting of a PEDOT:Hep + HA‐ALD + GC hydrogel, which can adhere to the surface of a glass petri dish, and K) when holding the dish vertically. Aldehyde groups were introduced to HA through reaction with sodium periodate, as previously reported. 24 Fourier transform infrared spectroscopy (FTIR) was used to confirm the reaction, with a newly formed peak at 1733 cm −1 corresponding to the stretching vibration of the C=O bond in HA‐ALD (Figure S3, Supporting Information). The polymerization of EDOT with Heparin or PSS as counterions was validated using 1 H NMR and FTIR ( Figure 2 C and Figure S1, Supporting Information). For example, for PEDOT:Hep the peaks at 1310 cm −1 and at 1520 cm −1 have emerged, corresponding to the C—C and C=C stretching vibrations of the quinoidal structure of the thiophene rings. The peaks at 1085 and 1144 cm −1 correspond to the C—O—C bond stretching in the ethylene dioxy ring. The peaks at 972, 830, and 681 cm −1 correspond to the C—S bond stretching vibrations in the thiophene ring. 25 The particle sizes of conductive polymers can affect the rheological properties of self‐assembled hydrogels. 26 We tuned the ratios between EDOT and sulfated polymer to control the particle sizes of different conductive polymers within a similar range. The particle sizes of PEDOT:Hep and PEDOT:PSS in water were characterized with dynamic light scattering (DLS), showing average sizes of 214. 13 ± 5. 29 and 217. 87 ± 0. 58 nm (mean ± sd, n = 3) respectively (Figure 2 A, B). The UV–vis–NIR spectra of PEDOT:PSS and PEDOT:Hep exhibited high absorbance in the NIR a wavelength (700–1000 nm) typical for PEDOT materials (Figure S2, Supporting Information). 27 Figure 2 A) Particle size distribution of PEDOT:PSS and PEDOT:Hep solutions, determined by DLS. Inset: The solutions of PSS or heparin in MilliQ kept in microtubes and change from transparent to black after the polymerization of EDOT. B) The Z‐average particle size of PEDOT:Hep and PEDOT:PSS. Mean ± sd ( n = 3). C) Fourier transform infrared (FT‐IR) spectra of Heparin, PEDOT:Heparin, HA‐ALD+GC and HA‐ALD+GC+PEDOT:Hep. D) PEDOT:Hep hydrogels (the loss factor plotted against the time). E) The loss factor at 10 min of gelation for hydrogels with PEDOT:Hep. The legend color is the same as Figure (D). F) Degradation assay of PEDOT:PSS or PEDOT:Heparin hydrogels. Inset: Images of PEDOT:PSS and PEDOT:Hep hydrogels during the degradation assay at 37 °C. The dynamic covalent hydrogel was formed by mixing HA‐ALD with GC, while dynamic noncovalent hydrogel was formed by mixing PEDOT:Hep or PEDOT:PSS with GC. Scanning electron microscope (SEM) image of the cross‐sections of the covalently cross‐linked hydrogel (HA‐ALD/GC) showed a porous network consisting of lamellar structures with large and smooth surfaces (Figure 1 G). For the noncovalently cross‐linked hydrogel, SEM images displayed a dense network of fibers (Figure 1 H). The hybrid dynamic electroconductive hydrogel was formed by adding GC to a premixed solution of PEDOT:Hep and HA‐ALD at room temperature (25 °C), resulting in dark‐blue hydrogels. The hybrid hydrogel with 0. 5% PEDOT:Hep (Figure 1 I), 0. 5% HA‐ALD, and 1% GC showed a hybrid network possessing features from both the covalent and noncovalent hydrogels, with lamellar structures and fibers. This network is less dense and has a larger pore size than the noncovalent PEDOT:Hep/GC hydrogel (Figure 1 G). Similarly, the hybrid hydrogel could also be formed using PEDOT:PSS (Figure 1 H). All three types of hydrogels were stable in aqueous solution, without detectable degradation over a time period of one month. Because of the biodegradability of HA (by hyaluronidase) and GC (by lysozyme), hydrogel degradation was tested by incubating the hybrid hydrogels in a mixture of both enzymes. As shown in Figure 2 F, 17. 5% and 18. 5% of weight loss of the PEDOT:PSS/GC/HA‐ALD and PEDOT:Hep/GC/HA‐ALD hybrid hydrogels were measured after 21 d, respectively. The degradation was also visually apparent, as the intact hydrogels broke gradually into amorphous pieces. Interestingly, the PEDOT:Hep hybrid hydrogel became amorphous clusters, caused by the aggregation of degraded products. The gelation of hydrogels was evaluated by measuring the storage ( G ′) and loss ( G ′′) moduli over time. The covalent, noncovalent, and hybrid hydrogels exhibited fast gelation (Figure 2 D). As compared to the covalent hydrogel, both noncovalently cross‐linked hydrogels (PEDOT:PSS/GC and PEDOT:Hep/GC) showed higher loss factors (δ = G ′′/ G ′), indicating that the hydrogels were more viscous (Figure 2 E). Incorporating dynamic covalent crosslinking (through adding HA‐ALD) into the network led to lower δ value. As the amino group of GC can either form a neutral imine bond with HA‐ALD or form electrostatic interaction with heparin or PSS, reducing the PEDOT:PSS or PEDOT:Hep content further increased the elastic portion ( Figure 3 A) and Young's modulus (Figure 3 B), as the dynamic covalent crosslinking became more dominant in the network. Figure 3 Rheological properties of the electroconductive hydrogels. A) Frequency sweep performed with a shear strain of 1% and the frequency increasing from 0. 01 to 100 Hz. B) The Young's modulus of hydrogels at 10% shear strain (mean, error bar, n = 3). C) Stress relaxation curves of the hydrogels at a constant shear strain (1%). D) The stress relaxation half‐times (τ 1/2 ) of the hydrogels (Statistically significant differences are shown with asterisks * p < 0. 05, ** p < 0. 01, *** p < 0. 001, mean, error bar, n = 3). E) Continuous flow experiments showing viscosity of the hydrogels plotted against the shear rate. F) The dual cross‐linked hydrogels were syringe‐injectable and thin filaments (indicated by the arrows) were be extruded into PBS solution. G) Self‐healing property of the hydrogels when the alternate step strain was switched from 1% to 1500%, indicated by the recovery of the elastic modulus. H) The self‐healing of the hydrogel highlighted by an electric circuit with an LED. (i) The electroconductive hydrogel closed the circuit and the LED lit up. (ii) The hydrogel was cut into two pieces and the circuit opened. (iii) The hydrogel pieces were placed together and left to self‐heal into one block (after 5 min) and the circuit was closed, as indicated by the LED. I) Scheme of the self‐healing process of the electroconductive hydrogel. The reversible imine bond led to a dynamic network and the HA‐ALD/GC hydrogel showed stress relaxation property with a T 1/2 of 120 s (Figure 3 C, D), while the irreversibly crosslinked hydrogels normally exhibited T 1/2 > 2000 s. 28, 29, 30 As expected, incorporation of noncovalent crosslinking accelerated the network dynamics and stress relaxation, as shown by the gradual reduction of T 1/2 upon increasing the PEDOT:Hep concentration. The full noncovalent PEDOT:Hep/GC hydrogel possessed the lowest T 1/2 of 2 s. The hydrogels also showed the non‐Newtonian behavior of shear thinning. As expected, with the increase of PEDOT:Hep concentration in the hybrid network, the shear‐thinning behavior became more remarkable (Figure 3 E). The hybrid materials exhibited the rheological properties associated with the dynamic network, while the loss factor ( G ´´/ G ´), stress relaxation, and shear‐thinning could be tuned by varying the ratio between the two types of crosslinking, as the covalent and noncovalent bonds possessed different bond energy. Self‐healing property of hybrid hydrogels was tested by performing step‐strain measurements (Figure 3 G). At high strain of 1500% both hybrid hydrogels yield, with G ' and G ” showing rapid decrease and at low strain of 1%, the hydrogels rapidly recovered to their original state. This behavior was stable over many cycles, demonstrating the stable self‐healing ability of the hybrid network. Interestingly, under 1500% shear strain, the loss factors were increasing over the cycles (Figure S20, Supporting Information), reflecting enhanced shear‐thinning property. We speculate that because the dynamic network is reformed in each self‐healing cycle, the viscosity of the initial pre‐gel solution can be affected by the network rearrangement under very high shear stress treatment. The self‐healing property can also be highlighted by an electric circuit experiment containing the hybrid PEDOT:Hep hydrogel (Figure 3 H). The hydrogel was cut into two pieces. The LED light could fully recover after putting the two pieces in proximity without applying extra force. The shear‐thinning and self‐healing hybrid conductive hydrogels could be easily extruded through a microsyringe needle with an inner diameter of 200 µm (Figure 3 F, Movies S1–S3, Supporting Information). Moreover, the materials could also be printed on glass and tissue using a 3D microextrusion printer (custom‐made) (Figure 1 J, K, Movies S4 and S10, Supporting Information). Injection under biological wet conditions is extremely attractive for many biomedical utilities. Therefore, we tested the injection of the hybrid hydrogels into aqueous solution (Movies S1 S2, Supporting Information). The extruded filaments remained intact. Strikingly, the filaments of hybrid PEDOT:Hep hydrogel stuck to each other and formed an entangled cluster, while the PEDOT:PSS filaments remained separately and sunk to the bottom of the test vessel filled with water. This surprising observation suggested that the hybrid PEDOT:Hep hydrogel could fulfill the requirements not only for underwater injection but also for biological tissue‐relevant wet surfaces adhesion. The adhesive properties of the hydrogels were investigated by the use of pull‐off tests, measuring the maximum force at the peak of the force–distance curve (Movies S5 and S6, Supporting Information). The noncovalent hydrogels showed weak adhesive strength <1 kPa, while the dynamic covalent hydrogel GC/HA‐ADL exhibited remarkably enhanced adhesiveness of 4. 9 kPa ( Figure 4 A, B). Interestingly, by adding PEDOT:PSS or PEDOT:Hep to the GC/HA‐ADL hydrogel, the resulting hybrid hydrogels exhibited the highest adhesive strength. Further increasing the PEDOT:PSS or PEDOT:Hep concentration caused reduced adhesiveness. The pull‐off strength measurements are in good agreement with the elastic moduli obtained from tensile tests (Figure S16, Supporting Information), which measure the cohesion strength of materials. Similarly, increasing the noncovalent proportion of the network by increasing PEDOT:PSS or PEDOT:Hep concentration also reduces the cohesion strength. Shifting the hybrid network into the direction of noncovalent crosslinking caused weakened cohesion (Figure 4 B, C), as these hydrogels have also shown reduced storage modulus. The gelation process involves the formation of the double network including dynamic covalent and dynamic noncovalent networks. As the amino group of GC can either form a neutral imine bond with HA‐ALD or form electrostatic interaction with heparin or PSS, increasing the PEDOT:PSS or PEDOT:Hep content reduces the covalent portion of the network. Thus, these two dynamic interactions will affect both the gel–substrate surface interaction (surface adhesion) and the mechanical strength of gels (cohesion), while neither covalent hydrogel nor noncovalent hydrogel shows high adhesiveness. In comparison to adhesion under dry conditions, a fully hydrated substrate prevents the contact (adhesion) of hydrogel associated with the noncovalent interaction. Therefore, it is notoriously difficult to develop underwater adhesives based on hydrogels. In addition to the observation from the underwater injection and the pull‐off measurements, manual handling of the materials also suggested that the hybrid PEDOT:Hep hydrogel was the most adhesive material of all our fabricated and tested hydrogels. Polysaccharides have been found in many adhesive materials, also including natural ones, 31, 32, 33 as they present a rich source of hydrogen bond donors and acceptors. By replacing the hydrophobic polystyrene backbone with polysaccharide chain, the hybrid PEDOT:Hep hydrogel allowed for more contact adhesion to highly hydrated surfaces. Figure 4 The adhesive property of electroconductive hydrogel. A) Adhesion pull‐off testing of PEDOT:Hep hydrogels, where the adhesive strength is plotted against the extension. Inset: Detail of the experimental setup where the hydrogel adhered between two glasses that were pulled apart at a constant speed. B) The adhesive strength of the hydrogels. (Statistically significant differences are shown with asterisks * p < 0. 05, ** p < 0. 01, *** p < 0. 001, mean, error bar, n = 3). C) Scheme of the adhesion measurement, which shows the behavior of the hydrogel as the extension increases in correspondence with the adhesion force curve plotted against the extension (0. 5% PEDOT:Hep 0. 5% HA‐ALD 1% GC). D) Double peeling force per band‐shaped adhesive width obtained while peeling on porcine myocardium tissue band from the other glued by the different conductive hydrogels. (Inset: Photo of the double peeling test setup). E) The calculated interfacial toughness corresponding to (D) (n = 4). Statistically significant differences are shown with asterisks * p < 0. 05, ** p < 0. 01, *** p < 0. 001, mean, error bars, n = 4. F) lap‐shear strength of hydrogels and commercial fibrin glue sheared on flat porcine myocardium tissue (n = 4). G) Image of the lab‐shear experimental setup and adhesion mechanism of hybrid covalent/non‐covalent dynamic network in conductive hydrogels. H) (i) Adhesion of the electroconductive hydrogel (0. 5% PEDOT:Hep 0. 5% HA‐ALD 1% GC) on muscle tissue (chicken heart), (ii) the hydrogel remains adhered to the muscle tissue under flushing with water and (iii) under water. The adhesive properties were additionally characterized using peeling tests and lap‐shear (Figure 4 D–F) with hydrogels on porcine myocardium tissues. In both tests, clinically used fibrin glue and hybrid PEDOT:Hep/GC/HA‐ALD hydrogel showed similar adhesive performance, which is remarkably higher than that of hybrid PEDOT:PSS/GC/HA‐ALD hydrogel or GC/HA‐ALD hydrogel. We have also performed repeated adhesion tests with lap‐shearing of the same sample (Figure S15, Supporting Information). After three cycles, a moderate loss of adhesion strength was observed for hybrid PEDOT:Hep/GC/HA‐ALD hydrogel. As shown in Figure 4 G, electrostatic interactions, hydrogen bonds, hydrophobic interactions, and Schiff‐base formation occur in the hydrogel network, as well as at the interface between the substrate (e. g. , mammal skin tissue) and hydrogel. The successive disruption, rearrangement, and reformation of the network when pulling a viscoelastic bulk material result in the adhesive property, as all the reactions are dynamic and reversible during several repeated cycles of attachment and detachment. Because of the difference in bond energy, to break the dynamic imine covalent bond is more difficult than disrupting the noncovalent interactions. Because the stability of an adhesive in aqueous environment is very important for its clinical suitability, we also measured hydrogel shear stress of hydrated samples. As shown in Figure S17 (Supporting Information), both fibrin glue and hybrid PEDOT:Hep/GC/HA‐ALD hydrogel show a moderate decrease of adhesive strength after immersing the materials in water for 6, 12, and 24 h. The reduction in shear stress after 6 h was mainly caused by the fact that the tissues swelled and became fragile after the long treatments in aqueous solution. After either the reversible test or underwater treatment, the hybrid PEDOT:Hep/GC/HA‐ALD hydrogel still possesses enough adhesive strength (>800 Pa in the lap shear test on porcine myocardium) for biomedical applications. The underwater adhesion of the hybrid PEDOT:Hep hydrogel to tissues was further evaluated after gelling in situ on the surface of the chicken myocardium (Figure 4 H‐i–iii). After rinsing with water and floating in a water bath with strong stirring (240 rpm min −1 ), the hydrogel remained intact and adhered tightly to the chicken heart tissue without detachment, suggesting a considerable potential for applications in vivo. Remarkably, the excellent binding strength of the material was able to withstand heavy water flushing of hydrogel on the tissue (Figure 4 H; Movies S7 and S8, Supporting Information), suggesting that the hydrogel as a potent adhesive for a wide range of applications. The hydrogels were electrochemically studied in a three‐electrode system. The hydrogels showed a stable electrochemical behavior after many cycles (>10) of voltage ramping. It is notable that the cyclic voltammetry (CV) curves of the hybrid PEDOT:Hep or PEDOT:PSS hydrogels were typical of PEDOT containing polymers ( Figure 5 A), showing the characteristic oxidation and reduction peaks. The hybrid PEDOT:Hep hydrogels showed an anodic peak (Epa) at ≈0. 63 V and a cathodic peak (Epc) at 0. 03 V. Increasing the concentration of PEDOT:Hep from 0. 2% to 0. 5% resulted in the increase of the current at both peaks. In the case of PEDOT:PSS hydrogels, the Epa was observed at 0. 65 V and the Epc at 0. 02 V. Increasing the PEDOT:PSS concentration from 0. 2% to 0. 5% also caused enhanced current at both peaks (Figure 5 D). The electrochemical impedance spectroscopy (EIS) was used to confirm the electrical performance of the materials. The EIS is presented as Nyquist plots (Figure 5 B). Upon increasing the concentration of conductive polymer (either PEDOT:Hep or PEDOT:PSS), reduced impedance for all the frequencies was measured. The conductivity of hybrid PEDOT:PSS hydrogel was slightly higher than the hybrid PEDOT:Hep hydrogel when same concentrations of PEDOT polymer were used (Figure 5 C). Figure 5 Electrochemical characteristics of the dual cross‐linked electroconductive hydrogels. A) Cyclic voltammograms (current density vs the potential) of hydrogels with different concentrations of PEDOT:Hep, in PBS at a scan rate of 50 mV s −1. B) Electrochemical impedance spectroscopy ( Z vs frequency) of hydrogels with different concentrations of PEDOT:Hep (inset: Nyquist plot of the hydrogels). C) The conductivity of hydrogels (means, error bars, n = 3). D) The current density at the cathodic and anodic peak of the voltammograms of hydrogels with different concentrations of PEDOT:Hep. E) The relative current variation as function of time under a continuous cyclic tensile loading−unloading. Inset: Images of the compression and relaxation processes of the hydrogel, which undergoes a rapid recovery of its original shape to its a compressive deformation. F) Changes in the relative resistance during the bending of a finger at of angles at 0° and 90°. Owing to the hydrogels' adhesiveness, electrical conductivity, and self‐healing capability, we investigated their response to deformation, for the potential applications as sensors in wearable devices and implants. As illustrated in Figure 5 E, we measured the response of conductivity to loading/unloading cycles. Compressing the hybrid PEDOT:Hep hydrogel resulted in an instant increase of measured current, while the change was reversible immediately after removing the stress. In another experimental setup (Figure 5 F), hybrid PEDOT:Hep hydrogel was attached to a trigger finger covered by gloves, in order to monitor finger movements electronically. Bending the finger at angles from 0° to 90° caused stretching of the hydrogel, which led to instant increase in resistance, while the change could be reversed immediately after moving back to the original position. The hybrid PEDOT:Hep hydrogel and PEDOT:PSS hydrogel exhibited various features associated with the dynamic network, to be printable and injectable, possessing stress relaxation and self‐healing properties, while PEDOT and polysaccharides led to electroconductivity and degradability, respectively. The hybrid network also generated new material properties. The hydrogels were mechanically stable, remained intact over 10 cycles of compression or bending, as well as water flushing. In contrast, both the noncovalent PEDOT:polymer/GC hydrogel and the covalent GC/HA‐ALD hydrogels were more fragile and broke easily into small pieces. Moreover, the hybrid materials have shown strong adhesion to various surfaces, including animal tissues. As their stiffness could match that of soft tissues, we investigated their cytocompatibility for their potential clinical applications, such as adhesives or implants. Strong adhesion requires not only noncovalent and covalent bonding but also physical interpenetration, thus involving interfaces with cells in both 2D and 3D. Therefore, we investigated the culture of myoblast cells either encapsulated in the soft conductive adhesives or seeded on the surfaces. Mouse skeletal myoblast line C2C12 cells were embedded in dynamic covalent hydrogel GC/HA‐ALD, noncovalent hydrogels PEDOT:PSS/GC and PEDOT:Hep/GC, and hybrid PEDOT:PSS and PEDOT:Hep hydrogels. The covalent hydrogel led to the lowest cell growth, while the cells embedded in noncovalent PEDOT:Hep/GC hydrogel exhibited the highest cell density after culturing for 7 d ( Figure 6 A). Interestingly, the myoblasts formed large clusters in the heparin‐containing hydrogels, especially in the noncovalent PEDOT:Hep/GC hydrogel. Cell viability was assessed by live/dead cell staining 3 and 7 d after the set up (Figure 6 B). All hydrogels exhibit high cytocompatibility, while a few cell deaths could be observed in the covalent GC/HA‐ALD hydrogel. Figure 6 LSM images of electroconductive hydrogel induced myogenesis of C2C12 myoblast cells. A) Immunofluorescence staining of the cells at day 3 and day 7 of the nuclei (DAPI) and actin (phalloidin). B) Live–dead staining of mouse myoblasts cultured in hydrogels at day 3 and day 7, with Calcein‐AM (viable) and propidium iodide (dead). C2C12 cells could also adhere and proliferate on all types of hydrogels. Upon subjected to differentiation condition, myogenesis into mature and multinucleated myotubes was evaluated by immunofluorescence staining for the myogenic differentiation marker troponin T. Remarkable differences in terms of myogenic differentiation were observed among the myoblasts cultured on different substrates. The C2C12 myoblast cells seeded on noncovalent hydrogel films differentiated into multinucleated myotubes, revealing tube‐like morphology. 34 On covalent nonconductive GC/HA‐ALD hydrogel, differentiated cells did not show myotube‐like morphology ( Figure 7 A). Interestingly, cells differentiated into mature myotubes with a long myotube‐like morphology on conductive hydrogels. Fusion index, troponin T‐positive area, and myotube length were determined with ImageJ software (Figure 7 B–E). The fusion index was significantly higher when the C2C12 myoblasts were seeded on the noncovalent PEDOT:Hep/GC hydrogel, as compared with other materials. The fusion index agreed quite well with the calculated myotube number and length as well as with the troponin T‐positive area. For both noncovalent and hybrid hydrogels, heparin‐containing materials led to a higher fusion index as compared to the PSS‐containing materials. While conductive polymer, such as PEDOT, cannot be avoided to generate electroconductive matrix, by replacing PSS with heparin as counterion in the PEDOT synthesis, the PEDOT:Hep material could probably provide a more ECM‐like environment for myogenesis by maximizing the polysaccharide content. Figure 7 Electroconductive hydrogel induced myogenesis of C2C12 myoblast cells. A) Troponin T (TnT) and nuclei immunofluorescence staining of mouse skeletal myoblasts at day 7 of differentiation. Scale bar: 100 µm. B–E) Morphometric parameters of C2C12 cells at day 7 of differentiation on the hydrogels and quantified with ImageJ software. B) Fusion index. C) Myotude number. D) Myotube length. E) Myotube cover area. Statistically significant differences found for HA‐ALD + GC hydrogel with other electroconductive hydrogels (###) p < 0. 001. Statistically significant differences were also found between PEDOT:PSS + HA‐ALD + GC and PEDOT:Hep + HA‐ALD + GC for (§§§) P < 0. 001, (§§) P < 0. 01. Columns represent means and error bar standard deviation ( n = 9). To investigate the biocompatibility of the electroconductive soft adhesives, we studied the hydrogels either as injectable materials or for topical application with immunocompetent mice. The hybrid PEDOT:Hep hydrogel extruded from a syringe adhered very well on mouse skin (Figure S11A, Supporting Information). When the mouse was running in the cage, the hydrogel touched and stuck to the wall, and was then torn apart by the mouse, as the electroconductive hydrogel also has very good adhesive property to plastic surface (Figure S11B and Movie S9, Supporting Information). This was observed several times in a time period of 30 min. After that, the material was removed by a tweezer, no adverse skin reaction occurred in the following 2 weeks. GC/HA‐ALD hydrogels, hybrid PEDOT:PSS hydrogels and hybrid PEDOT:Hep hydrogels were injected subcutaneously in immunocompetent nude mice, in groups of three mice each (Figure S18, Supporting Information). Degradation in vivo represents an important parameter to evaluate the bioadhesives. While a quickly degradable material can be used as temporary adhesive, slow degradation rate would suggest its application for long‐term uses. We applied magnetic resonance imaging (MRI) to follow the degradation ( Figure 8 A, B). Interestingly, the degradations of hydrogels in vivo were remarkably faster than those observed in the in vitro experiments (Figure 2 F). Hybrid PEDOT:Hep hydrogel degraded quickly, with only 4% of the initial volume left 1 d after the injection and full degradation after 7 d. Hybrid PEDOT:PSS hydrogel degraded remarkably slower, while 10% of the initial volume remained in the tissue after 11 d. Degradation of GC/HA‐ALD hydrogel was slower than that of hybrid PEDOT:Hep hydrogel and faster than that of hybrid PEDOT:PSS hydrogel. For all hydrogels, the degradations are faster than the observed degradation by lysozyme and hyaluronidase in vitro. This could be explained by the presence of heparinase in the skin, 35, 36 which is able to degrade the heparin component in hybrid PEDOT:Hep hydrogel and leads to the fast degradation. In contrast, hybrid PEDOT:PSS hydrogel is more stable, most likely caused by the synthetic PEDOT:PSS component. The overall faster degradation in vivo compared to the in vitro situation could also be caused by the mechanical stimulation induced by movements of the mice. Figure 8 In vivo evaluation of hydrogel degradation and biocompatibility. A) Localization of hydrogels and inguinal lymph nodes by MRI. B) Volume determination of hydrogels by MRI, */° p < 0. 05; **/°° p < 0. 001; n = 3; Mean ± SD; one‐way ANOVA, Bonferroni post‐hoc test, * HA‐ALD + GC versus HA‐ALD + GC + PEDOT:PSS, HA‐ALD + GC + PEDOT:Hep; ° HA‐ALD + GC + PEDOT:PSS versus HA‐ALD + GC + PEDOT:Hep. C) Volume determination of inguinal lymph nodes at hydrogel injection site by MRI, compared to negative (untreated) and positive (TPA injection) control, n = 3; Mean ± SD; one‐way ANOVA, Bonferroni post‐hoc test. D) Comparison of volumes of lymph nodes at the injection site compared to site without hydrogel injection. E) Representative immunohistological images of markers for pan‐macrophages (CD68), M2 macrophages (CD206), inflammation (COX‐2), matrix remodeling (TG‐2), and angiogenesis (VEGF), cell nuclei in blue and positive immunohistological staining in red, blue line indicates hydrogel–tissue interface. F) Quantification of immunohistological stainings (positively stained area related to cell nuclei area). To evaluate the foreign body reaction of the mice to the injected materials, volumes of inguinal lymph nodes at hydrogel injection site and reference site without injection were determined by MRI (Figure 8 A, C, and Figure S19C, Supporting Information). The volumes were compared to negative control (untreated) and positive control (inflammation induced by 12‐O‐tetradecanoylphorbol‐13‐acetate, TPA). Lymph nodes at the hydrogel injection sites tended to be larger than the negative control, but smaller than the positive control. Lymph nodes at the reference site were not enlarged. Among the three hydrogels, hybrid PEDOT:Hep hydrogel and GC/HA‐ALD hydrogel induced only minor increase of lymph node size, while hybrid PEDOT:PSS hydrogel caused relatively more lymph node enlargement. Samples for histological analysis were prepared 11 d after the injections. At this time point, GC/HA‐ALD hydrogel has been completely degraded. Hybrid PEDOT:PSS hydrogel was degraded to 10% of the initial volume and hybrid PEDOT:Hep hydrogel was degraded, but left some dark spots in the tissue (Figure S19D, E, Supporting Information). The dark PEDOT component has not been fully degraded, though no intact hydrogel remained, as seen by MRI. No fibrous capsule formation was induced by the hydrogels, as capsule thickness around hybrid PEDOT:PSS hydrogel (154 ± 30 µm) was lower than negative control (245 ± 46 µm) (Figure S19E, Supporting Information). Study of specific tissue response showed an accumulation of macrophages (CD68+) around and in the hybrid PEDOT:PSS hydrogel, which mainly displayed alternative activation (CD206+) and, therefore, act in an anti‐inflammatory manner and support tissue repair (Figure 8 E, F). GC/HA‐ALD hydrogel and hybrid PEDOT:Hep hydrogel did not induce macrophage accumulation at this time point, as their degradation has already been completed (Figure 8 E, F). Accumulation of macrophages during hydrogel degradation has been observed previously. 37 None of the three hydrogels induced an adverse inflammatory reaction (cyclooxygenase‐2 negative (COX‐2‐)), while no signs of matrix remodeling (transglutaminase‐2 negative (TG‐2‐)) and angiogenesis (vascular endothelial growth factor negative (VEGF‐)) were detected (Figure 8 E, F). Covalent crosslinking could improve the mechanical strength and cohesive force of a noncovalently assembled hydrogel; however, it could also impair various properties associated with a dynamic network, e. g. , self‐healing, shear‐thinning, and injectability. In this work, we have developed a hybrid dynamic hydrogel system composed of noncovalent network and reversible imine crosslinking. The use of conductive polymers, not only led to electroconductivity, but also to a plethora of hydrophobic moieties, which are absent in most synthetic hydrogels. In many natural adhesives, such as proteins from burrowing ground frogs of the genus Notaden, sericin from the silkmoth Bombyx mori, and protein/saccharide associations from echinoderms, adhesion results from the synergy among different types of interactions, including hydrophobicity, electrostatics, and hydrogen bonds. The synthetic hybrid hydrogels present a rich source of positive and negative charges, hydrophobic/aromatic groups, and hydrogen bond donors/acceptors, while the dynamic network allows them to adapt and adhere to various surfaces. Moreover, the hybrid network of noncovalent and covalent crosslinking results in enhanced cohesion strength of the network, leading to adhesive materials with high mechanical stability. In summary, we have created an electroconductive hybrid polymer network, in which both covalent and noncovalent crosslinking are reversible and dynamic. While the electroconductivity can function as an electronic sensor, as well as a biophysical cue to promote myogenesis, the cytocompatible adhesive can also be used to encapsulate cells for cell‐based therapy as injectable therapeutics. The materials were also characterized in vivo using immunocompetent mice, both for topical application and as injectable materials. The materials showed high biocompatibility in both studies. Interestingly, by replacing PEDOT:PSS with a polysaccharide‐doped bio‐electroconductive polymer PEDOT:Hep, the resulting hydrogel has shown remarkably enhanced adhesive strength on tissues, as well as fast degradation in vivo. These soft and stress relaxing materials can match the mechanical properties of soft tissues and would be particularly attractive for applications in heart, muscle, and neuron diseases, as well as wound dressings. 3 Experimental Section Animal experiments were performed in accordance with the guidelines of German Regulations for Animal Welfare. The protocol was approved by the local Ethical Committee for Animal Experiments (reference number DD24. 1‐5131/450/16). Details of the materials and experimental methods used are available in the Supporting Information. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201802112
| 2,019
|
Advanced Science
|
A Reversibly Responsive Fluorochromic Hydrogel Based on Lanthanide–Mannose Complex
|
Abstract Fluorochromic materials that are dynamic in response to external stimuli are of great interest for the development of advanced sensors and luminescent materials. Herein, a design based on a lanthanide‐containing polymeric hydrogel possessing characteristic emission of lanthanides (Eu and Tb) and showing response to stimuli of metal ions is reported. The fluorochromic hydrogel is prepared using a lanthanide–mannose complex in gelation matrix. The lanthanide–mannose complex shows tunable fluorescent emission in response to Fe 2+, due to the inhibition of the “antenna effect” between metal ions and ligands upon stimulation. The fluorescent hydrogel shows reversible “On/Off” fluorochromic response to Fe 2+ /ethylenediaminetetraacetic acid (EDTA). Remarkably, the fluorescent hydrogel is proven nontoxic and biocompatible; and a proof‐of‐application as in situ 3D cell culture extracellular matrix with reversible fluorochromic “On/Off” switch upon Fe 2+ /EDTA is demonstrated. This reversibly responsive fluorochromic hydrogel demonstrates a way to fabricate smart optical materials, particularly for biological‐related applications where reversible response is required.
|
1 Introduction Adaptive optical materials that can adapt or actuate in response to external stimuli such as light, temperature, pH, metal ions, mechanical stress, and chemical stress are critical to the development of smart material systems. 1 In recent years, there are profound interests to develop new fluorochromic materials (light emission) for a broad range of applications in probes, sensors, indicators and display devices. 2 One typical fluorochromic materials are organic chromophores such as spirothiopyran and diethyl‐2, 5‐bis(4‐(trifluoromethyl)phenylamino)terephthalate, which showed isomerization and changeable conformations upon external stimuli such as mechanical force and temperature, resulting in fluorescence change. 3 Another important and promising fluorochromic materials are lanthanide (e. g. , Eu and Tb) based materials by virtue of the advantageous performance such as long luminescence lifetime, and sharp emission bands. 4 As lanthanide ions coordinate with ligands, ligands act as an “antenna” to absorb energy and then transfer the energy to lanthanide ions, which result in characteristic luminescence of lanthanide. 5 The fluorescence change of organic chromophores upon stimuli relies on the breakage of covalent bonds; in contrast, the fluorescence of lanthanide/ligand complexes is more sensitive to stimuli due to the dynamic noncovalent coordination between lanthanide and ligands. [[qv: 5c, 6]] Most lanthanide‐based materials were synthesized in organic solvents, which restricted the applications, in particular in biomedical engineering. Instead, water is considered as the most environmentally friendly solvent, which has been extensively used as media to synthesize biocompatible materials. 7 Moreover, water associating with molecular network generates three‐dimensional networked materials, known as hydrogels, which are increasingly applied in a variety of important applications such as tissue engineering due to their high water content, biocompatibility, and the gel‐like character. 8 Recently, the preparation of lanthanide‐based hydrogels has been successfully demonstrated. [[qv: 1g, 9]] For example, Gunnlaugsson and co‐workers prepared lanthanide–cyclen hydrogels and demonstrated application in diagnosis of urinary tract infection. [[qv: 9b]] He and co‐workers synthesized lanthanide–iminodicetate hydrogels which showed fluorochromic response to multistimuli. [[qv: 1g]] Li and co‐workers reported lanthanide luminescent hybrid hydrogels which possessed reversible phase transition upon photoirradiation. 10 Herein we create new lanthanide‐containing polymeric hydrogels possessing characteristic luminescence of lanthanides (Eu and Tb) and showing response to stimuli of metal ions. The fluorochromic hydrogel is prepared by using lanthanide–mannose complex in gelation matrix. It is noteworthy that both mannose and gelatin are biomolecules from life systems which ensures the nontoxicity and biocompatibility. Lanthanide–mannose complex shows tunable fluorescent emission upon Fe 2+, due to the inhibition of the “antenna effect” between metal ions and ligands upon stimuli. The luminescent hydrogel shows reversible “On/Off” fluorochromic response to Fe 2+ /EDTA. Remarkably, the luminescent hydrogel is proven nontoxic and biocompatible; and a proof‐of‐application as in situ 3D cell culture extracellular matrix with reversible fluorochromic “On/Off” switch upon Fe 2+ /EDTA for high contrast observation is demonstrated. 2 Results and Discussion Some early works have demonstrated that saccharide could form complexes with lanthanide mostly in organic solvents. 11 In our design strategy, we utilized saccharides as a category of ligands to coordinate with lanthanide ions (Eu 3+ and Tb 3+ ) and formed fluorescent Ln–Sac complexes in water ( Figure 1 a). Then, we incorporated Ln–Man complex into gelatin molecular network, realizing the formation of lanthanide–mannose–gelatin (Ln–Man–Geln) hydrogels for the reversible fluorescence “On/Off” switch response to stimuli (Figure 1 b). And the Ln–Man–Geln hydrogels can as a 3D cell culture matrix to realize reversible “On/Off” fluorescence of matrix during the observation of cell culture system (Figure 1 c). Figure 1 a) Scheme of the formation of lanthanide–mannose (Ln–Man) complex through the coordination of mannose with lanthanide ions (Tb 3+ /Eu 3+ ). b) Scheme of the formation of lanthanide–mannose–gelatin (Ln–Man–Geln) hydrogel by introducing Ln–Man into gelatin network, and the property of fluorescence “On/Off” upon Fe 2+ /EDTA. c) Illustration of the Ln–Man–Geln hydrogel as a 3D cell culture matrix for reversible fluorochromic “On/Off” switch upon Fe 2+ /EDTA. Herein we explored five typical monosaccharides including glucose, sorbose, galactose, fructose, and mannose for the formation of Ln–Sac complexes (Figure S1, Supporting Information), and tested their performance of fluorescence with varying the concentration of saccharides. All of them had a characteristic green emission of Tb 3+ at λ = 545 nm ( 5 D 4 → 7 F 5 ) upon 312 nm UV light irradiation. Ln–Man complexes had relatively high fluorescence intensity and as the concentration of mannose increased, the fluorescence intensity of Ln–Man complexes increased first, and tended to be stable (Figure S2 and Figure S3, Supporting Information). We therefore chose mannose as the ligand to coordination with lanthanide ions, and the concentrations of mannose and lanthanide ions were optimized. Fourier transform infrared (FTIR) spectroscopy was performed to study the molecular interactions between mannose and lanthanide ions (Figure S4, Supporting Information). Upon the formation of Tb 3+ ‐mannose (Tb–Man) complex, the peaks of mannose at 3433 (ν ‐OH ), 1417 (ν ‐CH ), and 1066 cm −1 (ν ‐CO ) shifted to 3417, 1384, and 1064 cm −1, and the peak of mannose at 2935 cm −1 (ν ‐OH ) produced two peaks at 2972 and 2975 cm −1 ; upon the formation of Eu 3+ ‐mannose (Eu–Man) complex, the peaks of mannose at 3433 (ν ‐OH ), 1417 (ν ‐CH ), and 1066 cm −1 (ν ‐CO ) shifted to 3423, 1641, 1384, and 1068 cm −1, and the peak of mannose at 2935 cm −1 (ν ‐OH ) produced two peaks at 2976 and 2920 cm −1. And from the matrix assisted laser desorption/ionization mass spectrometry (MALDI‐MS) spectrum of Tb–Man complex and Eu–Man complex, the lanthanide ions are able to coordination with three mannose ligands simultaneously ( Figure 2 a; Figure S5, Supporting Information). The Ln–Man complexes exhibited characteristic emission of Ln 3+ upon the irradiation of 312 nm UV light: Tb–Man complex exhibited strong green emission at λ = 545 nm ( 5 D 4 → 7 F 5 ) (Figure S6a, Supporting Information), and Eu–Man complex exhibited strong red emission at λ = 615 nm ( 5 D 0 → 7 F 2 ) (Figure S6b, Supporting Information). The fluorescent excitation spectra of Tb–Man complex and Eu–Man complex showed a broad band in the range of 300–325 nm, which showed the occurrence of ligand to lanthanide ions energy transfer (Figure S6c, d, Supporting Information). The luminescent lifetime (τ) of Tb–Man complex and Eu–Man complex was 1. 16 and 0. 41 ms respectively (Figure S7, Supporting Information). Figure 2 Ln–Man complex. a) MALDI‐MS spectrum of Tb–Man complex. b, c) Fluorochromic response of Tb–Man complex b) and Eu–Man complex c) upon different metal ions. The concentration of metal ions was 2 × 10 −3 m. The quenching efficiency of d) Tb–Man complex and e) Eu–Man complex upon different metal ions with varied concentrations. f) The illustration of the fluorescence “On/Off” response of Ln–Man upon Fe 2+. g, h) Fluorescence images of g) Tb–Man complex and h) Eu–Man complex upon Fe 2+ under 312 nm UV light. The concentration of Fe 2+ was 2 × 10 −3 m. We next examined the fluorochromic response of Ln–Man complexes upon metal ions by considering the competitive coordination with mannose between lanthanide ions and other metal ions. We investigated seven typical metal ions including Ca 2+, K +, Mg 2+, Mn 2+, Fe 2+, Ni 2+, and Na +, and varied the concentrations of metal ions (Figure 2 b, c; Figure S8, Supporting Information). With the addition of other metal ions into Tb–Man or Eu–Man complex, the fluorescence of Ln–Man complexes was quenched in different degrees. Fe 2+ showed the most remarkable quenching effect on the fluorescence of both Tb–Man and Eu–Man complexes (Figure 2 b–e). During this process of fluorescence quenching, Fe 2+ competed with Ln 3+. Fe 2+ as transition metal ion can provide more empty tracks to accept electronics of mannose; as a result, the “antenna effect” between mannose and lanthanide was weakened, and the emission intensity was strongly quenched. 12 Metal ions (M n + ) combine with ligand (L) to form complex (ML n ), and the equation is given as below (1) M + n L → ML n The stability constant of complex is formulated as follows (2) K = ML n M L n K Fe 2 + is larger than the stability constant of other metal ions (Ca 2+, K +, Mg 2+, Mn 2+, Ni 2+, and Na + ), and as a result Fe 2+ showed the most quenching effect on the fluorescence of Ln–Man than other metal ions. Furthermore, with the increase of the concentration of Fe 2+, the fluorescence intensity of Tb–Man and Eu–Man complexes gradually decreased (Figure S9, Supporting Information). With this mechanism, we presented the feature of fluorescence “On/Off” response upon Fe 2+ and the illustration was shown in Figure 2 f. Tb–Man and Eu–Man complexes that originally exhibited distinctive green and red fluorescence respectively, was quenched to colorless state when the Fe 2+ ion in the final complexes was 2 × 10 −3 m (Figure 2 g, h). We further incorporated Ln–Man into gelatin molecular network, realizing the formation of Ln–Man–Geln hydrogels for the reversible fluorescence “On/Off” switch response to stimuli. To construct the effective fluorescence and response system of hydrogel, we optimized the content of gelatin to 18 wt% according to the fluorescence intensity (Figure S10, Supporting Information). Tb–Man–Geln hydrogel exhibited characteristic green emission at λ = 545 nm, and Eu–Man–Geln hydrogel exhibited characteristic red emission at λ = 615 nm upon 312 nm UV light excitation ( Figure 3 a). The Tb–Man–Geln and Eu–Man–Geln hydrogels showed well‐distributed green and red fluorescence under fluorescence microscope, respectively (Figure S11, Supporting Information). The microstructures of Tb–Man–Geln and Eu–Man–Geln hydrogel were further investigated by scanning electronic microscopy (SEM), which exhibited a typical porous structure (Figure 3 c). We then studied the mechanical properties of Tb–Man–Geln and Eu–Man–Geln hydrogels with a rotational rheometer. Rheology data confirmed the formation of hydrogel, as the shear‐storage modulus ( G ′) was constantly higher than the shear‐loss modulus ( G ″) (Figure 3 d), showing similar mechanical strength to the hydrogel without complexes (pure gelatin, Figure S12a, Supporting Information). Gel stability of Tb–Man–Geln and Eu–Man–Geln was also confirmed via strain sweep mode. The G ′ of Eu–Man–Geln was dominant over G ″ throughout the strain range before 55. 8%, indicating that the hydrogel was stable in this strain range (Figure S12b, Supporting Information). The luminescent lifetime (τ) of Tb–Man–Geln hydrogel and Eu–Man–Geln hydrogel was 1. 18 and 0. 42 ms respectively (Figure S13, Supporting Information). Compared with the luminescent lifetime (τ) of Tb–Man complex and Eu–Man complex, these results suggested that the luminescence can be well retained in gelation hydrogels. Figure 3 Ln–Man–Geln hydrogels. The fluorescent spectra and images of a) Tb–Man–Geln and b) Eu–Man–Geln hydrogels under daylight and UV light (312 nm). c) SEM image of Tb–Man‐Geln. d) The rheology properties of Tb–Man–Geln. After the successful construction of the Ln–Man–Geln hydrogel systems, we tested the fluorochromic response property of the hydrogels upon metal ions ( Figure 4 ). Seven metal ions including Ca 2+, K +, Mg 2+, Mn 2+, Fe 2+, Ni 2+, and Na + with varied concentrations were tested. Both Tb–Man–Geln and Eu–Man–Geln hydrogels showed different degrees of fluorescence quenching upon metal ions (Figure S14 and Figure 15, Supporting Information), among which, Fe 2+ ion had the maximal fluorescence quenching efficiency. In addition, with the increase of the Fe 2+ concentration, the fluorescence intensity of Tb–Man–Geln and Eu–Man–Geln hydrogels showed continuous decrease (Figure 4 c, d). The fluorochromic response of Tb–Man–Geln and Eu–Man–Geln hydrogels upon Fe 2+ ion was reversible by adding EDTA based on the chelate effect of EDTA with Fe 2+, during which process, the coordination between mannose and lanthanide was strengthened, and the “antenna effect” was recovered. The coordination and dissociation of mannose/lanthanide in Ln–Man (Tb–Man) was monitored by FTIR. As shown in Figure S16 in the Supporting Information, the peaks of mannose at 3411 cm −1 (ν ‐OH ) shifted to 3446 cm −1 upon the addition of Fe 2+, and shifted back to 3424 cm −1 upon the addition of EDTA, indicating the dissociation and recoordination between ‐OH of mannose and lanthanide. The amount of Fe in the Ln–Man–Geln (Tb–Man–Geln) supernatant after adding Fe 2+ ( m 1 ) and after adding EDTA ( m 2 ) was quantified by spectrum analysis of inductively coupled plasma (ICP). Compared to the initial added amount of Fe 2+ (m 0 = 11. 2 µg), m 1 decreased to 2. 09 µg, indicating the binding replacing of lanthanide by Fe 2+ and the dissociation of mannose/lanthanide; while m 2 recovered to 5. 36 µg, indicating the formation of Fe 2+ –EDTA and the recoordination of mannose/lanthanide. We further optimized the concentration of Fe 2+ and EDTA to quench and recover the fluorescence of Tb–Man–Geln and Eu–Man–Geln hydrogels (Figure S17, Supporting Information), realizing the reversible fluorochromic “On/Off” switch. As demonstrated in Figure 4 b, we displayed the dynamic fluorochromic response by fabricating Tb–Man–Geln and Eu–Man–Geln hydrogels in test tubes, respectively. Upon adding of Fe 2+ and EDTA, the fluorochromic “On/Off” switch of Tb–Man–Geln and Eu–Man–Geln hydrogels could be clearly observed. We further prepared Tb–Man–Geln hydrogel in a Petri dish, and demonstrated the reversible process by pressing fingerprint with Fe 2+ ion and erasing the fingerprint with EDTA (Figure S17b, Supporting Information). The quench‐recovery process of fluorescence intensity of Tb–Man–Geln and Eu–Man–Geln hydrogels was monitored via fluorescent spectra, and the efficiency of recovery reached 88. 5% and 85. 2%, respectively (Figure 4 e, f). The fluorescence intensity was not fully recovered due to the dilution by adding aqueous solution. The fluorochromic “On/Off” switch upon Fe 2+ /EDTA was further examined in cycles, and the recovery efficiency maintained 63. 2% and 55. 7% after three cycles, respectively (Figure 4 g, h). Although there was a decrease in the fluorescence intensity, the “On/Off” switchable emission is reversible in response to Fe 2+. Figure 4 Reversible response of Ln–Man–Geln hydrogels upon Fe 2+ /EDTA. a) Illustration of the Ln–Man–Geln hydrogels as matrices for reversible fluorochromic “On/Off” switch upon Fe 2+ ions. b) Photographs of the fluorescence “On/Off” of Tb–Man–Geln and Eu–Man–Geln upon UV light (312 nm) irradiation. Gradual decrease in the fluorescence intensity of c) Tb–Man–Geln and d) Eu–Man–Geln by increasing the concentration of Fe 2+. Fluorescent spectra of the hydrogel, the hydrogel treated with Fe 2+, and the recovered hydrogel treated with EDTA about e) Tb–Man–Geln and f) Eu–Man–Geln. Fluorescence quenching and recovery cycles of g) Tb–Man–Geln and h) Eu–Man–Geln. Encouraged by the above demonstration of Ln–Man‐Galn hydrogels with remarkable fluorochromic responsiveness upon Fe 2+ /EDTA, we for the first time exploited the application of Ln–Man–Geln hydrogels as fluorochromic matrix for 3D cell culture. Since clear fluorescent visualization of the interactions between the cells and hydrogel during 3D cell culture was meaningful, 13 it is of great significance to develop a system of dynamically tunable fluorochromic hydrogel, realizing the “On/Off” switch to eliminate and recover the fluorescence of matrix during the observation. To test the biocompatibility of our hydrogels, we cultured smooth muscle cells (SMCs) with hydrogel extracts and carried out the cytotoxicity test. Both Tb–Man–Geln and Eu–Man–Geln groups showed the similar or even higher cell viability compared with control group, indicating that Tb–Man–Geln and Eu–Man–Geln hydrogels were noncytotoxicity to SMCs ( Figure 5 a). We then demonstrated the application of 3D cell culture utilizing the Eu–Man–Geln hydrogel, and verified the function of its switchable fluorescence “On/Off” response during the observation of cell culture. The SMCs suspension was added to Eu–Man–Geln hydrogel precursor solution and then resuspended in a Petri dish. Afterward, the cell‐containing Eu–Man–Geln hydrogel was constructed and the SMCs were encapsulated in the hydrogel. The illustration of Eu–Man–Geln hydrogel as a 3D cell culture matrix and its switchable fluorescence “On/Off” response upon Fe 2+ /EDTA during the fluorescence microscopy observation was shown in Figure 5 b, and the corresponding images were shown in Figure 5 c. The image in red channel exhibited the fluorescence of Eu–Man–Geln hydrogel, and green channel exhibited the SMCs stained with Calcein AM; by merging the two channels, clear visualization of the interactions between the cells and hydrogel was achieved. Furthermore, as a unique system of fluorochromic hydrogel, the fluorescence of Eu–Man–Geln hydrogel could be quenched by adding Fe 2+ ion and recovered by further adding EDTA. In Figure 5 c, the images demonstrated the process of fluorochromic “On/Off” switch of the hydrogel. We expect that this unique fluorochromic “On/Off” system has a potential in the high contrast observation of 3D cell culture with multiple fluorescence labeling: “fluorescence ON” of the hydrogel to verify the position of cells in the matrix; and “fluorescence OFF” of the hydrogel to eliminate the background fluorescence and observe the multistained subcellular structures. Figure 5 3D cell culture in Ln–Man–Geln hydrogel. a) Cytotoxicity test of Tb–Man–Geln and Eu–Man–Geln hydrogels. b) Illustration of the Eu–Man–Geln hydrogel as a 3D cell culture matrix for reversible fluorochromic “On/Off” switch upon Fe 2+ /EDTA. c) Fluorescence microscope images of Eu–Man–Geln hydrogel as a 3D cell culture matrix for reversible fluorochromic “On/Off” switch upon Fe 2+ /EDTA observed in red channel, green channel and merged. Scale bar represents 100 µm. 3 Conclusion To summarize, novel lanthanide–mannose complexes were designed and synthesized which showed fluorochromic response to stimuli of Fe 2+ in solution. Moreover, a gelatin polymeric network was introduced into the lanthanide–mannose complex to fabricate a luminescent hydrogel. The resulting lanthanide hydrogel possessed switchable “On/Off” fluorescence upon Fe 2+ /EDTA, which was realized by the dynamic coordination between lanthanide and mannose controlled by the stimuli of external ions. Furthermore, the fluorochromic hydrogel with good biocompatibility was demonstrated as a 3D cell culture matrix, and realized reversible “On/Off” fluorescence of matrix during the observation of cell culture system. We expect that those stimuli responsive luminescent complexes and hydrogels provide a basic and prospective platform to build functional materials for more applications in biological‐related fields. 4 Experimental Section Materials : All commercially reagents and solvents were used without further purification. All of the solvents used were analytical‐reagent grade. Eu 2 O 3 (99. 9%) and Tb 4 O 7 (99. 9%) were purchased from Aladdin. Europium nitrate and terbium nitrate were obtained by dissolving Eu 2 O 3 and Tb 4 O 7 in concentrated nitric acid. Glucose, galactose, fructose, sorbose, and mannose were purchased from Solarbio. Gelatin was purchased from Sigma. Iron chloride tetrahydrate (FeCl 2, >99. 7%), magnesium chloride hexahydrate (MgCl 2, >98%), manganese chloride (MnCl 2, >99%), potassium chloride (KCl, >99. 5%), sodium chloride (NaCl, >99%), cadmium chloride (CaCl 2, >74%), and nikel chloride (NiCl 2, >98. 5%) were purchased from Tianjin guangfu science and technology Ltd. NH 4 OH and HCl were purchased from Aladdin. Synthesis of the Ln–Man Complexes : In a typical procedure, a solution of Ln (NO 3 ) 3 (0. 1 mol L −1 ) (Ln = Tb, Eu) was added to the solution of ligand (20 wt% mannose) and then by the addition of NH 4 OH until the pH of mixture reached 8. Synthesis of the Ln–Man–Geln Hydrogels : For the fabrication of Ln–Man–Geln hydrogels, Ln–Man complexes were mixed with gelatin solution (18 wt%), and let stood for 10–15 min at room temperature. SEM Characterization : SEM experiments were performed by Hitachi S4800. Samples were put on conducting resin after drying with a freezer dryer, and then the samples were tested after spraying gold. Optical Characterization : Fluorescence emission spectra were recorded by BioTek Synergy/H1 microplate reader at room temperature. Rheological Tests : Rheological tests were carried out on an AR2000 Rheometer (TA Instruments). Strain sweep tests were carried out from 0. 1% to 100% at 20 °C and a fixed frequency (1 Hz). Stimuli–Response Study : For Ln–Man complexes, 50 µL metal ion solutions were added into 200 µL Ln–Man complexes, and 50 µL ddH 2 O was added into 200 µL Ln–Man complex as control. For Ln–Man–Geln hydrogels, 200 µL metal ion solutions were added into 800 µL Ln–Man–Geln hydrogels, and 200 µL ddH 2 O was added into 800 µL Ln–Man–Geln hydrogels as control. Each sample had three repetitions. Luminescent Lifetime Tests : The Luminescence decay time of the Ln–Man complexes and Ln–Man–Geln hydrogels was measured on an Edinburgh Instruments FS920P. Fluorescence Microscopy Images : High‐resolution images of Ln–Man–Geln hydrogels and cell culturing were obtained by fluorescence microscopy (Nikon). Samples were placed on Petri dishes. Cell Culture : For the 3D cell culture, 200 µL SMCs solution was added to 1 mL Eu–Man–Geln hydrogel precursor solution and then resuspended in a Petri dish. After several minutes of standing, the cell/hydrogel was constructed and cultured for 24 h at 37 °C in a CO 2 incubator. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201802342
| 2,019
|
Advanced Science
|
Gas‐Shearing Fabrication of Multicompartmental Microspheres: A One‐Step and Oil‐Free Approach
|
Abstract Multicompartmental microparticles (MCMs) have attracted considerable attention in biomedical engineering and materials sciences, as they can carry multiple materials in the separated phases of a single particle. However, the robust fabrication of monodisperse, highly compartmental MCMs at the micro‐ and nanoscales remains challenging. Here, a simple one‐step and oil‐free process, based on the gas‐flow‐assisted formation of microdroplets (“gas‐shearing”), is established for the scalable production of monodisperse MCMs. By changing the configuration of the needle system and gas flow in the spray ejector device, the oil‐free gas‐shearing process easily allows the design of microparticles consisting of two, four, six, and even eight compartments with a precise control over the properties of each compartment. As oils and surfactants are not used, the gas‐shearing method is highly cytocompatible. The versatile applications of such MCMs are demonstrated by producing a magnetic microrobot and a biocompatible carrier for the coculturing of cells. This research suggests that the oil‐free gas‐shearing strategy is a reliable, scalable, and biofriendly process for producing MCMs that may become attractive materials for biomedical applications.
|
Multicompartmental microparticles (MCMs) are under investigation for applications such as multidrug delivery systems, 1 in cell culturing technologies, 2 for multienzyme tandem reactions, 3 as micromotors, [[qv: 1f, 4]] for multitarget detection, 5 as multifunctional encoded materials, 6 etc. 7 The full potential of MCMs remains to be explored, which is partly due to limitations in the production of MCMs. Several technologies have been developed to fabricate compartmentalized particles, including sputter deposition technology, 8 using pickering emulsions, 9 layer‐by‐layer self‐assembly, 10 using microfluidics, 11 protonation deprotonation cycling, 12 electrohydrodynamic cojetting, 13 and centrifugation‐based methods. [[qv: 2b, 5a, 14]] Recently, microfluidic technology has been extensively explored to fabricate MCMs, as it allows the highest control over the morphology and complexity of particles. 15 However, a serious restriction of microfluidic technology often comes when sensitive biological molecules have to be encapsulated in MCMs, since the use of oils, photoinitiators, crosslinkers, surfactants, and UV‐irradiation is inevitably required. [[qv: 11b, c, 16]] To overcome these limitations, strenuous efforts including centrifugation‐based methods, [[qv: 2b, 14c]] multiplex coaxial flow focusing, 17 and in‐air microfluidics 18 have been explored to fabricate MCMs for biomedical purposes. Despite this, fabricating such “biofriendly” MCMs with a super compartmentalized and controllable morphology in a one‐step, green, and high‐throughput process remains challenging. Inspired by some studies that report on the use of gases to fabricate microcapsules or particles, 19 we took on the challenge of designing MCMs by gas‐shearing to avoid oils and make use of a homemade coaxial needle system ( Figure 1 ). As shown in this report, MCMs with a size ranging from tens to hundreds of micrometers can be easily obtained and the size can be precisely controlled by adjusting the gas (nitrogen) flow. The production of the MCMs can be scaled up by increasing the flow rate of the polymer solution. Importantly, we can easily fabricate MCMs containing as many as eight compartments (known as “eight‐faced” microspheres). As far as the authors are aware, current technologies are limited to six‐faced microspheres. [[qv: 14c]] The MCMs are produced through a “cytocompatible” oil‐free process, and cells can be encapsulated in the various compartments of the microparticles. Figure 1 Schematic illustration of the formation of the multifaced microspheres with different spray ejector devices (SEDs) and the assembly of the needle system in the SED‐8 configuration. The device for fabricating MCMs consists of four major parts (as shown in Figure S1a and Movie S1, Supporting Information): an injection digital pump, a collecting bath, a gas holder, and a homemade coaxial needle system known as a spray ejector device (SED). The SED is the most critical part of the device. As shown in Figure S1, Supporting Information, the liquid‐flow needles are inserted coaxially in a shell. The nitrogen gas is transported through the space between the needle and the shell, generating a shear force allowing the formation of droplets, termed “gas‐shearing” in this study. Based on the number of liquid‐flow needles, we designed the SED‐1, SED‐2, SED‐4, SED‐6, and SED‐8 configurations (containing 1, 2, 4, 6, and 8 needles, respectively). A central holder was added to optimally align the needles coaxially in the SED‐6 and SED‐8 configurations (and to keep the needle system stable). To evaluate the feasibility of the method, we first fabricated one‐faced microspheres, as schematically shown in Figure 2 a and Figure S1a, Supporting Information. The SED‐1 configuration was coupled to the syringe loaded with sodium alginate (Na‐Alg) solution. The flow rate of the polymer solution was controlled by the digital pump. A nitrogen flow (controlled by a rotameter) was used to induce dripping of the Na‐Alg droplets. In this process, the growing droplet experiences two competing forces (assuming the effect of gravity to be neglected): shear forces from the gas pulling the droplet down and forces which arise from the surface tension holding the droplet on the tip. This force balance is given by ŋ gas u gas d drop ∼ γd tip, where ŋ gas is the viscosity of the gas, u gas is the mean velocity of the gas, d drop is the mean diameter of the droplet, d tip is the diameter of the inner needle, and γ is the surface tension. 20 Initially, the surface tension dominates, though, as the droplet grows the shear force by the nitrogen flow becomes comparable. When the shear force by the nitrogen flow overcomes the resistance force by the surface tension, a droplet is detached from the liquid flow 21 (Movies S3 and S4, Supporting Information). The CaCl 2 aqueous solution in the collection bath solidified the liquid droplets into Ca‐Alg microspheres (Movie S2, Supporting Information). Figure 2 b shows good agreement between the outcome of the CFD (computational fluid dynamics) simulations for droplet formation and experimental images for a nitrogen flow of 0. 4 L min −1 and a liquid flow of 0. 1 L min −1 (Movies S3 and S4, Supporting Information). Furthermore, examples of the velocity field droplet formation are presented in Figure 2 b (bottom panel) (Movie S3, Supporting Information) to show the generation of the droplet due to the shear force caused by the velocity gradient. As shown in Figure 2 c, d, microspheres with a very low polydispersity can be fabricated. Figure 2 e, f clearly reveals that the nitrogen flow significantly dominates the size of the microspheres. Increasing the nitrogen flow results in smaller microspheres. Microspheres with a size between 55 and 1400 µm can be prepared simply by increasing the nitrogen flow from 0. 1 to 1 L min −1 without any modification to the device. We note that other methods do not allow the production of such a broad size range of particles. [[qv: 2b, 22]] In addition to the nitrogen flow, we also assessed the influence of the receiving angle and receiving distance, the space between the needle (core) and the shell, the flow rate of the Na‐Alg solution, and the concentration of the Na‐Alg and CaCl 2 solutions. Interestingly, as shown in Figure S2, Supporting Information (panel a), pendant and regular microspheres can be obtained using a receiving angle of 0° and 90°, respectively. As expected, using devices with a smaller space between the (needle) core and the shell, which increases the shear by the gas flow, results in smaller microspheres (Figure S2, Supporting Information, panel b). Generally, we can fabricate high‐quality microspheres using a receiving distance larger than 9 cm, a flow rate of the Na‐Alg solution between 2 and 8 mL h −1 and concentrations of Na‐Alg and CaCl 2 ranging between 0. 4–3. 0% and 2. 0–8. 0%, respectively. To score the production throughput of the microspheres, we collected Na‐Alg droplets on A 4 paper (Figure S3, Supporting Information), which indicated that about 2000 particles could be produced in only 20 s (at a pump speed of 3 mL h −1 ). We concluded that the throughput from the gas‐shearing method is at least competitive with microfluidic and centrifugation‐based methods reported on in previous studies. [[qv: 14c, 23]] Figure 2 The generation of one‐faced microspheres by gas‐shearing. a) Schematic illustration of an SED‐1 generating isotropic microspheres and the crosslinking reaction of the Na‐Alg droplets by Ca 2+ ions. b) Top panel: high‐speed snapshots of the droplet formation; the dashed yellow lines indicate the profiles of the droplet; Middle panel: CFD simulation of the droplet formation; Bottom panel: simulation of the velocity (m s −1 ) field. c) Optical image of the obtained microspheres loaded with Fe 3 O 4 nanoparticles. d) Size distribution of the microspheres. e) Optical images of one‐faced microspheres fabricated at various nitrogen flow rates. f) The relationship between the particle size and the nitrogen flow. The scale bars in all panels are 400 µm. As shown in Figure 3, the SED‐2, SED‐4, SED‐6, and SED‐8 configurations were designed to evaluate if the gas‐shearing approach would allow MCMs. Interestingly, two‐, four‐, six‐, and eight‐faced MCMs could be successfully produced from Na‐Alg solutions containing green or red polystyrene nanospheres (200 nm), as shown in Figure 3 and Figures S4 and S5, Supporting Information. The fluorescence intensity profiles indicate that the various compartments within a single microparticle are physically well separated (Figure 3 and Figure S6, Supporting Information). Figure 3 Generation of anisotropic multicompartmental microspheres, selectively loaded with green or red polystyrene nanospheres (200 nm) and produced by gas‐shearing of Na‐Alg solutions. The scale bars are 400 µm. To interpret the versatility of the gas‐shearing approach, we tested if microspheres could be obtained from other water‐soluble polymers (such as chitosan; CS), and organic‐soluble polymers such as poly‐acrylonitrile (PAN), cellulose‐acetate (CA), ethyl‐cellulose (EC), poly‐caprolactone (PCL), cellulose‐acetate‐phthalate (CAP), and polyurethane (PU). As Figure 4 a illustrates, CS microsphere formation occurs through ionic crosslinking (such as for alginate microspheres), whereas for organic‐soluble polymers, solidification of the droplets occurs through the exchange of the dimethylformamide solvent with water (Figure 4 b). Panel c in Figure 4 confirms the versatility of the gas‐shearing approach as various types of microparticles are easily obtained. Although the same fabrication conditions were used, the size of the microspheres was polymer dependent and could be ascribed to differences in viscosity and/or surface tension among these polymer solutions. It is interesting to observe that nanofibers can also be fabricated by simply increasing the nitrogen flow (Figure S7, Supporting Information). Figure 4 Versatility of gas‐shearing for the production of microspheres. Note that the SED, the nitrogen flow, and the flow rate of the polymer solution were identical in all the experiments. a) Schematic illustration of the ionic crosslinking (using thiamine pyrophosphate) of CS aqueous droplets into CS microspheres. b) Schematic illustration of the solvent exchange process resulting in PAN, CA, EC, PCL, CAP, and PU microparticles. c) The diameter of the obtained microspheres. The scale bars in panel a) and b) are 400 µm. To illustrate the functionality of the multiple compartments in the MCMs, we took on the challenge of encapsulating magnetic Fe 2 O 3 nanoparticles and cells in specific compartments of the MCMs. As shown in Figure 5 a and experimentally shown in Figure 5 b, processing Fe 2 O 3 containing Na‐Alg solutions through the needles of configuration SED‐8 allows fabrication of up to seven types of “asymmetric” (anisotropic) microspheres. As shown in Figure 5 c‐iii and Movie S5, Supporting Information, the Ca‐Alg MCMs composed of alternating magnetic and nonmagnetic compartments can be rotated in a highly controlled manner using an external magnetic field. Such rotating “microrobots” may become of use in biomedicine and tissue engineering. Figure 5 a) Schematic illustration of the fabrication process of anisotropic multifaced microspheres. b) Experimentally obtained anisotropic multifaced microspheres with various magnetic and nonmagnetic compartments. c) Schematic illustration of the fabrication process for eight‐faced symmetric microspheres (i). Microscope images of the eight‐faced symmetric microspheres (ii). Microscope images of the response of the eight‐faced microspheres in a rotating magnetic field (iii). The scale bars are 400 µm. Currently, cell‐loaded gel microspheres are widely generated through microfluidics where water–oil phases are processed. 24 However, harmful organic reagents are often used as the continuous phase. 25 Although efforts have been undertaken to minimize the exposure time of the cell‐loaded microgels to these harsh conditions, 26 cytotoxicity often remains a challenge. Therefore, for 3D cell culturing in microparticles, it is highly desirable to fabricate the microspheres in a simple high‐throughput way under oil‐free and surfactant‐free conditions. Therefore, we tested if the gas‐shearing approach allows mild encapsulation of cells in the various compartments of the MCMs. We first encapsulated HepG 2 cells into Ca‐Alg microspheres, as shown in Figure 6 a‐i. The cells are well distributed in each microsphere (Figure 6 a‐iii). To examine the biocompatibility of our method, we applied fluorescent staining to quantify the viability of cells in the microspheres. We observed cell clusters growing in the microparticles (Figure 6 a‐iii, v) and found that after 1 day, 96% of the HepG 2 cells in the microparticles were alive, which is superior to the cell survival measured using a nonaqueous processing system or other (such as the centrifugation‐based) methods. [[qv: 2b, 22a, 24a, 27]] Even after 7 days, the cell viability remained as high as 91% (Figure 6 a‐iv). Figure 6 a) One‐step fabrication of the one‐faced microspheres carrying cells i) Ca‐Alg microspheres with encapsulated HepG 2 cells. ii) Fluorescence microscopy images of the Ca‐Alg microspheres (the cells were stained with Calcein‐AM/PI) after 1, 4, and 7 days. iii) Number of cell clusters per microsphere. iv) Cell viability after 1, 4, and 7 days. v) Size of the cell clusters after 1, 4, and 7 days; *** p < 0. 001. b) Multifaced Ca‐Alg microspheres carrying cells. The HepG 2 and Hela cells were stained with DIO (green) and DiI (red), respectively. The scale bars in all images represent 400 µm. Finally, we aimed to evaluate if the gas‐shearing method allows the fabrication of MCMs containing different cell types (HepG 2 and Hela) in the various compartments of one single microparticle. Such highly controlled cell‐loaded MCMs with different cell types well separated from each other may be of interest for 3D cell coculturing. As shown in the left panel of Figure 6 b and experimentally shown in the middle panel of Figure 6 b and Figure S8, Supporting Information, HepG 2 and Hela cells (stained with DIO and DiI) were encapsulated in the Ca‐Alg MCMs and arranged for coculturing. The HepG 2 and Hela cells became well separated and ordered into complex geometries. These results indicate that our strategy can be easily used to obtain multiple microenvironments within a one single microparticle to precisely assemble different cell types within a confined micrometer‐sized volume. Note that, to our knowledge, the successful encapsulation of different cell types in eight‐faced MCMs has never been reported. We believe that the platform established here might provide an effective strategy to study the complex interactions between different cells. As optimal rheological properties of a cell matrix are needed to allow optimal cell growth, which may be cell type dependent, 28 one can anticipate that MCMs composed of compartments with different viscoelastic properties can further add value to the materials investigated in this study. In summary, a one‐step strategy, based on gas‐shearing, has been presented for the fabrication of MCMs composed of up to eight compartments. We show that this fabrication approach is highly versatile, as both aqueous and organic polymer solutions can be processed, whereas the morphology and size of the microspheres can be flexibly controlled using an appropriate SED and adjusting the gas flow. Our study suggests that the obtained MCMs may have highly versatile applications in bioengineering, especially as carriers for cells, which remains a key challenge to the progress of the field of tissue engineering. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201870020
| 2,018
|
Advanced Science
|
Tissue Engineering: 3D Fabrication with Integration Molding of a Graphene Oxide/Polycaprolactone Nanoscaffold for Neurite Regeneration and Angiogenesis (Adv. Sci. 4/2018)
|
In this article number https://doi. org/10. 1002/advs. 201700499, Yuanming Ouyang, Weien Yuan, Cunyi Fan, and co‐workers present 3D integration molding of a graphene oxide/polycaprolactone nanoscaffold. The nanoscaffold has high electrical conductivity, has excellent biocompatibility with multiple pores for free exchange of nutrients and water, and can support long‐term nerve regeneration in vivo after traumatic injury. This innovative complex is vital to functional neurite sprouting and angiogenic restoration in nerve tissue engineering.
|
No full text available
|
10. 1002/advs. 201900043
| 2,019
|
Advanced Science
|
DNA Nanotechnology Enters Cell Membranes
|
Abstract DNA is more than a carrier of genetic information: It is a highly versatile structural motif for the assembly of nanostructures, giving rise to a wide range of functionalities. In this regard, the structure programmability is the main advantage of DNA over peptides, proteins, and small molecules. DNA amphiphiles, in which DNA is covalently bound to synthetic hydrophobic moieties, allow interactions of DNA nanostructures with artificial lipid bilayers and cell membranes. These structures have seen rapid growth with great potential for medical applications. In this Review, the current state of the art of the synthesis of DNA amphiphiles and their assembly into nanostructures are first summarized. Next, an overview on the interaction of these DNA amphiphiles with membranes is provided, detailing on the driving forces and the stability of the interaction. Moreover, the interaction with cell surfaces in respect to therapeutics, biological sensing, and cell membrane engineering is highlighted. Finally, the challenges and an outlook on this promising class of DNA hybrid materials are discussed.
|
1 Introduction Embedded in a unique language, deoxyribonucleic acid (DNA) carries the lion's share of the hereditary information in living cells. Ever since Friedrich Miescher isolated DNA in 1869, 1 the scientific community extensively investigated its properties and possible applications. James Watson and Francis Crick identified the molecular structure of DNA in 1953, 2 starting the age of genetics and modern molecular biology. The Watson‐Crick base pairing rules provide DNA with unique self‐recognition and sequence programmability, which enabled DNA and DNA‐based materials to find their applications in biomedicine, which includes drug delivery, gene silencing, and diagnostics. Apart from that technologies have been developed to evolve DNA molecules, which strongly bind a wide variety of target molecules (aptamers) or exhibit catalytic activity (DNAzymes). 3, 4, 5, 6 As therapeutics, nucleic acids inhibit either DNA or RNA expression, thereby blocking the production of proteins related to a disease. 7 However, the clinical application of therapeutic nucleic acids (TNAs) is still facing limitations due to unsolved challenges regarding delivery. For instance, negatively charged cellular membranes act as a natural barrier to prevent entry of foreign polyanionic nucleic acids. Once inside the cell, DNases or RNases degrade foreign nucleic acids to prevent their integration into the genome. 8 TNAs further have to be delivered to the correct cells with minimal side effects to other cells. 9 When using TNAs as artificial receptors, the failed anchoring or insertion of the DNA in the cell membrane restricts its excellent recognition properties. These challenges potentially decrease the applicability of DNA reporting signals from the cell or tissue. The unique programmability gives DNA an edge over other molecules that interact with membranes, such as peptides, proteins, and small molecules. In order to realize successful insertion of DNA in the cell membrane and efficient delivery of TNAs both in vitro and in vivo, one of the most commonly used strategies is increasing the hydrophobicity of nucleic acids. To this end, DNA is chemically conjugated with hydrophobic moieties, resulting in DNA amphiphiles. Efficient and stable insertion into live cell membranes allows amphiphilic DNA conjugates to cross the cell membrane. 10, 11, 12, 13 Importantly, these DNA amphiphiles can be modified with additional functional groups that enable specific targeting and biocompatibility in vivo, providing them with a tremendous potential for biomedicine. 14, 15, 16, 17 To date, the synthesis and application of amphiphilic DNA conjugates have been well demonstrated and reviewed. 18, 19, 20 2 Synthesis of DNA Amphiphiles A DNA amphiphile is based on hydrophilic DNA that contains a covalently connected hydrophobic segment. 19 Usually, the hydrophobic moiety is a polymer or a small molecule. The lipophilic modifications of DNA can be achieved by conjugation at either the 3′‐ or 5′‐terminal, or within the DNA sequence, allowing the construction of complex structures. 21, 22, 23, 24 These hydrophobic moieties can be conjugated to DNA, either on a solid support during DNA synthesis or by coupling to already synthesized DNA units in solution. The first successful chemical synthesis of a dinucleotide was achieved in 1955. 25 Stable deoxynucleoside phosphoramidites were introduced as synthons in 1985, opening up the field. 26 Nowadays, solid phase synthesis (SPS) allows generating DNA fragments of up to 200 nucleotides. This technology allows functionalization or introduction of non‐natural nucleotides. 27 The fully automated synthesis can be precisely controlled, monitored, and is characterized by a high reproducibility. To broaden the scope of synthesis robots by introducing special solvents, catalysts, extreme reaction conditions or long reaction times, the automated process can be replaced by the syringe synthesis technique or in‐flask reactions to realize various modifications of the DNA with hydrophobic units. 20 Coupling of DNA with specific motifs in solution phase has been demonstrated as another highly versatile strategy, which was reviewed by our group before. 19 Solution phase synthesis is used for covalent bond formation between functional groups such as amines 28 or thiols, 29 with groups such as carboxylic acids 30 or maleimides. 31 However, aqueous solution coupling of DNA with hydrophobic molecules often results in low yields due to the solvent incompatibility of starting materials. To overcome this limitation, we reported a conjugation protocol for coupling of hydrophobic molecules to DNA with high efficiency. 32 By complexing DNA with positively charged quaternary ammonium surfactants, we neutralized the charge on the DNA, making it soluble in organic solvent. The organic phase coupling technique expands the number of possibilities to generate amphiphilic DNA hybrids. One of the most commonly used lipids in DNA amphiphiles is cholesterol. In addition to cholesterol or one of its derivatives, other synthetic single‐chain fatty acids, 33 steroid molecules, 34 α‐tocopherol, 35 hydrophobic polymers, such as poly(propylene oxide) (PPO), 21 or the π‐conjugated system porphyrin 36, 37 have been successfully introduced to DNA ( Figure 1 ). Hence, synthetic protocols to introduce a wide range of hydrophobic moieties into DNA at various positions are available, allowing for the exploration of new functionalities in nanotechnology. 38 Figure 1 Selected lipid‐oligonucleotide conjugates, exemplifying the variety of lipophilic residues that can be appended to DNA. Structures of DNA conjugated with, from top to bottom, cholesterol obtained via a 1, 3‐dipolar Huisgen's cycloaddition between alkyne modified cholesterol and 5′‐azido‐5′‐deoxythymidine 39 ; a single hydrocarbon chain obtained via a 1, 3‐dipolar Huisgen's reaction between alkyne modified C18 chain and 5′‐azido‐5′‐deoxythymidine 39 ; tocopherol obtained by covalent attachment to the 5′ end of the strand 40, 41 ; a single fluorocarbon chain obtained via a Huisgen's reaction between 5′‐azide deoxythymidine and propargylated fluorocarbon chain 42 ; a PPO chain obtained via a PPO phosphoramidite during SPS 21 ; double hydrocarbon chains obtained via a reaction of stearoyl chloride with 1, 3‐diamino‐2‐dydroxypropane 40, 43 ; and double fluorocarbon chains obtained by a diperfluorodecyl phosphoramidite during SPS. 44 3 Nanoscale Assemblies from DNA Amphiphiles DNA amphiphiles can be designed to assemble into a variety of nanoscale structures. In general, nanoscale structures can be constructed “top‐down” or “bottom‐up”: The bottom‐up approach makes use of assembling single molecules into nanostructures by intermolecular interactions, yielding a level of molecular control that is out of reach to a top‐down strategy. DNA amphiphiles that contain both hydrophobic moieties and nucleic acids possess advantageous features derived from the DNA part as well as from the hydrophobic moieties combined in one molecule. The Watson‐Crick base pairing rules that govern DNA nanotechnology allow the rational design of complex nanostructures which result in novel functions. This molecular technology is based on bottom‐up self‐assembly, which was initiated by Nadrian Seeman in the early 1980s and has been growing rapidly ever since. 45 Depending on the design, the structures can be 1D, 2D, or 3D. In addition, single‐stranded overhanging sequences in the final structure enable further functionalization by hybridization with complementary sequences. More detail on the assembly of DNA nanostructures and their emerging applications in areas such as biophysics, drug delivery, synthetic biology, can be found in ref. 41, 46. On the other hand, hydrophobic units in amphiphiles tend to microphase separate due to hydrophobic interactions. 47, 48, 49 This structural concept can be further combined with assembly mechanisms relying on electrostatic forces, 50 π–π stacking interactions, 51 hydrogen bonding and Van der Waals interactions. Hence, DNA amphiphiles have the ability to self‐assemble into predictable morphologies ( Figure 2 ), such as spherical micelles, rods, vesicles, and bilayers. 52 An inspiring example of engineering such morphologies was reported by Baglioni and co‐workers in 2007 53 : They synthesized nucleolipids in which the choline headgroup of phosphatidylcholines was replaced by a nucleoside, either uridine or adenosine. The resulting mole‐cules had a negatively charged nucleotide group as polar head. Depending on the length of the alkyl chains, globular micelles, flexible cylindrical aggregates, or bilayers were obtained from these nucleolipids. The shape of the amphiphile dictates the obtained structures: a short hydrophobic chain provides an amphiphile with a conical shape, resulting in globular micellar aggregates, while a long alkyl chain gives a cylindrical shape that results in wormlike micellar aggregates. The latter morphology is further modulated by improved orientation of the bases that interact with each other. Figure 2 Schematic models of self‐assembled lipids. A) Micelles are preferentially formed by lipids with a conical shape. B) Vesicles are composed of spherical lipid bilayers with a water core. C) Planar lipid bilayers are formed by lipids with a cylindrical shape. Reproduced with permission. 52 Copyright 2014, American Chemical Society. 3. 1 Micelles from DNA Amphiphiles When above its critical micelle concentration, DNA amphiphiles self‐assemble into micellar systems with nanometer dimensions. 54 This occurs spontaneously because the amphiphiles phase separate in aqueous media. Micellar structures are composed of a hydrophobic core and a hydrophilic DNA shell. 3. 1. 1 Formation and Structure of DNA Amphiphile Micelles DNA amphiphiles form spherical micelles with a diameter from 6. 7 to 36. 4 nm, as measured by atomic force microscopy (AFM) and dynamic light scattering (DLS). 33, 54, 55 Similar to inorganic nanoparticles, 56, 57, 58 the size of the spherical micelles can be regulated by adjusting the DNA or hydrophobic segments. AFM revealed that such micelles deform, depending on the hydrophobic segments attached to the DNA molecules. Amphiphiles with different DNA lengths or different lipids form micelles with tunable size, indicating a relationship between micelle size and length of the constituent segments. In this context, DNA polymerase can be utilized to control the size of micelles: Treatment of micelles consisting of DNA‐ b ‐PPO (PPO block covalently connected to the 5‐end of a 22 nt single‐stranded DNA) with the enzyme terminal deoxynucleotidyl transferase (TdT) increases the size from 10 to 23 nm, depending on the incubation time ( Figure 3 A). 59 Similarly, the use of enzymes to digest and ligate nucleic acids resulted in DNA amphiphiles containing dsDNA with molecular weights of up to three million Daltons. 60 These strategies offer post‐synthetic control over the growth of DNA nanostructures in aqueous medium. Furthermore, the size and stability of DNA amphiphile micelles is determined by the number of hydrophobic moieties: Increasing the number of nucleotides containing dodec‐1‐ynyl chains attached to the nucleobases resulted in smaller micelles with increased stability. The position of the hydrophobic nucleotide units in the short sequences proved to have little influence on micelle structure and stability. 33 Figure 3 A) Enzymatic growth of DNA‐ b ‐PPO micelles. Reproduced from ref. 59. B) Schematic representation of hybridization of DNA‐ b ‐PPO micelles with different DNA molecules. a) Base pairing with a short complementary sequence yields micelles and maintains the overall shape of the aggregates. b) Hybridization with long DNA templates results in rod‐like micelles. Reproduced from ref. 61. C) DNAzyme induced reversible transformation of the aggregate shape of a DNA‐brush block copolymer. Reproduced from ref. 62. D) Schematic of the mixed micelle architecture. Two amphiphilic block copolymers, DNA‐ b ‐PPO and PEO‐ b ‐PPO‐ b ‐PEO with the trade name Pluronic F127, form mixed micellar structure and this micelle can be stabilized by formation of a semi‐interpenetrating network in its core. Reproduced with permission. 63 Copyright 2010, Royal Society of Chemistry. E) Schematic representation of the drug delivery system based on DNA amphiphiles. a) Targeting units (red dots) that are connected to the complementary sequence of the micelles are hybridized to equip the nanoparticle surface with folic acid units. b) The anticancer drug (green dots) is loaded into the core of the micelles. Reproduced from ref. 55. F) Schematic of DNA micelle‐templated VC formation. Loading hydrophobic molecules (top, green) into micelle core and hybridization of a complementary DNA connected to functional moieties (bottom, red) to the DNA micelle. Reproduced with permission. 66 Copyright 2010, American Chemical Society. G) Photoinduced cross‐linking of self‐assembled DNA‐methacrylamide‐lipid micelles. Green dot between DNA and lipid represents methacrylamide molecules which can be crosslinked. Reproduced from ref. 65. Hybridization allows precise post‐synthetic control over the shape of a DNA micelle (Figure 3 B). The shape of micelles can be changed from spheres to rods by addition of complementary single‐stranded DNA to the DNA amphiphiles, forming double‐stranded DNA. 61 Morphology can be controlled reversibly with for example DNA‐brush amphiphiles that assemble into spherical micelles (≈25 nm) and contain a RNA nucleotide as an enzymatic cleavage site (Figure 3 C). 62 Mixing spherical micelles with a DNA‐based phosphodiesterase that is specific for the DNA sequence and cuts at the RNA site, resulted in a long cylindrical structure (>1000 nm in length). To facilitate a subsequent cylinder‐to‐sphere transformation, a 19‐base ssDNA sequence was added, which forms a 9 nt duplex with the truncated DNA of the cylinder shell. The reverse sphere‐to‐cylinder transition was achieved again by the addition of a complementary 19‐base ssDNA designed to invade into the shorter nine‐base duplex in the micelle shell. Thus, DNA is a superb tool for encoding supramolecular structure information allowing exquisite control over morphology of DNA amphiphiles. Our group synthesized an additional type of structure, based on a mixed hybrid micellar architecture (Figure 3 D). 63 Here, DNA‐ b ‐PPO and Pluronic F127 (a triblock copolymer with a PEG (polyethylene glycol)‐ b ‐PPO‐ b ‐PEG architecture) were combined. In this construct, the PPO from both DNA amphiphile and Pluronic copolymer formed the core of the micelles, while DNA from DNA‐ b ‐PPO and PEG from Pluronic were located in the corona. The resulting self‐assembled structures were finally cross‐linked by forming a semi interpenetrating polymer network in the micelle core. The PEG domain did not undermine the hybridization of DNA and the hydrophobic core could be loaded with hydrophobic drugs. The resulting aggregates exhibit the potential for combining block copolymers of different nature, facile functionalization of DNA amphiphiles by hybridization and the possibility for stabilization of such aggregates by polymer network formation within the micelle core. As a result, micelles were obtained that are stable in regard to dilution, temperature increase and the possibility for attaching conveniently targeting units. Likely, such a PEG corona shields the DNA backbone and improves the biocompatibility and immune compatibility of the mixed hybrid micelles, vide infra. 3. 1. 2 Functionalization and Features of DNA Amphiphile Micelles DNA amphiphile micelles can be functionalized to introduce new properties. Micellar aggregation of DNA amphiphiles aligns the single‐stranded DNA in its corona, which allows DNA‐templated organic reactions to proceed in 3D space. Therefore, the ssDNA of the corona needs to be hybridized with sequences, which are equipped with reactants. 55 (Figure 3 E) Moreover, DNA amphiphiles were functionalized to a high degree for combined mRNA detection and gene therapy in molecular beacon micelle flares (MBMFs), which are self‐assembled diacyllipid‐molecular‐beacon DNA conjugates. 64 The MBMFs showed efficient cell uptake, enhanced enzymatic stability, excellent target selectivity, and superior biocompatibility compared to pristine DNA. Diperfluorodecyl‐DNA conjugates allow further improvement of target binding affinity and enzymatic resistance by virtue of the physicochemical properties of fluorination. 44 However, loss of integrity of micelles compromised the recognition ability of the aptamer when interacting with cells. Therefore, the same group developed a more stable cross‐linked DNA‐methacrylamide‐lipid micelle (X‐DLM) system (Figure 3 F), which incorporates a methacrylamide functionality between the hydrophilic and hydrophobic portions of DNA‐lipid amphiphiles that can be cross‐linked after self‐assembly in aqueous solution. 65 This X‐DLM system offers further improved stability in the cellular environment and better specificity regarding cell recognition. Besides cross‐linking of DNA amphiphiles, these nano‐objects can be encapsulated via a facile self‐assembly process. Therefore, the nucleic acid micelles were incubated with virus capsid (VC) proteins (Figure 3 G). 66 In this approach, the negatively charged DNA particles induced capsid formation, allowing the entrapment of oligonucleotides as a constituent part of the micellar template. The preloading of entities in the core or by hybridization of micelles enables encapsulation of various small molecules inside VCs, which marked a significant step forward in chemical virology due to the flexibility of loading these protein nanocontainers with various payloads. Thus, DNA amphiphiles form micelles that are tunable, versatile, and allow realization of functions. 3. 2 Liposomes from DNA Amphiphiles Next to micelles, amphiphilic DNA molecules can be aligned to form liposomes or bilayers, similar as indicated for conventional surfactant molecules in Figure 2 : Liposomes are flat bilayer sheets folded to form closed spherical objects, with the structure of the assembly determined by the conical shape of the DNA amphiphiles. 3. 2. 1 Formation and Structure of DNA Amphiphile Liposomes Nucleic acid functionalization of lipids allows additional control over lipid self‐assembly through specific interactions among the polar heads. As in micelles, the hydrophobic lipid tail and hydrophilic DNA head combined determine the phase behavior and aggregate microstructure. 53 DNA amphiphiles that form vesicular structures can be made for example by linking poly(butadiene) covalently to poly‐cytidine during solid phase synthesis. 67 The resulting amphiphilic copolymer self‐assembled into 80 nm vesicles as demonstrated by TEM and confocal microscopy. By using a functional DNA moiety as head group, one can induce more complex behavior: Conjugation of the lipid tail with a DNA sequence that forms an i‐motif renders the liposome structure pH sensitive upon acidification ( Figure 4 A). 68 The C‐rich DNA segment undergoes a structural change from random coil ssDNA to an i‐motif structure upon acidification (pH = 5), triggering the transformation of the vesicles into an entangled 3D network. This process was reversed when the pH was increased to 7. 3. This structure allowed the encapsulation of a hydrophobic molecule and a pH‐triggered release, showing that these DNA amphiphile systems can be engineered to be sensitive to external stimuli. Figure 4 A) Illustration of working principle of reversible pH‐responsive DNAsome. At pH 7. 3, C rich DNA‐PE spontaneously forms a DNAsome. When pH is lowered to 5, the i‐motif structure forms and the morphology of the DNAsome transforms to entangled 3D networks. Reproduced from ref. 68. B) Schematic of the frame‐guided assembly process with a DNA origami scaffold. DNA origami cuboid with A20 sequences protruding from the surface is folded by a template and corresponding staple strands. Then, D T DOEG dendron is anchored on DNA origami by hybridization. When G 2 Cl‐18 is added, hydrophobic groups on the DNA origami guide G 2 Cl‐18 dendrons to form hetero‐vesicles around the DNA frame. Reproduced from ref. 70. C) Size‐controlled liposome formation through a DNA scaffold. A DNA‐origami ring (red) with multiple single‐stranded empty handles is constructed first. Then DNA antihandles (oligonucleotides with complementary sequence to handle sequence that are chemically conjugated to DOPE, shown as green curl with orange head) are hybridized to the DNA ring. Afterward, this lipid‐modified ring is mixed with extra lipid and detergent, and dialysed to allow vesicle formation. After purification and release, uniform liposomes with sizes being determined by the DNA template are generated. Reproduced with permission. 71 Copyright 2016, Nature Publishing Group. D) Schematic figure of pathogen DNA delivery to protocell by DNA‐mediated fusion. When anchoring a set of complementary DNA on a protocell and an artificial pathogen membrane, DNA hybridization brings the two membranes in close proximity to enable fusion. Thereby, pathogen DNA is released into the protocell. Reproduced with permission. 73 Copyright 2018, American Chemical Society. E) Illustration of reversible control over the assembly of liposomes. When the liposome surface is equipped with self‐complementary DNA bearing a terminal azobenzene moiety, the vesicles undergo reversible assembly and disassembly in response to multiple stimuli including UV light, salt addition and temperature. Reproduced with permission. 78 Copyright 2016, American Chemical Society. 3. 2. 2 Templated Vesicle Formation by DNA Amphiphile Assembly Moreover, vesicles can be prepared with programmed geometry and dimensions using ssDNA‐modified gold nanoparticles or DNA origami as scaffolds. 69, 70 The ssDNA on the scaffold hybridizes with corresponding DNA amphiphiles and the resulting frame allows generation of the desired bilayer upon mixing with additional DNA amphiphiles (Figure 4 B). Strikingly, a variety of vesicle shapes was obtained by templating the DNA amphiphile assembly, i. e. , cuboids and dumbbells. In a similar way, DNA origami can be used to template vesicle formation in the interior of the origami structure. This allows size‐controlled liposome formation with the added feature that the origami can be removed. 71 In this case, the inner surface of the DNA origami ring is decorated with ssDNA extensions, which can hybridize with lipid‐DNA conjugates, thus acting as an exoskeleton for liposome formation (Figure 4 C). Using this approach, a series of highly monodisperse sub‐100 nm (29, 46, 60, and 94 nm) liposomes with a variety of different lipid compositions were produced. Thus, DNA amphiphile vesicles with desired sizes or shapes can be synthesized using templated vesicle formation. 3. 2. 3 Amphiphilic DNA Mediated Vesicle Fusion and Assembly Besides exclusively preparing vesicles from DNA amphiphiles, liposomes formed from other lipids can be functionalized by nucleic acids with the help of amphiphilic DNA conju‐gates. Thereby, the hydrophobic unit of the DNA amphiphile pierces into the lipid membrane. In this context, DNA amphiphiles are excellent tools for controlled vesicle fusion and formation of multivesicle assemblies. 72 For vesicle fusion, bilayers are brought into close proximity after which the lipid head‐groups from one vesicle insert into the other, creating the basis for the fusion pore. DNA hybridization connects vesicles and brings them together to initiate fusion. Using vesicles modified with double cholesterol terminated DNA strands that were complementary to each other, Höök and co‐workers reported for the first time amphiphilic DNA induced fusion of lipid vesicles. 73 The hybridization occurs in a zipper‐like fashion by forcing the vesicles into close contact, enabling opening of the fusion pore between the two vesicles. DNA‐induced fusion was more efficient with liposomes that consisted of cone shaped lipids such as DOPE (1, 2‐dioleyl‐ sn ‐glycero‐3‐phospho‐ethanolamine) and cholesterol, showing the importance of the geometry of those lipids for efficient fusion. In a separate study involving DNA conjugated to 1, 2‐O‐dioctadecyl‐rac‐glycerol at either the 3′ or 5′ end, it was shown that both lipid and content mixing of the vesicles took place, indicating vesicle fusion. 74 The fusion kinetics depended on the DNA sequence and the average number of lipid‐DNA per vesicle. Notably, vesicles without lipid‐DNA or ones presenting noncomplementary sequences underwent lipid mixing or exchange of membrane molecules, but no content mixing. To test the effect of membrane‐membrane spacing on fusion, a series of amphiphilic conjugates was synthesized by adding 2–24 noncomplementary nucleotides at the membrane‐proximal ends of the two complementary sequences. It was found that increasing the lengths of the linkers reduced lipid and content mixing, but increased vesicle docking rates. 75 To further improve vesicle fusion, we employed DNA modified with four hydrophobic chains, which resulted in stable incorporation of DNA into the liposomal bilayer with limited dissociation, which allowed for an efficient full fusion of the two liposome populations with complementary sequences. 76 Increased affinity of the hydrophobic domain of the DNA amphiphiles or stronger mechanical coupling between the anchor and the oligonucleotides may improve fusion further. In a striking example of the application of vesicle fusion between an artificial pathogen and a protocell, as shown in Figure 4 D, DNA templated docking and subsequent fusion induced by the oppositely charged membranes resulted in gene delivery. 77 Another excellent example of DNA‐programmed membrane fusion deals with efficient intracellular protein delivery on both suspended and adherent cells. 78 Thereby, DNA hybridization provides targeting and spatiotemporal control of the fusion between protein‐loaded liposomes and cell membranes, resulting in fast release of proteins into the cytoplasm. Docking of vesicles in the absence of fusion may lead to vesicle assemblies, which can be controlled by the design of the amphiphilic oligonucleotides. This assembly process, to some extent, is similar to the assembly of DNA‐inorganic nano‐particle conjugates, which was initiated in the 1990s by Mirkin et al. 79 In contrast to DNA‐covered inorganic nanoparticles, the assembly of multiple vesicles received much less attention. DNA‐controlled assembly of vesicles in solution and on solid supported membranes has been reported however, 80 using for example a lipid‐DNA conjugate in which ssDNA is coupled to two lipid membrane anchors at either end, with both ends inserting into the lipid membrane while the ssDNA protrudes into the solution. Upon treatment with a complementary DNA strand, the increased stiffness of the double‐stranded DNA releases one of the anchors into the solution, which allows binding to another liposome. Further inter‐liposomal membrane anchoring occurs, which leads to aggregation of the vesicles. This process provides sharp and reproducible thermal aggregation‐disaggregation transitions. The authors proposed that this system might be used to detect biologically relevant polynucleotides. Further optimization of the oligonucleotides and hydrophobic anchor parts allowed detection of DNA sequences at nanomolar concentrations and enabled sensitive mismatch discrimination of target sequences. 81 Next to thermal disaggregation, liposome assemblies were disconnected into the single vesicle state by means of light (Figure 4 E) 82 : A self‐complementary ssDNA bearing a terminal switchable azobenzene moiety was anchored on vesicles and hybridization of the DNA induced vesicle aggregation. Upon irradiation with UV light, the azobenzene isomerizes from the trans to a less hydrophobic cis isomer, decreasing its anchoring efficacy in the lipid membrane. As a result, the assembly of vesicles was destabilized. Hence, several means of control are present to reversibly assemble and disassemble multivesicle architectures aided by DNA. 4 Interactions of DNA Amphiphiles and Their Assemblies with Cell Membranes The cell membrane, or plasma membrane, plays an essential role in separating the cytoplasm from the extracellular environment, and consequently determines the size of a cell and is involved in cell signaling. 83 The most common components of the plasma membrane are phospholipids. Another major component is cholesterol, which localizes between the phospholipid molecules and regulates membrane stiffness and stability. Other types of lipids such as glycolipids take up a minor fraction, while membrane proteins occupy a significant portion of the surface. The individual phospholipid molecules are in a dynamic state in which they rotate freely around their long axes and diffuse laterally within each leaflet, thus providing cell membrane fluidity. The cell membrane is not a homogeneously mixed lipid bilayer but displays heterogeneity of the spatial arrangement of lipids and proteins. In some cases, even lipid rafts may be formed. They consist of cholesterol, sphingomyelin and tightly packed saturated phospholipids forming a liquid ordered phase, which is more stable and less fluid than the liquid disordered phase constituting the rest of the membrane. 84 Here we discuss the interaction of DNA amphiphiles with cell membranes, which provide biological applications from diagnostics to biomedicine. 4. 1 Anchoring DNA Amphiphiles on Cell Membranes DNA amphiphiles interact with cell membranes by hydrophobic interactions. In model membranes, DNA amphiphiles dissociate from their micellar aggregates and integrate in model membranes spontaneously. 54 Next to model membranes, DNA micelles have a strong affinity toward the cell membrane. The hydrophobicity of the DNA amphiphile influences the anchoring on cell membranes, as illustrated by a series of oligonucleotides conjugated to alkyl chains with either 12, 18, or 26 carbons, tested in a range of mammalian cell types. 85 A strong correlation exists between lipid length and the efficiency with which the amphiphiles are incorporated: Nonfunctionalized DNA shows negligible incorporation, while for DNA with C12 and C18 tails an intermediate insertion efficiency is observed and best piercing into cell membranes is detected for C26. Thereby, ssDNA strands conjugated with fatty acid tails are in a dynamic equilibrium with the culture medium, but when hybridized with its complementary ssDNA that is conjugated with a fatty acid as well, the construct remains in the cell membrane. 86 Due to double anchoring of the duplex, its interaction with membrane lipids is enhanced, hence the construct remained incorporated into the lipid bilayer. An alternative mechanism for interaction of amphiphilic DNA with cell membranes is through receptor‐mediated ligand binding. In general, two types of receptor‐mediated ligand interactions are involved: Direct and indirect ones. The DNA segment of the amphiphile can be an aptamer, which selectively targets a cell membrane receptor. 12 Thus, the amphiphilic DNA attaches to the cell membrane directly. Another receptor‐mediated ligand interaction occurs through an indirect pathway: The lipophilic tail of amphiphilic DNA binds to lipoproteins or other proteins, which are subsequently recognized by the corresponding receptors on the cell membrane. For instance, cholesterol conjugated siRNA can be bound to lipoprotein after intravenous injection into mice. 87 The high binding affinity of lipoprotein to cellular scavenger receptor SR‐BI facilitates the interaction of cholesterol‐siRNA amphiphiles with the cell membrane. Similarly, octadecyl tails of amphiphilic DNA bound with albumin and the resulting aggregate was recognized by cell surface albumin receptors Gp18 and Gp30. 88 4. 2 Factors Influencing the Interaction between DNA Amphiphiles and Cell Membranes When investigating the interaction efficiency of DNA amphiphiles with cell membranes, one key factor is the structure of the DNA in the amphiphile. Amphiphilic DNA with long DNA sequences incorporates slower into the cell membrane than ones with short nucleic acid chains, because longer DNA forms large micelles with a more densely charged corona, which reduces the availability of the hydrophobic domain. 54 Another possibility is that longer oligonucleotides contain more anionic phosphate groups, which are repelled by the anionic glycocalyx on cell surfaces. 85 Next to this, the hydrophobic tail of DNA amphiphiles influences the interaction: Diacyllipids DNA have a high affinity for insertion into the cell membrane, single chain C18 lipid DNA shows modest incorporation, while cholesterol modified DNA exhibits the lowest affinity. 43 Related to these experiments, a single acyl chain DNA‐mediated membrane anchoring is insufficient to mediate cell–cell adhesion, but the cell–cell interaction is achieved when diacyllipids are used. 89 Moreover, different lipid tails show preference for various lipid domains: In liposome membranes, diacyllipids mainly anchor to liquid or liquid‐ordered domains, while tocopherols anchor exclusively to liquid‐disordered domains. Cholesterols incorporate into membranes depending on the lipid composition of the membrane. Thus, DNA amphiphiles show preference for lipid domains on cell membranes. 84 Due to the fact that the lipid composition varies with cell type, different lipid tails can direct amphiphilic DNA to different cell types. 87 The interaction between amphiphile and cell membrane also depends on the amount of amphiphilic DNA. The number of amphiphilic DNA molecules that can be anchored to the cell membrane depends on the initial concentration of DNA amphiphile in the culture medium. 12, 85 A higher starting concentration leads to a higher density of DNA tethering. For interactions driven by aptamer recognition, densely packed aptamers on an amphiphilic micelle induce a multivalent effect, which leads to higher affinity for the cellular membranes. 10 Moreover, the cell culture medium influences the interaction of DNA amphiphiles with the membrane. The culture medium affects the anchoring efficiency in the decreasing order: PBS > DMEM > PBS with 10% FBS (Fetal Bovine Serum) > DMEM with 10% FBS. 85 The components in the cell culture medium alter the interaction between amphiphilic DNA and cell membranes. For instance, albumin in albumin‐rich culture medium binds the lipid domain and forms a complex that prevents amphiphilic DNA inserting into the cell membrane. 88 Hence, cell membranes display additional features that influence their interaction with DNA amphiphiles. It is important to consider their more complex structure compared to model membranes when applying DNA amphiphiles in living systems. 4. 3 Stability of the Complex between DNA Amphiphiles and Cell Membranes After binding to the cell membrane, DNA amphiphiles or their assemblies are not static in space and time. Instead, they are in a dynamic exchange with the medium, they can be degraded and they can be subjected to endocytosis. All DNA amphiphiles are in equilibrium between the aqueous medium and the cell surface. Even though a DNA sequence is connected to a long lipid tail, like C26, it still displays characteristic re‐equilibration. A gradual loss of lipid DNA on the membrane occurs when replacing the cell culture medium. 85 This loss is a result of adjusting a new equilibrium between DNA amphiphile on the cell membrane and the culture medium. DNA conjugated to an alkyl chain showed a gradual decay on the cell surface. 90 After the first hour of incubation, only < 20% loss was observed. However, after 2. 5 h only 50% of the initial amount of DNA was present on the cell surface. When incubated for 24 h, only a very weak signal originating from the DNA remained. This gradual decay is temperature‐dependent 89 : Surface anchored DNA decayed to 86% of its initial concentration after 160 min at 25 °C, while 67% of its initial concentration was left after the same time period at 37 °C. Amphiphilic DNA anchors to the outer leaflet of the cell membrane and is subjected to slow endocytosis. C18 and cholesterol modified oligonucleotides are taken up by cells via an energy‐dependent mechanism rather than by passive diffusion. 39 Indeed, some of the DNA amphiphiles enter cells via endocytosis, while the majority possibly flips and translocates from the cell surface to the organelles during membrane recycling. As micelles, amphiphilic DNA locates initially close to the cell membrane, then disassembles and fuses with the cell membrane. 91 This cellular uptake of a DNA amphiphile micelle represents a similar uptake mechanism compared to other amphiphilic molecules. 54 For interactions driven by receptor binding, endocytosis is suggested as the subsequent step after binding of the amphiphilic DNA or amphiphilic DNA embedded lipoprotein with the receptors. These complexes are recognized by corresponding receptors on the cell membrane and subsequently enter cells via receptor‐mediated endocytosis. 92 4. 4 Characteristics of DNA Amphiphiles Interacting with Cell Membranes Amphiphilic DNA and its assemblies interact efficiently with cell membranes and hence offer a facile strategy for further manipulating the cell surface. A major characteristic is that amphiphilic DNA allows convenient cell surface modification. Other common strategies for presenting DNA at cell surfaces, such as expression of a DNA binding domain of a protein at the cell surface, 93 covalent attachment of DNA to functional groups at the membrane, 94, 95 or a DNA aptamer that binds membrane target sites, 96 either involve complicated stepwise processes or can only be applied to very limited membrane targets. Instead, employing amphiphilic DNA to modify a cell surface is simple and quick. Coincubating amphiphilic DNA with cells allows spontaneous insertion of the amphiphiles into the cell membrane. This process is fast and can be performed within only 3 min. 90 Moreover, amphiphilic DNA can be anchored to different cell types, including natural killer cells, 43 T cells, 12 and cancerous cells. 11 This quick modification procedure results in stable anchoring of the DNA in the membrane: The majority of diacyllipid‐DNA locates on the outer leaflet and remains even after 2 h incubation with cells at 37°. 89 The easily accessible DNA on the membrane is a highly versatile technology platform in vitro and in vivo. To target cell membranes in vivo DNA amphiphiles can be administered locally. DNA amphiphiles injected into mice remained 72 h at the injection site, which reduced to 4 h with DNA that does not contain a hydrophobic tail. 43 More important, compared with nucleic acids coated on nanoparticles, 97 modifying cell membrane by amphiphilic DNA insertion is noninvasive and does not involve inorganic components. 5 Applications Hydrophobic domains within nucleic acids allow their easy incorporation into lipid bilayers and facilitate their uptake by living cells. Here we will discuss the biomedical functions of amphiphilic DNA structures, which can be derived from this behavior. 5. 1 Drug Delivery Both amphiphilic micelle and liposome nanostructures can be exploited for drug delivery. Until now, a number of examples have been reported demonstrating highly efficient drug delivery with DNA amphiphiles and their assemblies in vitro and in vivo. 12, 28, 30, 47 Our group loaded the hydrophobic anticancer drug (Doxorubicin) into the interior of DNA‐ b ‐PPO micelles, 15 which were taken up through receptor‐mediated endocytosis and significantly inhibited growth of Caco‐2 cancer cells. The cellular uptake of the micelles strongly depended on the density of the recognition elements, i. e. , folic acid, on the micellar surface. Moreover, DNA amphiphile micelles were very well suited for loading another hydrophobic anticancer drug: Paclitaxel. 39 Recently, we tackled in vivo functionality of the DNA amphiphiles micelle even with human tissue in the context of ophthalmology for treating eye infections. 16 Therefore, different antibiotics were loaded into the DNA amphiphile micelles ( Figure 5 A). Aptamers were complexed with an aminoglycoside, i. e. , DNA aptamer for Kanamycin B or RNA aptamer for Neomycin B, and subsequently conjugated at the 3′ end of the DNA amphiphiles through hybridization. Compared with pristine drugs, the DNA amphiphile micelles showed extended residence time on the ocular surface and improved efficiency on the cornea in vitro and in vivo. This study highlights the potential applicability of amphiphilic DNA‐based materials in the clinic. Figure 5 A) Schematic of DNA micelles for treating eye infections. Lipid‐modified DNA strands form micelles and were then hybridized with antibiotics loaded aptamers. Upon administration to the eye, DNA micelles adhere to the cornea and release antibiotics to treat infections. Reproduced with permission. 16 Copyright 2018, Elsevier. B) Illustration of targeting cancer cells (green) with aptamer‐modified immune cells (red). Immune cells were equipped with lipid modified DNA aptamer, which targets the cancer cell surface. When cancer cells are mixed with normal cells, immune cells only recognize cancer cells by their surface anchored aptamer and then kill cancer cells. Reproduced from ref. 12. C) Schematic of lipid DNA micelles as CpG carrier. Lipid conjugated DNA forms micelle and was then hybridized with different amounts of CpG sequences to form immunostimulatory nanoparticles (INPs) with different degree of CpG functionalization. These INPs activated immune responses both in vitro and in vivo. Reproduced with permission. 98 Copyright 2018, Elsevier. D) Assembly of liposomal spherical nucleic acids by anchoring tocopherol modified DNA to DOPC small unilamellar vesicles. Reproduced with permission. 22 Copyright 2014, American Chemical Society. Most recently, a lipid‐conjugated drug‐incorporated oligonucleotide was developed for hitchhiking with endogenous serum albumin for cancer chemotherapy. 88 By incorporating a hydrophobic lipid tail, floxuridine homomeric oligonucleotides inserted into the hydrophobic pocket of albumin to form complexes which accumulate at the tumor site by the enhanced permeability and retention (EPR) effect and internalize into the lysosomes of cancer cells after intravenous injection. Upon enzymatic degradation, the cytotoxic floxuridine monophosphate is released and inhibited cancer cell proliferation. 5. 2 Immunotherapy Furthermore, DNA amphiphiles and their self‐assembled structures find application in immunotherapy. Surface anchoring of DNA amphiphiles directed immune cells to their target cells 12 : Modification of natural killer (NK) cells with an aptamer named KK1B10 (Figure 5 B) resulted in specific targeting of cancer cells, i. e. , chronic myelogenous leukemia cell line K562. This resulted in 50% increased killing efficiency of NK cells toward K562 cancer cells, compared with unmodified NK cells. The higher killing efficiency was likely due to the better targeting efficiency of NK cells when the DNA aptamer amphiphile is attached. Moreover, the selectivity of the aptamer modified NK cells was demonstrated when the target K562 cells are mixed with an excess of nontargeted cells. In a different approach, the immunological effects of DNA amphiphile micelles decorated with the immune adjuvant (CpG) were studied in vivo recently. 98 Different amounts of immunostimulatory adjuvants were established on the surface of spherical micelles through simple stoichiometric incorporation (Figure 5 C). After that, a full immunological assay, including phagocytosis, the expression of costimulatory molecules, and the production of proinflammatory cytokines in spleen dendritic cells (DCs) was evaluated and analyzed. As a result, dose‐dependent activation of spleen DCs by CpG‐conjugated micelles was observed, which was accompanied by the pronounced up‐regulation of costimulatory molecule and cytokine production. In addition, labeling 50% of the DNA amphiphile micelles with the CpG segment can fully induce the activation of spleen DC. The straightforward functionalization by DNA duplex formation makes the DNA amphiphile micelles a biocompatible and scalable delivery platform for immunostimulation and immunotherapy. Since such DNA micelles still exhibit single‐stranded DNA on the surface ready for hybridization, these sites could be easily exploited for the incorporation of antigens to boost the generation of humoral and cellular vaccine‐specific immune responses. 5. 3 Gene Silencing Gene silencing offers the potential to cure certain diseases by down‐regulating the disease‐causing gene expression and protein production. 8 One of the most widely used gene silencing strategies is exogenously derived single‐stranded antisense oligonucleotides (ASOs). As discussed in the introduction part, the intrinsic physicochemical properties of ASOs, such as negative charges, high hydrophilicity, and high molecular weight, prevent their efficient delivery to the intracellular target site. 99 To this end, conjugation of hydrophobic moieties to ASOs has been used as a safer and straightforward strategy to assist their cellular uptake. 100, 101, 102 Early studies from the 1980s used cholesteryl conjugated oligonucleotides to inhibit HIV infections 103, 104 or targeted the intercellular adhesion molecule‐1 gene. 105 Later, hydrocarbon lipids were conjugated to oligonucleotides to assist antisense efficiency: Barthélémy et al. proposed an example involving lipid moieties that were connected to oligonucleotides via click chemistry, which promoted cellular uptake. 39 As a result, the hepatitis C virus (HCV) internal ribosome entry site (IRES)‐mediated translation was effectively suppressed. Interestingly, when the ASO was conjugated to a C18 lipid or cholesterol unit, a dose‐dependent reduction of the translation was measured in the Huh7 cell line. More importantly, the biological activity of the oligonucleotide was not affected by the lipid conjugation and toxicity was negligible at relevant concentrations. In another notable example, Mirkin and coworkers synthesized a spherical nucleic acid nanostructure, which consists of a liposomal core (30 nm) stabilized with a dense shell of tocopherol‐modified DNA that intercalates between the phospholipids and defines the liposomal structure (Figure 5 D). 22 By using commercially available and FDA‐approved building blocks, they demonstrated that such monodisperse DNA‐functionalized vesicles remain stable with no change in dispersity for at least 4 days at 37 °C. This behavior is contrary to native nonfunctionalized vesicles, which tend to fuse and form large poly‐disperse structures under such conditions. The obtained spherical nucleic acid architecture did not only stabilize the liposomal constructs but rapidly entered multiple cell lines and resulted in effective gene knockdown of HER2 in SKOV‐3 cells. 5. 4 Sensing the Extra and Intracellular Environment Tracking cell functions, metabolism, and cell–cell signaling in their native cellular environment has enormous implications for cell biology and regenerative medicine. 106 For the past few decades, molecular sensors 107, 108 or nanoparticles 109 tethered on the membrane surface have been utilized to monitor such cell activities. However, these sensors exhibit several drawbacks, such as limited targets, a need for complicated chemical modification, allowing measurements only under model conditions, or they do not monitor in real‐time. 106 Fortunately, as a relatively new cell surface biosensor, amphiphilic DNA outperforms other methods in several aspects. First, aptamers can be selected via a process called systematic evolution of ligands by exponential enrichment (SELEX) to specifically bind to certain target molecules, such as metal ions, small organic molecules or proteins with high affinity. Second, the straightforward functionalization of DNA with fluorophores facilitates signal readout by means of photoluminescence. Furthermore, hydrophobic tags permit anchoring of the biosensor to the cell membrane. Finally and importantly, DNA hybridization or the fast response of DNA aptamers for their targets render monitoring in real time and in situ with high spatiotemporal resolution feasible. However, sometimes the action of aptamers is compromised by nuclease degradation, variability of pharmacokinetics or rapid renal filtration in native environments. 110 To overcome these limitations, their activity or persistence under physiological conditions were optimized during selection. 3 Another means of stabilization represents the introduction of chemical modifications to decrease enzymatic digestion, and PEGylation to prolong circulation times. 111 So far, amphiphilic DNA has been used to monitor metal ions, 112 pH 113 and chemical transmitters 41 in cellular environment. Another notable example is the measurement of formation of lipid membrane domains to monitor and understand the dynamic signaling interactions on the cell surface ( Figure 6 A). 40 To achieve this, a ssDNA strand named S1 was anchored to the cell membrane via a hydrophobic lipid unit and was partially hybridized with a blocking strand B. Similarly, a S2 strand was anchored at a second anchor site and was partially hybridized with a walker strand W. An initiator strand can completely remove the blocking strand from the S1 strand by a strand displacement reaction, leaving S1 free for hybridization. Because strand W from the S2 site hybridizes preferentially with the free S1 strand, it will translocate once both sequences are in close proximity. To observe this displacement, strand S1 and strand W were labeled with a fluorescence resonance energy transfer (FRET) pair, leading to quenched fluorescence. The FRET efficiency becomes a measure of the lipid domain encounter rate since the DNA amphiphiles were anchored in different lipid domains. Three lipid tails were attached to the nucleic acid moie‐ties, i. e. , diacyllipid, cholesterol, and tocopherol, to specifically locate DNA strands in different cellular lipid domains. This method transduces transient encounters of nanodomains into a cumulative cell surface fluorescence signal and thus allows to detect signaling events on live cell membranes. Figure 6 A) Schematic illustration of using a DNA probe to measure the encounter rate of lipid domains on live cell membranes. The S1 strand is anchored to the cell membrane and is partially hybridized with a blocking strand B. Similarly, the S2 strand is anchored at a second anchor site and partially hybridized with a walker strand W. When an initiator stand I is introduced to the system, it removes strand B from S1, leaving S1 free for hybridization. Since the walker strand from the S2 site has priority to hybridize with this free S1 strand over its own S2 strand, it will translocate from the S2 site to the S1 site once both strands meet each other. Since the S1 and S2 stands are labeled with FRET dyes, once they encounter each other, the fluorescence is quenched. Different hydrophobic moieties attached the nucleic acid units introduce selectivity of the DNA strands for certain lipid domains. Thus the quenching rate is a measure to evaluate the encounter rate of different lipid domains. Reproduced with permission. 40 Copyright 2017, Nature Publishing Group. B) Working principle of switchable aptamer micelle flares for ATP imaging inside living cells. On the left, aptamers are folded, while upon binding the target molecule, the aptamer unfolds leading to a dequenching of fluorescence (right side). Reproduced with permission. 91 Copyright 2013, American Chemical Society. C) Two populations of cells exhibit anchored DNA on their membranes. When mixed together, hybridization of membrane‐embedded DNA induces cell–cell contact. Reproduced with permission. 114 Copyright 2009, National Academy of Sciences. D) Illustration of microtissues constructed by DNA hybridization. a) Illustration of cell adherence by DNA hybridization. One type of cell is anchored with a particular DNA strand on the membrane. The other cell type is functionalized with the complementary DNA strand. b) When the two cell types are mixed, hybridization induces cell‐to‐cell aggregation. c) Formation of microtissue by DNA hybridization. Iteration of this process allows assembling microtissues into the third dimension. Reproduced with permission. 115 Copyright 2015, Nature Publishing Group. Apart from probing the cell surface, amphiphilic DNA was utilized for imaging and detecting intracellular parameters such as the level of ATP. 91 A switchable aptamer‐containing micelle flare allowed detection of ATP within cells (Figure 6 B). This design implicated three segments with a DNA layer that folds into an aptamer loop against ATP. The hydrophobic segment was a diacyllipid tail, with a PEG unit as spacer between the DNA and the hydrophobic tail. A fluorophore and a quencher were covalently attached to 3′ and 5′ ends. Once ATP is binding, the DNA loop opens, leading to an increase of fluorescence. Due to the fact that the micelles interacted with the cell membrane and were internalized into cells, ATP in both membrane and cytosolic environment could be detected. 5. 5 Cell Capture and Assembly Similar as the DNA amphiphile mediated liposome assembly discussed in Section 3. 2. 2, when tethered onto the cell membrane, the amphiphilic DNA facilitates cell capture and assembly through the specific and fast recognition properties of the nucleic acids. The length of the DNA strands is of crucial importance for the successful cell to cell contact by hybridization. 89 A 20‐mer DNA strand on the cell surface cannot hybridize with its complementary sequence due to steric hindrance provided by the dense glycocalyx layer. However, a 60‐ and 80‐mer poly(dT) spacer inserted between the lipid anchor and the DNA recognition element that will hybridize, significantly increases the cell adhesion to other surfaces. Eventually, DNA‐anchored on cell surfaces can be linked to surface‐anchored complementary DNA. The accessibility of cell membranes with anchored DNA amphiphiles also facilitates cell assembly and microtissue formation. In one example, Bertozzi et al. linked nonadherent Jurkat cells together by employing DNA anchored on their surfaces (Figure 6 C). 114 This group found that the most important parameters for cell assembly are the cell concentration, DNA density on the cell surface, and DNA sequence complexity. Since the cells are attached to each other through DNA hybridization, this process can be reversed by DNase addition or thermal melting. This allows the construction of microtissues with defined cell composition and stoichiometry. This approach can be extended in a bottom‐up strategy that uses a DNA‐patterned substrate as a template and temporary DNA‐based cellular adhesions as synthetic linkages between cellular building blocks for tissue engineering in 3D (Figure 6 D). 115 In this way, the construction of arrays of 3D cell cultures with many tunable parameters was feasible. In the same study, template DNA was linked to a glass slide to form DNA patterns. Then, a PDMS flow channel was placed on top of the DNA pattern. A cell population functionalized on the surface with complementary DNA to the template DNA was added to the flow channel, which directed the cells to the designed 2D pattern. The formed cell pattern could be released by enzymatic cleavage of the DNA. Embedding such microtissues constructed from DNA in gels allows to study the influence of tissue size, shape and composition on cell behaviors in 3D. 5. 6 Complex DNA Nanostructures on and in the Cell Membrane The extraordinary self‐recognition and hybridization properties of DNA can be applied for creating various programmable nanostructures. 116 An exceptional form of DNA amphiphiles are DNA‐based nanopores. DNA‐based nanopores open exciting opportunities in the field of bio‐nanotechnology, as shown by their protein‐based counterparts. 117 Single‐stranded nucleic acid scaffolds together with staple strands or short oligonucleotides can fold into DNA‐based nanopores. When conjugated to hydrophobic units, the otherwise hydrophilic nanopores insert into synthetic lipid membranes. 118, 119, 120, 121 Moreover, these nanopores interact with biological membranes ( Figure 7 A). 122 A notable example is a DNA nanopore with a 2 nm opening and an outer diameter of 5. 5 nm and a height of 14 nm, which contained a hydrophobic belt with 72 ethyl phosphorothioates at the bottom of the pore to direct insertion into the cell membrane. After incubating these nanostructures with cervical cancer cells, the DNA nanopores mainly located at the membrane and caused cell death. Nanopores that did not contain a hydrophobic belt were mostly internalized by the cancer cells. The cytotoxic effect of DNA‐based nanopores could allow for anticancer activity, albeit for true applications the selectivity needs to be improved. Figure 7 A) A membrane‐spanning DNA nanopore (NP) with cytotoxic activity and three negative control nanostructures. a) The NP‐EP pore is composed of a six‐duplex bundle (blue) and a hydrophobic belt (purple) made up of 72 ethyl phosphorothioate (EP) groups. b) Inserting of NP‐EP pores into cellular membrane induces cell death. c) NP‐P features phosphorothioate groups but no hydrophobic ethyl modification. d) NP contains native phosphate groups. e) NNP contains EP groups but lacks three of the six strands required to generate the six‐duplex bundle nanopore. Reproduced from ref. 122. B) Design of the lipid‐scrambling DNA nanostructure. Reproduced with permission. 123 Copyright 2018, Nature Publishing Group. C) Illustration of the sequential MBB functionalization steps. Oligos (MIO) are first integrated into the cell membrane, then bridge oligos hybridize with MIO strands followed by bridge fortifier oligo hybridization. Lastly, membrane bound breadboard (MBB) binds to the cell membrane by hybridizing with bridge oligos. Reproduced from ref. 124. Apart from a cytotoxic effect, DNA nanostructures on cell membranes enable the transport of membrane lipids. A lipid‐scrambling DNA nanostructure, consisting of only eight DNA strands, which were modified by tetraethylenglycol (TEG)‐cholesterol (Figure 7 B), 123 spontaneously inserts into biological membranes by forming a toroidal pore that connects the inner and outer leaflets of the membrane. The inserted nanostructure facilitates the exchange of lipid molecules between the inner and outer bilayer leaflets rapidly equilibrating the lipid composition. The rate of lipid transport catalyzed by the DNA nanostructure is three orders of magnitude higher than that reported for lipid transport catalyzed by natural enzymes. The stable DNA‐induced toroidal lipid pore likely induces this exceptional transport behavior. The DNA‐based artificial scramblase also showed translocation of phosphatidylserine lipids from the inner membrane leaflet to the outer leaflet of human cancer cells. Besides insertion, DNA‐origami nanodevices can be placed onto the surface of living cells (Figure 7 C). 124 The membrane can be functionalized by anchoring DNA to the cell surface via cholesterol insertion into the membrane, followed by binding of a bridge‐oligonucleotide that partially hybridizes with this surface DNA. The bridging oligo allows binding of the membrane‐bound breadboard (MBB) binding sites, but also offers the possibility of removal of this MBB from another surface via a strand displacement reaction. Several cell types can be functionalized with MBBs, including primary, endothelial, and lymphoma cells. Furthermore, the MBB can be released from cell surfaces when a detachment strand is added. By using DNA origami nanodevices as engineering tools, MBB acts as a mediator for either homotypic or heterotypic cell–cell interactions, which mimic complex biological processes on the cell membrane. 6 Conclusions and Perspective DNA‐based materials have exceptional properties in regard to structural design. Compared to other building blocks like peptides, proteins, and synthetic macromolecules, DNA allows the bottom‐up construction of complex architectures and tuning the interaction energy between complementary DNA strands. Recent progress in the design and functionalities of DNA amphiphiles builds on these remarkable properties to implement DNA hybrid materials into the application areas of diagnostics and biomedicine. These efforts are enabled by well‐established protocols to synthesize amphiphilic DNA molecules and their commercial availability. Moreover, the topology and interactions of amphiphilic DNA is highly controllable, and their aggregation behavior into superstructures such as micelles or vesicles, but also many other geometries can be precisely adjusted. It is possible to tune their size, switch their assembly state, and modify their surfaces at will through duplex formation. With their hydrophobic units, amphiphilic DNA hybrids further provide a simple and efficient strategy for membrane modification of living cells. This simple functionalization procedure allows further cell surface engineering, cell assembly, and facilitates potential sensing applications. Despite these many favorable properties of DNA amphiphiles, certain challenges need to be overcome before translating them further toward the clinic. One of the most critical issues is the biological stability. Although enhanced enzymatic stability was reported for DNA amphiphile micelles, 64 it remains a challenge to minimize nuclease degradation, especially in vivo. Next to this, upon exposure to biological medium, amphiphilic DNA structures are encapsulated by a protein corona, 125 which possibly shields recognition elements on the surface and compromises its targeting efficiency. Another big challenge represents maintaining the solubility of amphiphilic DNA in biological media and its activity on membranes. Proteins from serum, like albumin or lipoproteins, are well known to form stable complexes with amphiphilic DNA, 126, 127 thus preventing their desired functions. Approaches to prevent such interactions of amphiphilic DNA with serum proteins are urgently needed for extending biomedical applications. 128 Furthermore, amphiphilic DNA molecules in micelle assemblies are always in a dynamic equilibrium within their environment: Strong dilution after intravenous injection might result in concentrations below the CMC, which leads to disassembly of micelles and drug release before reaching the target. 129, 130 To prevent this, the biological stability of amphiphilic DNA micelle needs to be adjusted to the desired delivery function. It has been shown that covalent cross‐linking of the lipid DNA molecules can for example increase the stability of the assembled nanostructures, 65 and therefore we foresee that this challenge will be overcome in the near future. Although the introduction of a hydrophobic segment into the nucleic acid amphiphiles is essential for their function, the biocompatibility and biosafety of these hybrids should be taken into consideration, especially toward clinical translation. For example, too many lipid‐DNA insertions in the membrane will lead to cell membrane disturbance, damage, and cell death. 12, 85 Since the insertion mechanism into membranes, which is mediated by hydrophobic interactions, is not specific for a given cell type, it is essential that additional features for selective incorporation are introduced. A notable example of such an effort is labeling the DNA amphiphiles with folic acid to target cancer cells. 15 At this stage, most studies on the biocompatibility of DNA amphiphiles were conducted using cell cytotoxicity evaluation, while only a few studies investigated their local or systematic toxicity in vivo. 16, 43 As more and more DNA amphiphiles are developed for biomedical applications, these activities need to be extended for more comprehensive toxicological evaluations, such as cell membrane damage, cell signaling interference, oxidative stress, genotoxicity, etc. , which are required for predicting long‐term biosafety. Since a lot of knowledge and control over synthesis and assembly mechanisms of DNA amphiphiles have been gained, we are in an excellent position to explore the unique properties of DNA amphiphiles when combined with hydrophobic molecules. Similar as native protein clusters on cell membranes, DNA nanostructures (not limited to nanopores) might act as artificial gate for intracellular/extracellular transportation, as means for cellular environment regulation and as tool to regulate cellular signaling. From the perspective of synthetic biology, the exciting examples of interfacing DNA amphiphiles with membranes will fuel further activities regarding artificial cell engineering, cell assembly and novel tissue formation. 131 Moreover, DNA amphiphiles might find potential applications in cell‐based therapy. In addition to immune cells, amphiphilic DNA nanostructure hitchhiked on other circulatory cells merits more investigations. 132 Taken together, DNA amphiphiles are at a stage where a large variety of nucleic acid materials is readily available, hence several structural designs were investigated in combination with living sytsems, especially addressing potential biomedical applications. We predict a further growth in this area addressing more complex functions including the fields of oncology, vaccination and theranostics. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201900099
| 2,019
|
Advanced Science
|
Intensified Stiffness and Photodynamic Provocation in a Collagen‐Based Composite Hydrogel Drive Chondrogenesis
|
Abstract Directed differentiation of bone‐marrow‐derived stem cells (BMSCs) toward chondrogenesis has served as a predominant method for cartilage repair but suffers from poor oriented differentiation tendency and low differentiation efficiency. To overcome these two obstacles, an injectable composite hydrogel that consists of collagen hydrogels serving as the scaffold support to accommodate BMSCs and cadmium selenide (CdSe) quantum dots (QDs) is constructed. The introduction of CdSe QDs considerably strengthens the stiffness of the collagen hydrogels via mutual crosslinking using a natural crosslinker (i. e. , genipin), which simultaneously triggers photodynamic provocation (PDP) to produce reactive oxygen species (ROS). Experimental results demonstrate that the intensified stiffness and augmented ROS production can synergistically promote the proliferation of BMSCs, induce cartilage‐specific gene expression and increase secretion of glycosaminoglycan. As a result, this approach can facilitate the directed differentiation of BMSCs toward chondrogenesis and accelerate cartilage regeneration in cartilage defect repair, which routes through activation of the TGF‐β/SMAD and mTOR signaling pathways, respectively. Thus, this synergistic strategy based on increased stiffness and PDP‐mediated ROS production provides a general and instructive approach for developing alternative materials applicable for cartilage repair.
|
1 Introduction In regenerative medicine, cartilage repair remains a challenging task, because articular cartilage exhibits a poor self‐repair ability after damages. 1 In an attempt to develop an appropriate method for cartilage defect repair, stem cell therapy that usually seeds bone‐marrow‐derived stem cells (BMSCs) in scaffolds for enabling the lineage‐specific differentiation toward chondrogenesis has attracted increasing interest. 2 However, due to inappropriate scaffold design, this strategy inevitably suffers from poor oriented differentiation tendency and low differentiation efficiency. 3 Typically, the stiffness of scaffolds exerts robust influences on the directed differentiation of BMSCs, e. g. , low stiffness can inhibit the chondrogenic differentiation of BMSCs, resulting in rapid degradation. 4 As well, exogenous scaffolds are also known to trigger inflammation and produce immune repulsion‐derived side‐effects, further suppressing cartilage repair. Thus, to effectively direct the lineage‐specific BMSCs differentiation toward chondrogenesis, rationally designing scaffolds featuring appropriate chemical, topographic and mechanical properties is of great importance, yet to be achieved. To address these issues, we have established a combined strategy that integrates intensified stiffness with augmented reactive oxygen species (ROS) production to facilitate the chondrogenic differentiation of BMSCs as well as cartilage repair. To this end, an injectable collagen‐based composite hydrogel was designed and fabricated, wherein the collagen served as a scaffold to accommodate BMSCs and cadmium selenide (CdSe) quantum dots (QDs) via chemical crosslinking by a natural and nontoxic crosslinker (i. e. , genipin), 5 as indicated in Figure 1. Genipin can guarantee excellent fluidity and injectability, 6 reduce inflammatory responses, 7 and simultaneously address the high toxicity and rapid gelation that traditional crosslinking agents encounter. 8 Collagen features excellent injectability, inherent biocompatibility, favorable degradability and flexible environmental responsiveness. 9 Herein, the used collagen type I is preferable than collagen type II, because it can repress inflammation and immune repulsion‐derived side‐effects, 10 and facilitate chondrogenic differentiation of stem cells. 11 To overcome its inherent low stiffness and poor fracture toughness, 12 the crosslinking of collagen with biocompatible CdSe QDs by genipin was adopted to reinforce the stiffness of collagen, 13 and benefit lineage‐specific differentiation of seeded BMSCs into chondrogenesis (Figure 1 ). 14 Figure 1 General schematic illustration of the fabrication process and implementation of collagen–genipin–quantum dot (CGQ) composite hydrogels. As well, appropriate ROS levels are advantageous for cell growth and differentiation of BMSCs, 15 and simultaneously fail to induce inflammation. 16 Thus, CdSe QDs' introduction is also expected to further promote cell survival, proteoglycan secretion and cartilage repair by accelerating the lineage‐specific differentiation of BMSCs toward chondrogenesis, 17 since CdSe QDs can serve as photosensitizers to produce ROS (e. g. , singlet oxygen) via the photodynamic provocation (PDP). 18 In vitro and in vivo experiments were carried out to confirm that the intensified stiffness and increased ROS production in this composite hydrogel can synergistically promote in vitro chondrogenesis and in vivo ectopic/orthotopic cartilage regeneration after seeding BMSCs. 2 Results and Discussion 2. 1 Design and Fabrication of Collagen‐Based Composite Hydrogels The synthesis scheme of injectable collagen‐based composite hydrogels is shown in Figure 1, wherein collagen was chemically crosslinked with CdSe QDs via genipin crosslinker. Digital photos of collagen (C), collagen–genipin (CG), and collagen–genipin–QDs (CGQ) gels are shown in Figure 2 a. The dark green in CG and CGQ gels is attributed to the successful crosslinking of genipin with CdSe QDs and collagen through reacting with amines, suggesting that genipin can serve as an indicator of crosslinking based on its color change. Irregularly shaped CdSe QDs with an average diameter of 11. 2 nm are found to be randomly oriented and uniformly distributed in CGQ and no size variation is observed (Figure 2 b and Figure S1a, Supporting Information). Dynamic light scattering (DLS) also shows no evident size variation of QDs (Figure S2, Supporting Information). Much stronger characteristic peaks of selenium and cadmium elements in energy‐dispersive spectrometry (EDS) analysis confirm the successful linkage of more CdSe QDs in CGQ hydrogels (Figures 2 c) in comparison to C and CG hydrogels (Figure S1b, Supporting Information). Noticeably, after crosslinking with genipin, the pore sizes of CG and CGQ hydrogels, which can be measured according to scanning electron microscopy (SEM) images, 19 decreases from ≈100 to ≈50 µm (Figure 2 d). Additionally, their Fourier transform infrared spectra (FTIR) further demonstrate the successful crosslinking of collagen with genipin and QDs in CGQ (Figure 2 e). Figure 2 Characterizations of CGQ hydrogels. a) Photographs showing the formation of collagen‐based hydrogels. b) TEM image of QDs in CGQ nanocomposite, scale bar = 20 nm. c) EDS spectrum of CGQ hydrogels, and the Y axis indicates the signal intensity of the characteristic peak of related elements. d) SEM images of C, CG, and CGQ hydrogels (scale bar = 200 µm). e) FTIR spectra of C, CG, and CGQ composite hydrogels. f) Fluorescence intensity (excitation at 595 nm and emission at 630 nm) of C, CG, and CGQ hydrogels after crosslinking at different time points. g) The absorbance of collagen, CG and CGQ hydrogels at 595 nm after crosslinked at different time points. Data were expressed as the mean value ±standard deviation (SD), n = 5 (C = collagen, CG = collagen crosslinked with genipin, CGQ = collagen crosslinked with genipin and QDs). In UV–vis spectra, a characteristic absorbance peak at 595 nm emerges in CG and CGQ gels due to the genipin crosslinking‐induced color change after 24 h incubation (Figure S3, Supporting Information). To further verify this result, the peak value at 630 nm in their photoluminescence spectra was recorded using a 595 nm laser as the excitation source. No obvious fluorescence signal harvested at 630 nm is observed in C hydrogels within 120 h (Figure 2 f). In contrast, the fluorescence intensity in either CG or CGQ significantly increases as the incubation time proceeds and ultimately reaches a plateau after 12 h, confirming the characteristic peak of CG and CGQ hydrogels at 595 nm. Interestingly, the stronger UV–vis absorbance intensity in CG and CGQ hydrogels than that in C hydrogels at 595 nm further validates this point (Figure 2 g). The minimum gelation time essential for completing the gelation process was obtained in Table S1 (Supporting Information), wherein the three hydrogels can keep fluid at 4 °C for over 22 h, and remaining fluid for over 7 min even at 37 °C, which sufficiently guarantees the injectability of these hydrogels. 2. 2 Evaluations on Structural Properties of CGQ Swelling ratio that is a routine concern for hydrogel scaffolds was investigated. Compared with CG hydrogels, the chelation of CdSe QDs in CGQ via covalently anchoring by the genipin crosslinker endows CGQ with a robust anti‐swelling property ( Figure 3 a). The degradation rate of scaffolds is another concern in tissue engineering, and an ideal degradation rate, in principle, is approximately consistent with the rate of tissue regeneration. Interestingly, the incorporations of QDs and genipin via mesh crosslinking with collagen endow CGQ with improved mechanical property and stronger resistance to degradation catalyzed by type I collagenase than C and CG (Figure 3 b), which will be beneficial for rendering CGQ degradation matched with cartilage regeneration. To comprehensively explore their degradation behaviors, in vitro degradation profiling at a concentration of 10 ng mL −1 that is close to that in serum in vivo shows that the degradation ratios of all the three gels approach 100% after 60 days (Figure S4, Supporting Information). Figure 3 Mechanical and photodynamic characterizations of CGQ hydrogels. a–c) The swelling ratio (a), degradation rate at a type I collagenase concentration of 100 µg mL −1 (b) and Young's modulus obtained at 10–20% linear curve (c) of C, CG, and CGQ hydrogels. d) Absorption value of DPBF probe after incubation with C, CG, or CGQ for different irradiation durations. Data are expressed as mean ± SD ( n = 6); ** indicates p < 0. 01; *** and ### indicate p < 0. 001 (C = collagen, CG = collagen crosslinked with genipin, CGQ = collagen crosslinked with genipin and QDs). Scaffold stiffness is known to exert a significant influence on the differentiation of stem cells. 20 Despite being a common scaffold in cartilage tissue engineering, collagen‐based hydrogels still suffer from unmatched mechanical properties and rapid degradation that are unfavorable for chondrogenesis. Given that chemically doped nanoparticles and crosslinking can improve the stiffness and stability of hydrogels to resist denaturation and enzymatic degradation, 14 CdSe QDs and biocompatible genipin crosslinkers are also anticipated to reinforce the stiffness of CGQ and drive lineage‐specific differentiation of BMSCs toward chondrogenesis. Indeed, contributed by the robust crosslinking between collagen scaffolds and CdSe QDs by genipin, the stiffness of CGQ hydrogel (28. 7 ± 2. 6 kPa) is considerably improved (Figure 3 c) with 14‐fold and threefold larger than that of C (1. 9 ± 0. 3 kPa) and CG hydrogels (9. 53 ± 2. 2 kPa), respectively. Although the equilibrium modulus (28. 7 ± 2. 6 kPa) of CGQ hydrogel is much lower than that of native cartilage (1–10 MPa), the soft CGQ scaffold is expected to enable cells to sense stiffness change and benefit cartilage regeneration, since most work has shown a stiffness sensing mechanism for evaluating cell responses on matrices much softer than native cartilage. 21 2. 3 ROS Production Furthermore, the ability of QDs in CGQ hydrogels to produce ROS was investigated since appropriate ROS level is also responsible for regulating BMSCs differentiation. 15 CdSe QDs have been well documented to generate singlet oxygen via the PDP process. 18 Herein, 1‐ 3‐diphenylisobenzofuran (DPBF) as the sensitizer of singlet oxygen was used to monitor the in vitro ROS production in CGQ hydrogels upon exposure to 808 nm laser irradiation. 22 As expected, the production level of singlet oxygen representing ROS in CGQ hydrogels is much higher than that in C and CG hydrogels, as demonstrated by the drastic decline of the absorbance intensity of DPBF in the CGQ group (Figure 3 d). This result indicates that CdSe QDs embedded in CGQ indeed generated singlet oxygen via the 808 nm laser‐activated PDP to oxidize DPBF. Furthermore, intracellular ROS production was evaluated using a dihydroethidium (DHE) probe. A larger power density brings about a stronger fluorescence signal, suggesting more ROS production in cells ( Figure 4 a), since the pink fluorescence intensity positively correlates with the level of ROS production. In particular, upon exposure to 167 mW cm −2, the fluorescence intensity is approximately identical to that in the ROS inducer (i. e. , EAtB)‐treated group, suggesting the same ROS level. However, the signal was extinguished after adding the ROS scavenger, i. e. , N ‐acetyl cysteine (NAC), which decreased ROS production despite an identical PDP. 23 These results sufficiently validate ROS production via the PDP in CGQ hydrogels, which enables the facilitated differentiation of BMSCs toward chondrogenesis. Afterward, in vitro safety of CGQ‐mediated ROS was evaluated to determine the optimal power density under which the viability can reach the largest value. As indicated in Figure 4 b, CGQ treatment with 808 nm laser at a power density of 16. 7 mW cm −2 corresponding to 3 J cm −2 gives birth to the largest cell viability, denoting that the optimal in vitro laser fluence is 3 J cm −2. Figure 4 In vitro chondrogenesis of BMSCs induced by intensified stiffness and ROS production in CGQ. a) Laser confocal microscopy images of cells embedded in CGQ after treatment with PDP at various laser power densities for 3 min, wherein ROS production was assessed with DHE probe (pink color) and stronger pink fluorescence signal means more ROS; NAC (a ROS scavenger) and EAtB (1 × 10 −3 m Elesclomol+100 × 10 −3 m AAPH+100 × 10 −3 m tBHP, the ROS inducer) were used; scale bar = 20 µm. b) In vitro viabilities of BMSCs cultured in CGQ scaffold after irradiation with different laser fluences for 3 min. c) Cell viability of BMSCs cultured in the different scaffolds (i. e. , C, CG, and CGQ) with or without PDP for a period of 21 days (scale bar = 30 µm). d–f) mRNA expression and g) GAG content of BMSCs after different treatments for 7 days (d), 14 days (e), and 21 days (f). h) Cell skeleton staining of the cells cultured in the scaffolds with or without PDP for 21 days (scale bar = 100 µm). Mean ± SD, n = 5; *, ** and *** indicates p < 0. 05, 0. 01, and 0. 001, respectively (C = collagen + BMSCs, CG = collagen crosslinked with genipin + BMSCs, CGQ = collagen crosslinked with genipin and QDs + BMSCs, C + PDP = collagen + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CG + PDP = CG scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CGQ + PDP = CGQ scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min). 2. 4 Improved In Vitro Proliferation and Differentiation of BMSCs Using CGQ in Combination with Laser Irradiation It is reported that increased ROS production may intensify cell proliferation in CGQ. 24 Indeed, after incubation with CGQ for 21 days, BMSCs in the CGQ+PDP group attain the highest proliferation rate, as evidenced by the qualitative observation (Figure 4 c) and quantification analysis (Figure S5, Supporting Information). Besides promoting BMSC proliferation, ROS can also promote differentiation and proteoglycan secretion of BMSCs, 25 thus yielding a hypothesis that increased ROS together with intensified stiffness may produce a synergistic effect in promoting BMSC differentiation. To validate this hypothesis, quantitative real‐time polymerase chain reaction (qRT‐PCR) was used to analyze cartilage‐specific gene expression including Acan, Sox9, Col2a1, and Col1a1 (Table S2, Supporting Information) which regulate the differentiation of BMSCs. 26 In all periods examined, higher expression levels of cartilage‐specific genes (i. e. , Acan, Sox9, and Col2a1 ) in BMSC‐seeded CGQ hydrogels are observed when compared with that in C or CG hydrogels (Figure 4 d–f). This result validates that the intensified stiffness induced by weaved QDs and crosslinker in CGQ is responsible for the pro‐chondrogenic outcome. Furthermore, upon exposure to laser irradiation, ROS arising from QD‐mediated PDP in CGQ+PDP enables CGQ hydrogels to induce the highest expression of Acan, Sox9, and Col2a1 regardless of the length of incubation time (Figure 4 d–f). Concurrently, the intensified stiffness and ROS production in CGQ+PDP result in significant down‐regulation of a fibrocartilage marker (i. e. , Col1a1 ). 26 These intriguing results indicate that the synergistic effect of appropriate ROS and intensified stiffness will facilitate the BMSCs differentiation and cartilage repair through the upregulations of Acan, Sox9, and Col2a1 and downregulation of Col1a1. Next, the secretion of GAG as a primary component in the extracellular matrix (ECM) of cartilage was monitored to indirectly evaluate the regeneration degree of ectopic cartilage. Consistent with the ranking order of ROS and stiffness, the CGQ+PDP group results in the largest CGA content, and a continuous increase over time is also observed but not in C and CG hydrogels (Figure 4 g). To visualize the differentiation of BMSCs in CGQ+PDP, the cytoskeletal morphology of BMSCs after phalloidin (PI)/DAPI co‐staining was evaluated using laser confocal scanning microscopy (LCSM). Elongation of cells with F‐actin fibers in the cytoplasm that is a typical characteristic of undifferentiated stem cells is observed in all groups (Figure 4 h). In contrast, round‐shaped chondrocytes that severely lose stress fibers are mainly present in CGQ hydrogels, suggesting the successful differentiation of BMSCs in the CGQ group. Upon further laser irradiation, PDP‐mediated ROS trigger more aggregates of round cells, further enhancing the differentiation of BMSCs into chondrocytes. 2. 5 In Vivo Hyaline Cartilage Regeneration Based on Intensified Stiffness and ROS Production in CGQ in a Noncartilaginous Environment To validate the synergistic effect in vivo, BMSCs seeded in C, CG, and CGQ hydrogels were subcutaneously injected into nude mice, and in vivo ectopic cartilage photos were captured after 4 and 8 weeks post‐transplantation, respectively. Comparing to C and CG hydrogels with loose structures, CGQ yields compact tissues with elasticity similar to native articular cartilage ( Figure 5 a), validating the ability of intensified stiffness in CGQ scaffolds to drive cartilage regeneration. Upon exposure to laser irradiation, QDs in CGQ responded to laser and generated ROS which further promoted BMSCs differentiation into chondrocytes, resulting in massive neotissue formation in CGQ+PDP (Figure 5 a). Taken together, the in vivo chondrogenesis results sufficiently demonstrate that the intensified stiffness and augmented ROS production in CGQ+PDP indeed promoted BMSCs differentiation into chondrocytes and gave birth to new and compact cartilage tissues. Interestingly, the dark color representing CG and CGQ is still detectable after the 4th week, but fades away after the 8th week, indicating the concurrent emergence of nascent cartilage‐like tissues and scaffolds' degradation. Figure 5 In vivo evaluations on BMSCs differentiation toward chondrogenesis in the noncartilaginous environment based on the intensified stiffness and ROS production in CGQ. a) Macroscopic observation of subcutaneously implanted tissues in nude mice with or without PDP for 4 and 8 weeks. b) In vivo fluorescence imaging showing ROS generation in the mice implanted with four composites subcutaneously with or without PDP. c) The fluorescence intensity based on data from (b). d, e) Expression levels of chondrogenesis‐specific genes, i. e. , Acan and Col2a1, in regenerated tissues after subcutaneous implantation in nude mice for 4 weeks (d) and 8 weeks (e). f) GAG content in regenerated tissues with subcutaneous implantation in nude mice with or without PDP for 4 and 8 weeks (normalized to the DNA content). Mean±SD, n = 6; *, ** and *** indicate p < 0. 05, 0. 01, and 0. 001, respectively. g) Hematoxylin and eosin (HE) staining for tissues with subcutaneous implantation in nude mice with or without PDP for 4 and 8 weeks (scale bar = 40 µm). Note: C = collagen + BMSCs, CG = collagen crosslinked with genipin + BMSCs, CGQ = collagen crosslinked with genipin and QDs + BMSCs, C + PDP = collagen + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CG + PDP = CG scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CGQ + PDP = CGQ scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min. To comprehensively understand the roles of intensified stiffness and ROS production in facilitating cartilage regeneration, deep principle was explored. ROS levels in the subcutaneously implanted noncartilaginous environments that received different treatments were first examined. As shown in Figure 5 b, c, evident fluorescence signal is observed in the group of CGQ combining with laser (i. e. , CGQ+PDP), but not in other groups, indicating that in vivo CdSe QD‐mediated ROS production occurred only in the CGQ+PDP group. Similar to in vitro results, more expression of Acan and Col2a1 was induced by intensified stiffness in BMSC‐seeded CGQ hydrogels, which was further strengthened by ROS after laser irradiation (Figure 5 d, e). The largest GAG secretion in the transplanted tissues also occurred in the CGQ+PDP group (Figure 5 f). These intriguing results further support that intensified stiffness and ROS production cooperatively contributed to the accelerated chondrogenesis. Furthermore, pathological examination by HE staining shows fibrocartilage‐like tissues in C or CG hydrogels, but spherical chondrocyte cells embedded in the lacuna that are the typical representative of hyaline cartilage phenotype are found in CGQ alone or CGQ combined with laser irradiation (Figure 5 g). This phenomenon definitely indicates the occurrences of chondrogenesis differentiation of embedded BMSCs and cartilage regeneration in CGQ alone or CGQ+PDP. The complete disappearance of residual scaffolds represented by blank (i. e. , no staining) after 8 weeks post‐injection suggests an excellent degradation rate in vivo that is consistent with the cartilage regeneration rate, akin to in vitro result (Figure S4, Supporting Information). In addition, CGQ in combination with PDP demonstrates an excellent safety profile as no obvious body weight loss of mice is observed in any of the groups (Figure S6, Supporting Information). 2. 6 Accelerated Cartilage Regeneration in Defects using CGQ in Combination with Laser Irradiation (i. e. , CGQ+PDP) Inspired by above success in facilitating BMSCs differentiation into a chondrocyte phenotype and accelerating cartilage regeneration, such a synergistic effect for cartilage defect repair by regenerating cartilage‐like tissues was explored. Identical to aforementioned in vitro results, the laser setting, i. e. , 3. 0 J cm −2 (16. 6 mW cm −2 ) for 3 min, is also determined as the optimal dose of laser fluence in this in vivo evaluation (Figure S7, Supporting Information). Digital photos show that the boundaries between uneven reparative tissues and original cartilage in C, CG, and CGQ hydrogels are distinguishable after 4 weeks post‐transplantation. However, the reparative tissues in the C and CG groups with or without PDP are loose and fibrous, while the regenerated tissues in the CGQ group are more compact and interconnected with the adjoining cartilage especially after exposure to laser irradiation (i. e. , CGQ+PDP) ( Figure 6 a). After 8 weeks, the regenerated tissues in the C and CG groups are still loose and nonintegrated with the adjacent normal cartilage. In contrast, the newborn tissues in defects treated with CGQ evolve into smooth and well‐integrated ones with surrounding tissues (Figure 6 a). Interestingly, the boundary between regenerated tissues and adjacent cartilage in CGQ+PDP vanishes, suggesting no obvious difference between repaired cartilage and adjacent cartilage. Figure 6 In vivo cartilage regeneration in defects after implanting BMSC‐seeded CGQ scaffolds. a, b) Gross macroscopic observation (a) and O'Driscoll score (b) of engineered cartilage tissues subjected to different treatments after 4 and 8 weeks, respectively. c) Mechanical property evaluation of the engineered cartilage tissues subjected to different post‐treatments after 8 weeks. d) Histological score of the engineered cartilage tissues subjected to different post‐treatments after 4 and 8 weeks, respectively. e) HE (top) and Safranin O (bottom) staining of the engineered cartilage tissues subjected to different post‐treatments (scale bar = 200 µm). Mean ± SD ( n = 6); *, ** and *** indicates p < 0. 05, 0. 01, and 0. 001, respectively. (C = collagen + BMSCs, CG = collagen crosslinked with genipin + BMSCs, CGQ = collagen crosslinked with genipin and QDs + BMSCs, C + PDP = collagen + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CG + PDP = CG scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CGQ + PDP = CGQ scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min). Noticeably, two characteristics of inflammation or synovitis, i. e. , osteophytes formation and cartilage erosion, are not evident in all groups (Figure 6 a), which suggests that a moderate ROS production in CGQ+PDP is insufficient to induce inflammation that may damage articular cartilage. This result is consistent with previous reports that the dominant skeleton (i. e. , collagen) in CGQ as a normal composition of human body fails to induce inflammation. 27 We further used O'Driscoll scoring to directly assesses the progress of repaired cartilage tissues. CGQ+PDP harvests the highest scores (Figure 6 b), further confirming that CGQ+PDP performed the best in cartilage repair. As a load‐bearing material of diarthrodial joints, articular cartilage can absorb mechanical shocks and distribute high joint loads more evenly, simultaneously maintaining minimal friction and wear. 28 Thus, detecting the equilibrium modulus of regenerated cartilage tissues is of significance. Herein, the stiffness of regenerated cartilage tissues after 8 weeks post‐transplantation was evaluated using equilibrium modulus. Identical to the ranking order of aforementioned O'Driscoll scoring, the stiffness ranking of regenerated tissues in different groups follows the order: C < CG < CGQ, and 96% and 32% increases in comparison to C and CG, respectively, are obtained in CGQ, suggesting that the intensified stiffness of CGQ scaffold caused more robust regenerated cartilage. Furthermore, the synergistic effect involving intensified stiffness and ROS production in CGQ+PDP enables the regenerated tissues to acquire the largest stiffness with an increase of compression modulus by 124% (Figure 6 c). Notably, the stiffness value (above 2 MPa) of regenerated cartilage tissues in either CGQ alone or CGQ+PDP groups after 8 week post‐implantation is within the stiffness window (1–10 MPa) of native cartilage, which suggests that the regenerated tissues can perform the normal function of native cartilage. These results demonstrate the validity of this general combined approach for guiding the design of cartilage‐repairing materials. Histological evaluations by HE and safranin‐O staining show that cartilage defects are filled with nascent tissues in all the groups (Figure 6 e). In C and CG, fibrous and immature tissues with a loose boundary that displays a discontinuous connection with the surrounding cartilage are almost negatively stained by safranin‐O and dominant in defects after 4 weeks post‐operation. Subsequently, they further rounded into fibrocartilage‐like tissues after 8 weeks. In contrast, fibrocartilage‐like tissues are first observed in defects treated with CGQ hydrogels after 4 weeks post‐operation, but subsequently evolved into hyaline cartilage tissues that are positively stained by safranin‐O after 8 weeks. Concurrently, the regenerated cartilage tissues exhibit a tight boundary adjacent to cartilage tissues. In particular, after exposure to laser irradiation, the regenerated tissues that are positively stained by intense red color in CGQ+PDP manifest no evident difference from the surrounding normal tissues. Quantitatively, the regenerated tissues in CGQ+PDP receive the highest histological scoring (Figure 6 d), indicating the accelerated healing of cartilage defects due to the intensified stiffness and ROS production. Intriguingly, the ranking order of histological scoring in different groups is consistent with that of aforementioned O'Driscoll scoring. Noticeably, pathological examination after H&E staining also fails to detect inflammatory cells, which further validates no inflammation in all the groups and indicates the biosafety of CGG scaffolds and moderate ROS. As well, the complete disappearance of scaffolds that is represented by no blank in both HE and safranin‐O staining after 8 weeks post‐injection further demonstrates the excellent in vivo degradation of collagen‐based scaffolds. More significantly, the in vivo degradation rate of CGQ scaffold completely matches the time frame of cartilage formation. More intense positive immunohistochemical staining in the CGQ+PDP group indicates more type II collagen in the regenerated tissues (Figure S8, Supporting Information). In contrast, the mostly negative staining in the C and CG groups after either 4 or 8 weeks post‐transplantation suggests the absence of hyaline‐like cartilage, which is in accordance with the results in above experiments. Additionally, the histological slices of engineered cartilage tissues after 4 or 8 weeks post‐implantation were detected by LCSM, and no fluorescence signal of CdSe QDs suggests neglectable internalization by cells (Figure S9, Supporting Information). 2. 7 Signaling Pathways Activated in Chondrogenic Differentiation Promoted by Intensified Stiffness and ROS Production Next, we investigated the signaling pathways associated with BMSCs differentiation into chondrogenesis and cartilage repair promoted by intensified stiffness and ROS production. The influence of intensified stiffness on TGF‐β/SMAD signaling pathway associated with mechanotransduction, 29 was first investigated, since this pathway may regulate cartilage maintenance during embryonic evolution. 30 In this pathway, SMAD2 and SMAD3 are downstream transcription factors that are usually phosphorylated in chondrocytes deriving from articular cartilage via mechanotransduction‐mediated TGF‐β1 or β3 ligand/receptor binding. 31 Thus, it is anticipated that the intensified stiffness in CGQ might be able to increase the phosphorylation of SMAD 2/3 through activating the TGF‐β/SMAD signaling pathway. As expected, the expression levels of SMAD2/3 and phosphorylated SMAD2/3 (p‐SMAD2/3) in BMSC cells treated with CGQ hydrogels are considerably enhanced (>2. 6‐fold) compared with that treated with C hydrogel due to the augmented compression modulus ( Figure 7 a). Moreover, a specific inhibitor of TGF‐β signaling pathway, i. e. , SB431542, 32 was applied to significantly reduce the levels of SMAD2/3 and p‐SMAD2/3. Similarly, the expression level of COL2A1 that is another effector of chondrogenesis is also increased by 1. 7‐fold in the CGQ group, but effectively suppressed by SB431542 (Figure 7 b). These results validate the involvement of TGF‐β/SMAD signaling pathway in BMSCs chondrogenesis promoted by intensified stiffness in CGQ and suggest that matrix stiffness tuning can serve as an effective approach to augment chondrogenesis. Figure 7 Signaling pathway exploration of chondrogenic differentiation promoted by intensified stiffness and augmented ROS production in CGQ. a, b) Expression levels of SMAD2/3, phosphorylated SMAD2/3 (p‐SMAD2/3) (a) and COL2A1 proteins (b) in cells cultured in C and CGQ scaffolds in vitro for 14 days with or without SB (SB43154). c–e) ROS generation (c), expression levels of mTOR, phosphorylated mTOR (p‐mTOR) (d) and COL2A1 (e) proteins in the cells encapsulated in CGQ scaffold in vitro with or without PDP for 14 days (NAC was used). f) Schematic of the associated signaling pathways involved in mechanotransduction and laser irradiation. Mean ± SD, n = 5; ** and ## indicate p < 0. 01; *** and ### indicate p < 0. 001 (C = collagen + BMSCs, CGQ = collagen crosslinked with genipin and QDs + BMSCs, C + PDP = collagen + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min, CGQ + PDP = CGQ scaffold + BMSCs + irradiation with an 808 nm laser at fluence of 3 J cm −2 for 3 min). As another pivotal factor, ROS were also demonstrated to facilitate the chondrogenic differentiation of BMSCs and cartilage repair in aforementioned experiments. It has been well documented that ROS can mediate cellular alterations of crucial second messengers to promote differentiation through activating mTOR signaling pathway that is critical to chondrogenesis. 33 Thus, evaluations on ROS‐mediated mTOR signaling pathway were undertaken, wherein mTOR phosphorylation and COL2A1 expression were highlighted. In Figure 7 c, laser irradiation induces marked abundant ROS production in CGQ, which was accompanied by increased mTOR phosphorylation and COL2A1 expression in BMSCs embedded in CGQ+PDP by 60. 2% and 120. 5%, respectively, compared with CGQ alone (Figure 7 d, e). After adding a ROS scavenger (e. g. , NAC), 23 the expression of mTOR phosphorylation and COL2A1 are tremendously suppressed and no obvious difference is observed between CGQ and CGQ+PDP. These results confirm the role of ROS in promoting chondrogenesis via activating mTOR signaling pathway. Given the above results, we proposed a model that depicts the upstream and downstream signaling events in chondrogenesis associated with stiffness and ROS (Figure 7 f). In detail, laser irradiation can trigger BMSC‐seeded CGQ scaffolds to produce ROS via the QD‐mediated PDP process, which activates the mTOR signaling pathway to promote more expression of a cartilage‐specific gene, i. e. , Sox9. On the other hand, the intensified stiffness induced by crosslinking with QDs and genipin in the CGQ scaffold can exert considerable influences on the TGF‐β/SMAD signaling pathway to further induce Sox9 expression. The activated Sox9 subsequently activates the other two downstream cartilage‐specific genes, i. e. , ACAN and COL2A1 in sequence, facilitating chondrogenesis. Therefore, enhancing stiffness and ROS production can serve as a unified method to achieve highly efficient cartilage repair by driving BMSCs differentiation. As well, the flexible hydrogels can be engineered into various shapes objective to irregular‐shaped defects (Figure S10, Supporting Information). 3 Conclusion In summary, an injectable collagen‐based composite hydrogel (CGQ) was obtained using collagen as the platform for carrying BMSCs and crosslinking CdSe QDs. This composite hydrogel exhibits appropriate biodegradability and excellent biocompatibility. The crosslinking and QDs introduction made the stiffness of CGQ considerably improved, and CGQ could give rise to increased ROS in the presence of laser irradiation via the QD‐mediated PDP process. A synergistic effect involving intensified stiffness and augmented ROS production has been demonstrated to promote more expression of cartilage‐specific genes and accelerate BMSCs differentiation into chondrocytes both in vitro and in vivo, enabling cartilage repair in cartilage defects. More significantly, the two means have been demonstrated to engage different pathways to regulate sequential expression of cartilage‐specific genes, e. g. , intensified stiffness activates the TGF‐β/SMAD signaling pathway, while ROS production correlates with mTOR signaling pathway. Collectively, this synergistic strategy and the corresponding composite hydrogels lay a solid foundation to highly efficient cartilage regeneration for future clinical application and offer a new avenue to guide the design of new scaffolds capable of cartilage repair. 4 Experimental Section Materials, methods, all experimental details, and procedures are included in the Supporting Information. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201900218
| 2,019
|
Advanced Science
|
Expanding the Functional Scope of the Fmoc‐Diphenylalanine Hydrogelator by Introducing a Rigidifying and Chemically Active Urea Backbone Modification
|
Abstract Peptidomimetic low‐molecular‐weight hydrogelators, a class of peptide‐like molecules with various backbone amide modifications, typically give rise to hydrogels of diverse properties and increased stability compared to peptide hydrogelators. Here, a new peptidomimetic low‐molecular‐weight hydrogelator is designed based on the well‐studied N ‐fluorenylmethoxycarbonyl diphenylalanine (Fmoc‐FF) peptide by replacing the amide bond with a frequently employed amide bond surrogate, the urea moiety, aiming to increase hydrogen bonding capabilities. This designed ureidopeptide, termed Fmoc—Phe—NHCONH—Phe—OH (Fmoc‐FuF), forms hydrogels with improved mechanical properties, as compared to those formed by the unmodified Fmoc‐FF. A combination of experimental and computational structural methods shows that hydrogen bonding and aromatic interactions facilitate Fmoc‐FuF gel formation. The Fmoc‐FuF hydrogel possesses properties favorable for biomedical applications, including shear thinning, self‐healing, and in vitro cellular biocompatibility. Additionally, the Fmoc‐FuF, but not Fmoc‐FF, hydrogel presents a range of functionalities useful for other applications, including antifouling, slow release of urea encapsulated in the gel at a high concentration, selective mechanical response to fluoride anions, and reduction of metal ions into catalytic nanoparticles. This study demonstrates how a simple backbone modification can enhance the mechanical properties and functional scope of a peptide hydrogel.
|
Peptide hydrogels are solid‐like, biocompatible, and biodegradable supramolecular materials, which are hence especially suitable for biological and biomedical applications such as 3D cell culture, tissue engineering, and controlled drug release. 1, 2 Peptide hydrogels can also be utilized for other applications, such as sensing, catalysis, and optoelectronics, in which they act as structural scaffolds or directly perform desired functions. 3 Yet, the application of peptide hydrogels is still limited in many cases due to poor or marginally tunable rheological properties, lack of enzymatic stability, and absence of multifunctionality. 4, 5, 6 Moreover, in the process of generating application‐specific functionalities by mutating the amino acid sequence, the mechanical properties of peptide hydrogels may become impaired. 7, 8, 9 A recently emerging strategy to address these issues is the design of peptidomimetic hydrogelators, in which one or more of the typical amide bonds have been replaced by other chemical groups that tether the constituent amino acids. This molecular design has led to metabolically stable nanostructures with tuneable mechanical properties, which serve as platforms in therapeutic and biomedical applications. 4, 10, 11 Such studied peptidomimetic hydrogelators include depsipeptides, 12 oxazolidine‐tethered, 13 and cyclobutane‐tethered peptides, 14 which form hydrogels of improved rigidity and longer biostability compared to their native α‐peptide counterparts. Here, we aimed to further explore the beneficial effect of backbone modification on hydrogel properties and functionalities. To this end, we selected the well‐studied N ‐fluorenylmethoxycarbonyl diphenylalanine (Fmoc‐FF) hydrogelator as a model system. 15, 16, 17, 18, 19 Several approaches, such as co‐assembly, covalent modification of the side chain or protecting group, variation of solvent composition, and pH modulations, have been employed in multiple studies to fine‐tune the physical properties of the Fmoc‐FF hydrogel. 17, 19, 20, 21, 22, 23, 24, 25 Yet, to our knowledge, its backbone modification has been attempted only twice by modifying the Fmoc‐FF amide group into N ‐benzyl glycine (Nphe) or ester. These modifications resulted in decreased hydrogen bonding, which led, in turn, to decreased rigidity of the formed hydrogel, as compared to the Fmoc‐FF under the same conditions. 26, 27 Accordingly, it would be expected that increasing the hydrogen bonding capability of the hydrogelator by the incorporation of a urea moiety would enhance the rigidity of its hydrogels. Indeed, a related congruent report has shown that urea‐based nonpeptidic gelators self‐associate through N—H/O hydrogen bonds to form stable six‐membered rings based on two donors and one carbonyl acceptor, thereby resulting in stronger urea–urea α‐tape hydrogen bonding interactions. 28, 29 Moreover, the presence of a urea moiety could impart self‐assembled nanostructures with novel functional properties or increased mechanical strength and metabolic stability, as compared to their unmodified counterparts. 29, 30, 31 For these reasons, we selected the urea group as a backbone modification for Fmoc‐FF. We report the incorporation of a urea moiety such that it substitutes the amide bond of Fmoc‐FF to form Fmoc—Phe—NHCONH—Phe—OH (Fmoc‐FuF, Figure 1 a). Fmoc‐FuF proved to be an efficient hydrogelator, giving rise to a hydrogel of improved rigidity, as compared to the unmodified Fmoc‐FF hydrogel. The Fmoc‐FuF hydrogel also exhibited self‐healing, shear thinning, and in vitro biocompatibility. In terms of functionality, substrates coated with the Fmoc‐FuF xerogel reduced the accumulation of bacteria, demonstrating rudimentary antifouling properties. Furthermore, free urea was encapsulated at a high concentration within the Fmoc‐FuF hydrogel network, resulting in slow release of the urea into the environment and hence suggesting a potential application of Fmoc‐FuF in slow‐release urea fertilizers. Fmoc‐FuF also showed characteristics of nonpeptidic urea‐based gelators, such as mechanical responsiveness to anion stimuli and reduction of metal ions into nanoparticles. Importantly, these functional properties were not displayed by the unmodified Fmoc‐FF. Overall, we developed a multifunctional aromatic ureidopeptide hydrogelator of low molecular weight with potential envisioned applications in tissue engineering, chemical catalysis, antifouling, and agriculture. Figure 1 Characterization of the Fmoc‐FuF hydrogel. a) Chemical structure of Fmoc‐FF and Fmoc‐FuF hydrogelators, highlighting (yellow) the unmodified (Fmoc‐FF) or modified (Fmoc‐FuF) backbone. b) Left: OD kinetics at 400 nm. Right: Photograph of the semitransparent Fmoc‐FuF hydrogel. c) Frequency sweep measurement for Fmoc‐FuF and Fmoc‐FF hydrogels. Data represent mean ± standard deviation (SD, n = 3 hydrogels per condition). d) Flow sweep measurement showing shear‐thinning behavior of Fmoc‐FuF hydrogel. e) Five‐step loop time sweep measurement showing the thixotropic nature of Fmoc‐FuF hydrogel. f) TEM and g) HRSEM images of Fmoc‐FuF fibers. h) Partial 1 H NMR spectrum, showing the NH and aromatic regions, in the solution and gel states. i) Fluorescence emission spectrum of Fmoc‐FuF in solution and gel states (λ ex = 285 nm). The Fmoc‐FuF hydrogel was prepared using the solvent‐switch method by diluting a dimethyl sulfoxide (DMSO) stock solution of Fmoc‐FuF with water to a final concentration of 0. 5 wt% and 20% (v/v) DMSO. Thus, a turbid Fmoc‐FuF solution had formed, which transitioned within 1 min into a semitransparent hydrogel that became more optically clear over a period of 30 min as per turbidometry (Figure 1 b; Figure S1, Supporting Information). Compared with Fmoc‐FuF, a control of Fmoc‐FF presented higher initial solution turbidity, slightly slower gelation time (≈90 s), and considerably faster rate of optical clearance (Figure 1 b; Figure S1, Supporting Information), suggesting distinct gelation dynamics for the two hydrogelators. The Fmoc‐FuF hydrogel is elastic, as was evident by rheological measurements. Oscillatory strain sweep (0. 1–300%; Figure S2, Supporting Information) and frequency sweep (0. 1–100 Hz; Figure 1 c) showed that in the linear viscoelastic region, the storage modulus ( G ′) of the Fmoc‐FuF hydrogel is an order of magnitude higher than its loss modulus ( G ″), characteristic of elastic hydrogels. 32 Interestingly, the G ′ value (at 10 Hz) of the Fmoc‐FuF hydrogel is 2. 7‐fold higher than that of Fmoc‐FF at the same wt% value (Figure 1 c). In line with this observation, oscillatory strain measurements at a constant frequency of 1 Hz showed that the critical strain value γ, required for gel breaking, is higher by 20% for Fmoc‐FuF, as compared to Fmoc‐FF at the same wt% value (Figure S2a, b, Supporting Information). These data therefore show that the urea backbone modification results in the improvement of the hydrogel mechanical properties, as was also reported for a nonpeptidic hydrogelator. 31 In this context, it is also worth mentioning the reported effect of pH on the mechanical properties of Fmoc‐FF gels, 19 where higher pH, especially above 5, leads to lower G ′. Interestingly, the measured pH of Fmoc‐FuF hydrogels (5. 2 ± 0. 1) is higher than that of control Fmoc‐FF gels (4. 7 ± 0. 5) and yet G ′ of the former hydrogel is higher. Hence, the urea backbone modification, while leading to a higher gel pH, overall improves the mechanical properties of the hydrogel. We note that although the incorporation of a backbone urea group is expected to improve the hydrogel mechanical properties by enhancing hydrogen bonding, its beneficial effect may be exerted by additional mechanisms, such as modulation of other noncovalent interactions. Rheologically, Fmoc‐FuF hydrogel presented shear‐thinning behavior (Figure 1 d), similar to that of Fmoc‐FF, in line with a previous report on Fmoc‐FF 33 (Figure S2c, Supporting Information). Fmoc‐FuF hydrogels additionally showed thixotropic and self‐healing properties. 33 A five‐step loop time sweep test, at a low strain of 0. 1% ( G ′ > G ″, gel state) and a high strain of 500% ( G ″ > G ′, solution state), 32 demonstrated the thixotropic nature of Fmoc‐FuF although a 10% decrease in G ′ was observed as compared with the first step (Figure S2d, Supporting Information). Similar thixotropic behavior, with a smaller decrease in G ′ (<2%), was observed for Fmoc‐FF hydrogels (Figure S2e, f, Supporting Information). The self‐healing property of the Fmoc‐FuF hydrogel was demonstrated by cutting and rejoining two hydrogel monoliths, which then bridged a 2. 5 cm long elevated gap, were lifted vertically from the surface using forceps, and after 2 h did not show visible cut marks (Figure S3a–c, Supporting Information). Similarly, Fmoc‐FF rejoined hydrogel bridged an elevated gap, yet it could not be lifted using forceps (Figure S3d–f, Supporting Information). Similar to the rejoined Fmoc‐FuF hydrogel, rejoined Fmoc‐FF hydrogel did not show visible cut marks after 2 h. We next investigated the underlying nanoscale morphology and molecular organization of Fmoc‐FuF in the gel state. Morphologically, transmission electron microscopy (TEM) and high‐resolution scanning electron microscopy (HRSEM) revealed that the Fmoc‐FuF hydrogel consists of typically entangled, flat, and twisted fibers 10–20 nm in width (Figure 1 f, g). Complementing atomic force microscopy (AFM) imaging showed that the fiber height ranges from 10 to 30 nm (Figure S4, Supporting Information). At the molecular level, 1 H NMR spectroscopy showed significant line broadening of the signals for the gel state as compared with the sharp peaks of dissolved Fmoc‐FuF (Figure 1 h). This change is associated with a reduction in molecular‐scale mobility of the gelator and its conversion to a solid‐like state upon molecular self‐assembly. 34, 35 Since peaks associated with NH protons almost disappeared in the gel versus the dissolved state, a hydrogen bonding network presumably forms upon the self‐assembly of Fmoc‐FuF into fibers in the gel state. Further structural insights were gained by fluorescence spectroscopy. In the emission spectrum (300–500 nm, λ ex = 285 nm), the characteristic fluorenyl ring band showed a redshift from 315 to 321 nm in the gel versus dissolved state, likely due to the formation of fluorenyl excimer with antiparallel arrangement and π–π stacking of the fluorenyl group. 36 The tail in the visible region at 455 nm is indicative of the presence of an extensive J ‐aggregate, which may include both phenyl and fluorenyl rings 36 (Figure 1 i). To elucidate the structural features and self‐assembly mechanism of Fmoc‐FuF assemblies at the molecular level, we performed microsecond‐long coarse‐grained molecular dynamics (CG‐MD) simulations on systems consisting of 200, 400, or 600 Fmoc‐FuF molecules in aqueous solutions containing 20% (v/v) DMSO ( Figure 2 a). In the simulation of the two larger systems, starting from disordered states, the molecules first aggregated into small spherical clusters and large worm‐like irregular aggregates ( t = 0. 1 µs), which then started to fuse into a large branched aggregate showing the structural characteristics of gels. 37 This process was not observed in the smaller system. Figure 2 Self‐assembly mechanism of Fmoc‐FuF revealed by microsecond‐long CG‐MD simulations of 200, 400, and 600 molecules. a) Snapshots of aggregates at four timepoints. The CG representation of an Fmoc‐FuF molecule is shown for clarity. b) Time evolution of the SASA fraction of fluorenyl, phenyl, and main chain groups. c–k) The FEL of worm‐like and branched gel‐like assemblies as a function of the centroid distance and the angle between the two aromatic rings in three different ring pairs. The basins are marked by arrows. l–n) Stacking patterns of fluorenyl–fluorenyl, fluorenyl–phenyl, and phenyl–phenyl ring pairs. The roles of the fluorenyl, two phenyls, and main chain groups in the self‐assembly process were assessed by the fraction of their solvent accessible surface area (SASA; Figure 2 b; Figure S5, Supporting Information). The SASA fraction of the fluorenyl group rapidly dropped within the first 0. 1 µs and started fluctuating at around 0. 2 µs, whereas the SASA fraction of the main chain quickly increased, reaching a plateau. Interestingly, the SASA fraction of the phenyl group did not significantly change throughout the simulation. These data indicate that the fluorenyl group is mostly buried inside the aggregate forming the hydrophobic spine, while the main chain is generally solvent‐exposed. Additionally, the fluorenyl group appears to play a crucial role in the formation of the branched Fmoc‐FuF aggregates. To further examine the importance of aromatic stacking in the self‐assembly process, 36, 38, 39 free energy landscape (FEL) as a function of the centroid distance and the angle of the fluorenyl–fluorenyl, fluorenyl–phenyl, and phenyl–phenyl aromatic ring pairs was calculated (Figure 2 c–k). The basin located at 10° (0. 45 nm) of angle (centroid distance) for the 200/400/600 systems indicated a strong preference for parallel stacking between fluorenyl rings in the nonbranched worm‐like aggregate (Figure 2 c–e). The additional basin at 80°, 0. 45 nm, in the 400/600 systems (Figure 2 d, e) indicated that besides parallel, perpendicular (T‐shaped) stacking patterns are also preferred. The difference between Figure 2 c–e reveals the importance of T‐shaped stacking in the formation of branched gel‐like aggregates (Figure 2 l). A shallow minimum‐energy basin at 80°, 0. 6 nm, in the FEL of the 200‐Fmoc‐FuF system (Figure 2 f, i) corresponded to parallel stacking of fluorenyl–phenyl and phenyl–phenyl rings. This was much deeper and larger in the 400/600 systems (Figure 2 g, h, j, k), corresponding to T‐shaped, herringbone, and parallel stacking patterns (Figure 2 m, n). These results revealed different stacking pattern preferences in nonbranched worm‐like and branched gel‐like aggregates, and the crucial role of the T‐shaped fluorenyl–fluorenyl stacking pattern in the formation of branched gel‐like aggregates. Interestingly, the π–π stacking interaction of Fmoc groups also plays an important role in stabilizing the Fmoc‐AA supramolecular assemblies, as reported in a recent MD simulation study by Mu et al. 40 Following structural investigation, we explored five functionalities of Fmoc‐FuF and its hydrogel, in which we compared the performance of Fmoc‐FuF to that of Fmoc‐FF. First, the anion binding of Fmoc‐FuF was explored. Urea‐based compounds are prominent anion receptors, which bind anions in a specific manner via two directional hydrogen bonds. 29, 41, 42 In the case of Fmoc‐FuF, specific interaction with fluoride ions (F − ) generated blue fluorescence, attributed to the formation of a charge‐transfer complex 42 ( Figure 3 a). Blue fluorescence upon UV illumination was observable to the naked eye when an Fmoc‐FuF DMSO solution was supplemented with F −, but not with other anions (Figure 3 b, a 100 × 10 −3 m pH of ≈7 stock solution of tetrabutylammonium fluoride, TBAF, was used as the F − source). Correspondingly, the appearance of a 450 nm emission band upon 285 nm excitation was observed for Fmoc‐FuF DMSO solution supplemented with F − solution, but not with other anions (Figure 3 c). In contrast, when unmodified Fmoc‐FF was tested for interaction with F −, no change in the emission spectrum was detected (Figure 3 c). The sensitivity of the Fmoc‐FuF optical response to F − was quantified by measuring the fluorescence during gradual addition of aqueous F − solution into Fmoc‐FuF DMSO solution. The fluorescence intensity (FI) was found to increase in a concentration‐dependent manner (Figure 3 d). Examining the FI at 450 nm versus equivalents (equiv. ) of F − showed that a significant increase of the intensity is observed in the range of 0–20 F − equiv. (Figure S6a, b, Supporting Information). The linear fitting of the plot log(FI) versus log[F − ] in the range of 1–10 equiv. ( R 2 = 0. 99; Figure S6c, Supporting Information) and the corresponding plot of FI versus concentration (Figure 3 e) indicate that Fmoc‐FuF can be used to reliably determine F − concentration in the range of (4. 93–50) × 10 −6 m. The lower limit value of 4. 93 × 10 −6 m is similar to that of previously reported fluoride sensors. 42, 43 The observed interaction between Fmoc‐FuF and F − was confirmed using 1 H NMR, which showed that both NH resonances (NH1 = 6. 74 ppm; NH2 = 5. 98 ppm; Figure S6d, Supporting Information) shifted significantly after the addition of F −, in line with a previous report. 44 These data indicate that the F − interacts predominantly with NH protons of the urea moiety via hydrogen bonding. The response of Fmoc‐FuF to F − also extends to the mechanical properties of the hydrogel. A partial gel–sol transition was induced upon the addition of F − to a preformed Fmoc‐FuF hydrogel, presumably due to the disruption of hydrogen bonds 45 (Figure 3 f). This mechanical response was investigated by measuring G ′ of Fmoc‐FuF hydrogels (0. 5 wt%) prepared with different equiv. of F − (while maintaining the final gel volume fixed). Frequency sweep measurements demonstrated that G ′ decreased with increasing concentrations of F −, reaching a value of 100 Pa in the presence of 6. 0 F − equiv. (Figure 3 g). TEM analysis of this quasiliquid (Figure 3 h; Figure S6e, Supporting Information) confirmed the breakdown of the fiber network into very short fibrils and amorphous matter. In contrast, the Fmoc‐FF hydrogel did not interact with F −, as it remained intact in a vial inversion test following overnight incubation with a 6. 0 equiv. of F − solution (Figure S6f, Supporting Information). We note that neither Fmoc‐FuF nor Fmoc‐FF hydrogels dissolved when sodium phosphate buffer, of the same pH and concentration as the TBAF solution (100 × 10 −3 m, pH ≈ 7), was added to preformed hydrogels instead of TBAF (Figure S6f, Supporting Information). It is therefore concluded that, in contrast to the Fmoc‐FF hydrogel, the Fmoc‐FuF hydrogel is mechanically impaired by interaction with F −. Figure 3 Selective anion binding by Fmoc‐FuF. a) Schematic representation of selective binding of F − by Fmoc‐FuF, forming a charge‐transfer complex. b) Photograph of Fmoc‐FuF (5 × 10 −6 m ) DMSO solutions after the addition of 500 equiv. of various aqueous anion solutions under UV illumination (λ ex = 365 nm). c) Fluorescence emission spectra of Fmoc‐FuF (5 × 10 −6 m ) DMSO solutions upon the addition of 450 equiv. of various aqueous anion solutions (λ ex = 285 nm). d) Fluorescence intensity (FI) of Fmoc‐FuF (5 × 10 −6 m ) in DMSO solutions upon titration with 0–1000 equiv. of aqueous F − solution (λ ex = 285 nm). e) Plot of log[FI] versus F − concentration. Red line is a linear fit, R 2 = 0. 99. f) Schematic illustration of breakage of Fmoc‐FuF hydogel by F −. Dashed lines represent hydrogen bonds. g) Frequency sweep measurement of Fmoc‐FuF hydrogels (0. 5 wt%) containing increasing F − concentration (0–6. 0 equiv. of F − ). h) TEM image of Fmoc‐FuF gel (0. 5 wt%) containing 6. 0 equiv. of F −. The next functionality tested was the ability of the Fmoc‐FuF hydrogel to reduce metal ions into nanoparticles in situ. We selected to synthesize Au and Ag nanoparticles (AuNPs and AgNPs) in order to demonstrate gel‐mediated nanoparticle synthesis 46, 47, 48 and to test the involvement of the urea group, which was previously associated with metal nanoparticle synthesis. 49, 50, 51 Nanoparticle‐containing Fmoc‐FuF hydrogels were prepared by gelating either 5 × 10 −3 m HAuCl 4 or 10 × 10 −3 m AgNO 3 aqueous solutions (with DMSO at 20% (v/v)) and allowing the nanoparticles to form in the dark in situ. This process resulted in color change to dark pink within 4 h in the former case or dark gray after 3 days in the latter (Figure S7a, b, Supporting Information). Hydrogels containing AuNPs or AgNPs showed characteristic surface plasmon resonance (SPR) bands at 530 and 430 nm, respectively ( Figure 4 a). TEM imaging showed that in either case, nanoparticle formation did not alter the underlying fibrous morphology of the gel. Instead, nanoparticles appeared to aggregate along Fmoc‐FuF fibers (Figure 4 b, c). The presence of AuNPs or AgNPs on the fibers was confirmed by selective area electron diffraction (SAED; Figure S7c, d, Supporting Information) and energy‐dispersive X‐ray (EDX) spectroscopy (Figure S7e, f, Supporting Information). Notably, no external reducing agent was added, nor was any external activation applied, such as heat, light, or pH modulation. In striking contrast, Fmoc‐FF hydrogel of the same concentration and DMSO content failed to reduce these metal ions into nanoparticles even after prolonged co‐incubation (Figure S7g, Supporting Information). These data suggest that the backbone urea group either directly reduces metal ions or facilitates metal ion reduction by the solvent, and thus allows for the Fmoc‐FuF gel as a whole to act as a template for in situ synthesis of AuNPs and AgNPs. Metal nanoparticles are commonly used as catalysts in various chemical reactions. The catalytic function of the AuNP‐containing Fmoc‐FuF xerogel was tested using a model reaction, in which 4‐nitrophenol (4‐NP) is reduced to 4‐aminophenol (4‐AP). 52, 53 Time‐dependent UV–vis spectroscopy showed the characteristic decrease in intensity of the 400 nm band and the concomitant increase in absorption around 300 nm, associated with the conversion of 4‐NP to 4‐AP (Figure 4 d). The rate constant of the reaction calculated from the slope of the plot of log( A ) versus time was 0. 089 min −1 (Figure S8a, Supporting Information). Importantly, reusing the AuNP‐containing xerogel as a catalyst was possible following washing with water and drying, with the reaction constant remaining nearly identical throughout a total of four reaction cycles (Figure 4 e; Figure S8b, Supporting Information). Figure 4 Diverse functionalities of the Fmoc‐FuF hydrogel. a) UV–visible spectra showing SPR bands of AuNPs or AgNPs formed in situ in the hydrogel (hydrogels were diluted threefold prior to measurement). b, c) TEM images of b) AuNP‐ and c) AgNP‐decorated fibers. d) Time‐dependent UV–vis spectra and e) corresponding normalized rate constant for repeated reaction cycles of the reduction of 4‐NP to 4‐AP, catalyzed by AuNP‐containing Fmoc‐FuF xerogel. f) Cell viability as determined by XTT assay, performed at different timepoints on 3T3 fibroblast cells cultured in naïve medium (control) or medium preincubated with Fmoc‐FuF hydrogel. Data represent mean ± SD from three independent experiments. Each timepoint was compared to control using Student's t ‐test, P > 0. 05 for all comparisons, indicating no statistically significant differences (ns). g) Representative Live/Dead staining of 3T3 fibroblast cells after 48 h of incubation on an Fmoc‐FuF hydrogel scaffold, showing a high abundance of living cells (green) and the negligible presence of dead cells (red). h) Bacterial adhesion to bare glass and glass surfaces coated with either Fmoc‐FF or Fmoc‐FuF xerogels, as per CFUs' count. Difference between Fmoc‐FuF coating and bare glass is significant ( P < 0. 001), difference between Fmoc‐FF coating and bare glass is not significant (ns, P > 0. 05), as per Student's t ‐test. Data represent mean ± SD from three independent experiments. i) Comparison of G ′ values of Fmoc‐FuF and Fmoc‐FF hydrogels with or without encapsulated urea. Data represent mean ± SD ( n = 3 hydrogels per condition). j) Comparison of cumulative urea release from urea‐containing Fmoc‐FuF (black) or Fmoc‐FF (gray) hydrogels, measured over a period of 6 days. We next explored functionalities of the Fmoc‐FuF hydrogel related to interaction with mammalian and bacterial cells in order to evaluate its potential use in biomedical as well as antifouling applications. Compatibility with mammalian cells was tested using murine embryo fibroblasts (3T3). Initially, cells were incubated for 48 h with culture media pretreated by overnight incubation with Fmoc‐FuF hydrogels, and their viability was examined using 2, 3‐bis‐(2‐methoxy‐4‐nitro‐5‐sulphophenyl)‐2 H ‐tetrazolium‐5‐carboxanilide (XTT) cell viability assay. Treated cells showed over 90% viability compared with control cells grown in naïve culture medium (Figure 4 f), indicating that toxic molecules are not released from the hydrogel. Cellular compatibility was further tested by assessing the adhesion of cells to the hydrogel, a key requirement for biomedical applications. To this end, Fmoc‐FuF scaffolds were cultured with 3T3 cells and tested by Live/Dead analysis in situ, 48 h following seeding. Living cells were abundantly found on the gel scaffold (Figure 4 g; control condition is shown in Figure S9, Supporting Information), showing that the biocompatibility of the Fmoc‐FuF hydrogel is similar to that of Fmoc‐FF, 15 and in line with the reported biocompatibility of nonpeptidic urea–modified supramolecular hydrogels 29 and covalent polymeric materials. 51, 52 Hence, the Fmoc‐FuF hydrogel can support the attachment of cells, as required for biomedical applications. 54, 55 Interestingly, whereas mammalian cell adhesion was supported by the hydrogel, bacterial adhesion (but not viability) was inhibited by its xerogel coating. Considering the application of poly(urea)‐based substrates as antifouling coatings, 56 we examined similar functionality for the Fmoc‐FuF xerogel. Inhibition of bacterial adhesion was examined by a fouling test, in which glass surfaces, either bare, coated with Fmoc‐FuF, or coated with Fmoc‐FF, were incubated in Escherichia coli cultures at 37 °C overnight. Following removal of the surfaces from the culture, adhered bacteria were detached and plated, and the number of colony‐forming units (CFUs) was subsequently counted. Decent reduction in bacterial adhesion was achieved by Fmoc‐FuF xerogel coating, where 3 × 10 7 CFUs cm −2 were counted as compared with 1. 2 × 10 8 CFUs cm −2 for bare glass, representing 0. 6 log reduction. In contrast, for Fmoc‐FF xerogel coating 1. 1 × 10 8 CFUs cm −2 were counted, indicating the absence of an antifouling effect by Fmoc‐FF (Figure 4 h). Furthermore, as a hydrogel, Fmoc‐FuF did not present antibacterial activity as per quantitative growth analyses of hydrogel‐grown bacteria and bacterial Live/Dead carried out on planktonic bacteria incubated with the hydrogels (Figure S10, Supporting Information), whereas Fmoc‐FF did present such activity (Figure S10, Supporting Information), as reported. 57 Taken together, these results show that the urea backbone modification of Fmoc‐FuF ultimately promotes antifouling activity, albeit to a limited extent. Importantly, such activity was not demonstrated by Fmoc‐FF. Finally, considering its inherent urea modification, we tested if Fmoc‐FuF could encapsulate free urea in the gel matrix for subsequent slow release. Slow release of the nitrogen source urea is desirable in agricultural fertilizers, where various strategies have been applied to delay urea release from fertilizers in order to maintain a steady supply of nitrogen for crop growth as well as reduce environmental pollution by urea leakage into the soil. 58, 59 Fmoc‐FuF was found to gelate urea solutions such that a final concentration of up to 8 m urea in the gel was obtained. These urea‐containing gels were characterized and their ability to release urea was evaluated. Compared to typical Fmoc‐FuF hydrogel, the urea‐containing Fmoc‐FuF gel showed a denser network of fibers (Figure S11, Supporting Information) and a marked increase of 2. 5‐fold in its G ′ value (at 10 Hz; Figure 4 i). Interestingly, inclusion of urea in an Fmoc‐FF gel increased its G ′ value but only by less than 1. 5‐fold (at 10 Hz), such that it was lower than the G ′ value of the urea‐containing Fmoc‐FuF gel by more than an order of magnitude (Figure 4 i). Notably, G ′ value increased for both Fmoc‐FuF and Fmoc‐FF gels, despite 19 the increase in gel pH (5. 5 ± 0. 2 for Fmoc‐FF and 5. 6 ± 0. 3 for Fmoc‐FuF). These data indicate that encapsulated urea overall contributes to G ′, but to a greater extent for Fmoc‐FuF. We attribute this phenomenon to increased hydrogen bonding, which is expected to be more dominant for the urea‐modified 31 Fmoc‐FuF. The release of urea from the urea‐containing Fmoc‐FuF gel was evaluated next. To this end, gels were immersed in water for a period of 6 days, during which the release of urea into the aqueous environment was monitored daily by a colorimetric assay. 60 The concentration of released urea increased gradually over time, from 4% (day 1) to 91% (day 6) of its initial concentration in the gel (Figure 4 j). In contrast, urea release from urea‐containing Fmoc‐FF gel was considerably more rapid, with 91% of the initial concentration in the gel released over a period of 2 days (Figure 4 j). Hence, Fmoc‐FuF, but not Fmoc‐FF, may be utilized for encapsulation and slow release of urea. Gelation of urea by Fmoc‐FuF apparently allows for similar or improved urea release behavior as compared with other encapsulation methods, 61, 62 although a comparative study will be required to fully assess this. Furthermore, as the use of existing slow‐release fertilizers is typically limited by their relatively non‐biodegradable, toxic coating materials, 63, 64 the biocompatible Fmoc‐FuF hydrogel may provide a more ecofriendly means for achieving slow release of encapsulated urea in agriculture. In summary, by incorporating a urea backbone modification in an extensively characterized peptide hydrogelator, we have obtained a novel peptidomimetic that gives rise to a multifunctional, stimulus‐responsive, and biocompatible supramolecular hydrogel. This hydrogel forms by self‐assembly of Fmoc‐FuF molecules into spheres and worm‐like irregular clusters, which then fuse to give rise to gel‐like fibers in a process mediated by aromatic stacking and hydrogen bonding. Fmoc‐FuF formed a hydrogel of high mechanical rigidity, with shear thinning and self‐healing properties. This hydrogel possesses diverse functional properties, namely anion sensing, metal‐ion reduction, mammalian cell scaffolding, antifouling capabilities, and the ability to slowly release encapsulated urea. Excluding cell scaffolding, these functional properties clearly stem from, or become significantly enhanced by, a urea backbone modification, as evident by direct comparison with the unmodified peptide hydrogelator Fmoc‐FF. This work demonstrates the significant impact of simple amide bond modification, thus laying the basis for improved biomaterial design toward practical applications. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201900344
| 2,019
|
Advanced Science
|
3D Printing of Personalized Thick and Perfusable Cardiac Patches and Hearts
|
Abstract Generation of thick vascularized tissues that fully match the patient still remains an unmet challenge in cardiac tissue engineering. Here, a simple approach to 3D‐print thick, vascularized, and perfusable cardiac patches that completely match the immunological, cellular, biochemical, and anatomical properties of the patient is reported. To this end, a biopsy of an omental tissue is taken from patients. While the cells are reprogrammed to become pluripotent stem cells, and differentiated to cardiomyocytes and endothelial cells, the extracellular matrix is processed into a personalized hydrogel. Following, the two cell types are separately combined with hydrogels to form bioinks for the parenchymal cardiac tissue and blood vessels. The ability to print functional vascularized patches according to the patient's anatomy is demonstrated. Blood vessel architecture is further improved by mathematical modeling of oxygen transfer. The structure and function of the patches are studied in vitro, and cardiac cell morphology is assessed after transplantation, revealing elongated cardiomyocytes with massive actinin striation. Finally, as a proof of concept, cellularized human hearts with a natural architecture are printed. These results demonstrate the potential of the approach for engineering personalized tissues and organs, or for drug screening in an appropriate anatomical structure and patient‐specific biochemical microenvironment.
|
1 Introduction Cardiovascular diseases are the number one cause of death in industrialized nations. 1 To date, heart transplantation is the only treatment for patients with end‐stage heart failure. Since the number of cardiac donors is limited, there is a need to develop new approaches to regenerate the infarcted heart. 2 Cardiac tissue engineering provides an alternative approach by integrating cardiac cells and 3D biomaterials. 3, 4 The latter serve as temporary scaffolds, mechanically supporting the cells and promoting their reorganization into a functional tissue. 5 Following in vitro maturation, the engineered cardiac patch can be transplanted onto the defected heart. When full integration to the host commences, the biomaterials gradually degrade, leaving a functional living patch that regenerates the heart. 3 The biocompatibility of the scaffolding materials is a crucial factor for eliminating the risk of implant rejection, which jeopardizes the success of the treatment. Therefore, the material itself or its degradation products should be carefully selected. 6 Most ideally, the biomaterial should possess biochemical, mechanical, and topographical properties similar to those of native tissues. Decellularized tissue‐based scaffolds from different sources meet most of these requirements. 7, 8 However, to ensure minimal response of the immune system, completely autologous materials are preferred. 9 Recently, our group has shown a new concept for engineering fully personalized cardiac patches. In this approach, a biopsy of fatty tissue was taken from patients and the cellular and a‐cellular materials were separated. While the cells were reprogrammed to become pluripotent stem cells, the extra‐cellular matrix (ECM) was processed into a personalized hydrogel. Following mixture of the cells and the hydrogel, the cells were efficiently differentiated to cardiac cells to create patient‐specific, immunocompatible cardiac patches. 9 However, these cardiac patches did not contain blood vessel networks that match the anatomical architecture of the patient's vasculature. This pre‐engineered vasculature within the parenchymal tissue was previously shown to be critical for patch survival and function after transplantation. 10, 11, 12, 13, 14, 15, 16, 17 In recent years, the strategy of 3D tissue printing evolved, 18, 19, 20 allowing to create vasculature within hydrogels. 21 However, in most of the studies, the endothelial cells (ECs) that form the blood vessels were printed without the parenchymal tissue, which was later on casted on top of the vessels. 22 In other pioneering works, the researchers were able to print ECs together with thin surrounding tissues. 22 However, the obtained tissues were not thick, the ECs did not form open blood vessels and perfusion through them was not demonstrated. Different strategies include printing of the parenchymal tissue with open, a‐cellular channels in‐between, followed by external perfusion of ECs to form the blood vessels. 23, 24, 25 Finally, decellularized hydrogels were also used for printing nonvascularized tissues. 19, 26 Therefore, to the best of our knowledge, the aforementioned studies did not demonstrate printing of a full, thick vascularized patch in one step. Here, we report on the development and application of advanced 3D printing techniques using the personalized hydrogel as a bioink. In this strategy, when combined with the patient own cells, the hydrogel may be used to print thick, vascularized, and perfusable cardiac patches that fully match the immunological, biochemical and anatomical properties of the patient. Furthermore, we demonstrate that the personalized hydrogel can be used to print volumetric, freestanding, cellular structures, including whole hearts with their major blood vessels ( Figure 1 ). 2 Results and Discussion Figure 1 Concept schematic. An omentum tissue is extracted from the patient and while the cells are separated from the matrix, the latter is processed into a personalized thermoresponsive hydrogel. The cells are reprogrammed to become pluripotent and are then differentiated to cardiomyocytes and endothelial cells, followed by encapsulation within the hydrogel to generate the bioinks used for printing. The bioinks are then printed to engineer vascularized patches and complex cellularized structures. The resulting autologous engineered tissue can be transplanted back into the patient, to repair or replace injured/diseased organs with low risk of rejection. Omental tissues from humans ( Figure 2 a) or pigs were obtained. While cells were extracted from one piece of the tissue, the remaining material was decellularized (Figure 2 b), as previously described, 9 and processed to generate a 2. 5% (w/v) thermoresponsive hydrogel, serving as a bioink for 3D printing (Figure 2 c). The bioink, composed of collagenous nanofibers (Figure 2 d, e) behaved as a weak gel at room temperature (RT) and became stronger upon heating to 37 °C (Figure 2 f). 9, 27 Figure 2 Bioinks characterization. A human omentum a) before and b) after decellularization. c) A personalized hydrogel at room temperature (left) and after gelation at 37 °C (right). d) A SEM image of the personalized hydrogel ultrastructural morphology, and e) a histogram of the fibers diameter. f) Rheology measurements of 1% w/v and 2. 5% w/v omentum hydrogels, showing the gelation process upon incubation at 37 °C. g) Stromal cells originated from human omental tissues were reprogrammed to become pluripotent stem cells (red: OCT4, green: Ki67 and blue: nuclei). h) Differentiation to ECs as determined by CD31 (green) and vimentin staining (red). Differentiation to cardiac lineage: i) staining for sarcomeric actinin (red), j) staining for NKX2‐5 (red), and TNNT2 (green). Scale bars: (e) = 10 µm, (g, i, j) = 50 µm, (h) = 25 µm. Our experiments included two cellular models. One model was used as a proof‐of‐concept for the patient‐specific treatment, and included induced pluripotent stem cells (iPSCs)‐derived cardoimyocytes (CMs) and ECs. The second model relied on rat neonatal CMs, human umbilical vein endothelial cells (HUVECs) and lumen‐supporting fibroblasts, allowing large‐scale printing. Two cellular bioinks were generated. The cardiac cell bioink, which was used for printing the parenchymal tissue, and a bioink consisting of blood vessel‐forming cells. To this end, iPSCs were reprogrammed from the stromal cells of human omental tissues, used for hydrogels preparation. The cells exhibited the pluripotency markers OCT4 and Ki67 (Figure 2 g), and after exposure to endothelial and cardiac cells differentiation protocols, as previously described, 9 expressed CD31 and vimentin (Figure 2 h) and NKX2‐5, sarcomeric actinin, and troponin (Figure 2 i, j). We next sought to fit the 3D printing scheme to the anatomy of a patient. Computerized tomography (CT) of a patient's heart ( Figure 3 a) was used to identify the 3D structure and orientation of the major blood vessels in the left ventricle (Figure 3 b). In order to fit a 3D cardiac patch to the patient's left ventricle, the patch dimensions and the blood vessel geometry were designed by computer‐aided design (CAD) software that used anatomical data from the CT images. CT cannot provide images of small blood vessels, therefore, to ensure adequate exposure of the entire cardiac patch to oxygenated medium during in vitro cultivation and after transplantation, smaller blood vessels were added to the basic vasculature design, according to a mathematical model. The model took into consideration oxygen diffusion according to Fick's second law and consumption according to Michaelis–Menten equation, 28 allowing to design an optimal size, distribution, and orientation of the supplemented blood vessels (Figure 3 c, d; Figure S1, Supporting Information). Figure 3 Imaging of the heart and patch modeling. CT image of a) a human heart and b) left ventricle coronary arteries. c) A model of oxygen concentration profile in an engineered patch. d) Replanning of the model showed better oxygen diffusion, sufficient to support cell viability. The personalized hydrogel was next mixed either with iPSC‐derived CMs or neonatal cardiac cells, and was used to print the engineered cardiac tissue. iPSC‐derived ECs or a combination of mature GFP‐expressing human ECs with RFP‐expressing fibroblasts were mixed with gelatin, which served as a sacrificial bioink. A 3D printer, equipped with extrusion‐based print heads, was used to print the parenchymal cell‐containing hydrogel simultaneously with gelatin that contains blood vessel‐forming cells ( Figure 4 a, b) to create a thick (≈2 mm), 3D, patient‐specific, vascularized patch that had high cell viability (Figure 4 c, d; Movie S1, Supporting Information). Upon incubation at 37 °C, the blood vessel‐forming cells adhered to the edges of the omentum hydrogel, while the gelatin became liquid and was washed away from the construct, leaving open cellular lumens of ≈300 µm in diameter within the cardiac patches (Figure 4 e–h; Figure S2, Supporting Information). The cardiac patches were physically robust and could be handled easily so that they could be pulled out and returned to the medium without loss of shape (Movie S2 and Figure S3a, Supporting Information). Furthermore, liquid could be infused into the open lumens of the patches, indicating that their lumen structure was maintained (Movie S3 and Figure S3b, Supporting Information). On day 7, the ECs formed a continuous layer at the edges of the hydrogel (Figure 4 e, h), and the iPSC‐derived patches were stained against actinin and CD31 in order to detect CMs and ECs, respectively (Figure 4 f). Furthermore, the interactions of GFP‐expressing ECs and RFP‐expressing fibroblasts (for lumen stabilization) within the forming lumen were evaluated by a confocal microscopy (Figure 4 g, h; Movie S4 and Figure S4, Supporting Information). Figure 4 3D printing of personalized cardiac patches. a) A 3D model of the cardiac patch. b) A side view of the printing concept and the distinct cellular bioinks. c) A printed vascularized cardiac patch. d) Cell viability after printing. e) A printed blood vessel, continuously layered with GFP‐expressing ECs. f) A printed iPSCs‐derived cardiac patch where the blood vessels (CD31 in green) are seen in‐between the cardiac tissue (actinin in pink). g, h) Cross‐sections of a single lumen, showing the interactions of GFP‐expressing ECs and RFP‐expressing fibroblasts. i–k) Calcium imaging within a printed vascularized cardiac patch (separate regions of interest are represented in different colors. The white arrow represents signal direction). i) The lumen of a blood vessel can be easily observed in‐between the cardiac cells. j, k) Quantification of calcium transients across a lumen of the vascularized cardiac patch. l) Transplantation of the printed patch in between two layers of rat omentum. Dashed, white line highlights the borders of the patch. m–o) Sarcomeric actinin (red) and nuclei (blue) staining of sections from the explanted patch (panel (o) represents a high magnification of the marked area in (n). Scale bars: (c) = 1 cm, (e, g, h, l, m) = 100 µm, (f) = 500 µm, (n) = 50 µm, (o) = 25 µm. Next, the function of the printed vascularized cardiac patch was evaluated by measuring calcium transients in the contracting engineered tissue. Signal propagation exceeded 10 cm s −1 at the parenchymal tissue, however a short delay in propagation could be detected across a 300 µm blood vessel, probably due to the detour of the signal around the vessel (Figure 4 i–k; Movies S5–S7, Supporting Information). Finally, to assess the presence and morphology of the printed cells in vivo, the engineered patches were transplanted and secured between two layers of rat omentum (Figure 4 l). Seven days later, the printed patches were located using a fluorescence prestaining of the CMs. The cellularized patches were then extracted, fixed, sectioned, and stained against sarcomeric actinin. As shown, the cells were elongated and aligned, with massive striation, which indicated on their contractility potential (Figure 4 m–o; Movie S8, Supporting Information). Furthermore, the printed lumen could be seen within the tissue by locating the fluorescent ECs (Figure S5, Supporting Information). Overall, taken together, these results demonstrate the ability to use the patient's own cells and materials in 3D printing of fully personalized, contracting cardiac patches that closely fit to the patient's biochemical and cellular properties, as well as for the anatomy of the patient. This approach can yield several mm thick structures, which is sufficient for engineering clinically relevant cardiac patches. However, when printing whole organs or tissues with significantly larger dimensions and high complexity is required, the patient‐specific hydrogel cannot sustain its weight, and a different printing strategy is needed. Therefore, we next relied on the concept of printing in a supporting medium. 20 Previously, Hinton et al. developed a semi‐transparent support medium composed of gelatin microparticles, which allowed to print embryonic heart structure that did not contained cells. 29 Such strategy could support free‐form printing of structures composed of a variety of bioinks. However, as the extraction procedure of the support medium was by quick melting of the gelatin microparticles at 37 °C, such strategy would not allow long‐term support until the personalized hydrogel is fully cured. Other strategies include the use of synthetic, transparent granular support for printing cellular structures. 30 However, in this case the integrity of delicate structures and viability of sensitive cells can be jeopardized by the mechanical force needed for removal of the support medium. Inspired by these two elegant, pioneering works, we sought to formulate a support medium that better fits the printing of the personalized hydrogel. To this end, we developed a fully transparent, cell‐friendly, microparticulate formulation that allows free‐form printing and curing in a wide range of temperatures, and extraction by a controllable and delicate process ( Figure 5 a). This support material is composed of alginate microparticles in xanthan gum‐supplemented growth medium (Figure S6, Supporting Information) that can undergo safe enzymatic or chemical degradation for extraction (Figure 5 b, c), while maintaining high cell viability (Figure 5 d). Using this printing approach we were able to print accurate, high resolution thick structures from the personalized hydrogel (Figure 5 e–h; Figure S7 and Movies S9–S11, Supporting Information). Figure 5 Printing the personalized hydrogel in a supporting medium. a) A scheme of the 3D printing concept. The construct is free‐formed printed inside the support followed by incubation at 37 °C to crosslink the personalized hydrogel. Then, the structure can be safely extracted by an enzymatic or chemical degradation process of the support material and transferred into growth medium for culturing. b) A multilayered crisscross construct printed inside the support bath, and c) after its extraction. d) Cell viability before and after printing and after extraction. e) 3D confocal image of a double layered construct, printed in the support medium. f) A single strand of the personalized hydrogel within the support. g, h) Accurate, high resolution thick structures printed from the personalized hydrogel. Scale bars: (b) = 0. 5 cm, (e) = 1 mm, (f) = 100 µm, (g, h) = 1 cm. Following, we have utilized the technology to print blood vessels within thick tissues. Here, both cardiac and endothelial cells were mixed with the personalized hydrogel to create two distinct bioinks. Each layer of the printed construct consisted of the parenchymal tissue and the blood vessel‐forming cells, resulting in a thick vascularized tissue (Figure S8, Supporting Information). In this manner, perfusable patches with an actual triaxial lumen were printed. A top view of the entry to one lumen (green) within the cardiac tissue (pink) could be clearly observed by a confocal microscope ( Figure 6 a; Movie S12, Supporting Information). As the immunostaining of thick cellular constructs (7 × 7 × 7 mm) is not efficient for visualizing the core of the tissue, we next supplemented the bioinks with fluorescently dyed nanoliposomes (pink for the parenchymal tissue and orange for the blood vessels). As shown, the triaxial lumen within the thick structure could be visualized (Figure 6 b, c) and dye could be infused, indicating on the potential of the printed blood vessels to efficiently transfer blood (Figure 6 d; Movie S13, Supporting Information). Figure 6 Printing thick vascularized tissues. a) A top view of a lumen entrance (CD31; green) in a thick cardiac tissue (actinin; pink). b) A model of a tripod blood vessel within a thick engineered cardiac tissue (coordinates in mm), and c) the corresponding lumens in each indicated section of the printed structure. d) Tissue perfusion visualized from dual viewpoints. e–k) A printed small‐scaled, cellularized, human heart. e) The human heart CAD model. f, g) A printed heart within a support bath. h) After extraction, the left and right ventricles were injected with red and blue dyes, respectively, in order to demonstrate hollow chambers and the septum in‐between them. i) 3D confocal image of the printed heart (CMs in pink, ECs in orange). j, k) Cross‐sections of the heart immunostained against sarcomeric actinin (green). Scale bars: (a, c, h, i, j) = 1 mm, (g) = 0. 5 cm, (k) = 50 µm. Next, we sought to demonstrate the ability of the approach to print volumetric, complex anatomical architectures. As a proof of concept, we have fabricated small‐scale cellularized human hearts with major blood vessels, based on a digital design (Figure 6 e). The hearts (height: 20 mm; diameter: 14 mm) were printed within the support medium with two distinct bioinks, containing Cy5‐prestained CMs or RFP‐expressing ECs. For better visualization, constructs could be printed with bioinks supplemented with 1 µm blue (blood vessels) and red (cardiac tissue) polystyrene spheres (Figure 6 f, g; Figure S9 and Movie S14, Supporting Information). Following heating to 37 °C and extraction, blue and red dyes were injected into the right and left ventricles, respectively, to demonstrate the integrity of the different compartments (Figure 6 h). The ability to manipulate and perfuse the printed hearts was also demonstrated, indicating on their basic anatomical structure and mechanical stability and robustness (Movies S15 and S16, Supporting Information). The mechanical properties of the printed heart‐structured ECM revealed a close resemblance to the properties of decellularized rat hearts (Figure S10, Supporting Information), indicating on the suitability of the printed hydrogel to serve as a scaffolding material for this purpose. 3D confocal image of the printed heart reveals the initial spatial organization of the CMs and ECs, according to the layers of the printing plan (Figure 6 i; Movies S17 and S18 and Figure S11, Supporting Information). One‐day post printing, a cross‐section of the printed heart was stained against sarcomeric actinin (green) to reveal the internal compartmental structure (Figure 6 j). Higher magnification of the cells comprising the printed heart showed homogeneous distribution of the CMs (Figure 6 k). Future studies should focus on development of designated hardware and procedures to enable efficient and controlled cultivation of the printed organ. Long‐term culturing that provides the tissue with a constant supply of oxygen and nutrients, as well as biochemical, physiological, and electromechanical cues will allow natural maturation processes to take place. Among these processes are the proper assembly of muscle bundles with elongated cells and massive striation, as well as ECM remodeling. Both are essential for generation of functional, mechanically stable organ that may be amenable to clinical applications. 3 Conclusion Overall, in this paper we demonstrate, for the first time, the use of fully personalized, nonsupplemented materials as bioinks for 3D printing. In this approach, a fatty tissue is extracted from the patient and the cellular and a‐cellular materials are processed to form diverse personalized bioinks. We report on the potential of the technology to engineer vascularized cardiac patches that fully match the anatomical structure, as well as the biochemical and cellular components of any individual. As we have previously shown, since the bioinks originated from the same patient, the engineered patches will not provoke an immune response after transplantation, eliminating the need for immunosuppression treatment. 9 Furthermore, a customized formulation was developed to support the free‐form printing of the personalized hydrogel, allowing to fabricate large, complex biological structures. Thus, cellularized hearts with a natural architecture were engineered, demonstrating the potential of the approach for organ replacement after failure, or for drug screening in an appropriate anatomical structure. Long‐term in vitro studies and in vivo implantation experiments in animal models should be conducted in order to adequately evaluate the fate and therapeutic value of the printed tissues. Although 3D printing is considered a promising approach for engineering whole organs, several challenges still remain. 18 These include efficient expansion of iPSCs to obtain the high cell number required for engineering a large, functioning organ. Additionally, new bioengineering approaches are needed to provide long‐term cultivation of the organs and efficient mass transfer, while supplying biochemical and physical cues for maturation. The printed blood vessel network demonstrated in this study is still limited. To address this challenge, strategies to image the entire blood vessels of the heart and to incorporate them in the blueprint of the organ are required. Finally, advanced technologies to precisely print these small‐diameter blood vessels within thick structures should be developed. 4 Experimental Section iPSCs Culture : iPSCs were generated from omental stromal cells and were a kind gift from Dr. Rivka Ofir, Ben Gurion University. The undifferentiated cells were cultivated on culture plates, pre coated with Matrigel (BD, Franklin Lakes, New Jersey), diluted to 250 µg mL −1 in Dulbecco's modified Eagle medium (DMEM)/F12 (Biological Industries, Beit HaEmek, Israel). Cells were maintained in NutriStem (Biological Industries) medium containing 1% penicillin/streptomycin (Biological Industries) and cultured under a humidified atmosphere at 37 °C with 5% CO 2. Medium was refreshed daily and cells were passaged weekly by treatment with 1 U mL −1 dispase (Stemcell Technologies, Vancouver, Canada) followed by mechanical trituration. Cardiomyocyte Differentiation from iPSCs : Cells were differentiated as previously described. 9, 31 Briefly, growth media (NutriStem) was refreshed daily until iPSCs reached 100% confluence. At this point (day 0) medium was changed to RPMI (Biological Industries) supplemented with 0. 5% glutamine (Biological Industries), B27 minus Insulin (X50, Invitrogen, Carlsbad, California) and 10 × 10 −6 m CHIR‐99021 (Tocris, Bristol, UK). Medium was refreshed every other day. At day 2, CHIR‐99021 was removed from media. At day 4, 5 × 10 −6 m IWP‐2 (Tocris) was added to media and was removed on day 6. At day 8, contracting implants were observed and medium was changed to medium supplemented with 0. 5% glutamine, B27 minus retinoic acid (×50, Invitrogen), and 1 × 10 −6 m retinoic acid (Sigma‐Aldrich). After day 10, medium was changed to M‐199 (Biological Industries), supplemented with 0. 1% penicillin/streptomycin, 5% fetal bovine serum (FBS, Biological Industries), 0. 6 × 10 −3 m CuSO4 · 5H2O, 0. 5 × 10 −3 m ZnSO4 · 7H2O, 1. 5 × 10 −6 m vitamin B12 (Sigma‐Aldrich), this media was refreshed every other day. Endothelial Cell Differentiation from iPSCs : Cells were differentiated as previously described with modifications. 9, 32 Briefly, After iPSCs reached ≈90% confluence (day 0), medium was changed to 50% (v/v) Neurobasal (Invitrogen) 50% (v/v) DMEM/F12 (Biological Industries), supplemented with 1% penicillin/streptomycin, 1% glutamine, B27 minus retinoic acid, N2 supplement (×100, Invitrogen), 1% nonessential amino acids (Invitrogen), 10 × 10 −6 m β‐mercaptoethanol (Gibco, Welltham, Massachusetts), 8 × 10 −6 m CHIR‐99021 (Tocris), and BMP4 20 ng mL −1 (R&D, Minneapolis, Minnesota). On day 3, medium was changed to EGM‐2 (Lonza, Basel, Switzerland), supplemented as according to the manufacturer instructions, and was refreshed every other day. Cell Dissociation from Matrigel‐Coated Plates : Cells grown on Matrigel‐coated plates were dissociated by enzyme digestion with collagenase type II (95 U mL −1, Worthington, Lakewood, New Jersey) and pancreatin (0. 6 mg mL −1, Sigma‐Aldrich) in DMEM (37 °C, 30 min), followed by TrypLE express (STEMCELL) treatment. Mathematical Modeling : Anonymous CT Images of a human heart have been contributed by the courtesy of Tel Aviv Sourasky Medical Center, Israel. The digital data file was then analyzed using RadiAnt DICOM viewer (Medixant). The left ventricle major blood vessels were segmented and measured. Based on these measurements, a 3D model of the cardiac patch was generated using COMSOL Multiphysics software. Oxygen concentration profile was calculated based on Fick's second law, Michaelis–Menten equations and the following data: Maximum cellular O 2 consumption rate of 5. 44 × 10 −2 nmol s −1 cm −3, Michaelis–Menten constant for oxygen consumption of 3. 79 nmol cm −3, diffusion coefficient (oxygen in hydrogel) of 1 × 10 −9 m 2 s −1 (Figure S1c, d). The model was then supplemented with blood vessels, ensuring that no region reach critical oxygen concentration (2. 64 × 10 −3 mol m −3 ). 28, 33 Fluorescent Endothelial and Fibroblast Cell Culture : Red fluorescent protein‐expressing human neonatal dermal fibroblast (RFP‐HNDF) cells (Angio‐Proteomie, Boston, Massachusetts) were grown in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, 1% glutamine, 1% nonessential amino acids, and 0. 2% β‐mercaptoethanol. Green/red fluorescent protein‐expressing primary human umbilical vein endothelial cells (GFP/RFP‐HUVECs, Angio‐Proteomie) were maintained in EGM‐2. Neonatal Cardiac Cell Isolation : Neonatal cardiac cells were isolated according to Tel Aviv University ethical use protocols from intact ventricles of 1‐ to 3‐day‐old neonatal Sprague‐Dawley rats as previously reported. 34 Cells were isolated using 6 cycles (37 °C, 30 min each) of enzyme digestion with collagenase type II (95 U mL −1 ) and pancreatin (0. 6 mg mL −1 ) in DMEM. After each round of digestion, cells were centrifuged (600 g, 5 min) and resuspended in M‐199 culture medium supplemented with 0. 6 × 10 −3 m CuSO4 · 5H2O, 0. 5 × 10 −3 m ZnSO4 · 7H2O, 1. 5 × 10 −3 m vitamin B12, 500 U mL −1 penicillin, and 100 mg mL −1 streptomycin, and 0. 5% FBS. To enrich the cardiomyocyte population, cells were suspended in culture medium with 5% FBS and were pre‐plated twice for 45 min. Cell number and viability were determined by a hemocytometer and trypan blue exclusion assay. Bioinks Preparation : Omenta were decellularized as previously described. 9 Briefly, human omenta (Helsinky #0237‐16‐ASF, Assaf Harofeh Medical Center, Israel; a consent was obtained from all subjects), or omenta from the remains of healthy pigs (Kibutz Lahav – designated for the food industry), were washed with phosphate buffered saline (PBS) (at least three human and ten pig omenta were used). Then, transferred to hypotonic buffer (10 × 10 −3 m Tris, 5 × 10 −3 m ethylenediaminete‐traacetic acid (EDTA), and 1 × 10 −6 m phenylmethanesulfonyl‐fluoride, pH 8. 0) for 1 h. Next, tissues were frozen and thawed three times in the hypotonic buffer. The tissues were washed gradually with 70% (v/v) ethanol and 100% ethanol for 30 min each. Lipids were extracted by three, 30 min washes of 100% acetone, followed by 24 h incubation in a 60/40 (v/v) hexane: acetone solution (solution was exchanged three times in 24 h). The defatted tissue was washed in 100% ethanol for 30 min and incubated over‐night (O. N. ) at 4 °C in 70% ethanol. Then, the tissue was washed four times with PBS (pH 7. 4) and incubated in 0. 25% Trypsin‐EDTA solution (Biological Industries) O. N. The tissue was washed thoroughly with PBS and incubated in 1. 5 m NaCl (solution was exchanged three times in 24 h), followed by washing in 50 × 10 −3 m Tris (pH 8. 0), 1% triton‐X100 (Sigma‐Aldrich) solution for 1 h. The decellularized tissue was washed in PBS followed by double distilled water and then frozen (–20 °C) and lyophilized. The dry, decellularized omentum was ground into powder (Wiley Mini‐Mill, Thomas Scientific, Swedesboro, NJ). The milled omentum was then enzymatically digested for 96 h at RT with stirring, in a 1 mg mL −1 solution of pepsin (Sigma‐Aldrich, 4000 U mg‐1) in 0. 1 M HCl. Subsequently, pH was adjusted to 7. 4 using 5 m NaOH and either DMEM/F12 × 10 or PBS ×10 (Biological industries). The final concentration of decellularized omentum in the titrated solution was 1% (w/v). For the personalized bioink preparation, omentum gel 1% (w/v) was homogenized at 15 000 rpm for 2 min (Silent Crusher‐M with 8F generator probe, Heidolph Brinkmann, Schwabach, Germany) and then weighted. Subsequently, while constantly stirred, the gel was allowed to reduce under a jet of sterile air until reached 1/3 of its initial weight. The concentrated gel (2. 5% w/v) was then centrifuged at 300 g for 2 min to remove air bubbles and stored at 4 °C until use. Dissociated iPSC derived CMs or neonatal rat cardiac cells were then dispersed in M‐199 medium and mixed with the omentum gel, reaching a final hydrogel concentration of 1% w/v with cells concentration of 1 × 10 8 mL −1. The cell‐laden ink was loaded into a syringe and kept at 4 °C. Sacrificial ink: Gelatin hydrogel was prepared by dissolving 15% w/v gelatin (from porcine skin, type A, Sigma‐Aldrich) in 40 °C warmed EBM‐2 (Lonza). The solution was then filtered by 0. 22 µm syringe filter and kept at 4 °C until further use. Cell‐laden gelatin ink was generated by dispersing ECs in warm EGM‐2 medium, mixed with prewarmed gelatin ink at a 1:2 v/v ratio, reaching a final concentration of 10% w/v gelatin and 1. 5 × 10 7 cells mL −1. HNDF cells were added to the bioink to a final concentration of 3 × 10 6 cells mL −1. The cell‐laden ink was then loaded into a syringe and allowed to cool to room temperature (22 °C). In the support bath method, in order to form the blood vessel perimeters, the personalized hydrogel bioink was used, encapsulating ECs at a concentration of 2 × 10 7 cell mL −1. Support Medium Preparation : For the generation of the printing support medium, an aqueous solution containing 0. 32% (w/v) sodium alginate (PROTANAL LF 200 FTS, a generous gift from FMC BioPolymer), 0. 25% (w/v) Xanthan gum (XANTURAL 180, kindly provided by CP Kelco), and 9. 56 × 10 −3 m calcium carbonate (as suspension, Sigma‐Aldrich) was prepared. While constantly stirred, the mixture was supplemented with freshly prepared, predissolved d ‐(+)‐gluconic acid δ‐lacton (Sigma‐Aldrich) to reach a final concentration of 19. 15 × 10 −3 m. This results in a slow decrease in the pH and solubilization of the calcium carbonate and liberation of the calcium ion that crosslinks the alginate. When the solution's viscosity is increased to a level that prevents precipitation of the calcium carbonate, the stirring was stopped and the mixture was incubated at RT for 24 h. Double distilled water at four times the volume of the resulted hydrogel were then added, followed by homogenization at 25 000 rpm (HOG‐020 homogenizer with GEN‐2000 generator probe, MRC ltd, Israel). The homogenate was centrifuged at 15 800 g for 20 min. The pellet was resuspended in DMEM/F12 (HAM) 1:1 culture media (Biological Industries) and centrifuged again, after which the supernatant was discarded. The pellet was then supplemented with 1% (w/v) xanthan gum in DMEM/F12 (HAM) 1:1 media (reaching a final concentration of 0. 1%) followed by vigorous vortexing to homogenize the mixture. Cardiac Patches Printing Process : Cardiac patches were printed using 3DDiscovery printer (regenHU, Villaz‐Saint‐Pierre, Switzerland). The bioinks were extruded through 30G needles onto glass slides. First, the CMs cell laden omentum gel was extruded in a crisscross geometry, creating the two lower layers of the patch. The third layer was composed of omentum gel, creating the supporting walls between which ECs laden gelatin ink was deposited to generate the vascular network. On top, two layers of crisscross CMs cell laden omentum gel were extruded, encapsulating the printed vessels. The printed patches were then incubated at 37 °C for 30 min to crosslink the omentum gel and to liquefy the gelatin, followed by submerging in EGM‐2 media for further culturing. Printing in a Support Bath : Support medium was transferred into a transparent, open sterile plastic box immediately prior to printing. The a‐cellularized or cellularized constructs were then printed (3D Discovery printer) by extrusion (through 30G needles) according to designs generated by BioCAD drawing software (regenHU) or according to data from STL files (sliced and processed by BioCAM software (regenHU)), which were downloaded from Thingiverse ( www. thingiverse. com ) (“Spheres in sphere” by Syzguru11 (modified), under the Creative Commons – Attribution license‐ CC BY 3. 0 – https://creativecommons. org/licenses/by/3. 0/ ; “Hand” by Teak (unmodified), under the Creative Commons – Attribution license – CC BY 3. 0 https://creativecommons. org/licenses/by/3. 0/ ; “Anatomical Human Heart” by 517860 (modified), under the Creative Commons – Attribution – Share Alike license – CC BY‐SA 3. 0 https://creativecommons. org/licenses/by‐sa/3. 0/ ). The cellularized constructs were printed using two omentum bioinks containing CMs and ECs. To improve visualization, constructs could be printed with bioinks supplemented with 1 µm blue or red polystyrene microparticles (Sigma‐Aldrich) or with lipid nanoparticles encapsulating cy3 or cy5 molecules, which were a kind gift from Prof. Dan Peer, Tel Aviv University. Upon completion of the printing process, the box was incubated at 37 °C for 45 min to crosslink the personalized hydrogel. Then, support medium was gradually aspirated and replaced with EGM‐2 medium containing alginate lyase 1 U mL −1 (Sigma‐Aldrich). The printed construct was then cultured O. N. , allowing final, complete degradation of the alginate particles. Finally, the medium was changed to fresh EGM‐2 medium for further culturing. Rheological Properties : Rheological measurements ( n = 3) were taken using Discovery HR‐3 hybrid Rheometer (TA Instruments, DE) with 8 mm diameter parallel plate geometry with a Peltier plate to maintain the sample temperature. The samples were loaded at a temperature of 4 °C, which was then raised to 37 °C to induce gelation; during which the oscillatory moduli of samples were monitored at a fixed frequency of 0. 8 rad s −1 and a strain of 1%. Compression tests on the printed or decellularized 35 hearts were performed with 20 mm diameter parallel plate geometry which compressed the samples at a rate of 5 µm s −1. Immunostaining, Confocal Imaging, and Calcium Imaging : Cells/tissues were fixed in 4% formaldehyde, permeabilized with 0. 05% (v/v) triton X‐100 and blocked with PBS, 1% bovine serum albumin, 10% FBS, and stained with primary antibodies followed by secondary antibodies (as indicated in the antibody list). Cells/tissues were imaged using an upright confocal microscope (Nikon ECLIPSE NI‐E) and inverted fluorescence microscope (Nikon ECLIPSE TI‐E). Images were processed and analyzed using NIS elements software (Nikon Instruments). Representative images from at least three different experiments were chosen. For calcium imaging, the cardiac patches were incubated with 10 × 10 −6 m fluo‐4 AM (Invitrogen) and 0. 1% Pluronic F‐127 (Sigma‐Aldrich) for 45 min at 37 °C. Cardiac patches were then washed in culture medium and imaged using an inverted fluorescence microscope. Videos were acquired with an ORCA‐Flash 4. 0 digital complementary metal‐oxide semiconductor camera (Hamamatsu Photonics) at 100 frames s −1. Antibody and Dyes List : Antibodies for NKX2‐5 (ab91196, 1:500), Troponin (ab47003, 1:100), CD31 (ab32457, 1:100), OCT4 (ab27985, 1:100), Ki67 (ab16667, 1:250), and Cytopainter deep red (ab138894) were acquired from Abcam (Cambridge, MA). Antibodies for actinin (A7811, 1:500) were acquired from Sigma‐Aldrich. Antibodies for Vimentin (1117481A, 1:100) were acquired from Invitrogen. Secondary antibodies: FITC‐conjugated goat anti‐mouse (ab6785, 1:800) and Alexa Flour 555‐conjugated goat anti‐mouse (ab150118, 1:500) have been acquired from Abcam. Alexa 647‐conjugated goat anti‐mouse (115‐605‐003, 1:500) and Alexa Fluor 488‐conjugated goat anti‐rabbit (111‐545‐144, 1:500) have been acquired from Jackson (Pennsylvania). For nuclei detection, the cells were incubated for 3 min with Hoechst 33258 (1:100) (Sigma‐Aldrich). Viability Assay : Cell viability was determined using a Live/Dead fluorescent staining with fluorescein diacetate (Sigma‐Aldrich, 7 µg mL −1 ) and propidium Iodide (Sigma‐Aldrich, 5 µg mL −1 ) for 10 min at 37 °C. The number of live and dead cells was determined by manual counting using NIS Elements software (Nikon) from at least three different microscopic field ( n ≥ 3 in each experiment), visualized by inverted fluorescence microscope. Scanning Electron Microscopy (SEM) : Human omentum hydrogel samples were fixed with 2. 5% glutaraldehyde (24 h at 4 °C), followed by graded incubation series in ethanol–water solutions (25–100% (v/v)). All samples ( n ≥ 3) were critical point dried, sputter‐coated with gold in a Polaron E 5100 coating apparatus (Quorum technologies, Lewis, UK) and observed under JSM‐840A SEM (JEOL, Tokyo, Japan). Statistical Analysis : Statistical analysis data are presented as means ± s. d. Differences between samples were assessed by student's t ‐test. p < 0. 05 was considered significant. ns denotes not significant. Analyses were performed using GraphPad prism version 6. 00 for windows (GraphPad Software). Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201900520
| 2,019
|
Advanced Science
|
Adaptable Microporous Hydrogels of Propagating NGF‐Gradient by Injectable Building Blocks for Accelerated Axonal Outgrowth
|
Abstract Injectable hydrogels in regeneration medicine can potentially mimic hierarchical natural living tissue and fill complexly shaped defects with minimally invasive implantation procedures. To achieve this goal, however, the versatile hydrogels that usually possess the nonporous structure and uncontrollable spatial agent release must overcome the difficulties in low cell‐penetrative rates of tissue regeneration. In this study, an adaptable microporous hydrogel (AMH) composed of microsized building blocks with opposite charges serves as an injectable matrix with interconnected pores and propagates gradient growth factor for spontaneous assembly into a complex shape in real time. By embedding gradient concentrations of growth factors into the building blocks, the propagated gradient of the nerve growth factor, integrated to the cell‐penetrative connected pores constructed by the building blocks in the nerve conduit, effectively promotes cell migration and induces dramatic bridging effects on peripheral nerve defects, achieving axon outgrowth of up to 4. 7 mm and twofold axon fiber intensity in 4 days in vivo. Such AMHs with intrinsic properties of tunable mechanical properties, gradient propagation of biocues and effective induction of cell migration are potentially able to overcome the limitations of hydrogel‐mediated tissue regeneration in general and can possibly be used in clinical applications.
|
1 Introduction Injectable materials with highly integrated functionalities have considerable benefits for regenerative medicine owing to their potential ability to mimic hierarchical natural living tissue and guide tissue regeneration with minimally invasive implantation procedures. 1, 2 Because of their tunable properties and ability to transport cargo, the materials can fill complex shaped defects, regulate cell behavior and guide cell growth, as well as the less invasive delivery procedure. 3 Hydrogel, as a scaffold, one of the most attractive injectable materials, displays amenable elasticity and enables facile diffusion of biomolecules or therapeutic agents for constructing a 3D extracellular matrix (ECM). 4, 5, 6 Some hydrogels have been approved for clinical uses in tissue engineering, such as in autologous fat, bovine collagen, and avian‐derived Hyaluronic acid. 7, 8 Others have also been applied in ex vivo 3D living tissues, and disease modeling. 9, 10, 11 Despite recent advances in tissue engineering, challenges of injectable hydrogel still exist. 12 First, the limitations of injectable nonporous hydrogel are weak cellular permeability and do not precisely match the rate of material degradation to tissue development for cellular infiltration, proliferation, and vascularization. 13 Degradation of materials by hydrolytic and enzymatic mechanisms can be manipulated through the molecule design and crosslinking density. However, an imprecise match of the material degradation rate to the rate of tissue development leads to insufficient scaffolding or the promotion of fibrosis. 14, 15, 16 Decreased mechanical durability of materials caused by a decoupling of cellular infiltration is also unavoidable. Second, propagating the gradient of growth factor levels and directing cells with a highly organized ensemble in vivo for injectable hydrogels are still difficult. In addition, regenerative deceleration can be caused by the inappropriate mimicking by the hydrogel of the elastic and mechanical properties of native tissue for orchestrating cells. 17, 18 Thus, there is a strong need to develop a functionally and gradient spatial structurally optimized injectable hydrogel scaffold with tunable mechanical properties, 19, 20 and efficient to promote the acceleration of tissue growth for applications in biological research. A simple solution to these tissue regeneration problems is to engineer formulations of hydrogels with a suitable porosity and interface to instruct tissue formation. 21, 22, 23, 24 Porous scaffolds that serve as tissue regeneration templates can guide new tissue before degradation, yet most of these scaffolds are noninjectable and exhibit poorly interlinked pores. Toward this end, injectable building blocks with connecting pores have been applied routinely. These building blocks possess specific properties such as injectability, self‐healing, interconnected pores, physical and chemical crosslinking that can be tailored by controlling their external molecular forces and physicochemical properties. 25, 26 For example, Diba et al. reported highly elastic and self‐healing composite colloidal gel assembled from pH‐controlled oppositely charged nanosized particles. 27 However, the pores of nanosized building blocks are too small to guide cell proliferation. 29, 30 By controlling the components of microgels, the resulting dynamic and adaptive gel network serves as a promising strategy for bone and organ regeneration. 31 For example, Segura and co‐workers reported that the assembled injectable microporous gel facilitates cell infiltration and accelerates wound repair processes at the brain‐damaged cavity. 1, 3 Relevant applications unique to microporous gels also include distinct physiological niches, such as cardiac, skin, and neural niches. The development of annealed gel creates an opportunity for tissue engineering. Even with the current breakthroughs in injectable gels, interconnected porous scaffolds have found limited applications in nerve regeneration, e. g. , peripheral nerve injury (PNI), 32, 33 due to the potential properties such as low gel formation rates, weak propagation gradient of growth factors, and uncontrollable inter/intramolecular modulus. For rapid gel formation, adaptable hydrogels provide adaptable linkages that can be immediately broken and reformed in a reversible manner without external triggers, facilitating filling and fixing in the nerve conduit during surgery. Furthermore, the key problem of peripheral nerve regeneration is accelerating Schwann cell (SC) migration across the injury site to achieve the effective formation of bands of Büngner in a short time and directing of motor and sensory peripheral axon outgrowth at the intrinsic rate. 34, 35 In this regard, Lee et al. designed a biocompatible axonal guidance device and the in vitro study showed that a propagating gradient of IGF‐1 directs axonal outgrowth of up to 5 mm to promote axonal growth, indicating the importance of gradients. 36 However, a complex PNI with a certain axonal loss and sizable gap defect exhibits a slowly regenerative rate of the defected nerve and delays functional recovery. 37 Therefore, engineering a material that can provide directional axonal growth while maintaining SCs may be essential for rebuilding the electrical and chemical signals between the distal and proximal stumps after PNI. In this study, we propose an adaptable microporous hydrogel (AMH) to accelerate and direct peripheral nerves based on a unique type of microsized building block that spontaneously forms interconnected pores, propagates the gradients of neuron growth factors, tailors the stiffness, and controls the pore sizes in nerve conduits. Through microfluidic fabrication, building blocks are constructed by a bottom‐up synthesis employing photocrosslinkable gelatin methacrylate (GelMA) and chitosan oligmer‐methacrylate (ChitoMA) as negatively and positively charged building blocks, respectively ( Figure 1 a). GelMA is a photo‐crosslinking hydrogel composed of modified collagen components. With the benefit of denatured collagen, GelMA retains natural cell‐binding motifs, such as cell adhesive peptide (arginyl‐glycyl‐aspartic acid, RGD) as well as matrix metalloprotease peptide (MMP) sequences that allowed cell controlled material degradation and subsequent resorption. 38 Furthermore, chitosan degradation products have been documented to facilitate peripheral nerve regeneration. 39 This AMH is reshapable and reassembles through shear‐thinning force and strong cohesive properties (Figure 1 b), facilitating the formation of a stable 3D porous scaffold. Such an interconnected injectable porous scaffold with suitable micropores for prompt cell migration as well as offer mechanical support and transports biomolecular cues to manage cell adhesion and growth (Figure 1 c, d). The adaptable microporous scaffold constructing cell‐penetrable pore sizes in real time, integrated with a propagating gradient of a NGF in a nerve tube (Figure 1 e), directs axon outgrowth of up to 4. 7 mm in 4 days in vivo and well aligned axons with functional recovery within 30 days postsurgery. Such synergistic effects of injectable AMHs of rapid bonding, precise pore control, and tunable molecular cue gradient formation effectively create a new horizon for applications in tissue regeneration. Figure 1 Microfluidic fabrication of adaptable microporous hydrogels (AMHs) for the creation of injectable scaffolds with 100% interconnected micropores. a) Schematic illustration of monodispersed gelatin methacrylamide (GelMA) building block formation using a microfluidic water‐in‐oil emulsion system. GelMA is crosslinked through a photo‐activated radical initiator and UV light. b) Micro‐network scaffold is formed by mixing oppositely charged building blocks via adaptable interaction. c) AMHs are injectable and moldable to form complex and macroscale shapes through physical crosslinking. AMH can be performed in the presence of live human adipose‐derived stem cells (hADSCs). d) Fluorescence image of a cell‐laden micronetwork scaffold. e) Schematic illustration of AMH with a propagating NGF‐gradient for directed and accelerated axonal outgrowth in vivo. 2 Results and Discussions 2. 1 Synthesis and Physicochemical Characterization of AMH The GelMA and ChitoMA building blocks were prepared by a robust microfluidic water‐in‐oil (w/o) emulsion approach to precisely segment droplets, which subsequently undergo photopolymerization as previously described (Figure S1, Supporting Information). 40 To obtain the photopolymerizable building blocks, the amine groups of gelatin were substituted with methacrylate anhydrate (MA) ( Figure 2 a). Both high and low degree of substitution (DS) of methacrylation of GelMA were fabricated and the DS of GelMA was determined by NMR spectroscopy (high and low DS‐GelMA are 84% and 51%, respectively, (Figure S2 in the Supporting Information)). The zeta potentials of GelMA and ChitoMA exhibited pH‐dependent behavior and possessed oppositely charged at pH 7, indicating the attractive electrostatic interactions in neutral condition (Figure 2 b). Due to the presence of amino groups in chitosan, it is a cationic polyelectrolyte (p K a ≈ 6. 5). After emulsifying and curing on a microfluidic chip, the building blocks fabricated by GelMA and ChitoMA with narrow size distribution can be observed in Figure 2 c. Once the building block was redispersed in PBS solution, the structures remained intact and could be stained by various fluorescence dyes with high photostability (Figure S3, Supporting Information). Figure 2 Synthesis and physicochemical characterization of AMHs. a) Chemical reactions for the methacrylation of gelatin via the amino groups of gelatin with methacrylic anhydride. b) The effect of the degree of substitution (DS) on the zeta potentials of GelMA and ChitoMA. c) The operational regime for microfluidic building block generation displayed an order of magnitude in size at various conditions (with CVs < 6%). d) Light microscopic images of building blocks. The scale bar is 200 µm. e) The size distribution of building blocks in paraffin oil and PBS solution. f) The elastic modulus of single building blocks at different DS and concentrations. g) The in vitro protein release from 10 wt% GelMA building blocks with collagenase. h) Degradation profile of 10 wt% GelMA building blocks incubated with collagenase ( n = 5, mean ± s. d. ). i) GelMA and ChitoMA building blocks in the Eppendorf tubes before mixing. The gelation of mixing two types of building blocks with equal volume. j) The average void volume fractions of 125 and 175 µm of building blocks (left). The AMH composed by equal sizes of building blocks was immersed to a fluorescence solution (50 kDa of fluorescent dextran) and the unoccupied volume was filled by the solution (right). 3D reconstructed images revealed the characteristics of pores in the AMH ( n = 6, mean ± s. d. , * p < 0. 05, t ‐test) The scale bar is 200 µm. k) The void pore sizes of 115, 156, and 210 µm of building blocks ( n = 6, mean ± s. d. , * p < 0. 05, t ‐test). The scale bar in fluorescence image is 200 µm. By simply controlling the microfluidic channel size and the ratio of the aqueous/oil rates, both negatively and positively charged building blocks with an average diameter ranged from 50 to 310 µm with low polydispersity can be formed. Under the same flow rate in an identical chip, the difference in size of ChitoMA and GelMA building block is caused by their different viscosity. Under the optical microscope, the building blocks with narrow size distribution can be also observed in Figure 2 d, and these microspheres can be preserved in deionic water for 3 months without any server changes. Figure 2 e shows the size changes of building blocks in water and oil, suggesting the slight swelling in water. The mechanical properties of building blocks can be controlled by adjusting DS and polymer concentration (Figure 2 f). The elastic modulus of 5 wt% GelMA with DS of 51% and 84% is 1. 2 and 3. 6 kPa, respectively. Once the concentration of GelMA was increased to 10 wt%, the modulus of gel can reach to 4. 5 and 12 kPa for DS of 51% and 84%. As the literature documented that the mechanical properties of subtract and microenvironment are the crucial issue to regulate the cell adhesion, proliferation, and differentiation. 13 Concerning repair of axons in the peripheral nerves, the modulus of the matrix (3. 80 kPa for SC proliferation) has to fit the SCs to obtain the prevalent in vivo bipolar morphology to connect proximal and distal of nerves. 33 In our study, the elastic modulus of 10 wt% of GelMA with 51% DS is 4. 5 ± 0. 3 kPa, matching the preferred stiffness of the SCs. To investigate the formation of NGF gradient in a conduit, various concentrations of NGF including 200, 150, and 100 ng mL −1 were embedded in GelMA building blocks, and then, three types of AMH containing the NGF‐loaded GelMA and empty ChitoMA building blocks at equal volume were placed in a tube one by one to form three sections. The release profiles of each section were monitored under a 0. 5 U mL −1 collagenase solution. Both the loading efficiency and the amount of NGF release from the AMH were determined using an NGF rat ELISA Kit (Arigo Biolaboratories Corp). Through in situ microfluidic fabrication, the loading efficiency of NGF in the AMH is ≈95%. The stability was attributed by complexation between NGF (pI of 9. 3) and GelMA building blocks (pI of ≈5) in a neutral solution. 41 Without adding the collagenase to NGF‐loaded AMH, the sustained release of NGF reaches about 20% within 1 week. While applying the collagenase, the obvious NGF release from AMH was observed, indicating that the degradation of AMH was the main factor to cause NGF release. Figure 2 g exhibits three similar release patterns of NGF from building blocks but demonstrated the significant differences in release amounts under collagenase solution. While loading higher concentration of NGF in AMH, the stronger signal of NGF could be monitored in each time point. For example, after 100 min, the cumulative release amount of 200 ng mL −1 NGF embedded in AMH was threefold greater than that in 100 ng mL −1 NGF embedded in AMH. The cumulative release was linear dependence of the square root of time, revealing that the degradation and concentration gradients possibly dominated the release pattern based on the Fick's second law (Figure S4a, Supporting Information). Furthermore, with the assistance of collagenase, the degradation is less than 5% for 24 h, and the linear degradation rate can be observed in the following 196 h. After 8 days, the percentage of degradation reached 80% (Figure 2 h). The surface erosion and degradation of AMH actuated by collagenase gradually induced the NGF release. 42 These results exhibited that the loading various concentrations of NGF in AMHs affected the release patterns. To investigate the gradient in a conduit in vivo, brain‐derived neurotrophic factor (BDNF) as a model protein was applied to form a BDNF‐G‐AMH@conduit for evaluating the formation of protein gradient in a conduit since BDNF could be stained by fluorescent receptor kit (ab229395, Abcam) (NGF does not have a corresponding fluorescent antibody to label it). In in vivo study, the BDNF‐G‐AMH@conduit was implanted to SD mice in sciatic nerve defects, and then, the conduit was harvested from mice at 1 day and 4 days postimplantation. Figure S4b, c (Supporting Information) displayed the CLSM images and fluorescence intensity of BDNF from the proximal to distal in a conduit, where BDNF represented as red. As the CLSM image shown, the fluorescence intensity of BDNF was weak at the proximal section of the conduit, and the obvious increased intensity of BDNF could be detected from the proximal to distal, indicating the BDNF gradient formation in a conduit at 1 day and 4 days postimplantation. PC12 cells, derived from a pheochromocytoma of the rat adrenal medulla, were incubated with NGF‐loaded building blocks to evaluate the activity of NGF. The neurite outgrowth of PC12 cells can be induced by NGF through activating the receptor tyrosine kinase, G protein‐coupled receptors and heterotrimeric G proteins. After 2 weeks of incubation, the neurite outgrowth of PC12 cells can be clearly observed on the building blocks, revealing the considerable activity of NGF in AMH (Figure S5, Supporting Information). In addition, the building blocks could also be degraded by human adipose stem cells (hADSCs) in vitro without collagenase after 26 days (Figure S6, Supporting Information). By directly mixing equal volumes of GelMA and ChitoMA building blocks with identical sizes, the rapid in situ gelation can be accomplished to form a solid hydrogel, which could be further reshaped by an appropriate shear‐thinning force (Figure 2 i). This phenomenon implicates that the adaptable interactions can be spontaneously rebuilt through physical crosslinking, known as adaptable hydrogel. To evaluate the void space of AMH, the AMH was immersed to a fluorescence solution (50 kDa of fluorescent dextran) in advance to fill the void volume, and the CLSM images of AMH and unoccupied volume were analyzed to estimate a void fraction distribution (Figure S7a, Supporting Information). As shown in Figure S7b, c (Supporting Information), the various z‐stacks of AMH were analyzed by Zeiss Zen software (blue edition) to determine the total void volume for each layer, and then the results were converted to an average void volume. The result showed an average void volume of approximate 50% for both 125 and 175 µm of building blocks since both building blocks are randomly packed (Figure 2 j, left). Furthermore, the interconnected pores with similar patterns can be observed by fluorescence images (Figure 2 j, right), suggesting the easy penetration of solution into gel. With an increase of sizes of building blocks from 115 to 210 µm, the void pore diameters could be enlarged from 48 to 80 µm (Figure 2 k). Due to the highly porous structure and adaptable gelation, the hydrogels were termed as adaptable microporous hydrogel (AHM), which offered a compact structure and the interconnected pore (Movie S1, Supporting Information). 2. 2 Rheological and Self‐Healing Properties of AMH To evaluate the rheological characterization of building blocks and AMH, the bulky storage moduli of building blocks and AMH were investigated through an amplitude sweep (0. 01–10% strain) within the linear range. Both storage modulus ( G ′) and loss modulus ( G ′′) were evaluated by an oscillatory time sweep test for 10 min at a stress of 1 Pa with a constant frequency of 1 Hz using a rheometer. In Figure 3 a, either positively or negatively charged building blocks displayed a low storage modulus ( G ′ = 20 Pa) since the interactions between building blocks are merely weak van der Waals force in deionic water and PBS solution. Upon mixing two types of building blocks, the storage modulus ( G ′) of AMH was improved to 100 to 3100 Pa, indicating that the strong interactions of the spontaneous adaptability. At various ratios of oppositely charged building blocks (P/N ratio), the wide range of moduli could be observed. For example, the G ′ of the AMH exhibited 17. 5‐ and 155‐fold higher values, i. e. , 350 and 3100 Pa, than the negatively charged hydrogels along at P/N = 1/5 and 1/1, respectively. The reason of lower G ′ of AMH in PBS solution than that in water is dominated by the ionic strength. In PBS solution, the large amounts of ions would weaken the electrostatic forces between two oppositely charged building blocks. However, the adaptable electrostatic forces are strong enough to maintain the compact AMH in ion‐rich environment. Figure 3 Rheological properties and self‐healing behaviors of AMHs. a) G ′ of AMH in different volume ratio as a function of the ionic strength. b) Schematic illustration of adaptable process electrophoretic adhesion. Cationic ChitoMA and anionic GelMA are represented in brown and gray, respectively. (i–iv) Images demonstrating self‐healing process and reshapable property of the AMH during electrophoresis. The scale bar is 10 mm. c) Multiple cycles strain oscillatory measurement for self‐healing behavior of AMH. d) (i) Injection of a AMH from a syringe through a 26 G needle. (ii, iii) Five segments of AMH loaded by different dyes were placed into a conduit through a syringe injection. (iv) Nonadaptable gel injected to a conduit. (v, vi) The CLSM images of anionic and cationic building blocks in AMH stained by red and green fluorescent dye. (The scale bar is 200 µm. ) AMH exhibited specific elastic viscosity via the introduction of reversible noncovalent interactions without any assistance from an additional external trigger. As illustrated in Figure 3 b, the electrostatic forces between positive and negative building blocks across the interface autonomously self‐healing cracks and rejoin the two halves of the gel. This phenomenon could be observed by directly placing two AMHs together, were stained by green and yellow dyes, respectively (Figure 3 bi–iv). The resulting gels exhibited the excellent adaptability, stability, and reshaping ability. This AMH demonstrated the tunable degrees of cohesiveness by controlling the adaptable electrostatic forces, which make it possible to be packed in the complicated voids. By modifying the P/N ratios, AHM could performed distinct rheological properties. For example, AMH possessed an excellent stability in water at P/N = 1/1 (Movie S2, Supporting Information). The elastic viscosity of AMH (at P/N = 1/1) was validated by the multicycle step strain oscillatory measurements at constant frequency of 1 Hz (Figure 3 c). At the first cycle, AMH exhibited an elastic and viscous modulus of 1500 and 42 Pa, respectively, indicating the elastic regime dominated over the viscous regime. At 1% strain of oscillatory shear, elastic modulus ( G ′) is stronger than viscous modulus ( G ″), and they performed constantly at the time duration, suggesting the AMH remained intact at low shear strain. With an rapid increase of shear strain to 1000%, the viscous modulus is much higher than elastic modulus, implying that the viscous regime dominated over elastic at this time period. On the reversal of shear strain, the elastic regime restored its predominant over the viscous domain. The result significantly confirmed the reformation ability of AMH, which also demonstrated the self‐healing nature of AMH. Furthermore, the repeated high strain up to 1000% also displayed no obviously loss in moduli upon the abrupt retraction of elastic domain. AMH regains the original G ′ and G ″ within few seconds after deformation because of the rapid adaptable interactions. The repeated oscillatory shear strain of AMH causes low change in moduli, indicating the excellent self‐healing characteristic with a repeatable feature. AMH were able to be filled in a syringe to pretest its ability for injection through a needle. Figure 3 di exhibits that the AMH could successfully be extruded from a syringe through a 26 G needle. At a closer observation, the pearl‐like gel was formed due to the strong interactions between oppositely charged building blocks. Furthermore, five segments of AMH loaded by different dyes were filled in a syringe, and subsequently, injected to a conduit (Figure 3 dii). Under an ultraviolet lamp, five parts of AMH with clear boundary was observed. With flipping the conduit, the AMH was still intact without pouring (Figure 3 diii). However, once the conduit was filled by the nonadaptable building blocks (GelMA building block only), the poor stability in a conduit would lead to building blocks flow‐out from the conduit owing to the low viscosity as well as the mixing of building blocks in a conduit during the surgery, which was difficult to construct the multiple sections in a conduit for the gradient filling (Figure 3 div). The confocal microscopy images of AMH (Figure 3 dv, vi) revealed its inter‐connecting structures, where the electrostatic forces between positive (green) and negative (red) building blocks across the interface autonomously self‐heal cracks and rejoin the two halves of the gel. The results demonstrated that AMHs have great potential as injectable tissue constructs. 2. 3 In Vitro 3D Cellular Network in AMH The ability of AMH to facilitate cell proliferation and network formation was evaluated by incubating three cell types, including SCs, hADSCs, and fibroblasts in the gels. The three cells readily adhered directly on the building block surface within 3 h and proliferated without additional steps for protein adhesion or attachment ( Figure 4 a), demonstrating the innate cytocompatibility of the AMHs. As the incubation time increased to 2 and 6 days, the SCs and fibroblasts on AMHs demonstrated continued proliferation, and the observed cells in the scaffold displayed a spread and network morphology. However, hADSCs exhibited slower proliferation with an incubation time of more than 6 days. However, cells proliferating in nonporous hydrogels with identical properties (10 wt% GelMA) showed no increase in cell spreading, even after 4 days of culture (right panel of Figure 4 a). With increasing the incubation time to 6 days, most of the AMH's surfaces could be covered by hADSCs, indicating the excellent affinity between AMH and cells (Figure 4 b). Furthermore, the surface charge effects of AMH on cell viability were also evaluated. Positive building blocks (P), negative building blocks (N), and AMH were incubated with three cell lines for 24 h (Figure 4 c). P/N ratio indicates the weight ratio of positively charged building blocks to negatively charged building blocks. The cell viability was more than 90% for the building blocks and AHM for three cell lines, suggesting their low toxicity to cells even through the differences of surface charges of AMH. Figure 4 Microporous scaffolds facilitate 3D cellular network formation and proliferation in vitro. a) Proliferation of hADSC, SC and fibroblast cells in AMHs and nonporous gels for various time. The scale bar is 100 µm. b) Fluorescence images demonstrating the formation of 3D cellular networks in AMH in vitro for 6 days. The scale bar is 100 µm. c) Cell viability of survival hADSC, SC, and fibroblast cells in AMHs for 24 h. N and P means the negatively charged and positively charged building blocks, respectively. P/N ratio indicates the weight ratio of positively charged building blocks to negatively charged building blocks. No significant difference for each group ( n = 5). d) hADSC, SC, and fibroblast cells proliferate within the microporous scaffold for 6 days. Statistical significance performed using one‐way ANOVA with Tukey's multiple comparison test ( n = 6, * p < 0. 05, ** p < 0. 01). The proliferation of SCs in AMH and a 2D cell culture dish was also monitored for 6 days. The proliferation rates of SCs in AMH and cell culture dish were almost the same at the beginning of 4 days (Figure S8a, Supporting Information). However, at sixth day, the numbers of SCs in AMH was ≈1. 6‐folds greater than that in culture dish. When compared the proliferation of SCs and fibroblast cells in AMH and nonporous hydrogel, the cell numbers for both cell lines in AMH were higher than that in nonporous hydrogels (Figure S8b, Supporting Information). The improvement of cell proliferation in AMH was potentially attributed by the suitable void volume of AMH for cell adhesion and nutrient transportation. Furthermore, the SC and fibroblast cell lines exhibited continued proliferation over 6 days, with a threefold increase at 2 days (Figure 4 d). Cell proliferation in AHM could potentially overcome the limitations associated with 2D cell culture and boost cells for several folds for tissue engineering and regenerative medicine. 2. 4 In Vivo Study of Regenerated Nerve To evaluate in vivo study of AMH in a conduit, the conduit with a thickness of 1 mm was prepared in advance by photocrosslinked GelMA through a molding and freeze‐dried process (Figure S10 in Supporting Information). Then, three types of AMHs containing 200, 150, and 100 ng mL −1 of NGF‐loaded GelMA and empty ChitoMA building blocks at equal volume were injected into the conduit to propagate the NGF gradient, hereafter referred to NGF‐G‐AMH@conduit, for the implantation to a peripheral nerve regeneration after 5 mm of sciatic nerve transection in SD rats ( Figure 5 a). Furthermore, other groups including conduit (nonfilled), NGF‐AMH@conduit (filled by homogeneous NGF‐distributed AMH), AMH@conduit, NGF‐gel@conduit (filled by homogeneous NGF‐distributed nonporous gel), NGF‐G‐gel@conduit (filled by nonporous GelMA gel with NGF gradient) and NGF‐G‐CL‐beads@conduit (filled by crosslinked GelMA building blocks with NGF gradient) were also implanted in vivo for a peripheral nerve regeneration, where the total amount of NGF in each group is identical for comparison (the concentration of NGF for homogeneous NGF was 150 ng mL −1 ). At 4 days postimplantation, the harvested conduit with regenerative nerve was fixed overnight in 4% paraformaldehyde in PBS. The conduit with regenerative nerve was sectioned in optimum cutting temperature compound, sliced, and stained for immunohistochemical analysis. Figure 5 a) In vivo study of immunohistochemistry images of the peripheral nerve regeneration in sciatic nerve defects harvested from (i) conduit, (ii) NGF‐G‐AMH@conduit, (iii) NGF‐AMH@conduit (homogeneous distribution of NGF), (iv) AMH@conduit, (v) NGF‐gel@conduit, (vi) NGF‐G‐gel@conduit (filled by nonporous GelMA gel with NGF gradient), (vii) NGF‐G‐CL‐beads@conduit (filled by crosslinked GelMA building blocks with NGF gradient), and (viii) NGF‐G‐porous‐gel@conduit (filled by porous GelMA gel formed by freeze‐drying). ß‐III tubulin represented as green for regenerated axons and AMH showed in red. The white dashed lines mark the leading edges of the axon outgrowth. (The scale bar represents 1 mm. ) b) The immunohistochemistry images of AMH and cells in the conduit for each group, where DAPI displayed as blue for nuclei. The scale bar represents 1 mm. c, d) The axonal outgrowth of each treatment after 4 days. Statistical significance performed using one‐way ANOVA with Tukey's multiple comparison test ( n = 5 for each treatment, mean ± s. d. , ** p < 0. 01). Figure 5 a, b exhibits the immunohistochemistry images of sciatic nerve defects in a conduit, where ß‐III tubulin represented as green for regenerated axons, DAPI displayed as blue for nuclei and AMH showed in red. Several conclusions can be drawn from these results. First, the length of regenerated nerve in the conduit alone was the shortest compared to other groups since there was no any support for the SC migration or biocues for guiding (Figure 5 ai). Second, ≈4. 7 mm of axons was induced by the guidance of NGF‐G‐AMH group within 4 days (Figure 5 aii). The positive result could be potentially conducted to the gradient NGF propagation and the interconnected pores of AMH, which could actuate the function of the bands of Büngner for SC migration and guide the direction of axons to shorten the repair time. To understand the effect on NGF gradient, the nongradient and gradient NGF‐loaded AMHs in the conduits were compared with similar total amounts of NGF in Figure 5 aiii. Without the NGF gradient, there was only 1. 78 mm of nerve growth in the conduit, which the nerve length is ≈38% compared to the NGF gradient group (i. e. , NGF‐G‐AMH@conduit), indicating the guidance effect of gradient NGF propagation. Furthermore, the AMH@conduit group did not obviously improve for nerve regeneration without the assistance of NGF since it is a key biocues for nerve regeneration (Figure 5 aiv). Furthermore, in the NGF‐gel@conduit (nonporous gel) group, the cells aggregated at the stump edge near the interface between the nerve defect and bulk gel (Figure 5 av), indicating the potential inhibition of the infiltration of cells in the nonporous hydrogel. In Figure 5 b, the distribution of cells also exhibited the related cell intrafiltration in the conduit for each group. The cells demonstrated the wide‐ranged distribution through whole conduit in the NGF‐G‐AMH@conduit group. However, in the NGF‐gel@conduit group, most cells were accumulated at the proximal section. The nerve regeneration by NGF‐G‐gel@conduit and NGF‐G‐CL‐beads@conduit was also evaluated the NGF‐gradient effects in different pore structures. In NGF‐G‐gel@conduit group, about 1. 95 mm of nerve growth was observed in the conduit, which the nerve length is about 41% compared to the NGF‐G‐AMH@conduit (Figure 5 avi). As the image shown, the cells aggregated at the stump edge of conduit. The results were caused by the low infiltration of cells in the nonporous hydrogel. As literature documented, a mismatch between the material degradation and the rate of tissue development might restrict the cell migration in the tissue regeneration. 43 In Figure 5 avii, the NGF‐G‐CL‐beads@conduit did not exhibit clear improvement in nerve regeneration and the nerve length was ≈1. 2 mm after 4 days of implantation. After the crosslinking process, the rate of degradation of GelMA was decreased due to the formation of covalent bonds on the gel surface, which could reduce the release of NGF from gel. To evaluate the crosslinking effects, the NGF‐G‐CL‐beads were incubated with PC12 for 2 weeks. After the incubation, only few PC12 cells were differentiated with neurite (Figure S11 in the Supporting Information). Furthermore, NGF‐G‐porous‐gel@conduit (filled by porous GelMA gel formed by freeze‐drying) was also implanted to SD mice in sciatic nerve defects for a peripheral nerve regeneration (Figure 5 aviii). The nerve length for NGF‐G‐porous‐gel@conduit was ≈2. 3 mm. The possible mechanism might be caused by the short interconnected pores in the structure. In Figure 5 b, the distribution of cells also exhibited the related cell intrafiltration in the conduit for each group. The cells demonstrated the wide‐ranged distribution through whole conduit in the NGF‐G‐AMH@conduit group. However, in the NGF‐gel@conduit group, most cells were accumulated at the proximal section. The average lengths of nerve regeneration were also provided. Compared to other groups, the NGF‐G‐AMH@conduit with the interconnected porous channels had the longer infiltrated cell distance and exhibited sevenfold longer than that in the conduit group (Figure 5 c). In Figure 5 d, for the average lengths of nerve regeneration for various NGF‐gradient porous gels in the conduits, NGF‐G‐AMH@conduit with shear thinning/adaptable properties and the interconnected porous channels also showed the longer infiltrated cell distance than that in the other groups. To understand the densities of axons in the conduit, the cross‐sections of regenerated nerves including the stump, proximal, middle and distal parts in the NGF‐G‐AMH@conduit were evaluated by CLSM at 7 days postimplantation. The axons were stained by ß‐III tubulin as green, and AMH was exhibited in red. Figure 6 a revealed that the infiltrated cells were appeared in the proximal, middle, and distal sections in NGF‐G‐AMH@conduit because it offered the interconnected porous channels. In NGF‐G‐gel@conduit, the amounts of cells could also be observed in the proximal and middle section of conduit, but the fluorescence intensity from cells was not as strong as that in NGF‐G‐AMH@conduit. However, without the assistance of AMH or gel, the signals of axons became weaker in middle and distal sections. This finding corresponded to the previous in vitro results in Figure 4 a, suggesting the cell mobility in the porous network. Figure 6 a) CLSM images of axons by staining ßIII‐tubulin in green in the indicated regions of regenerated nerves in NGF‐G‐AMH@conduit, NGF‐G‐gel@conduit, and conduit at 7 days postimplantation. (The scale bar represents 250 µm. ) b) Quantification of axonal growth in NGF‐G‐AMH@conduit, NGF‐G‐gel@conduit and conduit at representative segments in the proximal, middle, and distal sections of the nerve gap ( n = 5, mean ± s. d. , * p < 0. 05, ** p < 0. 01, t‐ test). c) Relative area of axons from proximal to distal ( n = 5, mean ± s. d. , * p < 0. 05, t‐ test). d) Schematic illustration of SCs in AMHs (left). S100 staining displayed the SCs in purple in the (i) NGF‐GAMH@conduit and (ii) conduit. The scale bar represents 250 µm. e) Quantification of the orientation angle of axons from the proximal to distal. Immunohistochemistry images of the NGF‐G‐AMH@conduit at 7 and 30 days postsurgery. f) The axons in green penetrated through the negative space of degradable AMH at day 7. After 30 days, AMH was completely degraded and the axons paralleled in the distal section of a conduit. The scale bar represents 200 µm. The distribution of axons in the NGF‐G‐AMH@conduit for whole regenerated nerve displayed the effects of AMH for nerve growth (Figure S12 in Supporting Information). The nerve filament intensity in NGF‐G‐AMH@conduit exhibited approximately 1. 7‐ and 1. 4‐ folds greater than that in conduit and NGF‐G‐gel@conduit, respectively, from proximal to distal section, suggesting a robust and steady rate of axon outgrow in NGF‐G‐AMH@conduit (Figure 6 b). Consistent to the densities of axons, the area of axons in NGF‐G‐AMH@conduit was also larger than that in conduit and NGF‐G‐gel@conduit (Figure 6 c). Furthermore, at 4 days postimplantation, both conduit and AHM@conduit were stained by CD68, a maker for macrophage, to track the immune response in the area of proximal stump and nerve gap. As shown in Figure S13 in Supporting Information, the lower fluorescence signal of CD68 was present in the AMH@conduit than that in the nerve gap (conduit), indicating that the positive charged building blocks would not cause serious inflammatory. The result is consistent with previous studies displaying the tolerable immune response of gelatin‐ and chitosan‐based materials in vivo. 1, 43, 44 SCs play an important role in creating a path as bands of Büngner for promoting axon penetration (Figure 6 d). Before the nerve regeneration, the SCs have to migrate into the subtract and form a fibrin‐like structures to guide the axons. Therefore, the proliferation of SCs in the NGF‐G‐AMH@conduit was also investigated. At 7 days postimplantation, the fibrin‐like morphology of SCs with high density was observed in NGF‐G‐AMH@conduit by confocal microscope in Figure 6 d, where the SCs were stained by S100 in purple, indicating that the quick infiltration of SCs into the gradient AMH. However, the fibrin structure of SCs in conduit alone was not clear. Through such effective guidance of SCs, the orientation of axons at the distal sections revealed that most of axon morphology followed the direction of conduits (i. e. , the orientation angle was 0°) in Figure 6 e. At seventh day, the degradation of AMH was occurred with the infiltration and regeneration of cells (Figure 6 f). After one month, the NGF‐G‐AMH significantly was degraded in vivo, and axons preformed into a regular arrangement and uniform size, illustrating that the AMH had the ability to construct the bridge structure necessary for the peripheral nerve generation (Figure 6 f). The degradation of AMH was dominated by collagenase preparations, mammalian MMP‐2, and MMP‐9 through the enzymatic degradability of gelatin preserved from the polymerization and modification with methacrylate pendant groups. 42 The finding to the rapid adaptation of interconnected porous and gradient scaffold allows SC migration to AMH without compromising the nonporous integrity of hydrogel and promotes axons outgrow directly. 2. 5 Functional Recovery of a Regenerated Sciatic Nerve In Vivo To evaluate in vivo‐promoted axonal regeneration of a transected sciatic nerve and functional recovery, the toe spreading of an injured hind paw and the static sciatic functional index (SSFI, a typical walking footprint analysis of toe spreading 45 ) were measured at 30 days postimplantation. The wider toe spreading of injured hind paw (yellow arrowhead in Figure 7 a) was observed in the NGF‐G‐AMH@conduit group than that in the conduit group, indicating the better nerve regeneration with the assistance of NGF‐G‐AMH. By analyzing the SSFI (0 and −100 for the healthy and transected sciatic nerves animals), the average value of the NGF‐G‐AMH@conduit was higher than that of the conduit group at 30 days postsurgery (Figure 7 b). The observations revealed that the function‐promoting effect of NGF‐G‐AMH@conduit was better than that in conduit in the nerve regeneration. Furthermore, the nerve fiber density in the NGF‐G‐AMH@conduit was also higher than that in the conduit and NGF‐G‐gel@conduit (Figure 7 c). The quantitative mean intensity of axons in the distal and middle parts were 3. 1‐ and 1. 7‐fold higher for the NGF‐G‐AMH@conduit groups than for the conduit group (Figure 7 d). Figure 7 Functional recovery of a regenerated sciatic nerve in vivo. a) Plantar views of the hind paws of animals receiving the implantation of the conduit and the NGF‐GAMH@conduit after 1 month. (The yellow arrowheads indicate the injured hind limbs. ) b) Static sciatic function index (SSFI) after implantation for 1 month and 2 months ( n = 5, mean ± s. d. , * p < 0. 05, ** p < 0. 01, t‐ test). c) ßIII‐tubulin (green) staining represents the regenerative axons. The segments in the middle and distal sections of the nerve gap were collected for analysis. d) The mean intensity of axonal growth in the distal section represents the numbers of axon fibers of the regenerative nerve. Statistical significance performed using one‐way ANOVA with Tukey's multiple comparison test ( n = 6, * p < 0. 05, ** p < 0. 01). e–g) Electrophysiological assessments of the regenerated nerves in different treatment groups after 2 months of implantation. Nerve conduction velocity (NCV) and CAP‐1st values were recorded at different intervals after surgery. The normal group refers to data collected from the normal limb ( n = 5, mean ± s. d. , * p < 0. 05, ** p < 0. 01, one‐way ANOVA with Tukey's multiple comparison test). The gastrocnemius muscle wet weight is another critical factor to evaluate the peripheral nerve regeneration after sciatic nerve transection. Once sciatic nerve injury occurred without treatment, the denervation would cause the atrophy of the gastrocnemius muscle. The relative gastrocnemius muscle wet weights of the injured limbs to the contralateral limbs after one month and two months of the implantation of autograft, NGF‐G‐AMH@conduit and conduit were investigated (Figure S14a, Supporting Information). The gastrocnemius muscles recovered in the NGF‐G‐AM@conduit group after 30 days were close to that in the autograft (gold standards in today's methods of nerve gap repair), indicating the positive effects of NGF gradient and AMH guidance. The fast nerve regeneration of autograft was directly actuated by the nerve bundle and supported tissue, but the limitation was still induced by the critical length and a donor nerve was also required. 46 The weights of relative gastrocnemius muscle after implanting NGF‐G‐AMH@conduit at various times also exhibited the gradual recovery of nerve and gastrocnemius muscle (Figure S14b, Supporting Information). To confirm the functional recovery of regenerated sciatic nerves, the nerve conduction velocity (NCV) and the compound action potentials (CAP 1st, an algebraic sum of all individual fiber action potentials of the nerve) was monitored in Figure 7 e. The CAP value is the algebraic sum of all individual fiber action potentials of the nerve and is highly influenced by the axon diameters and thickness of myelin around the nerve fibers. Furthermore, the value of NCV is based on the first peak of CAP and correlated to the mature of regenerative nerve fiber. When compared to normal group (uninjured control), the electronic signal recovery based on NCV was about 89% for the NGF‐G‐AMH@conduit and 67% for conduit alone, respectively (Figure 7 f). Similarly, a significantly higher CAP 1st signal in the NGF‐G‐AMH@conduit group was detected than that in conduit alone (Figure 7 h). These results demonstrated that the NGF‐G‐AMH@conduit not only enhanced the effective restoration of nerve conductivity but also offered the additional functional improvement through NGF‐loaded AHM. 2. 6 Regenerated Myelinated Nerve Fibers The myelination process of SCs plays an important role in axonal regeneration after PNI since the developing nervous system was surrounded by a myelin sheath. Distal SCs undergo atrophy owing lose axonal contact for a long duration leads to the inhibition of nerve regeneration. Typically, a higher number of distal axons and well myelination, the targeting tissue/organ would seem to perform the best functional recovery for the injured nerve. Transmission electron microscopy (TEM) of the cross section distal regenerative nerve segment at 60 days postimplantation revealed the well‐laminated myelin sheathes of regenerated nerve fibers in the normal and NGF‐G‐AMH@conduit group, whereas those in the conduit group had only small and thin myelin sheathes ( Figure 8 a). Furthermore, the myelin sheathes of regenerated nerve fibers in NGF‐G‐AMH@conduit was very close to that of the autologous nerve graft group (gold standard) after 2 months of repair. To compare the treatments, the layers of regenerated myelinated fibers were clearly calculated in the cross‐sections of the regenerated nerve fibers. Mean axon diameter and myelin layers were significantly greater in the NGF‐G‐AMH@conduit group (4. 76 ± 1. 11 µm and 39. 4 ± 1. 69 layers, respectively) than in the conduit group (2. 41 ± 0. 87 µm and 24. 0 ± 2. 42 layers, respectively) as shown in Figure 8 b, c. Both mean axon diameter and myelin layers in the NGF‐G‐AMH@conduit group were also close to the autologous nerve graft group (gold standard). The finding indicated that the NGF‐G‐AMH@conduit could enhance not only the diameter of regenerative axon but also the numbers of layers of myelin sheath through the gradient and interconnected pores, which amplified the nerve function recovery and the cell‐infiltrated ability in early nerve‐repairing stage. Figure 8 Transmission electron microscopy (TEM) analysis of regenerated distal nerves. a) TEM images of cross sections of regenerated distal nerves taken from types of nerve conduits implanted in rats after 60 days. b) Quantification of average axon diameter of regenerated myelinated nerve fibers. c) Average layers of regenerated myelinated sheath ( n = 5, mean ± s. d. , ** p < 0. 01, one‐way ANOVA with Tukey's multiple comparison test). 3 Conclusions In summary, a versatile adaptable hydrogel allowing incorporation of living cells was developed to serve as an injectable matrix with interconnected pores, tunable stiffness, and controllable pore sizes for spontaneous assembly into a complex shape in real time. These results were made possible by the use of microsized building blocks with opposite charges through microfluidic fabrication. Adaptable hydrogels with suitable micropores provide mechanical support for rapid cell migration and transport of bioresponsive cues to direct cell adhesion and growth. Furthermore, after loading gradient concentrations of growth factors into the building blocks, the propagated gradient of the nerve growth factor, combined with the cell‐penetrative connected pores constructed by the building blocks in the nerve conduit, effectively promotes SC migration and induces dramatic bridging effects on peripheral nerve defects, achieving axon outgrowth of up to 4. 7 mm and twice axon fiber intensity within 1 week in vivo. Such adaptable microporous hydrogels with intrinsic properties of tunable mechanical properties, a high drug payload, effective induction of cell migration and biocompatibility are potentially able to overcome the limitations of hydrogel‐mediated tissue regeneration in general and can possibly be used in clinical applications. 4 Experimental Section Microfluidic Chip Fabrication : The network of chips was designed by using geometric modeling software (AutoCAD, Autodesk Inc. , Sausalito, CA, USA). Designed patterns were generated on the surface of PMMA substrates (Kun Quan Engineering Plastics Co. Ltd. , Taiwan) by using a CO 2 laser micromachining system (LES‐10, Laser Life Co. Ltd. , Taiwan). Figure S1 (Supporting Information) shows the layout of the microchip (6. 5 cm ( L ) × 3. 5 cm ( W ) × 0. 4 cm ( H )), which consisted of two separated PMMA plates. The channel features were observed by using an optical microscope (FS‐880ZU, Ching Hsing Computer‐Tech Ltd. , Taiwan). Several access holes were drilled in the PMMA substrates. Each plate was immersed in D. I. water with ultrasonic agitation for 30 min and then dried with a stream of N 2(g). The cleaned plates were compressed together and bonded at 105 °C for 0. 5 h. Finally, poly(etheretherketone) (PEEK) tubes (Upchurch Scientific Inc. , Oak Harbor, WA, USA) were inserted into the holes and fixed by using an AB glue. 47 Fabrication of Building Blocks By a Microfluidic Chip : The aqueous phase (GelMA/ChitoMA solution with 0. 5 wt% Irgacure 2959 in D. I. water) and oil phase (5 wt% Span80 in paraffin oil) were injected, and the flow rates were well controlled by syringe pumps. The experimental setup was placed under an inverted optical microscope to monitor the process of droplet formation. The droplets, which flowed through the outlet and the connected glass tubing, were crosslinked to building blocks by UV light (365 nm, Series 1500, OmniCure). The size of the building blocks was determined using microscope images and Nikon software. These building blocks were washed with hexane three times and then centrifuged in PBS to remove the oil and surfactant. After removing the upper organic phase, the building blocks were washed with PBS three times and preserved at 4 °C until further use. Because of the gelling temperature of the prepolymer, these experiments should be performed at room temperature (not below 20 °C). For the fluorescence imaging, the building blocks were dyed red by Rhodamine B isothiocyanate (RITC). Briefly, after the building blocks were polymerized on a microfluidic water‐in‐oil (w/o) chip, they were collected and washed by the excess of deionic water for several time. Then, the building block was redispersed in PBS solution, 0. 01% RITC was added in the solution at 4 °C for 3 days. After 3 days, these building blocks were washed with PBS for three times, and then, centrifuged to remove the unreacted RITC and preserved at 4 °C until further use. Elastic Modulus Measurement : The cylindrical GelMA/ChitoMA hydrogels at different concentrations were prepared as previously described. 48, 49 Then, the hydrogels were compressed at a rate of 20% strain min −1 on an Instron 5542 mechanical tester. The elastic modulus was determined as the slope of the linear region corresponding to 0−5% strain of a stress−strain curve. Degradation of The GelMa Building Blocks : Lyophilized GelMA building blocks were weighted in 2. 0 mL Eppendorf and then swelling in 2. 0 mL of 0. 5 U collagenase type II solution at 37 °C. Collagenase solution was replaced 1. 0 mL every day to ensure constant enzymatic activity. At each time point, the solution was removed from Eppendorf and the building blocks were washed three times with PBS, lyophilized and weighed. The equation, D % = ( w 0 − w t )/ w 0 × 100%, was used to calculate percentage degradation ( D ), where w 0 was the initial lyophilized GelMA building blocks dry weight and w t was the dry weight after time t. In Vitro NGF Release Study : To prepare NGF‐loaded GelMA building blocks with 200, 150, and 100 ng mL −1 NGF, 2, 1. 5, and 1 µL of NGF solution (100 µg mL −1 ) was added to a premade solution containing 10. 0 wt% GelMA and 0. 5 wt% Irgacure 2959, respectively. After washing, the crosslinked building blocks were 175 µm in diameter in 0. 5 U collagenase solution. The loading efficiency and the amount of NGF released from the GelMA microspheres were determined using an NGF rat ELISA Kit (Arigo Biolaboratories Corp). The release profile of NGF from the GelMA building blocks in vitro was investigated for 8 days until the MPs were all degraded. GelMA building blocks (1. 0 mL) were pipetted into Eppendorf tubes containing 1. 5 mL of 0. 5 U collagenase solution at 37 °C. At each time point, 500 µL of storage solution was collected for measurement, and 500 µL of fresh 0. 5 U collagenase solution was added to the remaining solution. Rheology Technique for The Gel Measurement : All measurements were performed using a flat steel plate geometry (20 mm diameter) at 25 °C with a gap distance of 0. 5 mm using a rheometer (AR 2000ex, TA Instruments). To determine the bulk storage modulus of unmixed building blocks and mixed AMH scaffolds, an amplitude sweep was performed (0. 01−10% strain) to find the linear amplitude range for each. An amplitude within the linear range was chosen to run a frequency sweep (0. 5–5 Hz). For AMH scaffold measurements, first, each sample of oscillatory stress and strain sweeps was performed on a linear viscoelastic region. Subsequently, the storage modulus G ′ and loss modulus G″ were determined using an oscillatory time sweep test for 10 min at a constant stress of 1 Pa and constant frequency of 1 Hz. Cell Lines and Animals : Fibroblast cells (HIG‐82) were obtained from ATCC. Schwann cells (SCs) were isolated from SD rat (National Laboratory animal center, Taiwan). Human adipose stem cells (hADSCs) were kindly provided by Dr. Hui‐Yi Hsiao (Chang Gung Memorial Hospital, Center for Tissue Engineering, Taiwan). HIG‐82 cells were cultured in F‐12 medium containing 10% fetal bovine serum and 100 U mL −1 penicillin–streptomycin. The SCs were cultured in DMEM medium (Gibco) containing 0. 094 g L −1 d ‐valine (Sigma‐Aldrich), 10% (v/v) FBS, 1% (v/v) N 2 supplement (Gibco), 20 µg mL −1 (wt/v) bovine pituitary extract (Sigma‐Aldrich), 5 × 10 −6 m forskolin (Sigma‐Aldrich), and 1% (v/v) penicillin/streptomycin (Gibco). The hADSCs were cultured in DMEM/F‐12 medium (Gibco) containing 10% fetal bovine serum and 100 U mL −1 penicillin–streptomycin. All cells were maintained in a 37 °C incubator (Water‐Jacketed 3010 CO 2 ) under a humidified atmosphere with a 5% CO 2 supply. All experiments were performed in the logarithmic phase of cell growth. All animals (SD rats) were purchased from National Laboratory Animal Center, Taiwan. Cell Culture on Building Blocks : The building blocks were incubated in media with serum before being dried and pelleted by centrifugation and aspiration of supernatant. One hundred microliters of building block was added to a 100 µL stock solution of 1. 5 × 10 4 cells mL −1 at 37 °C for 90 min in situ for adherence rather than being transferred to the culture dish. Afterward, the cells were stained with DAPI for the nucleus and F‐actin for the cytoskeleton, washed twice with PBS after each staining, and imaged using a laser scanning confocal microscope (ZEISS LSM‐780). Cell Proliferation Analysis : Proliferation was assessed by counting the number of cell nuclei present in the AMH scaffold or nonporous gel constructs after 90 min and 2 and 6 days of culture in vitro. Nuclei were stained with a 2 µg mL −1 DAPI solution in PBS for 2 h, followed by visualization on a laser scanning confocal microscope (ZEISS LSM‐780). Specifically, each scaffold was imaged by taking 30 z slices over a total z height of 250 µm. At each time point, the total number of cells for 10 z ‐stack images was counted. These data were presented for all three cell lines. The cell viability was evaluated by 100 µL of cells containing 5000 cells per well coincubated with the building blocks on 96‐well plates for 24 h. Then, the cell viability was evaluated by MTS assay and determined by normalizing the results of treated to untreated with building blocks. Preparation of GelMA Nerve‐Guiding Conduit : Prepolymer solution was formed by dissolving 30% (w/v) GelMA with 0. 5 wt% Irgacure 2959 in D. I. water at 60 °C. The solution was injected into a conduit mold cavity in a 60 °C oven, which avoided gelation and bubble formation. After 5 min, the molds were removed and quickly cross‐linked by UV light. The resulting cross‐linked NGCs were hydrated by lyophilization and then stored at −20 °C until further use. Before surgery, the NGCs were immersed in dH 2 O for 1 h and autoclaved for sterilization. The swelling ratio, diameter, and SEM morphology are shown in Figure S10 in Supporting Information. Morphology of Regenerated Distal Nerves : The morphology at the surface of the GelMA NGCs was observed by using scanning electron microscopy (SEM, Hitachi S4800) with an accelerating voltage of 5 kV. Prior to observation, the samples were sputter‐coated with gold for 90 s before SEM imaging. Transmission Electron Microscopy of the Regenerated Nerves : The regenerative nerve was fixed at 4 °C with 3% glutaraldehyde, washed in 0. 1 m PBS, postfixed with 1% osmium tetroxide (Fisher Scientific, Pittsburgh, PA), dehydrated in graded ethanol solutions, and embedded in Araldite 502 (Polysciences Inc. ). Ultrathin sections (60 nm) were lifted onto formvar‐coated grids, poststained with lead citrate and uranyl acetate, and subsequently imaged using electron microscopy. Relative Gastrocnemius Muscle Weight (RGMW) : After performing the experiments, the animals were sacrificed and regenerating nerve was removed for further analysis, the gastrocnemius muscle of both hind limbs was excised and cleaned with PBS. Both right and left gastrocnemius muscle were weighted in order to determine relative gastrocnemius muscle weight. Tissue Section Immunofluorescence at 4 And 7 Days Postinjection : At 4 and 7 days postsurgery, the harvested conduit with regenerative nerve was fixed overnight in 4% paraformaldehyde in 0. 1 m phosphate buffered saline (PBS) at pH 7. 4. The conduit with regenerative nerve was sectioned and embedded in optimum cutting temperature compound (OCT; Surgipath FSC22, USA). Horizontal cryostat sections (10 µm in thickness) were sliced and then stained for immunohistochemical analysis. In brief, the frozen sections were incubated at room temperature to melt frozen section compound for 30 min. Then Immersed slide in the ethanol at 4 °C to remove frozen section compound for 10 min. Then, the sections were stained overnight at 4 °C with the primary antibody solutions. The obtained slides were stained as follows: 1) ß‐III tubulin (1:200, rabbit IgG1, Abcam) for regenerated axons. 2) S100 (1:200, rabbit polyclonal, Abcam) for Schwann cells. The secondary antibody was stained for 2 h by using Alexa 488 (goat anti‐rabbit IgG1) and then washed three times with PBS. Next, stained the section with DAPI for 10 min and then washed three times with PBS. The morphology of all the stained sections was observed using a laser scanning confocal microscope (ZEISS LSM‐780). The intensity and the alignment of regenerating axons at each segment (proximal, middle and distal) were analyzed by image J software. Statistical Analysis : Statistical analysis was performed using GraphPad Prism software (version 5. 0). Unpaired Student's t‐test (two‐tailed) was used to compare mean values of two groups. One‐way ANOVA followed by Tukey's post hoc analysis was used for the mean comparison of three groups and more. Values are expressed as means ± standard deviation. A value of p < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201900566
| 2,019
|
Advanced Science
|
In Situ Subcellular Detachment of Cells Using a Cell‐Friendly Photoresist and Spatially Modulated Light
|
Abstract Dynamic adhesion and detachment of subcellular regions occur during cell migration, thus a technique allowing precise control of subcellular detachment of cells will be useful for cell migration study. Previous methods for cell detachment were developed either for harvesting cells or cell sheets attached on surfaces with low resolution patterning capability, or for detaching subcellular regions located on predefined electrodes. In this paper, a method that allows in situ subcellular detachment of cells with ≈1. 5 µm critical feature size while observing cells under a fluorescence microscope is introduced using a cell‐friendly photoresist and spatially modulated light. Using this method, a single cell, regions in cell sheets, and a single focal adhesion complex within a cell are successfully detached. Furthermore, different subcellular regions of migrating cells are detached and changes in cell polarity and migration direction are quantitatively analyzed. This method will be useful for many applications in cell detachment, in particular when subcellular resolution is required.
|
1 Introduction Regulated adhesion and detachment of cells is important for various processes in life, including development, homeostasis, wound healing, and immune responses. 1 Engineered surfaces controlling cell adhesion have been developed for implantable devices, tissue engineering, and fundamental study of cell biology. 2 A number of cell adhesive (e. g. , RGD peptide) and cell repellent (e. g. , poly(ethylene glycol (PEG)) moieties were identified and used to promote or prevent cell adhesion on surfaces. [qv: 2b, 3] Furthermore, by controlling spatial distribution of cell adhesive and repellent moieties using microfabrication, morphologies, and spacing between neighboring cells were modulated. [qv: 2b, 3, 4] While these methods enabled precisely controlled cell–extracellular matrix (ECM)/cell–cell interactions to fine tune cell fates, differentiation, and activation, 5 methods based on cell immobilization have limitations in recapitulating dynamic nature of life involving active migration and detachment of cells, which occurs under various physiological/pathological circumstances. 6 To achieve dynamic modulation of cells on engineered surfaces, various stimuli‐responsive materials were used. Temperature‐responsive materials, including poly( N ‐isopropylacrylamide) (PNIPAAm), have been widely used to detach cells by switching temperatures to harvest intact cell sheets and aggregates for further applications. 7 However, spatially regulated cell detachment was challenging with temperature‐responsive materials. Electrical stimulation can be an alternative method for spatial modulation of cell adhesion and detachment, 8 but this method requires predefined regions patterned with conducting materials, thus may not be as flexible as light stimulation. Near infrared (NIR) absorbing materials were used to selectively detach cells near light‐illuminated regions by converting light into heat. 9 While NIR is a biocompatible light source with low toxicity, critical feature size/resolution achievable by this method is tens of micrometers, comparable to mammalian cell sizes, thus cannot be used for subcellular level control of cells. Ultraviolet (UV) light is attractive for high‐resolution light stimulation and direct photochemical conversion of chemical moieties, but can be toxic for cells. 10 To achieve high‐resolution control of cell dynamics with minimal cytotoxicity, our group developed a cell friendly photoresist poly(2, 2‐dimethoxy nitrobenzyl methacrylate‐r‐methyl methacrylate‐r‐poly(ethylene glycol) methacrylate) (PDMP). 11 PDMP undergoes photochemical reaction in response to 365 nm wavelength of light illumination ( Figure 1 A) and becomes soluble in buffer solution with pH ≈ 7. 4, including phosphate buffered saline (PBS) and standard cell culture media. Thin films of PDMP immersed in PBS or cell culture media are spontaneously dissolved by several seconds of brief light illumination through a 4′, 6‐diamidino‐2‐phenylindole (DAPI) filter of a standard fluorescence microscope, a widely used light source for live cell imaging for cell nucleus, with minimal cytotoxicity. 12 Original PDMP thin films are cell repellent due to the presence of PEG side chains, 11, 13 thus by selectively removing PDMP thin films in certain regions by light illumination through a photomask, we can precisely control various dynamic cellular processes such as cell adhesion, spreading, and migration, and perform quantitative analysis. 12, 13, 14 However, the previous methods primarily focused on triggering cell spreading and migration by removing PDMP thin films adjacent to cells to form new adhesion due to the cell repellent nature of PDMP. Methods that allow precise detachment of subcellular regions would be useful for the study of cell polarity and migration where subcellular adhesion plays an important role. 15 Figure 1 Characteristics of fibronectin‐coated PDMP. A) Photochemical reaction of PDMP. Blue circle: PEG side chain, red circle: organic soluble group (left side) to water soluble group (right side) conversion by the photochemical reaction. B) Effects of plasma treatment on the fibronectin adsorption on PDMP surfaces. The amount of fibronectin on the PDMP surfaces was measured by immunofluorescence microscopy. Fluorescence intensity is arbitrary unit (a. u. ). C–E) Effects of plasma treatment and fibronectin coating of PDMP surfaces on cell adhesion. Cells seeded on each type of the surface were incubated for 3 h, washed to remove unbound cells, and DIC images were acquired. Representative DIC images of HeLa cells on different types of surfaces are shown in C (Scale bar: 50 µm). Using DIC images, D) cell density and E) cell area of four different cells (HT1080, MDCK, HeLa, and NIH 3T3) were measured on various types of surfaces. Data are shown as mean ± s. e. m. [two‐sided Student's t ‐test] ns: not significant, * p < 0. 05, *** p < 0. 001. In this study, we developed a new method that allows in situ detachment of cells with subcellular resolution using the cell friendly photoresist PDMP. First, we converted cell‐repellent properties of PDMP thin film surfaces to cell‐adherent by plasma treatment. Then, spatially modulated light (SML) generated by a digital micromirror device (DMD), which allowed fabrication of ≈1. 5 µm features, was illuminated on cells adhering on surface‐modified PDMP to selectively detach single cells, multicellular clusters, and single focal adhesions. Using this new method, effects of different subcellular region detachment on cell polarity/migration were investigated. 2 Results and Discussion 2. 1 PDMP Surface Modification To create surfaces that initially promote cell adhesion, but can trigger partial detachment of cells by light illumination, the surfaces of PDMP, a cell friendly photoresist polymer previously developed for dynamic cell micropatterning (Figure 1 A), 11, 12, 13, 16 was modified. Original PDMP thin films have bioinert surfaces with minimal protein and cell attachment due to PEG side chains (a blue circle in Figure 1 A). 11 Therefore, additional surface treatments were required to coat adhesion molecules such as fibronectin on PDMP surfaces. To activate PDMP surfaces, they were treated with air plasma for 1 min prior to fibronectin coating. Physicochemical properties of untreated and plasma‐treated PDMP surfaces were extensively characterized by cross‐sectional scanning electron microscopy (SEM), water contact angle (WCA) measurement, atomic force microscopy (AFM), and X‐ray photoelectron spectroscopy (XPS), and shown in Figure S1 in the Supporting Information. Plasma treatment slightly etched PDMP thin films (cross‐sectional SEM) with almost no changes in surface roughness (AFM), and slightly increased hydrophilicity (WCA) with oxygen incorporation (XPS). Importantly, detailed analysis of C1s XPS peaks revealed substantial reduction of the C–O peak (≈286 eV) that are mostly generated from the (CH 2 CH 2 O) n occurred for plasma‐treated PDMP surfaces, indicating PEG side chains critical for protein resistance were damaged by plasma treatment. Fibronectin coating on the plasma‐treated PDMP surfaces was assessed by immunofluorescence microscopy (Figure 1 B). In the absence of plasma treatment, undetectable amounts of fibronectin binding occurred (plasma‐/fibronectin + sample in Figure 1 B), whereas at least tenfold increased fluorescence intensity was detected when the PDMP surfaces was treated with plasma prior to fibronectin coating (plasma + /fibronectin + sample in Figure 1 B). Next, we examined cell adhesion on the modified PDMP surfaces (Figure 1 C). Four different types of cells, including HT1080 (human fibrosarcoma cell), MDCK (Madin–Darby canine kidney epithelial cell), HeLa (human cervical cancer cell), and NIH 3T3 (murine fibroblast), were used. Cells in cell culture media supplemented with 10% fetal bovine serum (FBS) were seeded on various surfaces for 3 h and gently washed to remove nonadhering cells. Then, differential interference contrast (DIC) images were acquired in randomly selected positions (Figure 1 C), and average cell density was calculated (Figure 1 D). In the absence of plasma treatment, no cell adhesion was observed for all cell types. In contrast, plasma treatment was sufficient to induce substantial cell adhesion in terms of cell density, presumably by promoting adhesion molecule binding in FBS on the plasma‐treated PDMP surfaces. 17 However, cells on fibronectin‐coated PDMP surfaces exhibited more spread morphologies (Figure 1 C) with significantly larger areas (Figure 1 E) compared with cells on uncoated surfaces (or only plasma‐treated surfaces), meaning fibronectin coating on PDMP surfaces further enhanced cell adhesion. 2. 2 In Situ Detachment of Cells on Fibronectin‐Coated PDMP Surfaces Using Spatially Modulated Light In situ detachment of cells adhering on fibronectin‐modified PDMP thin films was achieved by following procedure schematically shown in Figure 2 A: 1) a digital image of cells was acquired (Figure 2 Ai), 2) a region for detachment was defined in the digital image (Figure 2 Aii), and 3) PDMP thin films underneath the cell in the predefined regions were dissolved by illuminating spatially modulated light (SML, Figure 2 Aiii). In order to implement this procedure, we integrated a DMD to a fluorescence microscope (Figure S2, Supporting Information). 18 Each micromirror in the DMD can be titled to two different angles, thus we can generate a beam with a desired shape by adjusting the tilting angle of each mirror. Using this instrumental setup, we next tested whether we can perform micrometer‐scale micropatterning on fibronectin‐coated PDMP thin films by illuminating SML with an array of circles with various diameters (1. 5–10 µm). Dissolution of PDMP thin films and generation of fibronectin micropatterns were confirmed by DIC and fibronectin immunofluorescence microscopy for all diameters of SML (Figure S3, Supporting Information). Therefore, we could generate micropatterns with critical feature size of 1. 5 µm, which corresponds to subcellular length scale, on fibronectin‐coated PDMP surfaces. Figure 2 In situ detachment of cells using spatially modulated light (SML). A) Schematic procedure for in situ cell detachment. B, C) Representative DIC images of HeLa cells on a fibronectin‐coated B) PDMP or C) PMMA surfaces before (left) and after (right) SML illumination. Scale bar: 20 µm. Time at SML illumination is set to “0. ” D, E) Representative time‐lapse DIC and IRM images of a D) HeLa and E) MDCK cell monolayers before and after SML illumination. SML illuminated regions were marked with yellow circles on DIC and IRM images acquired prior to SML illumination. Scale bar: 50 µm. Time at SML illumination is set to “0. ” F) Representative MDCK cell cluster detached by SML illumination. Scale bar: 50 µm. Time at SML illumination is set to “0. ” With this micropatterning capability, we next tested whether we can detach cells attached on fibronectin‐coated PDMP surfaces by dissolving the PDMP thin films underneath the cells by SML illumination. First, SML completely covering single HeLa cells attached on the fibronectin‐coated PDMP surfaces was illuminated for 3 s, and behaviors of the HeLa cells were observed by time‐lapse microscopy (Figure 2 B; Movie S1, Supporting Information). HeLa cells, initially spread on the surfaces with flat morphologies, rounded up with peripheral dark rings in DIC images, which occurs when cells were slightly out of focus, 19 meaning cells were detached from the surfaces by SML illumination. To rule out the possibility that cell detachment occurred by brief light illumination, identical experiments were performed using fibronectin‐coated poly(methyl methacrylate) (PMMA) surfaces, which has identical backbone structure to PDMP but lacks photoresponsive moiety. As expected, no detectable detachment was observed for cells on PMMA surfaces, even with 10 s of SML illumination (Figure 2 C; Movie S2, Supporting Information), suggesting cell detachment on PDMP surfaces was due to the dissolution of light illuminated‐PDMP thin films, not due to direct light‐mediated adverse effects on cells. Next, we observed detachment behaviors of cells forming confluent monolayers. To clearly observe dissolution of PDMP underneath cell monolayers, interference reflection microscopy (IRM) images, which generates dark spots for regions with cell–substrate contact <100 nm, 20 were acquired in conjunction with DIC images. Behaviors of detached cells varied for different types of cells presumably due to differences in cell–cell adhesion strengths. 14 When we detached cells in the middle of monolayers of HeLa cells, which tends to form weak cell–cell junctions, by illuminating circular SML with 50 µm diameter, the majority of junctions between detached cells were broken, and the detached cells were drawn toward undetached cells (Figure 2 D; Movie S3, Supporting Information). In sharp contrast, when we performed identical experiments with MDCK cells, which forms tight cell–cell junctions, junctions in the detached cells remained intact, and the detached cells remained on the SML‐illuminated region with substantial relocation (Figure 2 E; Movie S4, Supporting Information). IRM images confirmed that PDMP thin film on the SML‐illuminated region was dissolved, but MDCK cells on the PDMP‐dissolved region remain suspended on the surfaces as shown in DIC images, presumably by tight cell–cell junctions. Using this property of MDCK cells, we could harvest intact small‐sized multicellular clusters of MDCK cells by illuminating SML covering entire multicellular clusters (Figure 2 F; Movie S5, Supporting Information). The capability of detaching single cells and multicellular clusters can be useful for harvesting specific cells in cell mixtures if they adhere onto fibronectin‐coated PDMP surfaces. 2. 3 Detachment of a Focal Adhesion Complex in a Cell As the critical feature size for our patterning technique is close to the size of a focal adhesion complex (FAC), 21 we next attempted to detach a single FAC. HeLa cells transfected with paxillin–mCherry were used to visualize FACs, 22 and filamentous actin (F‐actin) was stained using an SiR actin probe. 23 FACs are dynamic supramolecular assemblies of macromolecules connecting integrins bound to extracellular substrates and F‐actin. Therefore, we thought by selectively dissolving regions underneath certain FACs, we could detach cell adhesions mediated by the specific FACs. We illuminated light on a circular spot with 3 µm diameter (white circles in Figure 3 A, B) that covered a FAC using SML to selectively dissolve fibronectin‐coated PDMP films underneath the FAC, and monitored behaviors of cells and FACs by time‐lapse microscopy. Interestingly, FACs detached by SML translocated inward to the cell body rather than disassembled for the majority of the cases (13 out of 15, Figure 3 A; Movie S6, Supporting Information). Force sensing and transmission occurs through FACs, 24 thus detachment of integrin associated with certain FACs could lead to the movement of FACs toward cell bodies to release tensions applied to the FACs via F‐actin. When FACs were connected with stress fibers, which generated strong F‐actin fluorescence signals due to the formation of bundles of F‐actin and myosin II, detachment of the FACs from the substrates leads to substantial shrinkage of the cells (Figure 3 B; Movie S7, Supporting Information) due to the release of traction forces mediated by myosin II in stress fibers. [qv: 24a, b, 25] Indeed, FACs in lamellipodia, which were not associated with stress fibers, could be detached with minimally influencing cell adhesion area (≈5%), whereas FACs located at cell sides slightly backward of lamellipodia, which were associated with stress fibers, were detached with substantial cell shrinkage (≈20%, Figure 3 C). Figure 3 Detachment of focal adhesion complex (FAC). A, B) Representative time‐lapse images of FAC‐detached cells. FAC located A) in lamellipodia and B) associated with stress fibers located in the cell side were detached. FAC was labeled with paxillin–mCherry and F‐actin was labeled with an SiR actin probe. White circles are SML illuminated regions. White arrows: detached FACs. Yellow arrows: initial position of the detached FAC. Blue arrows: undetached FACs. Scale bar: 20 µm. Time at SML illumination is set to “0. ” C) Effects of FAC detachment on cell area. FACs located in lamellipodia or associated with stress fiber were detached, and normalized cell areas, defined by cell area at time t divided by initial cell area, of cells after cell detachment were measured and plotted. Data are shown as mean ± s. e. m. of five cells in each condition. 2. 4 Influence of Subcellular Detachment on Cell Polarity and Migration In the previous section, FACs at the cell peripheries were visualized by paxillin–mCherry and selectively detached using SML illumination. Visualization of entire FACs in the cells was technically challenging with a wide‐field fluorescence microscope used in our experimental setting due to limited resolution and overwhelming fluorescence signals of paxillin–mCherry in cytoplasm that were not associated with FACs as shown in Figure 3 B. 26 Therefore, instead of detaching a specific FAC, we detached adhesions in a specific subcellular region in a polarized and migrating cell, and observed how cell polarity and migration were affected by the local detachment. Indeed, different types of FACs exist in a polarized cell depending on the subcellular regions: nascent FACs formed in the lamellipodia with leading edge protrusion gradually grow and mature as they move backward of the cells by cell migration. [qv: 1c, 27] Therefore, partial detachment of a certain subcellular region is likely to alter cell polarity and migration by influencing force balance within a cell. To identify and trace cell polarity, PH‐Akt‐YFP, which probe phosphatidylinositol (3, 4, 5)‐trisphosphates (PIP 3 ) distribution on the cell membrane, 28 was transfected into HeLa cells. Phosphoinositide 3‐kinases (PI3K) activity at the leading edge of migrating cells generates PIP 3 at the plasma membrane to modulate cytoskeleton organization for membrane protrusion, [qv: 1a, 29] thus PIP 3 is a good marker for cell front. PH‐Akt‐YFP‐transfected HeLa cells were further labeled with CellTrace Far Red (CTFR), which labels cell cytoplasm. Based on DIC and CTFR images, front/rear of the migrating cells were identified: typically, migrating HeLa cells on fibronectin‐coated PDMP surfaces exhibited half‐moon shape morphology with arc at the cell front by thin membrane lamellipodia formation and thick cytoplasm located at the cell center/rear ( Figure 4 A). Then, regions for front/side/rear/center detachment were determined as schematically shown in Figure 4 A. For center detachment, a circular region of diameter 14 µm was used, whereas for front/side/rear detachment, a circular region of diameter 20 µm that overlap with ≈50% of cells, thus detach effectively the same area as the center region, was used. Time‐lapse imaging of DIC/PH‐Akt‐YFP/CTFR was performed before and after partial cell detachment to monitor cell shape/polarity/migration. PIP 3 orientation and migration direction were determined using the PH‐Akt‐YFP and CTFR images by following the procedure described in Figures S4 and S5 in the Supporting Information, respectively. Front/side/rear detachment significantly reduced cell areas, whereas center detached cells exhibited comparable cell area to control (undetached) cells (Figure 4 B), indicating detachment of cell peripheries induce cell shrinkage regardless of regions due to release of adhesion‐mediated traction forces. Time‐lapse images of normalized PH‐Akt‐YFP images (obtained by the ratio of PH‐Akt‐YFP images and CTFR images) for a front‐detached cell along with PIP 3 orientation at different time points (red arrows in Figure 4 C; Movie S8, Supporting Information) clearly showed drastic changes in PIP 3 orientation and migration direction by partial cell detachment. PIP 3 orientation change at t min after partial cell detachment was measured by angle differences between initial PIP 3 orientation ( P 0 → ) and PIP 3 orientation at time t ( P t → ), defined as θ t in Figure 4 C, and the distribution of θ 30 and θ 60 for control (undetached) and front/rear/side/center detached cells were plotted in Figure S6 in the Supporting Information and Figure 4 D, respectively. In addition, migration direction change at time t, ϕ t, was measured by measuring angle between P 0 → and migration direction at t, and distributions of ϕ 30 and ϕ 60 for different cell detachment regions were plotted in Figure S7 in the Supporting Information and Figure 4 E, respectively. In the absence of partial detachment (control), the majority of cells maintained initial polarity and minimally changed migration direction over 1 h. Rear detachment minimally affected PIP 3 orientation and migration direction, whereas front/side/center detachment significantly altered PIP 3 orientation and migration direction. Changes in PIP 3 orientation and migration direction occurred gradually for the majority of cells, similar to the case of the front‐detached cells shown in Movie S8 in the Supporting Information, rather than abruptly. Migration direction change can be induced by changes in protrusion direction as well as shrinkage of cells mediated by detachment, whereas PIP 3 orientation change reflect altered polarity within cells due to partial detachment. Therefore, migration direction change and PIP 3 orientation change correlated, but did not completely agree with each other. Perturbation in force balances in lamellipodia/lamella regions located in front/side of migrating HeLa cells may be a major cause for PIP 3 orientation and migration direction changes for front/side detached cells. In addition, prevention of new adhesion formation on detached regions, as shown in IRM images in Figure 2 D, E, may bias new adhesion sites to alter cell polarity and migration, as center detachment caused substantial changes in PIP 3 orientation and migration direction without any significant cell shrinkage. While detailed mechanisms need to be further investigated, potentially by performing high‐resolution fluorescence microscopy of detached components and polarity‐regulating molecules in conjunction with subcellular detachment, our method enabled us to precisely detach subcellular regions and quantitatively analyze cell behaviors after detachment. Figure 4 Effects of subcellular region detachment on cell polarity and migration. A) Representative DIC/CellTrace Far Red (CTFR) overlay image marked with different regions for detachment. White line: cell boundary. F: front; S: side; R: rear; C: center. Scale bar: 20 µm. B) Normalized cell area of subcellular detached cells. Normalized cell area was measured by the ratio of cell area 10 min after detachment and cell area 2 min before detachment. Two‐sided Student's t ‐test was used. C) Normalized time‐lapse PH‐Akt‐YFP images with PIP 3 orientation (red arrows) for a front‐detached cell. Time at SML illumination is set to “0. ” Scale bar: 20 µm. D) Distribution of PIP 3 orientation change for control (undetached) and front/side/rear/center detached cells at 60 m after detachment. E) Distribution of migration direction change for control (undetached) and front/side/rear/center detached cells at 60 m after detachment. 18–23 cells in each case were analyzed, and Kolmogorov–Smirnov test D, E) was used for statistical analysis. * p < 0. 05, ** p < 0. 01, *** p < 0. 001. Previously, subcellular regions of migrating fibroblasts were detached by locating a pipet tip releasing a cell adhesion peptide GRGDTP either near front or rear of cells, and changes in cell area, traction force, and migration direction were monitored. [qv: 6b] Overall, the results of the previous study agreed well with ours: rear detachment caused substantial changes in cell area, and minimal changes in traction force and migration direction, whereas prolonged exposure of GRGDTP peptide near front caused changes in cell polarity and migration direction. However, precise control of cell detachment using a pipet tip is technically challenging, and spatial resolution of cell detachment is also limited. In contrast, our method allows us to detach precisely controlled subcellular regions, including regions within central areas of cells. Cell detachment techniques based on thermos‐sensitive polymer layers, 7 and NIR‐sensitive thin films 9 were developed, but their applications were mostly detachment of cell layers or small cell clusters due to limited resolution. Electrochemical release of adhesion molecules attached on electrodes can be used for high‐resolution subcellular region detachment, 8 but in this case, only regions predefined by electrodes can be detached. Compared with the previous methods for cell detachment, our method is superior in that we can achieve high‐resolution in situ subcellular region detachment by using SML and light‐responsive cell‐friendly photoresist PDMP. 3 Conclusion We developed a new cell detachment method that allowed in situ detachment subcellular regions of cells using a cell friendly photoresist PDMP and SML. To achieve this goal, we first modified PDMP surfaces by plasma treatment to allow cell adhesion on PDMP surfaces. Then, we integrated a DMD to a wide‐field fluorescence microscope to generate SML that allowed us to generate micropatterns with critical size ≈ 1. 5 µm, comparable to the size of a single FAC. Using this new method, we demonstrated detachment of a single cell, cells in monolayers, and a single FAC. In addition, we investigated how different subcellular region detachment influenced cell polarity and migration. Our method will be useful for wide applications for cell detachment, in particular for the cases when high‐resolution in situ partial cell detachment is required. 4 Experimental Section Fluorescence Microscopes : A modified Zeiss Axio Observer Z1 epifluorescence microscope with 40X (Plan‐Neofluar, NA = 1. 30) and 20X (Plan‐Neofluar, NA = 0. 5) objective lens and a Roper Scientific CoolSnap HQ charge‐coupled device (CCD) camera was used for fibronectin intensity measurement and cell adhesion assay. An XBO 75 W/2 xenon lamp (75 W, Osram) and GFP filter (EX BP 470/40, BS 495, EM BP 525/50) was used for fluorescence imaging for fibronectin intensity measurement. A modified Olympus IX 81 epifluorescence microscope with a 40X (UPlanFLN, NA = 1. 30) objective lens and a Roper Scientific Cascade camera was used for cell detachment and imaging experiments. A X‐Cite series 120 PC lamp (120 W, Excelitas) and DAPI filter (EX 365) were used for SLM. A LAMBDA LS xenon lamp (175W, Sutter instrument) and Texas Red (EX BP 559/34, BS 580, EM BP 630/69), Cy5 (EX BP 620/60, BS 660, EM BP 770/75) filter sets were used for fluorescence imaging after cell detachment. Both microscopes were automatically controlled by Metamorph, and stages were equipped with Chamlide TC incubator system (Live Cell Instrument, Korea) to maintain a cell culture condition (37 °C, CO 2 5%). Fibronectin‐Coated PDMP Thin Film Preparation : Random terpolymer PDMP was synthesized and characterized as described elsewhere. 11 Clean coverslips were coated with gelatin (Sigma) by incubating in a 0. 1% gelatin solution at room temperature for 30 min. Gelatin coated coverslips were spin coated with 3 wt% of PDMP in 1, 4‐dioxane (Sigma) at 2000 rpm for 2 min, and baked at 100 °C for 24 h. PDMP thin films were treated with air plasma using CUTE (Femto Science) for 1 min and coated with fibronectin by incubating in fibronectin solution (50 µg mL −1 in PBS) at 37 °C for 30 min. Fibronectin coating was validated by immunofluorescence microscopy using anti‐fibronectin rabbit antibody (EMD millipore, polyclonal) as a primary antibody and anti‐rabbit goat antibody tagged with Alexa fluor 488 (Abcam, polycolnal) as a secondary antibody. Cell Culture and Cell Adhesion Assay : HeLa, MDCK, HT1080, and NIH‐3T3 cells were cultured in Dulbecco's Modified Eagle's medium (DMEM) (Gibco) supplemented with 10% FBS (Gibco), 1% penicillin–streptomycin (Gibco). Cell suspension (1 mL, 5 × 10 4 cells mL −1 ) was applied on the PDMP surfaces and incubated for 2 h. Then, unattached cells were washed with PBS, and the surfaces were mounted on a microscope. DIC images of 10 randomly selected positions were acquired with a 20X objective lens, and number of cells in each position was manually counted and converted to cell density. DNA Preparation and Transfection : Paxillin–mCherry (addgene) and PH‐Akt‐YFP (gift from Prof. Sung Ho Ryu in POSTECH) plasmid DNA was prepared by maxi‐prep kit (Qiagen) and concentrated to 1 mg mL −1. Neon transfection system (Invitrogen) was used for DNA transfection. Briefly, DNA (8 µL) was added to HeLa cell suspension (10 6 cells) in 150 µL of R buffer (Invitrogen). Then, the cell and DNA mixture was loaded in 100 µL electroporation tip (Invitrogen) and 1 electric pulse was treated with 1400 V for 20 ms. Transfected cells were seeded in cell culture dishes filled with DMEM cell growth medium containing 10% FBS and 1% P/S, and cultured for 1 day in an incubator maintaining 37 °C, 5% CO 2. Instrumentation for Spatially Modulated Light Generation : Spatially modulated light (SML) was generated by the optical system schematically shown in Figure S2 in the Supporting Information. A DMD (DLP Discovery 4100 Development Kit, Texas instrument) was located at a diaphragm of a fluorescence microscope (Figure S2, Supporting Information). Light from a light source (Metal Halide, 120 W) was adjusted to reach the DMD after total reflection by the total internal reflection (TIR) prism. SML generated by the DMD passed through the TIR prism and was reflected by a dichroic mirror to an objective lens. MATLAB and ALP basic GUI (ViALUX GmbH) were used to control DMD to determine the shape of SML. Cell Detachment Experiments : Cells were seeded on fibronectin‐coated PDMP surfaces, incubated in a cell culture incubator (37 °C, 5% CO 2 ) for at least 3 h, and mounted on a microscope stage equipped with an incubator (37 °C, 5% CO 2 ). Digital images of the cells were acquired to select a region for detachment. The selected region was briefly illuminated with SML using a DMD, and time‐lapse imaging was initiated. F‐actin was labeled by adding 500 × 10 −9 m of an SiR‐actin (Cytoskeleton, Inc. ) in the media, and cytoplasm was labeled using CTFR (Invitrogen) by following the manufacturer's instruction. Image Analysis : Acquired images were analyzed using Image J (NIH). Cell area and center of mass of the cells were measured by manually drawing cell boundaries in DIC images. PIP 3 orientation and cell migration direction was measured using PH‐Akt‐YFP transfected cells labeled with CTFR as described in Figures S4 and S5 in the Supporting Information. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201900597
| 2,019
|
Advanced Science
|
Fine‐Tunable and Injectable 3D Hydrogel for On‐Demand Stem Cell Niche
|
Abstract Stem‐cell‐based tissue engineering requires increased stem cell retention, viability, and control of differentiation. The use of biocompatible scaffolds encapsulating stem cells typically addresses the first two problems. To achieve control of stem cell fate, fine‐tuned biocompatible scaffolds with bioactive molecules are necessary. However, given that the fine‐tuning of stem cell scaffolds is associated with UV irradiation and in situ scaffold gelation, this process is in conflict with injectability. Herein, a fine‐tunable and injectable 3D hydrogel system is developed with the use of thermosensitive poly(organophosphazene) bearing β‐cyclodextrin (β‐CD PPZ) and two types of adamantane‐peptides (Ad‐peptides) that are associated with mesenchymal stem cell (MSC) differentiation and that serve as stoichiometrically controlled pendants for fine‐tuning. Given that complexation of hosts and guests subject to strict stoichiometric control is achieved with simple mixing, these fabricated hydrogels exhibit well‐aligned, fine‐tuning responses, even in living animals. Injection of MSCs in fine‐tuned hydrogels also results in various chondrogenic differentiation levels at three weeks postinjection. This is attributed to the differential controls of Ad‐peptides, if MSC preconditioning is excluded. Eventually, the fine‐tunable and injectable 3D hydrogel could be applied as platform technology by simply switching the types of peptides bearing adamantane and their stoichiometry.
|
1 Introduction Stem cells constitute an emerging cell source for regenerative tissue engineering owing to their pluripotency, capacity for self‐renewal, and their potential to provide solutions for various human disorders, such as osteoarthritis and cardiac disease. 1 The potential of stem cell based tissue engineering could be enhanced with increased stem cell retention in designated lesions, enhancement of viability, and enforcement of desired control in their differentiation patterns. 2 Accordingly, these aforementioned issues are considered as challenges in establishing clinically useful stem cell derived tissue engineering technologies. Intravenous or intraperitoneal administration of stem cells without use of scaffolds could allow stem cells to be spread to the entire body within a short period. 3 Therefore, the use of biomaterials for tissue engineering is essential for increased stem cell retention, thereby avoiding repetitive stem cell administration. 4 To enhance the stem cell retention in a designated lesion site, the use of hydrogels is the best option for stem cell encapsulation owing to their capacity to absorb large amounts of water, their high biocompatibility, and their close resemblance to living tissue. 5 If injectable hydrogels (e. g. , with temperature‐ and pH‐dependent responsiveness) are employed for stem cell derived tissue engineering, considerable advantages will be evoked in terms of the minimal invasiveness and the convenience for recovering irregularly shaped lesions, such as uneven cartilage lesion site in osteoarthritis. 6 Increased retention of stem cells by using an injectable hydrogel also has resulted in successful tissue engineering outcomes, as reported previously by our group. 7 The survival rate of administrated stem cells in hydrogel is also a significant issue for superior tissue engineering outcomes. 8 Stem cell viability could be limited when these suffer immune rejection and unpredictable severe physiological condition even in the injectable hydrogel. Hence, there are several available ways to enhance the stem cell survival rate, such as the use of genetically engineered stem cells, 9 the employment of bioactive factors, 10 and the enhancement of biocompatibility of hydrogels. 11 However, genetic engineering of stem cells is associated with conflicts between the safety and the transduction efficiency when each viral or nonviral vector is used. 9 Therefore, using bioactive factors and biocompatible scaffolds is a fine choice for the enhancement of stem cell viability, and is free from safety risks. To obtain an ideal tissue engineering with hydrogels, the fine‐tuning of hydrogel with bioactive factors is required to achieve the desired differentiation patterns of stem cells. 12 Although there are various types of biocompatible and injectable stem cell scaffolds such as poly(lactic acid) derivate, 13 hyaluronic acid, 14 and supramolecular hydrogel, 15 the controlled differentiation of stem cells is of foremost importance to the attainment of the desired tissue construction. Depending on the use of specific differentiation inducible factors, including chemicals, proteins, and peptide, the fate of stem cells and ultimate phenotype may vary (e. g. , stem cells may differentiate into osteocytes, chondrocytes, myocytes, adipocytes, etc. ). 16 Therefore, fine‐tuning the scaffold is essential for the ultimate stem cell differentiation. 17 Although recently reported fine‐tuned hydrogels were prepared to regulate stem cell differentiation, it is difficult to use them in vivo system owing to the requirement for use of transdermal UV irradiation, which is carcinogenic. 18 Conversely, from the perspective of fine‐tuning hydrogels with bioactive molecules, the modification of hydrogel for the stem cell differentiation leads to unpredictable molecular substitutions 19 or physical mixing of bioactive factors. 20 In the construction of fine‐tunable and injectable hydrogels, the polymeric pendant modification with at least two types of bioactive molecules is also difficult without (meth)acrylate‐/UV‐mediated conjugation, which permits only in vivo transplantation. 21 This means that the preparation of a fine‐tuned and injectable hydrogel as a potential stem cell derived tissue engineering approach is associated with a conflict. Eventually, to fulfill the fine‐tuning of bioactive molecules and maintain the innate injectability of the hydrogel, host–guest interactions, noncovalent systems, could become wonderful alternatives compared to the use of a chemical conjugation system. To address these important issues, such as the retention of stem cells, stem cell viability, and the control of stem cell differentiation, we propose an integrated concept that accounts for the fine‐tuning and injectability issues of 3D stem cell scaffolds. To design a fine‐tunable and injectable 3D hydrogel to serve as stem cell scaffolds, we employed a hybrid system based on host–guest interactions and thermosensitive hydrogels. Given that/st interactions exhibit the strong self‐assembly characteristics between host and guest molecules, the fine‐tuning process is achieved by controlling the stoichiometric ratio of guest molecules. Among the various host molecules, such as cyclodextrins, cyclophanes, and curcubiturils, β‐cyclodextrin (β‐CD) is selected as the host owing to its high water solubility postmodification, low toxicity, and low immunogenicity. 22 Since a guest molecule of adamantane (Ad) exhibits a high binding affinity with β‐CD, it is also chosen as the guest molecule, 22 and is elongated with two types of stem cell differentiation induction peptides (Ad‐peptide). Despite the existence of many studies on the topic of host–guest interaction using β‐CD and Ad, their roles were limited in the gelation process based on host–guest interaction and the pendantly attached guest molecule carrier. 23 Beyond these issues, we approach the host–guest interaction as a means of fine‐tuning with maintaining injectability of hydrogel. Consequently, fine‐tuned hydrogels fabricated by the host–guest interaction based on simple mixing did not require further chemical synthesis processes. For several decades, various kinds of thermosensitive hydrogels (e. g. , poly( N ‐isopropylacrylamide), Pluronics, or Poloxamer) had been developed tremendously in the field of tissue engineering. However, the limitations of aforementioned hydrogels had also been reported such as nonbiodegradibility, toxicity, and lack of functional group. 24 Hence, in this study, the injectable hydrogel was originated from poly(organophosphazene) (PPZ), which benefits for biocompatibility, biodegradability, and thermosensitivity. Once we succeeded in synthesizing both the thermosensitive PPZ bearing β‐CD (β‐CD PPZ, host) and two types of mesenchymal stem cell (MSC) chondrogenesis inducing Ad‐peptides (guests, bioactive peptides derived from TGF‐β1 and N‐cadherin), the fine‐tuning process was conducted based on the stoichiometric control of Ad‐peptides in β‐CD PPZ ( Figure 1 ). Fine‐tuned hydrogels were prepared via simple mixing, and incorporated with various stoichiometric ratios of Ad‐peptides without any additional synthetic processes. Furthermore, various degrees of MSC differentiation were resulted from the control of the Ad‐peptide ratio. This system could be utilized as a platform technology by switching the types of guest molecules or their stoichiometry to achieve the fine‐tunable and injectable 3D construct. Figure 1 Schematic representation of the injection of mesenchymal stem cells (MSCs) encapsulated with 3D β‐cyclodextrin poly(organophosphazene) (β‐CD PPZ), adamantine‐TGF, and HAV (Ad‐TGF and Ad‐HAV). Aforementioned several molecules and cells were simply fabricated to form 3D thermosensitive hydrogels. After injecting them once, the solution containing the indicated mixture was transformed into a solidified gel. To control the fate of MSCs, Ad, guest molecule for β‐CD, which was elongated with TGF‐β1 peptides (Ad‐TGF) and N‐cadherin peptides (Ad‐HAV) was introduced, and it exhibited stoichiometric flexibility in this 3D hydrogel. 2 Results and Discussion 2. 1 Manufacturing β‐CD PPZ and Ad‐Peptides To synthesize β‐CD PPZ, the host molecule, PPZ that contained —COOH (acid PPZ) at the terminal end was prepared as described in our previous report. 25 Briefly, the thermally ring‐opened poly(dichlorophosphazene) was reacted with isoleucine ethyl ester (IleOEt), α‐amino‐ω‐methoxy‐poly(ethyleneglycol) (AMPEG), and 2‐aminoethanol in order under the completely moistureless condition. The intermediate PPZ product that contained the hydroxyl group (—OH) from 2‐aminoethanol was amphiphilic, owing to the retention of both hydrophilic AMPEG and hydrophobic IleOEt. This amphiphilic property of PPZ had aided the thermosensitive sol–gel transition behavior. 26 The PPZ exposing —OH groups were esterified to yield the —COOH groups, by allowing them to react with glutaric anhydride, and these —COOH groups were then conjugated with the primary amine (—NH 2 ) containing β‐CD, to allow the formation of β‐CD PPZ via 1‐ethyl‐3‐(3‐dimethylaminopropyl) carbodiimide (EDC) chemistry (Scheme S1, Supporting Information). The structure of β‐CD PPZ and its accurate conjugation, which have been confirmed with 1 H‐NMR and Fourier transform infrared spectroscopy (FT‐IR), were shown in Figure S1 in the Supporting Information. Eventually, each PPZ containing —COOH and β‐CD showed a thermosensitive sol–gel transition (Figure S2, Supporting Information). The diminution of T 0, the temperature that exhibited the initially increased viscosity, has been observed in the viscosity result of β‐CD PPZ, as compared to that of acid PPZ. This phenomenon had been caused by the decrease in the hydrophilic —COOH substitution to β‐CD. Furthermore, the chemical evidence for β‐CD conjugation was obtained by 1 H‐NMR measurement of the β‐CD conjugation specific anomeric proton generation; the FT‐IR peak change was observed because of C=O stretching in β‐CD PPZ, as compared to that in acid PPZ. Bioactive peptides derived from proteins could be a great alternative, after addressing the drawbacks resulting from the use of proteins, which exhibits a structural instability. We had chosen two different tracks for inducing MSC chondrogenesis, i. e. , stimulation of receptor mediated signaling (TGF‐β1) and cell‐to‐cell interactions (N‐cadherin). TGF‐β1 exists at the site of embryonic bone and cartilage development, and has a critical role in the intracellular signaling cascade that facilitates cartilage‐specific gene expression. 27 Out of all the peptide sequences in TGF‐β1, the binding site exhibiting the maximum reactivity to the TGF‐β1 receptor was selected for this study (CESPLKRQ). 28 The next protein source is N‐cadherin, which has a significant role in both cell‐to‐cell interaction and chondrogenesis. In a recent decade, the His‐Ala‐Val (HAV) motif derived from N‐cadherin has been mentioned in many studies, owing to its performance during chondrogenesis. 18, 29 Though the HAV sequence alone is enough to induce MSC chondrogenesis, a previous study revealed that an elongated HAV sequence such as CLRAHAVDIN was more effective than the short HAV motif. 30 These two kinds of selected proteins are commonly associated with both mitogen‐activated protein kinase (MAPK) signaling and the regulation of its subunits such as p38, extracellular signal‐regulated kinase‐1, and c‐Jun N‐terminal kinase, with regard to MSC chondrogenesis. 31 We prepared a couple of Ad‐peptides, by mimicking them from TGF‐β1 and N‐cadherin (Ad‐TGF and Ad‐HAV) (Scheme S2, Supporting Information). Adamantane acetic acids were elongated with 1. 0 kDa of diamine poly(ethylene glycol) (PEG), to produce monoamine adamantane PEG (Ad‐PEG‐NH 2 ). The byproducts of diadamantane PEG and PEG were eliminated using column chromatography. Such an Ad‐PEG‐NH 2 product was modified with methacrylate chloride, to yield methacrylate groups (Figure S3, Supporting Information). Ad‐TGF and Ad‐HAV were prepared through the Michel click chemistry between methacrylate containing adamantane and Cys (—SH) containing peptide sequences. These final products of Ad‐TGF and Ad‐HAV were characterized using 1 H‐NMR spectra. Their peptide specific guanidine and amide bond peaks of Ad‐TGF and Ad‐HAV were observed in 1 H‐NMR spectra (Figures S4 and S5, Supporting Information). Following the synthesis of host and guest molecules, the temperature‐dependent gelation profiles and the visualization of sol–gel transition with β‐CD PPZ molecules, which contained Ad‐TGF 100 and Ad‐HAV 100, were measured. The numbering of the following Ad‐peptides is associated with the Ad‐peptide contents percentage in relation to that of the entire molecules of β‐CD in β‐CD PPZ. For instance, Ad‐TGF 50 represents the presence of 50% saturated Ad‐TGF, as compared to that of fully saturated Ad‐TGF, in the inclusion complex with β‐CD. Furthermore, we had observed the viscosity properties of our fabricated hydrogels. Even if Ad‐TGF 100 and Ad‐HAV 100 were incorporated in β‐CD PPZ, both hydrogels had shown gelation properties and a sufficient level of viscosity, which impacted the body temperature ( Figure 2 a, b). T 0 values of β‐CD PPZ/Ad‐TGF 100 and β‐CD PPZ/Ad‐HAV 100 after fabrication, achieved via host–guest interactions, were slightly higher than viscosity of β‐CD PPZ only, owing to the possession of hydrophilic PEGs in Ad‐TGF and HAV. In particular, β‐CD PPZ, β‐CD PPZ/Ad‐TGF 100, and β‐CD PPZ/Ad‐HAV 100 hydrogels showed a certain viscosity at body temperatures of 268. 75, 387. 50, and 325. 00 pa s, respectively (Figure 2 a). Furthermore, the loss modulus ( G ″) was higher than the storage modulus ( G ′) at a cool temperature (4 °C) whereas G ′ value was greater than G ″ at a body temperature in the rheological results, irrespective of any hydrogels (Figure 2 b). It means that the sol state at cool temperature was changed to the gel state in all of the hydrogels by increasing a temperature up to 37 °C. Based on these proofs for the thermosensitive property from the viscosity and rheology results, all of the fabricated hydrogels were suitable for the administration into live animals. Figure 2 Fabrication of thermosensitive β‐CD PPZ/Ad‐TGF or HAV hydrogels, based on their assembly using host–guest interactions. β‐CD PPZ in all groups was dissolved with PBS buffer for the measurements of viscosity and rheology at the concentration of 10 wt%. a) Thermosensitive gelation details for β‐CD PPZ, β‐CD PPZ/Ad‐TGF 100, and β‐CD PPZ/Ad‐HAV 100. Viscosities at the body temperature (37 °C) were also elucidated for each group. b) Sol–gel transition visualization for β‐CD PPZ (top), β‐CD PPZ/Ad‐TGF 100 (middle), and β‐CD PPZ/Ad‐HAV 100 (bottom). c) The storage modulus ( G ′) and loss modulus ( G ″) of β‐CD PPZ, β‐CD PPZ/Ad‐TGF 100, and β‐CD PPZ/Ad‐HAV 100 at 4 and 37 °C for displaying sol–gel transition. d) 2D‐NOESY results that provide evidence of the occurrence of host–guest interactions between β‐CD PPZ and Ad‐TGF (left), and β‐CD PPZ and Ad‐HAV (right). 2. 2 Establishment of Fine‐Tuned and Thermosensitive Hydrogels Consisting of β‐CD PPZ and Ad‐Peptides To verify the host–guest interactions between β‐CD PPZ and Ad‐peptides, 2D nuclear Overhauser effect spectroscopy (NOESY) spectra were measured in the aqueous state with D 2 O, in the same conditions as those of fabricated hydrogels. 2D‐NMR is an effective method for observing the intermolecular interactions and (or) the conformation of the inclusion complex. 32 Cross peaks in 2D‐NOESY could be obtained from nuclei resonance connections that were spatially closer than those of the coupled bond. 33 Cross peaks in 2D‐NOESY spectra that showed the involvement of molecules in host–guest interactions; Ha, c, Hb (δ 0. 6–1. 3 ppm) of adamantane and H‐5′ (δ 3. 4–3. 5 ppm) of the β‐CD inner cavity, are fairly elucidated in Figure 2 c. Furthermore, to identify whether these guest molecules of Ad‐peptides had stoichiometrically and accurately been inserted into host molecules, dynamic light scattering (DLS) measurement were performed after setting the host:guest ratio = at 1:1, 1:1. 2, and 0:1. Above all, the increased hydrodynamic diameters for β‐CD PPZ complexed with Ad‐peptides (host:guest = 1:1) are demonstrated in Figure 3 b. The DLS results of using host:guest ratio of 1:1 and any peak involved in the measurement of the guest molecules (Ad‐TGF and Ad‐HAV) alone were not shown (Figure 3 c). This means that there were no breakaway guest molecules from β‐CD PPZ. Since the particle sizes of Ad‐TGF and Ad‐HAV alone (host:guest = 0:1) were 353. 63 ± 31. 71 and 460. 78 ± 49. 53 nm, respectively (Figure 3 d, e), which were much higher size than β‐CD PPZ alone, this phenomenon was resulted from the aggregation of hydrophobic adamantane molecules in the aqueous state. To sum up, all Ad‐peptides had been incorporated into the inclusion complex with β‐CD after complexing of β‐CD PPZ and Ad‐peptides (host:guest = 1:1) in an aqueous environment. On the other hand, when excessive amounts of Ad‐peptides were mixed with β‐CD PPZ (host:guest = 1:1. 2), the surplus Ad‐peptides formed other aggregates, whose occurrence was supported by the measurement of analogous peaks, while obtaining DLS results with Ad‐TGF and Ad‐HAV alone (Figure S6, Supporting Information). If there is no host–guest interaction between β‐CD PPZ and Ad‐peptides, other peaks for Ad‐peptide aggregates of DLS could be produced. Consequentially, stoichiometrically controllable Ad‐peptides based on host–guest interaction were proved using both 2D‐NOESY and DLS measurement. Figure 3 Host–guest interactions between β‐CD PPZ (host) and Ad‐peptides (guest). a) The illustration showing host–guest interactions between β‐CD PPZ and Ad‐peptides. b) Hydrodynamic diameter results for β‐CD PPZ, β‐CD PPZ/Ad‐TGF, and Ad‐HAV ( n = 3). c) Size distribution results and TEM images of β‐CD PPZ, β‐CD PPZ/Ad‐TGF, Ad‐HAV, and Ad‐peptides alone (Ad‐TGF and Ad‐HAV). Scale bar is 50 nm for β‐CD PPZ, β‐CD PPZ/Ad‐TGF, and Ad‐HAV, and 500 nm for Ad‐peptides alone. d) The illustration showing the aggregation of guest molecule alone in the aqueous state. e) Hydrodynamic diameter results of Ad‐peptides alone ( n = 3). We measured the long‐term maintenance of host–guest interaction with initially different amounts of incorporated Ad‐TGF and Ad‐HAV in β‐CD PPZ using an in vivo imaging system (IVIS) in living animals. In the molecular tails of Ad‐TGF and Ad‐HAV, rhodamine (Rho) and fluorescein isothiocyanate (FITC), respectively, were linked by using the conjugation process. These fluorescence expressing guest molecules were then injected after the fabrication of inclusion complexes with β‐CD PPZ. Since β‐CDs in PPZ were lost to the outside region of the 3D hydrogel, owing to PPZ capacities of biodegradability and dissolution of PPZ, Ad‐peptide signals containing fluorescence became lesser over time. Furthermore, the stoichiometric controlled patterns of Ad‐TGF and Ad‐HAV were observed in results during 21 days, even with the use of two different guest molecules ( Figure 4 a). For instance, the fluorescence intensities of Ad‐TGF were found to be in the increasing order for T100 H0, T75 H25, T50 H50, and T25 H75. This result was derived from stoichiometrically different guest molecules and was dependent on the host–guest interactions. Additionally, the region of interest (ROI) value for the expressed fluorescence was calculated as shown in Figure 4 b. As Ad‐peptides were mixed with acid PPZ (lack of β‐CD in PPZ), the rapid escape of Ad‐peptides from 3D hydrogels occurred within day 7. Although the ROI values decreased in processes over time in all the groups (T100 H0, T75 H25, T50 H50, T25 H75, and T0 H100), owing to the biodegradability and dissolution of β‐CD PPZ, the controlled level of Ad‐peptide fluorescence expression in β‐CD PPZ had been produced by initially adding different Ad‐peptides in all groups and at all time points, due to host–guest interactions. This result showed that host–guest interactions in our 3D β‐CD PPZ hydrogel were maintained for a considerably long period even in living animals. Figure 4 In vivo long‐term maintenance of host–guest interaction dependent on the gradient Ad‐peptides ratio was observed using IVIS. a) Images showing host–guest interacted Ad‐TGFs and Ad‐HAVs in mice; β‐CD PPZ and MSCs tagged with FITC and Rho, respectively. Within a period of 21 days, the gradient fluorescence levels of Ad‐TGFs and Ad‐HAVs were detected using IVIS. b) The average radiant efficiency values in mice Ad‐TGF (top) and Ad‐HAV (bottom) during 21 days ( n = 3). 2. 3 Fine‐Tuned and Injectable Hydrogels for In Vivo Ectopic Chondrogenesis To confirm the biocompatibility of the host molecule of β‐CD PPZ, the in vitro cytotoxicity tests were performed with the 2D plated cells and 3D hydrogel encapsulated cells. The viabilities of 2D plated MSCs treated with 1 wt% hydrogel and MSCs laden 10 wt% 3D hydrogel had been evaluated as 82. 3% and 91. 2%, respectively (Figure S7, Supporting Information). We assumed that the enhanced interaction between MSCs and biocompatible hydrogel had induced the superior MSC viability in the 3D hydrogels. Furthermore, the previous studies also had demonstrated that the viability of stem cells encapsulated in 3D hydrogel was greater than 2D plated stem cells. 34 It suggests that this polymer showed a promising level of biocompatibility for the next study, which would be conducted in animal test. Basically, the in vivo biocompatibility of postfabrication with β‐CD PPZ, and use of ratio‐controlled Ad‐peptides and MSCs was confirmed using hematoxylin and eosin (H&E) staining. According to the histological results obtained after H&E staining, any foreign body responses such as the generation of foreign‐body giant cells and toxicity were not observed in all groups of MSCs encapsulated with β‐CD PPZ/ratio‐controlled Ad‐peptides ( Figure 5 a). The neo‐chondrogenesis occurring in ectopically injected MSCs encapsulated with β‐CD PPZ/ratio‐controlled Ad‐peptide was then observed using safranin‐O staining. Ectopically generated cartilages and cytoplasm were stained in red and green color, by safranin‐O and fast green, respectively. As shown in Figure 5 a, the remarkable levels of cartilage stained colors as red were observed in Ad‐peptides contained groups. However, T0 H0 group had shown no staining of the cartilage, owing to the absence of peptide stimulation for chondrogenesis. Here, we showed that the qualitative analysis of chondrogenesis was performed using safranin‐O staining under conditions of stoichiometric control of Ad‐peptides. All Ad‐peptides incorporating groups had exhibited the MSC chondrogenic differentiation. Our results from safranin‐O staining, which showed occurrence of ectopically induced chondrogenesis, were similar to a previous study by another group and even their treatment of the chondrogenic induction media. 13 Figure 5 Biocompatibility and basic chondrogenesis capacity using the inclusion complex of β‐CD PPZ/various gradient Ad‐peptides/MSCs. a) The result of H&E staining (left) for biocompatibility verification and safranin‐O staining (right) for the demonstration of chondrogenic differentiation with MSCs (Scale bar = 100 µm). b) The maintenance of MSCs in thermosensitive β‐CD PPZ with various gradient Ad‐peptides measured using IVIS. c) The average radiant efficiency values of GFP tagged MSC in mice during 21 days ( n = 3). The * on the bar graph indicates the significance (*: p < 0. 05) of less fluorescence in T0 H0 compared with other groups. d) The basic illustration of the in vivo experimental schedule. Moreover, in vivo MSC maintenance was measured using IVIS for the reason that the local presence of MSCs over long‐term remained MSCs enhanced their therapeutic efficiency. 35 Hence, MSCs capable of expressing green fluorescent proteins (GFPs) were enveloped within the inclusion complexed β‐CD PPZ/Ad‐peptides, to measure the long‐term MSC maintenance in living animals. Locally injected MSCs were maintained at their injection sites for a period of 21 days. However, the significantly low maintenance of MSC fluorescence was monitored in T0 H0 group with an absence of both receptor‐mediated MSC attachment and factors promoting cell‐to‐cell interactions (Figure 5 b). Previous reports showed that cell‐to‐cell interactions and TGF‐β1 stimulation can enhance the survival rate of stem cells as well as chondrogenic differentiation. 36 The continuous decline of GFP tagged MSC in all groups resulted from biodegradability of β‐CD PPZ. Nevertheless, the considerably reduced level of MSC maintenance was shown in T0 H0 group, owing to an absence of both cell‐to‐cell interaction and TGF‐β1 stimulation. Evaluation of ROI value for the remained MSCs also substantiated the low level MSC maintenance in T0 H0, as compared to that in other Ad‐peptides containing groups (Figure 5 c). The presence of Ad‐peptides with MSCs had assisted local MSC maintenance at injected sites. In summary, these guest molecules, which were stoichiometrically incorporated in β‐CD PPZ hydrogels, were biocompatible at both in vitro and in vivo levels. Moreover, MSC chondrogenesis induction using β‐CD PPZ/Ad‐peptide complex was also observed in all groups except T0 H0 group. In the next section, we had thoroughly evaluated the chondrogenic differentiation levels in an environment with stoichiometrically controlled Ad‐peptides. Evaluation of exact levels of MSC chondrogenesis using the flexible, varied different guest molecule compositions had been performed with typically used chondrogenesis markers of aggrecan (Agg) and type‐II collagen (Col II). The consequences of chondrogenesis were evaluated by measuring the gene expression levels and by immunohistochemistry. Ectopically neo‐formed tissues resulted from the local and subcutaneous injection of MSCs encapsulated with β‐CD PPZ and Ad‐peptides were extracted in all experiment groups. Agg, the cartilage‐specific proteoglycan core protein, was fairly synthesized and expressed in all the Ad‐TGF and/or Ad‐HAV incorporating groups ( Figure 6 a). In particular, the highest Agg specific fluorescence and gene expression level were observed in T50 H50 compared to T100 H0, T75 H25, T25 H75, and T0 H100 (Figure 6 b, c). The level of gene and protein expression in T50 H50 is almost twice compared to other Ad‐peptide involving groups such as T100 H0, T75 H25, and T25 H75. In the absence of Ad‐peptides, there was no significant Agg expression during the measurement of both gene and protein fluorescence. Col II, the major protein in cartilage, was also selected as the specific marker for evaluating chondrogenesis. Col II was well developed under the incorporation of Ad‐peptides in β‐CD PPZ compared to T0 H0 group ( Figure 7 a). The expressed gene and fluorescence levels of Col II were also similar to the Agg expression pattern as shown in Figure 7 b, c. Notably, the highest expressions of Col II fluorescence and gene were shown in T50 H50 group. Because the contrary protein expression levels for Agg and Col II were observed in T0 H100 group, we assumed that TGF‐β1 alone could upregulate Col II expression. After all, the MSCs encapsulated with β‐CD PPZ and fine‐tuned Ad‐peptides had resulted in distinct levels of chondrogenesis. Among all of the groups, the highest levels of typical chondrogenesis markers were identified in T50 H50 group. Figure 6 MSC chondrogenic induction screening with the various combinations of Ad‐peptides with β‐CD/MSCs. Mice were sacrificed at day 21 postinjection. a) Immunohistochemistry analysis for Agg, the representative protein used for chondrogenesis detection (Scale bar = 20 µm). b) Analysis of fluorescence intensities of Agg ( n = 3). c) Agg gene expression levels in mice tissues ( n = 3). The * on the bar graph indicates the significance (*, **: p < 0. 05) of Agg expression in T50 H50 compared with other groups (*) and less protein expression compared with T100 H0, T75 H25, T50 H50, and T25 H75 (**). Figure 7 MSC chondrogenic induction screening with the various combinations of Ad‐peptides with β‐CD/MSCs. Mice were sacrificed at day 21 postinjection. a) Immunohistochemistry analysis for Col II, the representative protein used for chondrogenesis detection (Scale bar = 20 µm). b) Analysis of fluorescence intensities of Col II ( n = 3). c) Col II gene expression levels in mice tissues ( n = 3). The * on the bar graph indicates the significance (*: p < 0. 05) of Col II expression in T50 H50 compared with other groups. Fundamentally, TGF‐β1, which orchestrates the elaborate control of MAPK factors is significantly associated with the initiation of MSC chondrogenesis. 37 N‐cadherin had also notably induced the chondrogenesis of MSC via cell‐to‐cell interactions. [qv: 29a, 38] While different routes by using both Ad‐peptides were employed for inducing MSC chondrogenesis in this study, MAPK activation for the chondrogenic differentiation had been harmonized by both TGF‐β1 and N‐cadherin. 31 As aforementioned, MSC differentiations with TGF‐β1 and N‐cadherin are concentrated on the intracellular signaling or cell‐to‐cell interaction mediated chondrogenesis, respectively. Hence, the moderate chondrogenesis was resulted from the perfectly leant Ad‐peptide treatment groups such as T100 H0 and T0 H100 owing to the deficiency of opposite stimulation. Though T75 H25 and T25 H75 groups were treated with both kinds of Ad‐peptides, the modest levels of chondrogenesis had also been produced due to the one‐side strong Ad‐peptides MAPK stimulation. Outstanding levels of chondrogenesis were generated by fairly balanced Ad‐peptides group of T50 H50, which simultaneously stimulated MAPK factors. We assumed that well‐balanced MAPK stimulation in MSCs was stronger than the slanted Ad‐peptide stimulation for MSC chondrogenesis such as groups of T100 H0, T75 H25, T25 H75, and T0 H100. Previous studies had shown that the inductions of chondrogenesis were under the treatment of ≈10–20 000 times higher concentrations of TGF‐β1 or N‐cadherin than our T50 and (or) H50 Ad‐peptide employment. 31, 39 We assumed that stimulation with both Ad‐peptides at concentrations of 1. 23 × 10 −13 mole per cells (in T50 H50) had been sufficient for inducing MSC chondrogenesis even at the lower Ad‐peptide concentrations in our study. From these results, it can be observed that none of the factors was dominant; instead they were cooperative for the development in chondrogenesis. Eventually, these balanced two Ad‐peptides group such as T50 H50 had elicited an optimal level of chondrogenesis in this system. 2. 4 Fine‐Tuned and Injectable Hydrogels for Avoiding Osteogenesis Osteogenesis, also known as ossification, is a single continuous development process, for which cartilage formation acts as a precursor. 40 Thus, even if hypertrophic chondrocytes are differentiated into osteocytes in the presence of specific stimulators such as runt‐related transcription factor 2 (Runx2) and osterix, the use of Ad‐peptides should prevent the further process of chondrocyte to osteogenic termination. 41 It is important that the inhibition of osteogenesis occurs only because of the generation of incorrect terminal MSC differentiation products during the chondrogenesis process. Generally, Runx2 is considered as the typical key transcription factor for osteogenesis. Hence, Runx2 was a candidate gene assay marker for detecting the further progression of osteogenesis. Furthermore, we had tried to observe the calcium production, a significant marker for a final osteogenesis product, in the collected tissues via von Kossa staining. There was no calcium dot observed to be involved in any experimental group ( Figure 8 a). Furthermore, considerably excessive amounts of Runx2 genes were not detected in any groups, as compared to T0 H0 group (Figure 8 b). By all accounts, the final product of osteoblast development, such as calcium and Runx2, were clearly not observed. Eventually, the occurrence of osteogenesis was completely inhibited even using TGF‐β1 derived peptide in this study. Figure 8 Analysis of the evidence of nonosteogenesis of each Ad‐peptide combinations injection with β‐CD PPZ and MSCs after 21 days. a) Von Kossa staining with various Ad‐peptide compositions/β‐CD PPZ/MSCs (Scale bar = 100 µm). b) Runx2 gene expression levels, the typical gene marker for osteogenesis, was observed for the analysis of osteogenesis ( n = 3). 3 Conclusion We showed that injectable 3D hydrogel fine‐tuned with diverse peptides and its proportions had induced MSCs chondrogenesis. 3D MSC niche was constructed using one host molecule (β‐CD PPZ) and two kinds of guest molecules (Ad‐TGF and Ad‐HAV). Guest molecules in 3D hydrogel were under the stoichiometrically strict control based on host–guest interaction. Although these Ad‐peptides were formed in an inclusion complex with β‐CD PPZ, their gelation properties were still maintained at the body temperature. These fabricated hydrogels, which incorporated various ratios of Ad‐TGF/Ad‐HAV and MSCs, were subcutaneously injected into pockets of living animals to induce chondrogenesis. As two Ad‐peptides had been used to induce MSC chondrogenesis via different routes, a remarkable level of chondrogenesis was observed in T50 H50 group with stoichiometrically balanced Ad‐peptides, owing to balanced stimulation of MAPK. Furthermore, we showed that β‐CD PPZ/Ad‐peptide mediated chondrogenesis did not induce osteogenesis, as posthypertrophic stage of MSCs. We had performed the MSC differentiation process, independently of the method for MSC preconditioning, such as the employment of induction media. Nonetheless, one advance in this study is the facile fabrication of hydrogel with controllable guest molecule without further synthesis steps. Diverse peptides bearing adamantane, which are available for host–guest interactions, could be utilized in our system. Eventually, this technology could be used as a platform system by switching kinds of peptides bearing adamantane or their ratios to manufacture ideal 3D biomedical constructs. 4 Experimental Section Materials : Hexachlorocyclotriphosphazene (Aldrich) was purified by sublimation at 55 °C under vacuum (about 0. 1 mmHg). Poly(dichlorophosphazene) was prepared as described previously. 42 It was prepared from hexachlorocyclotriphosphazene using aluminum chloride (AlCl 3 ) as a catalyst at 250 °C during 5 hr. l ‐Isoleucine ethyl ester hydrochloride (IleOEt⋅HCl) was prepared from l ‐isoleucine (Aldrich) according to the literature. α‐Amino‐ω‐methoxy‐poly(ethylene glycol)s with molecular weights of 750 Da were prepared according to the literature. 43 Tetrahydrofuran (THF) and triethylamine (TEA) (Junsei Chemical Co. , Ltd. ) were purified under the dry nitrogen atmosphere by refluxing at the boiling point over sodium metal/benzophenone (Acros) and barium oxide (Acros). β‐Cyclodextrin purchased from Aldrich was used without further purification. Mono‐6‐OTs‐βCD and mono‐6‐diethylamino‐βCD (NH 2 ‐βCD) were synthesized according to the method reported in the literature. Acetonitirile (ACN), ethanol amine (AEtOH), 4‐(dimethylamino) pyridine (DMAP), isobutyl chloroformate (IBCF), and 1‐ethyl‐3‐(3‐dimethylaminopropyl) carbodiimide were obtained from Aldrich. Dichloromethane (DCM) was purchased from Daejung Chemical Company (Korea) with an extra pure quality and no further purification. N‐cadherin mimic peptide (CLRAHAVDIN) and TGF‐β1 mimic peptide (CESPLKRQ) were purchased from LifeTein, LLC (US). Synthesis of β‐Cyclodextrin Conjugated Poly(oranophosphazene) (β‐CD PPZ) : All reactions were processed under a dry nitrogen atmosphere using standard Schlenk‐line techniques. β‐CD PPZ was the same as that used in the previous study and the detailed synthesis protocol can be found in the paper. Briefly, IleOEt, AEtOH, and AMPEG750 in dried THF were added slowly to poly(dichlorophosphazene) in dried THF/TEA. After purification, glutaric anhydride and 4‐(dimethylamino) pyridine (DMAP) in dried THF were also added to the poly(dichlorophosphazene) solution to transform the hydroxyl group of AEtOH to carboxyl group. Finally, mono‐6‐diethylamino‐β‐CD was conjugated to the activated carboxyl group of PPZ using 1‐ethyl‐3‐(3‐dimethylaminopropyl) carbodiimide. Purification of synthesized β‐CD PPZ was performed using dialysis and lyophilization. Synthesis of Ad‐PEG‐CLRAHAVDIN (Ad‐HAV) and Ad‐PEG‐CESPLKRQ (Ad‐TGF) : Ad‐PEG‐MeAc ( M W : 1302. 6 Da, 200 mg, 0. 15 mmol) was dissolved in pH 10. 0 adjusted aqueous solutions. Tris (2‐carboxyethyl) phosphine) (TCEP, 44. 0 mg, 0. 15 mmol) and TGF‐β1 mimic peptide (CESPLKRQ, 960. 12 Da, 442. 2 mg, 0. 46 mmol) were added to Ad‐PEG‐MeAc solution at once. In case of Ad‐HAV, N‐cadherin mimic peptide (CLRA HAV DIN, 1111. 29 Da, 767. 8 mg) were added to Ad‐PEG‐MeAc solution. Reaction solution was purged by N 2 atmosphere for 10 min. And reaction was performed during 2 h at room temperature. After reaction, purification was carried out with dialysis and lyophilization. Characterization of β‐CD PPZ, Ad‐HAV, and Ad‐TGF : The structures of prepared β‐CD PPZ, Ad‐HAV, and Ad‐TGF were estimated by measuring 1 H NMR (Bruker avance III 400 MHz Fourier transform mode with DMSO‐d 6 and CDCl 3 ). The viscosity of the aqueous polymer solutions were assessed on a Brookfield RVDV‐III+ viscometer between 5 and 70 °C under a fixed shear rate of 0. 1. The measurements were carried out with a set spindle speed of 0. 2 rpm and with a heating rate of 0. 33 °C min −1. Moreover, the measurement of rheology was also performed with the representative samples of T0 H0, T100 H0, and T0 H100 at the concentration of 10 wt%. The rheometer (MSC 102, Anton Paar, DE) was equipped with a politer temperature‐controlled bottom plate and a 25. 0 mm parallel plate measuring system. All of the measurements were conducted with a gap length of 0. 3 mm at an oscillating frequency of 1 Hz, 0. 1% of the oscillating strain, and temperature in 4 and 37 °C. The storage modulus ( G ′) and loss modulus ( G ″) were calculated by the instrument's software. 2D‐NOESY Spectra : Spatial information was obtained from 2D‐NMR (NOESY) with a 1:1 molar mixture of β‐CD and adamantine containing peptide dissolved in D 2 O. A 2D‐NMR spectrum was recorded on a DD2 600 MHz FT NMR (Agilent Technologies). Dynamic Light Scattering : The sizes of β‐CD PPZ, β‐CD PPZ included with Ad‐peptides, and Ad‐peptides alone were measured by a Zetasizer Nano ZS (Malvern Instruments Ltd. , Malvern, UK) at room temperature. The final concentration of samples was 10 µg mL −1, and the samples were measured in triplicate. Transmission Electron Microscopy (TEM) : The shapes and sizes of β‐CD PPZ, β‐CD PPZ included with Ad‐peptides, and Ad‐peptides were observed using TEM (CM30 electron microscope, Philips, CA, USA). One drop of sample solution was placed on a copper grid, and the negative staining was performed with 2 wt% uranyl acetate. Evaluation of β‐CD PPZ Biocompatibility : Mesenchymal stem cells derived from Balb/c nude and mouse fibroblast cells with the consistent density (1 × 10 4 cells per well) were seeded in 96‐well tissue culture plate (SPL, Korea). Each cell line was incubated for 24 h with the well‐dissolved β‐CD PPZ (concentration: 0–20 000 µg mL −1 ). After incubation, the spent medium was discarded and cells were washed once with DMEM and fresh PBS. After adding fresh media (200 mL per well), the 3‐(4, 5‐dimethylthiazol‐2‐yl)‐2, 5 diphenyltetrazolium bromide (MTT) solution (100 mg per well) was added to the cells followed by incubation for 4 h at 37 °C in a humidified atmosphere of 5% CO 2. The formed formazan crystals were solubilized by incubating the cells with DMSO. The absorbance of the solution was measured at 570 nm, using a microplate reader (Bio‐Tek Instruments, USA). The cell viability (%) was calculated using the following formula: Cell viability (%) = [ab]test/[ab]control × 100%. Furthermore, 3D culture of MSCs was also performed to observe the viability of MSCs in the concentration of 10 wt% hydrogel with the 100 µL volume of each well. The groups of T0 H0, T100 H0, and T0 H100 were representatively selected for this test ( n = 3). The hydrogels encapsulating MSCs were incubated in 24‐well culture plates (SPL Life Science, KR) with cell insert (Corning, US). In the case of the CCK‐8 assay, all of the 3D hydrogels, including MSCs, were collapsed with media at 0, 7, and 14 days. Collapsed hydrogels with media were moved to 96‐well culture plates (SPL Life Sciences, KR), and 10 µL of CCK‐8 (Dojindo Molecular Technology, Inc. , JP) solution was added to each well. The CCK‐8 solution contained hydrogel and was placed in the cell incubator in a humidified atmosphere at 37 °C, 5% CO 2, for 2 h. After incubation, the absorbance was measured using a microplate reader (Bio‐Rad, Hercules, CA, US) at a wavelength of 450 nm. For live/dead assays, the culture media was removed and calcein AM/ethidium homodimer‐1 (live/dead assay kit, Thermo Fisher Scientific Inc. ) dissolved in DPBS solution was substituted at 0, 7, and 14 days. All images were obtained using a confocal microscope (Zeiss LSM 800, DE) in the 3D state. In Vivo Artificial Tissue Generation with β‐CD PPZ/MSC/Various Ratio Combinations of Ad‐peptides : All experiments with live mice were carried out in compliance with the relevant laws and institutional guidelines of the Institutional Animal Care and Use Committee (IACUC) in Korea Institute of Science and Technology (KIST), and IACUC approved the experiment (the approval number of 2017‐092). Balb/c nude mice (4 weeks old, 20–25 g, female) were purchased from Nara Bio Inc. (Gyeonggi‐do, Korea). Nude mice were anaesthetized with 3% isoflurane in the balanced oxygen and nitrogen. MSCs (passages below 8)/β‐CD PPZ (10 wt%)/pendently assembled 0–100% of Ad‐TGFs (T, adamantane‐PEG1000‐CESPLKRQ, 2248. 70 Da) and Ad‐HAVs (H, adamantane‐PEG1000‐CLRAHAVDIN, 2399. 87 Da) ( Table 1 ) were injected into subcutaneous pockets in mice on their right sides, lateral to the dorsal midline, using a syringe with a 31‐gauge needle. Each mouse received an 100 µL injection containing 2 × 10 6 cells mixed with 10 wt% β‐CD PPZ and a regulated Ad‐RGD amount. All tissues were collected 4 weeks postinjection, and were used for histological examinations and gene analysis. Table 1 Practically embraced guest molecules (Ad‐TGF and Ad‐HAV) to β‐CD PPZ hydrogel Abbreviation Exactly spent molecules [mg] a) Abbreviation Exactly spent molecules [mg] a) T b) H b) T H T100 H0 1. 10 0 T25 H75 0. 28 0. 90 T75 H25 0. 83 0. 30 T0 H100 0 1. 20 T50 H50 0. 55 0. 60 T0 H0 0 0 a) Exactly spent Ad‐peptides in 0. 01 g of β‐CD PPZ b) T and H denote the actual spent Ad‐TGF and Ad‐HAV, respectively. John Wiley & Sons, Ltd. Histological and Immunohistological Analysis : All collected tissues were embedded in paraffin, and sectioned with a microtome (thickness 6 µm). For histological evaluations, tissue sections were deparaffinized, rehydrated, and stained with H&E, safranin‐O, von Kossa, and immunohistochemistry. For immunohistochemistry, sectioned tissues were incubated overnight 4 °C with primary antibodies: antiaggrecan (1:500, Abcam, ab3778) and anticollagen II (1:500, Abcam, ab34712). After washing three times, the slides were incubated with appropriate secondary antibodies conjugated to fluorescent dyes such as goat antimouse IgG (TRICT, Abcam, ab6786) and goat antirabbit IgG (Alexa 488, Abcam, ab150077). Images were captured using a confocal laser‐scanning microscope (Zeiss) and a bright imaging microscope (Zeiss). Gene Assay from Harvested In Vivo Artificial Tissue : RNA extraction was prepared by using Trizol (Invitrogen, Carlsbad, CA). After tissues were treated with DNase (Invitrogen), 1 mg of RNA was used for cDNA synthesis (Superscript First‐strand synthesis system, GibcoBRL, Life Technologies). In brief, a reverse transcription reaction was carried out in a 20 mL mixture (1 RT buffer, 1. 25 × 10 −3 m MgCl 2, 5 × 10 −3 m DTT, 2. 5 g random hexamer, 0. 5 × 10 −3 m each of dATP, dCTP, dGTP, and dTTP, and 50 U of Superscript II enzyme) at 42 °C. After the reverse transcription reaction, RNA was degraded by 2U of Escherichia coli RNase H. PCR was performed in a 50 mL reaction buffer containing 2U of Takara Taq, 1 × PCR buffer, 0. 8 × 10 −3 m dNTP mixture, and specific primers at the concentration of 100 pmol. Standard PCR conditions were as follows: 3 min at 95 °C, followed by cycles of 5 s denaturation at 95 °C, 34 s annealing at 60 °C, and 1 min extension at 72 °C. Oligonucleotides used as primers are described in Table 2. The gene expression values were normalized against the housekeeping gene of β‐actin. Table 2 Primer used for real time RT‐PCR Name Forward Reverse Col II GCGGTGAGCCATGATCCGCC GCGACTTACGGGCATCCT Aggrecan GAAATGACAACCCCAAGCAC TCTCCGCTGATTTCAGTCCT Runx2 GCGTCAACACCATCATTCTG CAGACCAGCAGCACTCCATC β‐actin ACTCTTCCAGCCTTCCTTCC ACTCGTCATACTCCTGCTTGC John Wiley & Sons, Ltd. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201900849
| 2,019
|
Advanced Science
|
Targeting Antitumoral Proteins to Breast Cancer by Local Administration of Functional Inclusion Bodies
|
Abstract Two structurally and functionally unrelated proteins, namely Omomyc and p31, are engineered as CD44‐targeted inclusion bodies produced in recombinant bacteria. In this unusual particulate form, both types of protein materials selectively penetrate and kill CD44 + tumor cells in culture, and upon local administration, promote destruction of tumoral tissue in orthotropic mouse models of human breast cancer. These findings support the concept of bacterial inclusion bodies as versatile protein materials suitable for application in chronic diseases that, like cancer, can benefit from a local slow release of therapeutic proteins.
|
1 Introduction Bacterial inclusion bodies (IBs) are insoluble and discrete particles highly enriched by a single protein species, which deposits as interdigitated amyloidal fibers and nonamyloidal protein forms. 1 They organize into porous fibrilar networks 2 that confer mechanical stability, in which native or quasinative protein species are embedded. 3 These protein clusters, ranging between the nano‐ and micro‐scales, are built up in bacterial cells upon expression of a recombinant gene encoding the IB protein. Upon purification from bacteria, they behave as mechanically stable biomaterials, 4 easy to be handled in a diversity of platforms, presentations and interfaces. In this regard, IBs have been explored in the context of tissue engineering as soft topographies 4 since they are nontoxic to mammalian cells. 5 Being partially composed of functional polypeptides, 6 IB enzymes have found a role in industrial applications as self‐immobilized catalysts. 7 In addition, IBs tend to attach and penetrate mammalian cells without any deleterious effect, 8 via macropynocytosis. 9 Since part of the functional IB protein is released under physiological conditions (namely cell culture media, intracellular environment upon uptake, or in organic tissues), 9 IBs act as slow release protein platforms, potentially useful in therapeutic approaches (at both cell or organism levels) for protein replacement therapies or protein drug delivery. 8, 10, 11 If the IB protein is targeted to a biological marker, for instance, through the incorporation of a tumor‐homing peptide, IBs can remotely deliver IB proteins through the bloodstream to tumoral tissues upon subcutaneous administration. 12 The bulk IB material, however, is partially stable in the administration site and it remains detectable for weeks long. 13 Then, once implanted, IBs act as local protein depots, mimicking the natural amyloid repositories of human hormones in the endocrine system, 14 found them listed among examples of the expanding catalogues of nontoxic functional amyloids. 15 In this context, extracellular activities of IB proteins, either attached to IBs or upon release, are responsible for downregulation of cell surface receptor expression (by the display of a peptidic ligand of a cytokine receptor 13 ) or for enhanced cell growth (when formed by a growth hormone 8 ), through appropriate signaling. Whether and how IB proteins released in vivo can penetrate neighboring cells and interact with intracellular circuits for biological manipulation and therapeutic impact remains poorly explored despite its obvious pharmacological interest. 2 Results and Discussion To investigate this issue, we have here designed two modular proteins formed by two functional polypeptides (p31 and Omomyc, respectively) involved in cell cycle regulation, with therapeutic potential in antitumoral therapies as tumor‐targeting agents. The p31 protein consists of the C‐terminal fragment of p130cas that has shown the ability to promote apoptosis by disassembling focal adhesion complexes. 16, 17 The Omomyc protein is a Myc dominant negative 18 extensively validated transgenically and pharmacologically as an antitumoral agent. 19, 20, 21 Both proteins were tagged with the tumor‐homing peptide FNI/II/V (FN) which binds CD44, a glycoreceptor that promotes extracellular cell adhesion through hyaluronic acid interaction. CD44 is a well‐recognized tumoral marker, associated to tumor progression and metastasis, 22, 23 and that has been used to identify, along with low expression of CD24, the cancer stem cell population in breast cancer. 24 Therefore, the use of the tumor‐homing peptide FN seemed appropriate to test targeted nanodepots in a tumor type, the triple negative breast cancer, that currently lacks targeted therapies. 25 The modular proteins FN‐p31‐H6 and Omo‐FN‐H6 ( Figure 1 A) were produced in Escherichia coli as proteolytically stable full‐length polypeptides (Figure 1 B). FN‐GFP‐H6 26 was also produced as a fluorescent reporter, and this protein was found able to penetrate into 55. 6% of CD44 + MDA‐MB‐231 cells after 24 h of exposure, but not into CD44 − HepG2 cells (7. 4%, Figure 1 C). This data fully supported the CD44 targeting of the FN segment and the further exploration of FN‐empowered IBs as CD44‐targeted agents. Upon production, an important population of FN‐p31‐H6 and Omo‐FN‐H6 was found in the insoluble cell fraction, aggregated in the form of IBs (Figure 1 D). After purification from bacterial cell extracts, p31 and Omomyc IBs were clearly distinguishable in shape and morphology. While the first ones organized in a rod‐shape architecture (≈200 × 1000 nm) with a smooth surface, as previously described for some nonconventional IBs, 27 Omomyc IBs were smaller (≈200 × 500 nm), exhibiting a rough surface and an ellipsoid geometry that is much more common among bacterial IBs. 28 Figure 1 Proteins and protein materials. A) Schematic representation of the fusion proteins FN‐p31‐H6 and Omo‐FN‐H6, indicating the molecular mass of the products. Control proteins used in the study (FN‐GFP‐H6 and GFP‐H6, 26 ) are also included. Box sizes are only approximate. B) Coomassie blue staining of a SDS‐PAGE gel loaded with purified proteins. Numbers on the left indicate the molecular masses in kDa of the ladder marker. On the right, arrows indicate the position of the full‐length recombinant proteins. M indicates the molecular marker line, and p31 and O indicate FN‐p31‐H6 and Omo‐FN‐H6 lanes, respectively. C) Internalization of FN‐GFP‐H6 in cultured CD44 + MDA‐MB‐231 and CD44 − HepG2 cells, measured as the % of green fluorescent cells. Protein was added at 0. 5 × 10 −6 m and exposed to cells for 24 h. The percentage of CD44 + cells in each cell line is also indicated as a reference. D) Representative FESEM images of FN‐p31‐H6 and Omo‐FN‐H6 IBs at two different magnifications (zoom in the insets). Magnifications are equivalent in each micrograph pair to allow comparative visualization. Magnification bars represent 500 nm. To first explore the potential of these IBs to impact tumor cell biology, cultured breast cancer human MDA‐MB‐231 cells overexpressing CD44 24, 29 were exposed to either p31 or Omomyc IBs and appropriate GFP controls. Both p31 and Omomyc IBs showed a dose‐dependent cytotoxic effect on cell viability, not observed when exposing cells to GFP‐based IBs ( Figure 2 A). The lack of toxicity associated to exposure to GFP IBs fully confirmed the protein‐associated toxicity of IBs, as the IB material, per se, is not intrinsically cytotoxic. 1 Moreover, while p31 IBs showed an increasing cell killing potential with longer exposure time, Omomyc IBs showed the highest efficacy already at 48 h post exposure, indicating a faster mechanism of action (Figure 2 B). Similar cytotoxic effects of p31 and Omomyc IBs on cell viability were observed in multiple CD44 + cell lines, 23 reaching slightly different extents probably due to intrinsic differences in cell survival mechanisms, while GFP‐based IBs remained nontoxic (Figure 2 C). Microscopy images of MDA‐MB‐231 cultured cells exposed to active and nonactive materials, fully confirmed the cytotoxicity promoted by functionalized IBs (Figure 2 D). Indeed, functional IBs caused a clear reduction in tumor sphere size when compared to the GFP ones. Figure 2 Biological impact of IBs on cultured cells. A) Dose‐dependent loss of MDA‐MB‐231 cell viability upon exposure to IBs for 96 h. At 9 × 10 −6 m, background reduction of cell viability was observed in cells exposed to GFP‐based IBs (not shown). B) Time‐dependent loss of MDA‐MB‐231 cell viability upon exposure to IBs (3 × 10 −6 m ). C) Killing of different CD44 + cell lines by cytotoxic IB (3 × 10 −6 m ), upon exposure for 96 h. D) Representative bright field images of MDA‐MB‐231 tumor‐spheres upon addition of 3 × 10 −6 m protein and further incubation for 72 h. Qualitative assessment of sphere morphology and integrity when challenged with p31 and Omomyc IBs, compared to an equivalent GFP construct. Magnification bars represent 100 µm. One‐way ANOVA and a post‐hoc Dunnett test was performed comparing all groups to PBS‐treated control cells. p > 0. 05 (n. s. ); p < 0. 05 (*); p < 0. 01 (**); p < 0. 001 (***). Since both p31 and Omomyc perform their activities intracellularly, cell uptake of IBs can be inferred from the cytotoxicity exhibited by these materials on cultured cells, in contrast with the absence of effect shown by control GFP IBs (Figure 2 ). However, to effectively assess the cell penetrability of cytotoxic IBs, these materials were labeled with AF647. When MDA‐MB‐231 cells were exposed to labeled IBs, intracellular fluorescence increased in a dose‐dependent fashion when cells were exposed to FNI/II/V functionalized IBs ( Figure 3 A). However, GFP‐H6 IBs, which lack a cell ligand, did not accumulate in CD44 target cells (Figure 3 A). This confirms that cell targeting of IBs for specific protein delivery is a feasible concept, not restricted to the CXCR4‐binding T22 peptide, in which this event has been recently described for the first time. 30 Internalization efficacy of p31‐ and Omomyc‐ based IBs was comparable (Figure 3 A, B) and with similar intracellular distribution (Figure 3 C). Furthermore, FNI/II/V‐functionalized IBs accumulated inside target cells overtime, a fact that is compatible with moderate or absent lysosomal degradation (Figure 3 B). In this regard, the confocal analyses of the internalizing material revealed a punctuated fluorescence pattern indicative of an endosomal route of internalization, and a perinuclear localization of most of the engulfed protein particles that was more evident in the case of p31 IBs (Figure 3 C). Figure 3 Cell penetration of CD44‐targeted IBs. A) Quantification of protein internalization by fluorescence was quantified with flow cytometry into MDA‐MB‐231 cells incubated for 24 h at increasing concentration of IBs. B) Kinetics of IB penetration into MDA‐MB‐231 cells exposed to 3 × 10 −6 m IBs. C) Representative confocal microscopy images of protein internalization in MDA‐MB‐231 after 24 h treatment with 3 × 10 −6 m of IBs. Nuclei and membrane cells were labeled with Hoechst (blue) and WGA (light grey) respectively. Magnification bars represent 5 µm. Because of the high penetrability capacity of CD44‐targeted IBs in cultured cells, we decided to evaluate their potential macroscopic effect in vivo, in mouse cancer models. For a first screening of antitumoral effect, we chose to test Omomyc IBs, which had shown superior cytotoxicity in vitro compared to FN‐p31‐H6. Omomyc IBs were injected intratumorally in an orthotopic animal model of human breast cancer and the biological impact of the administered materials on the tumor progression was evaluated in repeated dose administration (weekly) for 4 weeks. Although the general tumor volume measured by caliper was not significantly affected during the test period ( Figure 4 A), the proportion of necrotic tumors (Figure 4 B) was clearly higher in the group of animals receiving Omomyc IBs compared to those in control groups (Figure 4 C). This observation indicated destruction of tumoral tissue, which, although not reflected by any significant change in tumor volume, was representative of a relevant biological effect. Hence, the detected antitumoral effect was indicative of the functional IBs having an impact on the biology of tumoral target cells. Figure 4 Antitumoral effect of IBs. A) Effect of Omomyc‐based IBs on tumor growth. Balb/c nude female mice bearing orthotopic tumors of MDA‐MB‐231 cell line were divided in 5 groups ( n = 7–8) and treated intratumorally (i. t. ) once a week. One group was treated with Omomyc IBs. GFP‐based IBs and PBS were included as treatment controls. No significant effect on tumor volume between the treatment groups was detected. B) Tumor necrosis was also evaluated at the experiment end point. Representative images are shown. C) The percentage of necrotic tumors for the groups treated with either IBs was measured. Omomyc IBs treated mice showed an increase in the number of necrotic tumors compared to GFP IB‐treated animals. In parallel, we continued the characterization of p31‐based IBs assessing their behavior in in vivo conditions. When these IBs were administered in a single dose, the fluorescent material was detected up to 7 d in the injection site, without apparent loss of emission ( Figure 5 A), confirming the stability of the protein particles. In this context, p21 overexpression (a cell cycle regulator usually associated to growth arrest and/or senescence) was also clearly observed by immunodetection in tumors treated with FN‐p31‐H6 IBs (Figure 5 B). Since the nuclear overexpression of p21 has been associated to the triggering of apoptotic cascades through the blocking of cyclin dependent kinases, 31 the presence of apoptotic cells was assessed in BT‐20 tumors upon single dose administration of FN‐p31‐H6 IBs, since these cells were the ones with the highest cell growth inhibition in vitro (Figure 2 C). As observed in Figure 5 C, a significant increase in apoptotic cell bodies was clearly visible in FN‐p31‐H6 IBs treated tumors when compared with control animals, and overexpression of p21 was confirmed by Western Blot in BT20 cells exposed to both IBs (Figure 5 D). Figure 5 In vivo biodistribution and efficacy of FN‐p31‐H6 IBs. A) Mice bearing orthotopic tumors of HCC1806 cell line were treated intratumorally (i. t. ) with p31 IBs labeled with the AF647 fluorochrome. In vivo imaging was performed with IVIS‐Spectrum 0, 4, and 7 days after administration (left panel) and quantified (right top panel). No fluorescence was observed outside the tumors (not shown). Ex vivo imaging (right bottom panel) confirmed that fluorescence was retained in the tumors, with higher intensities at the site of injection. B) Representative images of p21 protein expression by immunohistochemistry in tumors of animals treated i. t. with control GFP or p31 IBs, showing that only p31 containing IBs were able to induce p21 expression, in BT‐20 orthotropic tumors. Quantitative evaluation of the stained section confirmed that p21 staining was significantly higher in tumors of p31 IBs‐treated mice (right panel). C) Representative images of Hoechst staining of control FN‐GFP‐H6‐treated and p31 IBs‐treated mice. Arrows indicate the presence of characteristic apoptotic nuclei. Quantification of apoptotic nuclei in control and FN‐p31‐H6 treated tumors indicated that apoptosis was significantly increased upon treatment with p31 IBs (right panel). Magnification bars in (B, C) represent 100 and 50 µm, respectively. p < 0. 01 (**); p < 0. 001 (***). D) p21 expression in BT20 cells treated for 4 h with FN containing constructs. In summary, self‐assembling proteins and protein materials are of emerging interest because of their intrinsic lack of toxicity, biodegradability, functional and structural versatility, and because of their suitability for biological fabrication in a diversity of microbial cell factories. 32 Some of them, such as the bovine α‐lactalbumin 33 or the hen egg white lysozyme, 34 have already proved their utility as the basis of wide spectrum antitumoral drugs. Among the diversity of applications of protein constructs, bacterial IBs have found different promising niches in biomedical research, 35 as delivery systems of therapeutic proteins 8, 11 or as convenient agents in immunoprophylaxis and immunotherapies. 36 The utility of this platform is based on the high mechanical stability of IBs combined with the ability of the particles to release functional proteins under physiological conditions both in vitro and in vivo in cell culture media, 37 the intracellular environment 9 or as locally administered in organs or tissues of entire organisms. 13 Both the mechanical stability and capability of protein release are generic IB properties 38 supported by the particular sponge‐like architecture (combining amyloidal and nonamyloidal protein forms) of IBs. 3 Protein released from subcutaneous IBs can reach the bloodstream and be remotely delivered to target organs, provided the IB protein incorporates appropriate ligands for cell surface receptors. 12 In addition, IBs are also nontoxic in vitro when exposed to cultured mammalian cells 4 or upon systemic administration (i. e. , by oral delivery). 8 Therefore, these materials are promising as implantable protein depots for the release of functional polypeptides with therapeutic value aimed to the local or generic treatment of chronic diseases. A critical step in this direction would be to generate a proof of concept of the healing potential of the biomaterials in clinical contexts. In this regard, we have here proved for the first time that two recombinant proteins with recognized antitumoral effects, empowered with a potent ligand of the tumoral marker CD44 (Figure 1 ), are able to significantly impact tumor cell biology (Figures 2, 4, and 5 ) through the internalization of these biomaterials in target cells, both in vitro in cell culture (Figure 3 ) and in vivo in entire organisms (Figure 5 ). Indeed, we have demonstrated here that upon efficient internalization, engineered versions of p31 and Omomyc, formulated as CD44‐targeted IBs, promote selective target cell death through their respective impacts in regulatory cell circuits. Omomyc anti‐tumor activity is based on its direct interference with Myc function. Myc has been shown to be involved in multiple aspects of tumorigenesis, 39 both at the intracellular and extracellular level, being responsible for cell division, increased cell metabolism, immune tolerance, and survival to treatment of tumor cells. The benefit of inhibiting Myc has been demonstrated by multiple studies, 19 which indicated that interfering with Myc function does not only lead to cell growth arrest, but also to energetic crisis, anti‐tumor immune reprogramming, and cancer cell death. Omomyc has been instrumental in demonstrating this therapeutic opportunity, first as a transgene 20 and more recently as a therapeutic polypeptide, 40 while also demonstrating the complete safety of the approach for normal tissues. The potential use of Omomyc as a therapeutic mini‐protein has only emerged recently 40, 41 and, to date, this is the first and only report of its application in the IB format, which establishes the feasibility of its topical administration by intratumoral injection directly to the tumor site. P31, so far neglected regarding its potential uses as a therapeutic agent, has been directly linked to the control of p21 expression and pro‐apoptotic activity, especially in triple negative breast cancers (TNBC). 42 The present study proves that P31 is packable as IBs while keeping relevant antitumoral functionalities. This is not only pertinent regarding the incorporation of this protein to the growing catalogue of putative protein drugs, but also because it indicates that the engineering of antitumoral IBs is a general concept not restricted to a specific polypeptide. Importantly, the IB format itself is nontoxic, as CD44‐targeted GFP IBs do not cause any deleterious effects over exposed cells (Figure 2 ). 3 Conclusions Since the biological production of IBS is cost‐effective, using this material appears to be a convenient way of local administration of protein drugs. Considering the increasing spectrum of therapeutic proteins already approved or under development for cancer therapies, most of them produced by recombinant DNA technologies, 43 their potential packaging as IBs might largely expand their spectrum of activities and applicability in innovative and personalized medicines, for oncology and in other fields of human or animal clinics. 4 Experimental Section Protein Design, and IB Production and Purification : Two modular proteins were designed to contain the CD44 ligand FNI/II/V fused to either p31 (the C‐term fragment of p130cas) or Omomyc (the Myc‐derived bHLHZip domain mutant), appended with a His‐tag (Figure 1 ). The FNI/II/V domain was placed at the N‐ or C‐terminus of p31 17 and Omomyc‐containing modular protein respectively, considering preliminary results which showed no impact on protein activity. The amino acid sequences of the fusion proteins are MWQPPRARITGYIIKYEKPGSPPREVVPRPRPGVTEATITGLEPGTEYTIYVIALKNNQKSEPLIGRKKTGGSSRSSSGQYENSEGGWMEDYDYVHLQGKEEFEKTQKELLEKGSITRQGKSQLELQQLKQFERLEQEVSRPIDHDLANWTPAQPLAPGRTGGLGPSDRQLLLFYLEQCEANLTTLTNAVDAFFTAVATNQPPKIFVAHSKFVILSAHKLVFIGDTLSRQAKAADVRSQVTHYSNLLCDLLRGIVATTKAAALQYPSPSAAQDMVERVKELGHSTQQFRRVLGQLAAAGGSSRSSSKHHHHHH (FN‐p31‐H6) and MTEENVKRRTHNVLERQRRNELKRSFFALRDQIPELENNEKAPKVVILKKATAYILSVQAETQKLISEIDLLRKQNEQLKHKLEQLRNSCAGGSSRSSSWQPPRARITGYIIKYEKPGSPPREVVPRPRPG VTEATITGLEPGTEYTIYVIALKNNQKSEPLIGRKKTGGSSRSSSKHHHHHH (Omo‐FN‐H6) respectively. FN‐p31‐H6 and Omo‐FN‐H6 genes, together with the biologically inactive control gene FN‐GFP‐H6 and nontargeted GFP‐H6 control gene, were provided by GenScript (Hong Kong, China). Recombinant proteins were produced in E. coli strain BL21 (DE3), except Omo‐FN‐H6 that was produced in BL21 pLys cells. Protein production was usually induced at 37 °C for 3 h upon 1 × 10 −3 m IPTG addition. FN‐GFP‐H6 protein was produced overnight at 16 °C. Bacterial IBs were purified upon protein production; cells were incubated 2 h at 37 °C with a Protease Inhibitor Complete tablet, 10 × 10 −3 m PMSF and 1 µg mL −1 lysozyme. Then, cultures were treated with 0. 1% Triton X‐100 for 1 h at room temperature (RT). Cycles of freeze/thaw were applied daily to destroy remaining bacteria until reaching CFU mL −1 < 10 −2. Samples were incubated with 0. 02% NP‐40 for 1 h at 4 °C followed by 1 h incubation with 1 µg mL −1 DNAse at 37 °C. IBs obtained after centrifugation at 15 000 g were washed twice with PBS, separated in 1 mL aliquots and stored at −80 °C in PBS. A milder purification protocol was applied for GFP‐H6 IBs isolation. After the incubation with PMSF and lysozyme, cells were lysed using a French Press at 1200 psi. Lysates were treated with 0. 1% Triton X‐100 for 1 h at RT followed by 1 h incubation with 1 µg mL −1 DNAses at 37 °C. Cycles of freeze/thaw were applied daily until reaching CFU mL −1 < 10 −2. IBs were centrifuged, washed twice in PBS and stored at −80°C in PBS in 1 mL samples. IB Protein Quantification : Serial dilutions of 1 mL IB samples were prepared in water with Laemmli buffer and incubated at 98 °C for 45 min. IB dilutions together with a soluble GFP‐H6 standard of known concentration were loaded on SDS PAGE. Protein quality was analyzed by Coomassie blue staining, while protein quantification was calculated by Western Blot using an anti‐His antibody (sc57598, Santa Cruz Biotechnology, Santa Cruz, CA, USA). Protein bands were quantified from the standard curve fitting equation of GFP‐H6 using the Quantity One software. Cell Viability Assay : Cell viability of breast cancer cell lines (MDA‐MB‐231, MDA‐MB‐468, BT‐20, BT549, and HCC1187) upon treatment with IBs was measured by the MTT metabolic test (Roche, Basel, Switzerland) following manufacturer recommendations. MDA‐MB‐231, MDA‐MB‐468, BT‐20, BT549, and HCC1187 cells were plated at 2500 c/well, 5000 c/well, 6000 c/well, 1000 c/well, and 6000 c/well respectively for 24 h. Cells were incubated with 0. 5, 1, and 3 × 10 −6 m of IB protein. Viability was measured at 48, 72, and 96 h comparing luciferase signal against untreated control cells. MDA‐MB‐231 Mammosphere Production : MDA‐MB‐231 cells were seeded at 1000 c/well in low attachment cell culture plates (Nunc, Waltham, MA, USA) and cultured in a defined media without serum 44 at 37 °C and 5% CO 2. After 5 and 7 d of incubation mammospheres appeared and cultures were challenged by the addition of 3 × 10 −6 m FN‐GFP‐H6 and FN‐p31‐H6 IB proteins and Omo‐FN‐H6 IB protein for 72 h. Sphere integrity and morphology were assessed by bright field imaging in an Olympus BH2 microscope (Olympus Corporation, Tokyo, Japan). Field Emission Scanning Electron Microscopy : To characterize the morphometry (size and shape) at ultrastructural level of IBs at nearly native state, a rapid and easy method was used. Microdrops of 5 µL of diluted samples were deposited for 2 min on silicon wafers (Ted Pella Inc, Redding, CA, USA), liquid excess was blotted, air dried and immediately monitored without coating in a Field Emission Scanning Electron Microscope (FESEM) Zeiss Merlin (Carl Zeiss, Oberkochen, Germany) operating at 1 kV. IB images were collected with a standard secondary electron (SE) detector (4000–40, 000x magnifications). IB Labeling : AlexaFluor 647 NHS (Molecular Probes, Eugene, OR, USA) was used to label the inclusion bodies so that internalization assays could be normalized and compared at a unique wavelength. Conjugation was performed by resuspending IBs at 1 mg mL −1 in PBS with the addition of dye at a dye:protein molar ratio 2:1 and incubating for 1 h at RT with agitation. Serial cleaning steps were applied to remove free dye by centrifuging and resuspending the IBs with H 2 O. Fluorescence units/mass was determined by flow cytometry. Confocal Laser Scanning Microscopy : MDA‐MB‐231 cells were seeded on Mat‐Teck culture plate (Mat Teck Corporation, Ashland, MA, USA) at 200 000 c/well for 24 h. Medium was removed and changed to OptiPRO serum‐free medium supplemented with L‐glutamine and 1 × 10 −6 m of IBs was added and incubated for 24 h. After incubation, cells were washed with PBS and 3 µg mL −1 WGA 594 (Molecular Probes) and 0. 2 µg mL −1 Hoechst 33 342 (Molecular Probes) were added for 5 min in darkness to visualize the plasma membrane and the nuclei respectively. Later, cells were washed with PBS and complete medium was added. Stained cells were examined using a TCS‐SP5 confocal laser scanning microscope (Leica Microsystems, Heidelberg, Germany) with a Plan Apo 63 × /1. 4 (oil HC × PL APO l blue) objective. Hoechst 33 342, WGA 594 and Alexa Fluor 647 were excited by a blue diode (405 nm), a helium–neon laser (594 nm) and a helium–neon laser (633 nm) respectively. Z‐series were collected at 0. 5 mm intervals. Images were processed using Imaris version 6. 1. 0 software (Bitplane, Zürich, Switzerland). Flow Cytometry : MDA‐MB‐231 cells were cultured on a 24 well plate at 80 000 c/well and incubated at 37 °C and 5% CO 2 in a humidified atmosphere for 24 h. Cells were incubated with IBs at increasing concentrations and incubation times (internalization kinetics). After protein incubation, cells were treated with 1 mg mL −1 trypsin for 15 min to detach cells and remove unbound protein, followed by the addition of complete medium and centrifugation at 1400 g for 5 min. Collected cells were then resuspended in DPBS. Protein internalization was analyzed using a FACS‐Canto system (Becton Dickinson, Franklin Lakes, NJ, USA), with a 15 W air‐cooled red diode laser at 635 nm excitation. Alexa fluor 647 emission was measured with an FL4 detector (661/16 nm band pass filter). CD44‐positive cells were determined as described. 45 Human Cancer Animal Models : All the animal studies were performed in accordance with the ARRIVE guidelines and the 3 Rs (rules of Replacement, Reduction, and Refinement). Mice were housed and treated following the protocols approved by the Ethical Committee for the Use of Experimental Animals (CEEA) at the Vall d'Hebron Research Institute (VHIR), Barcelona. MDA‐MB‐231 cells, classified as mesenchymal‐like, 46 were resuspended in cold PBS at 15 000 000 c mL −1 and maintained on ice. Before surgery, mice were anesthetized with 2% isoflurane and buprenorphine (0. 75 mg kg −1 ) was administered subcutaneously. 100 µL (1 500 000 c/mouse were injected between the fourth and fifth right mammary fat pads of 8‐week‐old BALB/c nude female mice ( n = 40). Tumor size was evaluated 2 times a week by caliper measurements and tumor volume calculated using the following formula: volume = ( D × d 2 ) / 2, where “ D ” is the largest diameter and “ d ” the smallest one. Weight of the mice was measured at least twice a week. When tumors reached a volume between 100 and 350 mm 3, animals were randomized in the different groups ( n = 7–8): control buffer (50 × 10 −3 m KH 2 PO 4, 500 × 10 −3 m NaCl, 0. 8 m Urea, 100 × 10 −3 m GuHCl; pH 7. 4), PBS, FNI/II/V‐GFP‐H6, and Omo‐FNI/II/V‐H6. Protein stocks were diluted to 4 g L −1 for the treatments. Animals were treated intratumorally once a week and the volume/dose of intratumoral administration was adjusted according to tumor volume (1:10 of the tumor volume was inoculated up to a maximum of 50 µL, e. g. a tumor of 220 mm 3 was treated with 22 µL). Animals were treated up to 4 weeks, then euthanized by CO 2. For biodistribution and efficacy studies, exponentially growing HCC1806 (2. 5 × 10 6 ) or BT20 (10 × 10 6 ) cells, respectively, were orthotopically implanted into the fat mammary pad (i. f. m. p. ) and monitored weekly as described above. These cells are classified as basal like 47 and luminal androgen receptor (LAR), 48 respectively. In biodistribution assays, once the tumors reached an average size of 150 mm 3, mice were intratumorally administered with Alexa 647 labeled FN‐p31‐H6 IBs (200 µg per mouse, n = 3) and fluorescent signal was monitored in vivo at different time periods post‐administration by means of IVIS Spectrum equipment (Perkin Elmer, Tres Cantos, Spain). Quantification of the fluorescent signal (in Radiant Efficiency) was performed by using the Living Image software (Perkin Elmer). At the end point, tumors were excised and subjected to fluorescent imaging. Afterward, tissues were snap frozen and stored for further analysis or fixed in 4% formaldehyde and processed for histopathological analysis and evaluation. For efficacy assays, mice bearing BT20 tumors with an average size of 200 mm 3 ( n = 8), were administered with 600 µg of FN‐p31‐H6 or FN‐GFP‐H6 IBs. Administration was repeated twice in two week and animals were euthanized one week after the last treatment. Tumor p21 Immunohistochemistry and Apoptosis : The presence of p21 antigen in tumor sections was analyzed by pre‐treating paraffin embedded sections with 100 × 10 −3 m citrate buffer (pH 9) in a cooker. Sections were incubated with 10% normal goat serum (NGS) in antibody diluent (1% BSA in 100 × 10 −3 m Tris buffer) and then of 1:100 dilution of anti‐p21 antibody (M 7202, Dako, Santa Clara, CA, USA). Secondary antibody consisted in a HRP conjugated system (EnVision+ System‐HRP Labeled Polymer anti‐Mouse), which was later visualized with DAB and counterstained with Harris haematoxylin. The p21 signal intensity and extension were scored under the light microscope by two blinded observers. Intensity cores ranged from 0 to 3 (absence, low intensity, normal or high, respectively). For each tumor, two independent sections with p21 staining were evaluated. For the evaluation of the extension of the apoptotic nuclei, sections were stained with Hoescht (1:500, Sigma‐Aldrich, San Luis, MI, USA) mounted in Prolong (Invitrogen, Carlsbad, CA, USA). Two independent slides per tumor and five representative images per slide were analyzed after acquiring images at 10x magnification in an Olympus BX61 microscope (Olympus Corporation). In each image, the number of apoptotic nuclei versus the number of nonapoptotic nuclei was counted by two independent blinded observers. Results were presented as percentage of apoptotic nuclei. Anti‐p21 (MS‐891‐B Thermo Scientific) Western Blot was performed on cell extracts as described 49 using BT20 cells exposed to 9 × 10 −6 m protein for 4 h. Fold increase of p21 was referred to a β‐actin control. Statistical Analysis : Quantitative data were expressed as mean and standard error (x̄ ± SE). Depending on the type of data, parametric or nonparametric statistics were applied to compare divergences between all groups in relation to control or pairwise comparisons. In some data, one‐way ANOVA followed by a post‐hoc Dunnett was used, whereas Kruskall–Wallis and Mann Whitney or Wilcoxon tests were performed with the rest of data. Statistical significance levels were represented as n. s. p > 0. 05, * p < 0. 05, ** p < 0. 01, and *** p < 0. 001. All analyses were performed using SigmaPlot 10. 0 software. Conflict of Interest T. J. is an employee of Peptomyc S. L. ; L. S. and M. E. B. are co‐founders and shareholders of the same company.
|
10. 1002/advs. 201900867
| 2,019
|
Advanced Science
|
Osteochondral Regeneration with 3D‐Printed Biodegradable High‐Strength Supramolecular Polymer Reinforced‐Gelatin Hydrogel Scaffolds
|
Abstract Biomacromolecules with poor mechanical properties cannot satisfy the stringent requirement for load‐bearing as bioscaffolds. Herein, a biodegradable high‐strength supramolecular polymer strengthened hydrogel composed of cleavable poly( N ‐acryloyl 2‐glycine) (PACG) and methacrylated gelatin (GelMA) (PACG‐GelMA) is successfully constructed by photo‐initiated polymerization. Introducing hydrogen bond‐strengthened PACG contributes to a significant increase in the mechanical strengths of gelatin hydrogel with a high tensile strength (up to 1. 1 MPa), outstanding compressive strength (up to 12. 4 MPa), large Young's modulus (up to 320 kPa), and high compression modulus (up to 837 kPa). In turn, the GelMA chemical crosslinking could stabilize the temporary PACG network, showing tunable biodegradability by adjusting ACG/GelMA ratios. Further, a biohybrid gradient scaffold consisting of top layer of PACG‐GelMA hydrogel‐Mn 2+ and bottom layer of PACG‐GelMA hydrogel‐bioactive glass is fabricated for repair of osteochondral defects by a 3D printing technique. In vitro biological experiments demonstrate that the biohybrid gradient hydrogel scaffold not only supports cell attachment and spreading but also enhances gene expression of chondrogenic‐related and osteogenic‐related differentiation of human bone marrow stem cells. Around 12 weeks after in vivo implantation, the biohybrid gradient hydrogel scaffold significantly facilitates concurrent regeneration of cartilage and subchondral bone in a rat model.
|
1 Introduction Articular cartilage defects along with subchondral bone degeneration in the knee joint is a common clinical problem, which leads to knee joint dysfunction, significant pain, and even disability. 1, 2, 3, 4 What is worse, the degeneration of subchondral bone in cartilage defect may further aggravate osteoarthritis. 5, 6, 7, 8 Thus, the success of osteochondral regeneration is largely dependent upon constructing a scaffold with an ability to recapitulate niche cues. Hydrogels are emerging as a promising class of biomaterials for both soft and hard tissue regeneration. 9, 10, 11, 12, 13 Recently, with rapid development of 3D printing technology, increasing studies, including ours, are focused on fabricating scaffolds by 3D printing of hydrogel bioinks for osteochondral repair. 14, 15, 16, 17 Among a wide array of hydrogels for tissue engineering, gelatin hydrogel has been developed as a variety of bioinks for 3D printing due to its better biocompatibility, biodegradability, bioactivity, and abundance from diverse sources. However, physical gelatin hydrogel itself or extensively used chemically crosslinked methacrylated gelatin (GelMA) hydrogels are weak and brittle in mechanical properties, precluding their applications as load‐bearing scaffolds. 18, 19, 20 Recently, mechanically resilient gelatin hydrogels crosslinked by supramolecular host–guest interactions 21 or dual synergistic physical crosslinking 22 were reported by Bian. GelMA/β‐tricalcium phosphate composite hydrogel was also developed and explored as a bone graft material. 23 However, very few studies were devoted to fabricating high strength gelatin hydrogels and thus developing printable bioinks for osteochondral reconstruction. To treat osteochondral defect, our team constructed a biohybrid scaffold by one‐step thermal‐assisted extrusion printing of supramolecular poly( N ‐acryloyl glycinamide)‐based hydrogel bioink mixed with organic nanoparticles. 24, 25 However, to make the hydrogel printable, the hydrogen bonding interactions were unavoidably sacrificed, thus resulting in a marked decrease in the mechanical strength of printed constructs far below that of native cartilage. Moreover, this hydrogel was not degraded due to the stable hydrogen bonding interactions, and the printed bare scaffold did not support cell attachment either; thus, it exists as the foreign matter in the defect without integrating with the surrounding tissues. This will severely restrict their applications for osteochondral regeneration. Therefore, developing biodegradable high‐strength hydrogel scaffold for guiding the ingrowth of bone and cartilage tissue to achieve true osteochondral regeneration is highly desired. Motivated to address the above challenges, we aim to design and synthesize a novel biodegradable high strength supramolecular hydrogen bonding strengthened‐gelatin hydrogel, and thus fabricate a gradient hydrogel bioscaffold that is able to provide a mechanical support in the early stage of osteochondral repair, and eventually degraded along with the ingrowth of the new tissue. Previously, inspired by remarkable stimuli‐responsive changes in consistency of sea cucumbers, we prepared supramolecular poly ( N ‐acryloyl 2‐glycine) (PACG) hydrogel with side chain containing amide and carboxyl dual hydrogen bonds, which not only contributed to strengthening the hydrogels formed, but also served as dynamic bonds to modulate the mechanical properties of the hydrogels. 26 Nevertheless, at a physiological neutral pH, the PACG hydrogel was quickly autolyzed and disintegrated within several hours, making it unsuitable for tissue repair scaffold application. In light of strengthening mechanism and dynamic cleavability of PACG, we propose to develop a novel supramolecular hydrogen bonding strengthened GelMA chemical crosslinked hydrogel (PACG‐GelMA) by incorporating the reversible hydrogen bonds of ACG into the GelMA hydrogel system. The dual hydrogen bonds of PACG side chain can reinforce and stabilize the GelMA network, and in turn the chemical crosslinking of GelMA prolongs the degradation of PACG network. It is anticipated that the PACG‐GelMA hydrogel will be eventually degraded in vivo due to the final degradability of stabilizing GelMA. Furthermore, owing to the thermoreversible gel⇔sol transition behavior and tunable viscosity suited for 3D printing, gelatin mixed with ACG monomer is used as a bioink. In this work, a series of biodegradable preprogrammed biohybrid gradient PACG‐GelMA hydrogel scaffolds are printed, and stabilized with UV light irradiation. To enhance the repair efficacy of cartilage defect, the bioactive manganese ions (Mn 2+ ) are loaded into the top cartilage layer of gradient PACG‐GelMA hydrogel scaffolds, while the bioactive glass (BG) is incorporated into the bottom subchondral bone layer of the scaffold with an aim to facilitate new bone regeneration ( Scheme 1 ). The mechanical properties of the hydrogels and printed scaffolds are determined. We will focus on evaluating the bioactivity and osteochondral regeneration efficacy of the PACG‐GelMA hydrogel scaffolds in vivo. Scheme 1 Schematic illustration of 3D printing of the biohybrid gradient scaffolds for repair of osteochondral defect. A) The compositions of bioink A and bioink B, and 3D‐bioprinting method of the biohybrid gradient scaffolds assisted with a low‐temperature receiver; B) formation of stable hydrogel scaffold after UV light‐initiated polymerization and main hydrogen bonding interactions in the PACG‐GelMA network; C) the repair of osteochondral defects treated with the biohybrid gradient PACG‐GelMA hydrogel scaffold with Mn 2+ and BG being respectively loaded on the top layers and bottom layers in animal experiment. 2 Results and Discussion 2. 1 Preparation and Characterization of PACG‐GelMA Hydrogels Prior to printing biohybrid gradient scaffold for treatment of osteochondral defect, we first prepared a supramolecular polymer strengthened chemical crosslinking hydrogel ink by copolymerizing ACG with GelMA. To improve its 3D printability, we synthesized and selected GelMA with low degree of methacrylation (29%, determined by integration of 1 H NMR spectra, Figure S1, Supporting Information) due to its similar viscosity to that of the pristine gelatin (to be determined in the later rheological measurement). The Fourier‐transform infrared spectroscopy (FTIR) spectra suggested the successful formation of GelMA and PACG‐GelMA hydrogels (Figure S2A, B, Supporting Information). The problem for in vivo implantation of PACG hydrogel is its quick dissolution at a neutral pH due to the dissociation of carboxyl groups. 26 In this study, copolymerization of GelMA can maintain the swelling stability of the PACG‐GelMA hydrogel in aqueous solution because of chemical crosslinking networks formed. Figure S3 in the Supporting Information displays the swelling ratios of the hydrogels as a function of time. In general, with an increase of ACG content at fixed GelMA concentration or with an increase of GelMA content at fixed ACG concentration, the swelling ratio and the time to reach a swelling equilibrium of the PACG‐GelMA hydrogels decrease by a different extent. An interpretation is that increasing the contents of ACG and GelMA introduces more H‐bonds of PACG and chemical crosslinks between the PACG and GelMA chains, thus restricting the diffusion of water molecules into the network. Notably, the equilibrium swelling ratio of PACG10‐GelMA10 hydrogels (2. 49) is close to that of PACG35‐GelMA7 hydrogels (2. 24) possibly due to the similar crosslinking density. Importantly, after swelling equilibrium, the volume of the PACG‐GelMA hydrogels remains very stable without occurrence of further swelling, indicating excellent “nonswellable” properties of the PACG‐GelMA hydrogels, which are critical as an implantable tissue engineering scaffold for early stage of load bearing. The equilibrium water contents (EWCs) of the PACG‐GelMA hydrogels vary from 74% to 98% over the selected range of the monomer concentrations depending on ACG/GelMA ratio (Figure S4, Supporting Information). The PACG‐GelMA hydrogels happen to be noncovalently strengthened chemically crosslinked hydrogels that combines primarily the covalent GelMA‐PACG crosslinking with PACG hydrogen bonded crosslinking, though there exist PACG‐GelMA hydrogen bonding interactions. 27, 28 Figure 1 summarizes the mechanical properties of the fully swollen PACG‐GelMA hydrogels. It can be clearly seen that the mechanical properties of PACG‐GelMA hydrogels are significantly enhanced compared to those of pristine GelMA hydrogels. And when the GelMA content is fixed, the mechanical strengths are considerably increased with the increment in monomer content of ACG. While at a constant ACG content, the mechanical strengths exhibit first an increasing trend and then decrease with increasing GelMA content. Over a 4–7% GelMA range, the mechanical properties of the PACGX‐GelMA7 are better than those of PACGX‐GelMA4; when GelMA content continues to increase from 7% to 10%, the Young's modulus and compressive modulus are further increased, but the tensile strength, break strain, compressive strength, and compressive failure strain decrease slightly. In the selected range of initial concentrations of ACG, the PACGX‐GelMA4 hydrogels achieve 0. 018–0. 721 MPa tensile strength, 42–224% break strain, 38–252 kPa Young's modulus, 0. 048–8. 9 MPa compressive strength, 7–471 kPa compressive modulus, and 68–93% compressive failure strain. The PACGX‐GelMA7 hydrogels achieve 0. 029–1. 108 MPa tensile strength, 52–245% break strain, 49–281 kPa Young's modulus, 0. 11–12. 4 MPa compressive strength, 27–651 kPa compressive modulus, and 70–97% compressive failure strain. The PACGX‐GelMA10 hydrogels achieve 0. 025–0. 978 MPa tensile strength, 30–207% break strain, 74–320 kPa Young's modulus, 0. 125–9. 9 MPa compressive strength, 51–837 kPa compressive modulus, and 72–88% compressive failure strain. It has been proved that there is only covalent crosslinking in the pristine GelMA hydrogel network whose triple helix hydrogen bonds are too weak to act as a strengthening mechanism. 29, 30 Therefore, the pristine GelMA hydrogels are poor in mechanical properties. In comparison, for the PACG‐GelMA hydrogels, the higher mechanical strengths are primarily attributed to multiple hydrogen bonding reinforced mechanism from PACG. 26, 31 However, excessively high crosslinking density of GelMA may interfere with the formation of PACG hydrogen bonds, resulting in a decline of mechanical strength. It is noted that the PACG10‐GelMA10 and PACG35‐GelMA7 hydrogels demonstrate excellent mechanical properties with robust tensile strength (up to 0. 158 and 1. 1 MPa), large stretchability (up to 139% and 245%), high Young's modulus (up to 143 and 281 kPa), high compressive strength (up to 4. 1 and 12. 4 MPa), and compressive modulus (up to 421 and 651 kPa). We note that the compressive modulus of PACG10‐GelMA10 hydrogel is in the range of native cartilage's modulus (300–800 kPa). 20, 32, 33 Figure 1 A–C) Tensile stress–strain curves and D–F) compressive stress–strain curves of the PACG‐GelMA hydrogels with varied initial concentrations of ACG and GelMA in deionized water. The PACG‐GelMA hydrogels were shown to be sufficiently stiff and stable in an aqueous medium. Next, to imitate in vivo environment, we evaluated the degradation behaviors of PACG‐GelMA hydrogels in collagenase solution by monitoring the percentage of residual hydrogel mass as a function of time. As shown in Figure 2, the GelMA hydrogel is degraded rapidly, and is completely dissolved in solution within 11 h. In contrast, the PACG‐GelMA hydrogels show a slower degradation rate under the same condition. The degradation time varies from 7 to 100 days over the selected range of the initial contents. It is noted that with an increment of ACG at a fixed GelMA content, the degradation rate is decreased significantly. Similarly, raising GelMA feeding at a fixed ACG content also results in a lower degradation rate. It is not difficult to understand that copolymerization of ACG can protect cleavage sites on gelatin backbone from collagenase attack within the copolymer networks. In addition, the enhanced crosslinking density significantly limits the penetration of collagenases into the network, thus slowing down the degradation process. 34 It is noteworthy that the complete degradation times of PACG10‐GelMA10 and PACG35‐GelMA7 hydrogel are 35 and 63 days, respectively. This suggests that the PACG35‐GelMA7 hydrogel scaffold can provide a longer time of mechanical support before the defected tissue is repaired, meanwhile avoiding the collapse of the scaffold. Figure 2 In vitro degradation behaviors of PACGX‐GelMAY hydrogels with varied initial concentrations of ACG and GelMA in collagenase solution. 2. 2 Preparation and Characterization of PACG‐GelMA‐Mn 2+ and PACG‐GelMA‐BG Hybrid Hydrogels As aforementioned, the PACG10‐GelMA10 and PACG35‐GelMA7 hydrogels exhibit a similar swelling behavior, which can avoid the occurrence of delamination of the printed multilayer scaffolds upon contacting body liquid. In addition, these two hydrogels are sufficiently stiff and degraded within appropriate time. Considering their different degradation time, the PACG10‐GelMA10 hydrogel with a shorter degradation time was used as an ink for printing top cartilage layer, and PACG35‐GelMA7 hydrogel with a longer degradation time served as an ink for bottom bone layer. Taking into account the bioactivity of Mn 2+ ions in promoting cartilage synthesis, 35 Mn 2+ ‐doped PACG10‐GelMA10 hydrogel (termed as PACG10‐GelMA10‐Mn 2+ ) was prepared. We measured the mechanical properties of the PACG10‐GelMA10‐Mn 2+ hydrogels, and found no difference with those of PACG10‐GelMA10 hydrogel since only trace amount of ions (0. 125–4 ppm) was added, causing no influence on the mechanical strength. Bioactive glass (BG) has proven to be an important biocompatible bone‐repair material since its released bioactive ions can promote cell adhesion, proliferation, and differentiation. 36, 37 To facilitate bone regeneration, we then prepared bone layer ink by adding bioactive glass into the PACG35‐GelMA7 hydrogel. The FTIR spectra of BG and PACG35‐GelMA7‐BG hybrid hydrogels suggest the successful doping of BG into the hydrogel system (Figure S2C, D, Supporting Information). Figure S5 in the Supporting Information exhibits the mechanical properties of fully swollen PACG35‐GelMA7‐BG hydrogels with varied BG contents. It is clearly observed that the mechanical strengths decrease with the increment of BG content. It is rational to think that incorporation of BG does not affect the chemical crosslinking, but may interfere with the formation of hydrogen bonds of PACG. Nevertheless, at 1% BG content, 0. 71 MPa tensile strength, 7. 51 MPa compressive strength, 267 kPa Young's modulus, and 579 kPa compressive modulus can be maintained. Thus, in the subsequent experiments, we chose PACG35‐GelMA7 hydrogel doped with 1% BG (abbreviated as PACG35‐GelMA7‐BG without otherwise statement) as a bone layer ink. We found that only 1% BG exerted a negligible effect on degradation of the hybrid hydrogels. Cell affinity is a prerequisite for tissue regeneration. Currently reported high‐strength hydrogels rarely support cell adhesion, spreading, and proliferation. 38, 39 Herein, we examined the effects of different concentrations of Mn 2+ on the proliferation of hBMSCs on the surface of PACG10‐GelMA10 hydrogel. Figure S6A in the Supporting Information shows that there is no statistically significant difference for the cell proliferation over the concentration range of 0–1 ppm, whereas the cell proliferation ability in the group of 1 ppm is slightly lower than that in the 0. 5 ppm group. Comparatively, the cell proliferation on the Mn 2+ ‐doped hydrogels is significantly different from that on the pure PACG10‐GelMA10 hydrogel at 2 and 4 ppm concentrations of Mn 2+. Based on the above analysis, we chose PACG10‐GelMA10 hydrogel doped with 0. 5 ppm Mn 2+ (defined as PACG10‐GelMA10‐Mn 2+ for description simplicity) as the cartilage layer ink. Similarly, PACG35‐GelMA7 and PACG35‐GelMA7‐BG hydrogels can also support the cell proliferation very well, with no significant difference between these two groups (Figure S6B, Supporting Information). From the results of live/dead assay (Figure S6C–F, Supporting Information), we can see that the hBMSCs grow well and show spindle‐like morphologies on the surfaces of the PACG10‐GelMA10, PACG10‐GelMA10‐Mn 2+, PACG35‐GelMA7, and PACG35‐GelMA7‐BG hydrogels. The above results indicate that introducing GelMA can promote cell adhesion and proliferation on the gel surface. The rheological characteristics of the PACG10‐GelMA10‐Mn 2+ (cartilage layer ink) and PACG35‐GelMA7‐BG (bone layer ink) hydrogels were further examined. Figure S7A, B in the Supporting Information shows the frequency scans of PACG10‐GelMA10‐Mn 2+ and PACG35‐GelMA7‐BG hydrogels. As the frequency increases from 0. 1 to 20 Hz, the storage modulus G′ and the loss modulus G″ of both the PACG10‐GelMA10‐Mn 2+ and PACG35‐GelMA7‐BG hydrogels remain almost constant, and the G′ is far above the G″ (Figure S7A, B, Supporting Information), suggesting gelling properties. 40, 41 In addition, the G′ values of PACG35‐GelMA7‐BG hydrogels are much larger than those of PACG10‐GelMA10‐Mn 2+ hydrogels, which is consistent with the results of mechanical test above. Figure S7C, D in the Supporting Information shows the recovery ability of PACG10‐GelMA10‐Mn 2+ and PACG35‐GelMA7‐BG hydrogels under continuous alternate oscillatory shear strain measurements. Under the small shear strain (10%), the G′ of the gel is larger than G″, showing a typical elastic solid characterisitic. When the large shear strain (100%, 300%) is applied, the PACG‐GelMA hydrogel switches into a sol state with the G″ being greater than G′, implying the breakup of the network. Interestingly, the G′ and G″ of the gel can rapidly recover to the initial values after the large shear strain is removed, indicating the strong chemical crosslinks and reversible multiple hydrogen bonding interactions contribute to the excellent recovery ability of the network. 42, 43 This quick recovery capability owned by PACG10‐GelMA10‐Mn 2+ and PACG35‐GelMA7‐BG make them ideal inks for printing gradient scaffold for treating osteochondral defect. 2. 3 3D Printing of Biohybrid Hydrogel Scaffolds The gel–sol transition temperature, viscosity at the transition point and shear‐thinning behavior are important parameters of hydrogel inks. 44 These parameters provide a reference for subsequent determination of printing schemes. The modulus‐temperature curves ( Figure 3 A, B) show that the gel–sol transition temperature of pristine GelMA7 ink (36. 5 °C) is slightly lower than that of Gelatin7 ink (38. 5 °C). It is possible that methacrylation modification of gelatin affects the formation of triple‐helix hydrogen bonds. The gel–sol transition temperatures of ACG35‐GelMA7 ink and ACG35‐GelMA7‐BG ink occur around at lower temperatures of 22. 4 and 23 °C, indicating that the introduction of ACG further disturbs the formation of hydrogen bonds in gelatin network; while doping a small amount of BG has a negligible impact on gel–sol transition. As expected, Gelatin10 ink and GelMA10 ink respectively reveal higher gel–sol transition temperatures of 40 and 38. 5 °C due to denser physical crosslinking. As described above, methacrylation leads to a slight decrease in transition temperature. After copolymerization, the gel–sol transition temperature of ACG10‐GelMA10 ink is near 29 °C; while the gel–sol transition temperature of ACG10‐GelMA10‐Mn 2+ is identical to that of ACG10‐GelMA10 ink (data not shown), meaning that trace amount of Mn 2+ ions has no effect on transition temperature. Compared to ACG10‐GelMA10 ink, the lower transition temperature of ACG35‐GelMA7 ink suggests that copolymerizing more of ACG leads to a disturbance to the formation of hydrogen bonds in gelatin to a greater extent. Figure 3 A, B) Influences of the concentration of GelMA and ACG, and addition of BG on the rheological properties of the hydrogels. Variation in dynamic storage moduli G′ and loss moduli G″ as a function of temperature in a temperature amplitude sweep test, where the cross‐over points between the gel and sol state (G′ = G″) represent gel‐sol transition temperature; C, D) variation of viscosity as a function of temperature; E, F) viscosity measurement of different bioinks in a shear rate sweep from 0. 1 to 500 s −1, indicative of shear‐thinning behavior. A sufficiently large viscosity can maintain the shape fidelity of the printed filaments, and thus prevent the structures from collapsing before it is chemically crosslinked by the UV light irradiation. Figure 3 C, D clearly demonstrates that the viscosities of all ACG‐GelMA inks remain stable below 22 °C, and then drop with increasing temperature. Heat‐triggered gel–sol transition is favorable for 3D printing, and a relatively high viscosity at a low temperature is advantageous for improving the fidelity of the shape of filaments and obtained scaffold. Furthermore, all the ACG‐GelMA inks exhibit similar shear thinning behaviors (Figure 3 E, F), confirming that the ACG‐GelMA inks can be smoothly squeezed out of the printing nozzle. Comprehensively considering the above rheological characteristics, an air‐extrusion 3D printing method assisted with a low‐temperature receiver was adopted to precisely fabricate osteochondral regeneration scaffolds. ACG‐GelMA inks were first kept at 4 °C for 20 min to allow for full formation of triple helices in gelatin network to increase the viscosity, thus improving the printability of low concentration of GelMA bioinks. 45 Then the printing cartridges were increased back to 20 °C to prevent the ink from excessive physical crosslinking and clogging in the printing nozzle, making sure that the stable and smooth filament could be deposited on the receiving platform controlled at −10 °C at which the shape of the fiber can be well maintained. This ensures the fidelity of the scaffold. Then the printed scaffolds were subjected to UV light irradiation in a cooling environment to initiate polymerization to eventually fix the final architecture of the scaffolds. Herein, PACG10‐GelMA10‐Mn 2+, PACG35‐GelMA7‐BG, and gradient scaffold consisting of top cartilage layer of PACG10‐GelMA10‐Mn 2+ and bottom bone layer of PACG35‐GelMA7‐BG were printed. Figure 4 A, B displays the morphologies of the representative printed PACG35‐GelMA7 hydrogel scaffolds in which the interconnected macroscopic pores are orderly arranged and uniform. The scaffolds can still keep the intact geometry shape after reaching swelling equilibrium in deionized water, indicating the printed PACG‐GelMA hydrogel scaffolds can maintain a very stable swelling stability in aqueous solution, and thus structural integrity and the high resolution of the architecture. The obtained scaffolds demonstrate excellent compressive strengths (>1 MPa). The compressive strengths of PACG10‐GelMA10‐Mn 2+ and PACG35‐GelMA7‐BG and biohybrid gradient hydrogel scaffolds are measured to be 1. 24, 2. 51, and 2. 17 MPa, respectively. The corresponding compressive moduli are 163, 249, and 195 kPa, respectively (Figure 4 C). Subjected to 100 cycles of continuous compression loading and unloading, the overlapped successive cyclic force curves can be achieved, manifesting a high elasticity of the printed scaffold (Figure 4 D). Even after cyclic compression, the scaffold was able to remain its macro‐ and microstruture owing to its high strength and better elasticity. Figure 4 A) A photograph of ACG‐GelMA hydrogel scaffold printed by air‐extrusion method assisted with a low‐temperature receiver before UV light irradiation; B) microscope image of PACG‐GelMA hydrogel scaffold after UV light irradiation and reaching swelling equilibrium (stained with rhodamine); C) compressive stress–strain curves of the printed porous PACG10‐GelMA10‐Mn 2+ hydrogel scaffolds, PACG35‐GelMA7‐BG hydrogel scaffolds, and gradient hydrogel scaffolds; D) cyclic compressive stress‐strain curves for the printed gradient scaffold (the top three layers were printed by PACG10‐GelMA10 hydrogel, and the bottom nine layers were printed by PACG35‐GelMA7‐BG hydrogel under a maximum strain of 30%. The cycle numbers were set as 100. 2. 4 Ions Release and Biofunctions of the Printed Hydrogel Scaffolds The cumulative release behaviors of Mn 2+ from the printed PACG10‐GelMA10‐Mn 2+ hydrogel scaffolds and release of Sr 2+, Si 4+, and B 3+ from the printed PACG35‐GelMA7‐BG hydrogel scaffolds are shown in Figure S8 in the Supporting Information. Within the initial three days, a burst release is observed owing to the rapid leakage of ions near the surface. And then these ions diffuse out of the printed porous hydrogel scaffolds gradually. Finally, ions release from the printed porous hydrogel scaffolds level off around 25 days. Alkaline phosphatase (ALP), a well‐known marker for early osteogenic differentiation of hBMSCs, was assayed after hBMSCs cells were cultured for 7 and 14 days on PACG35‐GelMA7 and PACG35‐GelMA7‐BG hydrogel scaffolds (Figure S9, Supporting Information). The hBMSCs seeded on the PACG35‐GelMA7‐BG hydrogel scaffolds present stronger ALP activity than those on the PACG35‐GelMA7 hydrogel scaffolds and the hBMSCs on the culture plate. These results reveal that incorporation of BG can promote the osteogenic differentiation of hBMSCs. Chondrogenic and osteogenic genes expression levels at 7 and 14 days were further analyzed by real‐time quantitative polymerase chain reaction (RT‐qPCR) ( Figure 5 ). For the group of PACG10‐GelMA10‐Mn 2+ scaffolds, the expression levels of cartilage‐specific gene (COL II, AGG, and SOX‐9) are markedly upregulated compared with PACG10‐GelMA10 scaffold group at both 7 and 14 days, demonstrating that loading Mn 2+ in the PACG10‐GelMA10 scaffold can promote the chondrogenesis. Meanwhile, there is no significant upregulation expression of fibroblastic gene (COL I) even after 14 days of culture, showing no sign of fibrocartilage formation in Mn 2+ ‐doped scaffold (Figure 5 A, B). Compared with PACG35‐GelMA7 scaffolds, bone‐related gene (ALP, COL I, OCN, and RUNX2) expression also displays an upregulation at both 7 and 14 days in the PACG35‐GelMA7‐BG hydrogel scaffolds, implying that continuous release of ions from the BG loaded in the PACG35‐GelMA7 hydrogel scaffolds may be conducive to enhancing the osteogenic differentiation of hBMSCs (Figure 5 C, D). These results verify that combining Mn 2+ with PACG10‐GelMA10 scaffolds and BG with PACG35‐GelMA7 scaffolds can benefit the chondrogenic and osteogenic differentiation, respectively. Figure 5 Gene analysis for chondrogenic differentiation and osteogenic differentiation from hBMSCs grown in different scaffolds. A, B) The expression of cartilage‐associated gene (COL II, aggrecan, SOX‐9, and COL I) after incubated for 7 and 14 days, respectively; C, D) the expression of osteogenesis‐associated gene (ALP, OCN, COL I, and RUNX2) after incubated for 7 and 14 days, respectively. Mn 2+ ‐loaded PACG10‐GelMA10 hydrogel scaffold or BG‐loaded PACG35‐GelMA7 hydrogel scaffold significantly elevates the chondrogenic or osteogenic differentiation of hBMSCs, respectively (* p < 0. 05, # p < 0. 01, compared with control groups). 2. 5 In Vivo Osteochondral Repair Efficacy of the Gradient Hydrogel Scaffolds To evaluate the new subchondral bone formation, micro computed tomography (micro‐CT) images were taken at 4, 8, and 12 weeks after surgery. Representative 3D reconstruction images of each group are shown in Figure 6. Clearly, compared with the untreated control and PAG gradient scaffold without loading Mn 2+ or BG, the new subchondral bone formation in the PAG‐Mn‐BG (biohybrid gradient scaffold consisting of top layers of PACG10‐GelMA10‐Mn 2+ and bottom layers of PACG35‐GelMA7‐BG) group is significantly improved at 4 weeks postimplantation. This suggests that the release of bioactive ions from the PAG‐Mn‐BG scaffold may promote the early regeneration of subchondral bone. At 8 weeks postimplantation, it is seen in the PAG scaffold group that the new tissue grows along the alignment direction of filaments of the implanted interconnected porous scaffolds. In comparison, the repair effect of subchondral bone in the PAG‐Mn‐BG scaffold group is further improved. Interestingly, the newly formed subchondral bone is especially pronounced in the PAG‐Mn‐BG scaffold group at 12 weeks postimplantation, where the entire defect area is filled with new subchondral bone. The repair effect of subchondral bone is also improved in the PAG scaffold group, but to a lesser extent than in the PAG‐Mn‐BG scaffold group, since there is still a large cavity in the defect region. In contrast, the defects in the untreated group show only very little new subchondral bone formation along the edge of the implantation. All these results prove that the 3D printed stiff gradient hydrogel scaffold can play an important role in template‐guide and mechanical support, and can further speed up the regeneration of subchondral bone after the introduction of ions. Quantitative analysis data of the newly formed subchondral bone within the defect further confirm the micro‐CT findings (Figure S10, Supporting Information). For all the three groups, the values of the ratio of bone volume to tissue volume (BV/TV) and bone mineral density (BMD) in trabecular volume of interest (VOI) keep increasing with the extension of implantation time, but the trend of increase varies considerably. Both values are significantly higher in the PAG‐Mn‐BG scaffold group than in the untreated and PAG scaffold groups at the same postimplantation time, indicating that the PAG‐Mn‐BG scaffold containing various bioactive ions can boost osteochondral tissue repair in the defects. Figure 6 Characteristic 3D reconstruction images of micro‐CT analysis of the repaired subchondral bone at 4, 8, and 12 weeks in different groups. Histological staining analysis further confirms that the PAG‐Mn‐BG scaffold is able to simultaneously enhance the repair of articular cartilage and subchondral bone, relative to the PAG scaffold and the untreated control ( Figure 7 A). As shown in hematoxylin and eosin (H&E) and toluidine blue (T‐B) staining, new cartilage and bone in the PAG‐Mn‐BG scaffold group are observed at the edges of the defects at 8 weeks postimplantation. At 12 weeks postimplantation, a layer of new cartilage is observed in the chondral region treated with PAG‐Mn‐BG group, with a thickness similar to that of the adjacent cartilage. Similarly, the subchondral bone area is also filled with new bone. Both regenerated cartilage and subchondral bone are well integrated with the host tissues. In comparison, the defects treated with the PAG scaffold show less effective bone and cartilage repair with poor quality even at 12 weeks; while empty cavities and scarce regenerated tissue are present in the untreated defects, with the collapse of adjacent cartilage. Immunohistochemical (IHC) staining for both cartilage and bone‐specific proteins were also carried out to further evaluate the osteochondral repair. COL II and GAGs are positively and uniformly stained in PAG‐Mn‐BG scaffold group, and the protein expressions are increased with time. It is worth noting that the staining for these two proteins are much stronger in the PAG‐Mn‐BG scaffold group than in the PAG and the untreated control groups especially after 12 weeks postimplantation. Both proteins of COL I and OCN are positively stained at the subchondral bone region in the group treated with PAG‐Mn‐BG scaffold, with a time‐dependent increase in staining intensity, and remarkably stronger than those in the PAG and the untreated control groups. Furthermore, there is negligible COL I or OCN distributed in neo‐cartilage layer. These above results indicate that the 3D printed PAG‐Mn‐BG scaffold can simultaneously enhance cartilage and subchondral bone repair, benefiting from the force support at early stage and the function of various ions released continuously from the 3D printed PAG‐Mn‐BG scaffold and the degradation characteristics of the 3D printed PAG‐Mn‐BG scaffold. Figure 7 Histological assessment of repaired cartilage subchondral bone at 8 and 12 weeks postsurgery in different groups. A) Hematoxylin and eosin (H&E), toluidine blue (T‐B) staining, and immunohistological staining for Coll II, GAGs, COL I, and OCN, show simultaneously enhanced cartilage and subchondral bone repair in the PAG‐Mn‐BG group, compared to the PAG group and the untreated control (blank) (N: normal cartilage; R: repair cartilage; the arrows indicate the margins of the normal cartilage and repaired cartilage; Scale bar = 200 µm); B) compression destruction tests of the repaired knees at 12 weeks postsurgery in different groups ( # p < 0. 01, compared with the normal knees); C, D) quantification analysis of interleukin‐1 (1L‐1β) and tumor necrosis factor‐α (TNF‐α) in the serum of experimental rats at 1, 2, 3, 4, 8, and 12 weeks after surgery ( # p < 0. 01, compared with 1 week at the same group). High‐resolution scanning electron microscope (SEM) is further used to assess microstructure of the repaired cartilage (Figure S11, Supporting Information). Interestingly, at 12 weeks postsurgery, the microtopography of the repaired cartilage treated with PAG‐Mn‐BG scaffold is smooth and uniform, and has no noticeable difference with the normal cartilage. On the contrary, the surface is rough and even some cracks appear on the repaired tissues in the PAG scaffold group and untreated defect group. Figure 7 B shows the compression destruction test of the repaired knees at 12 weeks postsurgery. For the knees treated with PAG‐BG‐Mn scaffold, the value of maximum failure load is close to that of the normal knees, and significantly higher than those of the knees treated with PAG scaffold and untreated control groups. These results are in accordance with micro‐CT and histological staining analysis results. Inflammatory cytokines in the serum of experimental rats were quantitatively analyzed (Figure 7 C, D). All three groups present an initial increase in cytokines during the acute phase at 1 week after surgery. This is normal acute inflammatory reaction; 3 however, the levels of both inflammatory factors maintained a relatively low level during the repair process from 2 to 12 weeks. Although we did not make a comparison with the healthy normal group, the decrease in the levels of main inflammatory factors indirectly reflected that the inflammatory response was decreased with time. Collectively, the PAG‐Mn‐BG scaffold recapitulating osteochondral architecture and niche can efficiently accelerate the concurrent regeneration of cartilage and subchondral bone. 3 Conclusions Conventional gelatin hydrogels are poor in mechanical properties, precluding their applications as load‐bearing scaffolds. In this study, a novel biodegradable and supramolecular hydrogen bonding strengthened chemical crosslinked gelatin hydrogels were successfully prepared by copolymerization of ACG and GelMA. The cleavable dynamic hydrogen bonds of PACG could considerably strengthen and stiffen the inherently weak GelMA hydrogel to possess a high compressive strength (up to 12. 4 MPa), and compressive modulus (up to 837 kPa), much superior to those of reported GelMA hydrogels (≈200 kPa compressive strength and ≈100 kPa compressive modulus) without undermining its thermosensitive printability, and in turn, the chemical crosslinking of GelMA served to stabilize the intrinsically transient PACG network. The first reported high strength gelatin hydogel can provide a mechanical support in the early stage of osteochondral repair. In mimicking articular cartilage‐subchondral bone architecture, a bilayer biohybrid gradient hydrogel scaffold consisting of top cartilage layer of PACG‐GelMA‐Mn 2+ and bottom bone layer of PACG‐GelMA loaded with bioactive glass was precisely tailored by one‐step thermal‐assisted extrusion printing technique, followed by UV light irradiation to initiate polymerization at a cooled temperature to fix the formed high strength construct. Incorporating BG could improve the proliferation, ALP activities and differentiation of hBMSCs, and loading Mn 2+ facilitated chondrogenic differentiation of the hBMSCs. The resultant biohybrid gradient hydrogel scaffold showed superior performance for accelerating cartilage and subchondral bone repair simultaneously in rat knee osteochondral defect. It is our belief that this 3D printed high‐strength and biodegradable biohybrid gradient hydrogel scaffold can be extended to treatment of other load‐bearing tissue defects. Of course, the present system cannot be directly printed with cells since small molecular ACG monomer was mixed with GelMA to produce a ink. We believe this strategy can be extended to enhance the mechanical strength of other naturally occurring biomacromolecule hydrogels. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201901041
| 2,019
|
Advanced Science
|
Endoscopically Injectable Shear‐Thinning Hydrogels Facilitating Polyp Removal
|
Abstract Submucosal elevation, the process of instilling material in the submucosal space for separation of the surface mucosa and deeper muscularis layer, is a significant aspect of the endoscopic mucosal resection of large lesions performed to facilitate lesion removal and maximize safety. Submucosal injection, when applied, has historically been performed with normal saline, though this is limited by its rapid dissipation; solutions ideally need to be easily injectable, biocompatible, and provide a long‐lasting submucosal cushion with a desirable height. Here, reported is a new set of materials, endoscopically injectable shear‐thinning hydrogels, meeting these requirements because of their biocompatible components and ability to form a solid hydrogel upon injection. These findings are supported by evaluation in a large animal model and ultimately demonstrate the potential of these shear‐thinning hydrogels to serve as efficient submucosal injection fluids for cushion development. Given these unique characteristics, their broad application in mucosal resection techniques is anticipated.
|
1 Introduction Polypectomy remains the single intervention facilitating the interruption of polyp progression toward cancer. Approximately over 15 million colonoscopies are performed annually in the United States with ≈20–25% of these involving polypectomies many of which are performed with the aid of a submucosal injection to facilitate resection 1, 2, 3 Endoscopic mucosal resection is a commonly used minimally invasive technique applied in the removal of large polyps (≥2 cm) and early stage tumors because of its simplicity and safety. 4, 5 This is often assisted through an initial submucosal injection used for the establishment of a cushion between the surface mucosa and muscular tissue layers. Since its description in 1984, 6 normal saline (0. 9 wt% sodium chloride) has been the main injection fluid used for endoscopic mucosal resection. Recently, other fluids including hypertonic saline, hypertonic dextrose water, autologous blood, sodium hyaluronate, glycerol, hyaluronic acid, succinylated gelatin, hydroxypropyl methylcellulose, poloxamer, and fibrinogen have been applied to prolong cushion stability by increasing the viscosity of the fluid. 7, 8, 9, 10, 11, 12, 13 However, the application of these solutions has been largely restricted by unmet safety profiles and durations. Specifically, the heights of cushions elevated by hypertonic saline, dextrose water, and glycerol reduce to less than 50% in 30 min. 14, 15, 16 Additionally, injection solutions showing significantly prolonged duration can be associated with administration challenges. For example, carboxymethylcellulose solutions can require a special 18 gauge submucosal injection needle catheter to minimize injection resistance because of its high viscosity. 17, 18 Moreover, hyaluronic acid potentially stimulates the growth of residual tumor tissues. 19 Fibrinogen and autologous blood are biologic materials which may increase the risk of infection via contamination. 20, 21 Submucosal injection solutions play a critical role for the successful, safe, and intact removal of lesions as they not only lift up diseased mucosa but also provide a gap between the mucosal and deeper layer of tissues facilitating the resection of lesions. Ensuring complete, safe resection stands to mitigate the risk of local recurrence. 22 Hence, an ideal injection solution for submucosal elevation needs to be biocompatible, easily injectable, and provide durable submucosal cushions. One potential set of materials with enhanced biocompatibility and prolonged duration of lift is hydrogels due to their high‐water content and stiffness, which can simultaneously minimize toxicity and resist diffusion. 23 Unfortunately, conventional hydrogels crosslinked by chemical bonds or physical interactions are generally not amenable to injection through an endoscopic needle. 24 In situ formation of hydrogels is an effective approach to solve this problem and has been widely used to facilitate in vivo application of hydrogels. However, the formation of hydrogels by in situ chemical reactions requires the injection of two or more components simultaneously, 25 which is challenging for endoscopic submucosal injection. Additionally, the physical crosslinking of hydrogels such as thermo‐triggered gelation has risk of obstructing the endoscopic needle during injection. 26, 27 Here, we describe the development of endoscopically injectable shear‐thinning hydrogels (EISHs) that can be easily injected through the endoscopic needle and immediately recover their mechanical properties as a solid gel upon deployment in the submucosal compartment. These gels demonstrate their potential to serve as safe and easily injectable agents that can provide durable submucosal cushions in a large animal model. Given their unique characteristics, we believe EISHs could prove useful in endoscopic mucosal resection technique for accurate removal of polyps and early stage tumors. 2 Results and Discussion 2. 1 Design and Preparation of EISHs Shear‐thinning is a term used in rheology to describe non‐Newtonian fluids which demonstrate viscous flow under shear stress and subsequent recovery upon removal of the stress. 28 Recognizing this property, we hypothesized that shear‐thinning hydrogels could serve as a platform for endoscopic injection and cushion formation. We recognized Laponite, a layered nanosilicate with good biocompatibility and biodegradability, which is commonly utilized as a rheology modifier and additive to promote shear‐thinning and thixotropic behavior, as a material candidate for the synthesis of the gels. 29, 30 (Laponite is a trademark of the company BYK Additives Ltd. ) Alginate, an anionic polysaccharide extracted from seaweeds, is biocompatible and its aqueous solutions demonstrate non‐Newtonian fluid behavior with shear‐thinning properties. 31 Here, we use alginate to exfoliate and disperse Laponite nanosheets by mutual repulsion resulting from a possible site‐specific wrapping of their positively charged edge parts 29 with anionic alginate ( Figure 1 a). Laponite was briefly exfoliated in 0. 2 wt% sodium alginate aqueous solution and a transparent EISH was formed immediately after sonication (Figure 1 b). Transmission electron microscopy (TEM) images show that Laponite nanosheets were dispersed homogenously (Figure 1 c). The prepared EISHs could be easily injected through a 25 gauge needle and immediately reformed a solid gel after injection, as shown in Figure 1 d. Figure 1 EISH platform and preparation approach. a) Schematic illustration of exfoliated clay nanosheets (gray segments) by the interaction of their positively charged edges with anionic sodium alginate (red lines). b) Preparation of EISHs by dispersing clay nanosheets in 0. 2 wt% sodium alginate. c) TEM images of the exfoliated clay nanosheets. d) Feasible injection of EISHs through a 25 gauge needle and the immediate reformation of a steady gel after injection. 2. 2 Rheological Properties of EISHs We investigated the effect of Laponite concentration on the rheological behavior of EISHs. Oscillatory measurements indicate that both the storage modulus ( G ′) and loss modulus ( G ″) of EISHs increased with concentration. Time sweep experiments show that the G ′ and G ″ values of EISHs separately increased from ≈10 and ≈100 Pa to ≈70 and ≈1100 Pa with increasing Laponite concentration from 2 to 5 mg mL −1 ( Figure 2 a). Frequency sweep measurements display that EISHs exhibited constant G ′ values approximately 10–19 times higher than G ″ values throughout the frequency sweep from 0. 1 to 100 rad s −1, indicating the formation of stable hydrogels (Figure 2 b). Strain‐dependent oscillatory rheology experiments were further performed to examine the linear viscoelastic range of EISHs. As shown in Figure 2 c, the moduli of EISHs were independent of strain amplitude and showed linear viscoelastic behavior at low strain ranging from 0. 01 to 0. 1. Beyond their critical strains around 0. 1, the G ′ values of EISHs decreased rapidly with the increase of strain, suggesting the gels underwent gel–sol transition and behaved as liquids. The cross point of G ′ and G ″, representing the transition of the gel network to a liquid state (solution behavior: G ′ < G ″, solid behavior: G ′ > G ″), increased from ≈25 to ≈600 Pa with increasing Laponite concentration from 2 to 5 mg mL −1, respectively. These data demonstrate that the rheological behavior and shear‐thinning properties of EISHs can be conveniently tuned by adjusting Laponite concentrations. Figure 2 Rheological properties of EISHs. a) Oscillatory time sweeps of EISHs. Sweeps were performed at 0. 5% strain and 6. 3 rad s −1. b) Oscillatory frequency sweeps of EISHs. Sweeps were performed at 0. 5% strain. c) Oscillatory strain sweeps of EISHs. Sweeps were performed at 6. 3 rad s −1. d) Deformation and recovery of EISHs. Gels evolved over time from repeated cycles of 3 min low 0. 5% strain and 2 min high 500% strain oscillations at 6. 3 rad s −1. G ′ (filled symbols) and G ″ (empty symbols) represent storage and loss modulus, respectively. Step‐strain measurements were performed to verify the reversible gel–sol transition of EISHs. The deformation and recovery of EISHs were conducted at repeated cycles of 3 min low‐magnitude strain of 0. 5% and 2 min high‐magnitude strain of 500% oscillations at 6. 3 rad s −1. After applying alternative low and high strains, we monitored the moduli of EISHs during the strain changes. As shown in Figure 2 d, the gels underwent gel–sol transition and behaved as liquids upon increasing oscillatory strain from 0. 5% to 500%. Inversely, EISHs rapidly underwent sol–gel transition and recovered back to their initial moduli immediately with lowering the strain from 500% to 0. 5%. The gel–sol transition was reversible and all gels were capable of self‐healing to their original state without showing any signs that mechanical fidelity was compromised, irrespective of the number of times they were previously shear‐thinned. These data demonstrate the robust reversibility of the mechanical properties of EISHs. 2. 3 In Vitro Evaluation of EISHs We next studied the injection feasibility of EISHs by utilizing a standard 25 gauge endoscopic needle 32 that is widely used for in vivo submucosal injection in endoscopic procedures ( Figure 3 a). Representative formulations of EISHs with a Laponite concentration of 2 mg mL −1 could be injected as shown in Figure 3 b. The storage modulus of EISHs with Laponite concentrations of 2, 3, and 4 mg mL −1 decreased from its initial G ′ after passing through a 25 gauge needle with an injection speed of 0. 25 mL s −1 to 23%, 31%, and 43%, respectively (Figure 3 c). To elucidate the recovery capability of EISHs, oscillatory time sweep rheology measurements were performed immediately following the injection. As shown in Figure 3 d, the modulus of EISHs with Laponite concentrations of 2, 3, and 4 mg mL −1 increased by 2. 9, 2. 6, and 1. 9 times in 30 min, respectively. These results demonstrate the feasibility of injection of EISHs and their rapid conversion to a solid gel following injection. Figure 3 In vitro evaluation of EISHs. a) A photo of an endoscopic needle. b) Fluent injection of EISHs via the endoscopic needle. c) Storage modulus of EISHs before and after injection by a 25 gauge needle at an injection speed of 0. 25 mL s −1. d) Gel restoration kinetics of EISHs subjected to shear force induced by syringe injection ( G ′ of 2 mg mL −1 at 0 min was 28 Pa and at 30 min was 79 Pa, G ′ of 3 mg mL −1 at 0 min was 153 Pa and at 30 min was 391 Pa, and G ′ of 4 mg mL −1 at 0 min was 561 Pa and at 30 min was 1064). e) Gel erosion kinetics of EISH cushions in saline at 37 °C. The error bars show standard deviation ( n = 3). We then evaluated the stability of EISHs by measuring their erosion kinetics in a physiological environment. A volume of 0. 5 mL of EISHs was injected in saline and further incubated at 37 °C for predetermined time intervals. The volume of remaining gels at each time point was recorded to calculate the erosion kinetics of EISHs. As shown in Figure 3 e, the mass of EISHs with Laponite concentration of 2 mg mL −1 remained constant within 1. 5 h. While the mass of the gels decreased to 40% with further prolonging of incubation time to 2 h, which could be explained by the passive diffusion of both Laponite and alginate. However, EISHs with a higher Laponite concentration of 3 mg mL −1 maintained their mass up to 2 h. Interestingly, EISHs with a high concentration of 4 mg mL −1 swelled gradually and reached 1. 4 times of their initial mass after 2 h incubation. We speculate that the dispersion of a high Laponite content of 4 mg mL −1 in alginate aqueous solution forms steady hydrogels that can promote their water absorption. These erosion profiles suggest the potential of EISHs to resist passive diffusion and achieve relative long‐term submucosal cushions. 2. 4 Endoscopic Development of Submucosal Cushions Having confirmed the feasible injection and rapid recovery as well as high stability of EISHs, we then tested their performances for cushion development in vivo. Yorkshire pigs weighing 40–80 kg were used as a large animal model and endoscopic injection was utilized to develop submucosal cushions in the colon. As displayed in Figure 4 a, b, a clear cushion was easily formed by submucosal injection of 1. 5 cc of EISHs (2 mg mL −1 ) through an endoscopic needle. Four different injections were performed and each time a well‐formed cushion was observed. Additionally, endoscopic videography was used to observe the duration of cushions formed by EISHs. We found that the cushions created by normal saline flattened dramatically within 1 min (Figure 4 c, d), while the cushions produced by EISHs remained almost unchanged for up to 3. 5 min (Figure 4 e, f), showing the prolonged duration of cushions developed by these gels. Figure 4 Endoscopic development of submucosal cushions. a, b) Pig colon before (a) and after (b) submucosal injection of 1. 5 cc of 2 mg mL −1 EISH. c–f) Endoscopic images of the submucosal cushions developed by saline (c, d) and 2 mg mL −1 EISH (e, f) at different post‐injection time points, showing the significantly prolonged duration of the cushions developed by EISHs. To accurately evaluate the duration of cushions, we measured their heights by a digital caliper through a midline laparotomy during a terminal procedure. 2 cc of EISHs with different Laponite concentrations were submucosally injected in pig colon and the heights of the resulted cushions were measured at predetermined time intervals. Normal saline was used as a control. As observed in Figure 5 a–d, with the increase of incubation time, the heights of cushions created by EISHs decreased at a far slower rate than that of saline. At 20 min incubation, the height of cushions developed by saline decreased dramatically to less than 50%, while the heights of all cushions created by EISHs with different concentrations from 1 to 3 mg mL −1 remained as high as 82% to 91% (Figure 5 e). Even with incubation time prolonging to 2 h, the heights of cushions elevated by EISHs with a low Laponite concentration of 1 mg mL −1 remained around 50%. Notably, the cushions developed by EISHs with a Laponite concentration of 3 mg mL −1 maintained up to 68% of their initial heights after 2 h incubation, demonstrating the significantly prolonged durations. We further investigated the effect of injection volume on the duration of cushions. As illustrated in Figure 5 f, with increasing injection volume from 1 to 3 mL, no obvious differences were observed between the heights of cushions produced by EISHs with a Laponite concentration of 2 mg mL −1. We also evaluated the durations of cushions elevated in pig small intestine. As expected, their durations were quite similar to those created in pig colon (Figure 5 g), showing the capabilities of EISHs to develop durable submucosal cushions at different sites. Figure 5 In vivo submucosal cushion duration. a–d) Photographs of submucosal cushions in pig colon post‐injection at: a) 0, b) 30, c) 60, and d) 120 min, respectively. The cushions were lifted by 2 cc of saline (top right), 1 mg mL −1 EISH (bottom right), 2 mg mL −1 EISH (bottom left), and 3 mg mL −1 EISH (top right), respectively. e) Duration of cushions lifted by 2 cc of EISHs with different concentrations in pig colon ( t = 20 min, p < 0. 05). No further changes observed in saline injection site. f) Relationship between time post‐injection and height of cushions developed by 2 mg mL −1 EISHs with different volumes (no significance). g) Duration of cushions lifted by 2 cc of EISHs with different concentrations in small intestine ( t = 15 min, p < 0. 05). 2 cc of saline solution was injected as a control. No further changes observed in saline injection site. The error bars show standard deviation ( n = 3). Significance was assessed using Student's t ‐test. 2. 5 Local Toxicity of EISHs To complete our evaluation of the advantages of EISHs as submucosal injection agents, we evaluated their local toxicity by histological analysis. 33 Hematoxylin and eosin (H&E) staining was used to evaluate the toxicity of EISHs against the tissue of pig colon in vivo. 3 cc of EISHs with a Laponite concentration of 3 mg mL −1 was submucosally injected submucosally in the colon of a sedated pig and normal saline was used as a control. At 2 h post‐injection, the pig was euthanized, and the tissues were immediately harvested, fixed by formalin, and further embedded by paraffin. The resultant tissues were then sectioned and stained by H&E for microscopic imaging. As shown in Figure 6 a–c, no significant difference was observed between the tissues treated by EISHs and the control tissues injected with normal saline. Similar results were obtained by incubation of EISHs placed on the top of the mucus for 2 h (Figure 6 d–f), supporting the low local toxicity of these gels using as cushion development agents. Further evaluation of local effects as well as long‐term impact on adjacent regions including draining lymph nodes will be required for successful future human translation. Figure 6 In vivo toxicity. a) H&E staining images of pig colon without any treatment, b) submucosally injected with saline, and c) 3 cc of 3 mg mL −1 EISH, respectively. The tissues were harvested at 2 h post‐injection. d) H&E staining images of pig colon without any treatment, e) incubated with saline, and f) 3 cc of 3 mg mL −1 EISH on the top of the mucus for 2 h, respectively. 3 Conclusions In summary, we report the development and application of shear‐thinning hydrogels as safe and endoscopically injectable solutions capable of establishing durable submucosal cushions. We show that these shear‐thinning hydrogels can be rapidly prepared by dispersing commercially available Laponite into an aqueous solution of alginate and their rheological properties can be easily tuned by varying the concentrations of Laponite. We also show that these hydrogels can be injected through a standard endoscopic needle and further demonstrate their low toxicity as well as the significantly enhanced durations of cushions elevated by these gels. In sum, the hydrogel materials developed herein present 1) commercially available and inexpensive resources; 2) tunable shear‐thinning properties and endoscopically injectable capability; 3) good biocompatibility and significantly improved stability for the development of durable submucosal cushions. All these features make EISHs a promising set of hydrogel materials for broad application in mucosal resection techniques and potentially luminal constriction, drug delivery, and tissue engineering. 4 Experimental Section Materials : Sodium alginate, Laponite, Indigo carmine, methylene blue, and other chemical reagents were purchased from Sigma and used as received unless otherwise noted. Nanopure water (18 MΩ cm) was acquired by means of a Milli‐Q water filtration system, Millipore (St. Charles). TEM Measurements : TEM experiments were carried out on a JEOL 2100 FEG instrument at an acceleration voltage of 200 kV. The TEM sample was prepared by dropping the exfoliated Laponite solutions onto a 300‐mEISH carbon‐coated copper grid. Samples were blotted away after 30 min incubation at the room temperature and then washed twice with distilled water and air dried prior to imaging. Preparation of EISHs : 0. 2% sodium alginate aqueous solution was prepared as stock solution. Laponite was added into the stock solution with various concentrations and then sonicated for ≈2–5 min to obtain EISHs. EISHs with Laponite concentrations of 2, 3, 4, and 5 mg mL −1 were prepared accordingly and used directly for further measurements. Measurements of the Rheological Properties of EISHs : Dynamic oscillatory time, frequency, and strain sweeps were performed using an AR2000 stress‐controlled rheometer (TA Instruments, New Castle, DE) with 25 mm steel plate geometry at a 27 mm gap distance. Laponite was dispersed in 0. 2 wt% alginate solution by sonication to form EISHs with specified compositions and the gels were applied between the two plates of the rheometer. The top plate was lowered to a 27 mm gap distance and excess gel was scraped off. Care was taken to achieve a homogenous distribution of gel within the top and bottom plates of the rheometer. Dynamic oscillatory time sweeps were collected at angular frequencies of 6. 3 rad s −1 and 0. 5% strain. An initial strain amplitude sweep was performed at 25 °C at different frequencies to determine the linear viscoelastic range for the gels. Rheological properties were examined by frequency sweep experiments at fixed strain amplitude of 0. 5%. Experiments were repeated on three to four samples and representative data were presented. For shear recovery experiments at 6. 3 rad s −1, shear thinning was induced via application of 500% strain for 2 min. The strain was released to 0. 5% for 3 min to allow the gel to recover. Erosion Studies of EISHs : The erosion kinetics of the EISHs was measured in a physiological environment. A volume of 0. 5 mL of EISHs was injected in saline and further incubated at 37 °C for 30, 60, 90, and 120 min, respectively. The volume of remaining gels at each time point was recorded to calculate the erosion kinetics of EISHs. Ex Vivo Cushion Development in Pig Colon : Ex vivo cushion development was performed by injection of 0. 5 cc EISHs (2 mg mL −1 ) into the pig colon. The colon tissue was isolated from freshly procured intact gastrointestinal tracts from pigs from selected local slaughter houses. The top view and the side view of the developed cushions were shown in Figure S1 in the Supporting Information. In Vivo Cushion Development in a Pig Model : All pig experiments were approved by the Committee on Animal Care at the Massachusetts Institute of Technology. Female Yorkshire pigs (40–80 kg) were obtained from Tufts University and housed under conventional conditions. Animals were randomly selected for the experiments. The animals were placed on a liquid diet for 24 h prior to the experiment with the morning feed held on the day of the experiment. At the time of the experiment, the pigs were anesthetized with intramuscular administration of Telazol (tiletamine/zolazepam, 5 mg kg −1 ), xylazine (2 mg kg −1 ), and atropine (0. 04 mg kg −1 ). An endoscope (Pentax, US endoscopy) was inserted into the distal colon and a Carr–Locke needle was inserted through the channel of the endoscope into the colon. Subsequently, 1. 5 mL of saline and hydrogel were separately injected into the submucosal space, repeated three times. Videos were recorded to monitor the decrease of the size of the cushions lift. All animals were recovered from anesthesia. Measurements of In Vivo Cushion Duration : All procedures were conducted in accordance with protocols approved by the Massachusetts Institute of Technology Committee on Animal Care. Female Yorkshire swine, approximately 40–80 kg in body weight were anesthetized with intramuscular administration of Telazol (tiletamine/zolazepam, 5 mg kg −1 ), xylazine (2 mg kg −1 ), and atropine (0. 04 mg kg −1 ). Animals were intubated and maintained on 2–3% isoflurane in oxygen. As part of a terminal or nonsurvival procedure, a midline laparotomy was performed and the proximal jejunum or distal colon was accessed and stabilized with gauze. A longitudinal incision was made to access the luminal side and 2 cc normal saline solution and 1 mg mL −1 EISH, 2 mg mL −1 EISH, and 3 mg mL −1 EISH were injected into the submucosal space to form the cushions. The length, width, and the height of the cushions were measured at 0, 30, 60, and 120 min after the injection. 1, 2, and 3 mL 2 mg mL −1 EISHs were also injected to investigate the cushion properties. Animal were euthanized prior to anesthetic recovery with intravenous administration of 120 mg kg −1 of sodium pentobarbital. H&E Staining : The toxicity of EISHs was evaluated during an in vivo terminal experiment. All procedures were conducted in accordance with protocols approved by the Massachusetts Institute of Technology Committee on Animal Care. Pigs were intubated and maintained on 2–3% isoflurane in oxygen. A midline laparotomy was performed and the proximal jejunum was accessed and stabilized with gauze. 3 cc normal saline solution and 3 mg mL −1 EISH were submucosally injected to the pig colon to form the cushions. Meantime, multiple 4–5 cm incisions were made along the antimesenteric side of the colon. 3 cc normal saline solution and 3 mg mL −1 EISH were incubated on the top of the mucus using wells secured with carbopol and covered with an adhesive membrane. The pigs were euthanized with sodium pentobarbital (120 mg kg −1 ) intravenously prior to tissue collection. Tissues were harvested and placed into formalin (4%). After tissues were fixed in formalin, they were paraffin embedded, sectioned, and stained with H&E for analysis. Conflict of Interest Y. P. , J. L. , R. L. and G. T. are co‐inventors on a provisional patent application encompassing the technology described in this manuscript. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201901146
| 2,019
|
Advanced Science
|
3D Printing of Hot Dog‐Like Biomaterials with Hierarchical Architecture and Distinct Bioactivity
|
Abstract Hierarchical structure has exhibited an important influence in the fields of supercapacitors, catalytic applications, and tissue engineering. The hot dog, a popular food, is composed of bread and sausage with special structures. In this study, inspired by the structure of a hot dog, the strategy of combining direct ink writing 3D printing with bidirectional freezing is devised to prepare hot dog‐like scaffolds with hierarchical structure. The scaffolds are composed of hollow bioceramic tubes (mimicking the “bread” in hot dogs, pore size: ≈1 mm) embedded by bioceramic rods (mimicking the “sausage” in hot dogs, diameter: ≈500 µm) and the sausage‐like bioceramic rods possess uniformly aligned lamellar micropores (lamellar pore size: ≈30 µm). By mimicking the functions of hierarchical structure of bone tissues for transporting and storing nutrients, the prepared hot dog‐like scaffolds show excellent properties for loading and releasing drugs and proteins as well as for improving the delivery and differentiation of tissue cells. The in vivo study further demonstrates that both the hierarchical structure itself and the controlled drug delivery in hot dog‐like scaffolds significantly contribute to the improved bone‐forming bioactivity. This study suggests that the prepared hot dog‐like scaffolds are a promising biomaterial for drug delivery, tissue engineering, and regenerative medicine.
|
Hierarchical structure design is widely studied in various fields including supercapacitor, 1 catalytic applications, 2 and tissue engineering. 3 In the past decades, it is found that the micro/nanoscale hierarchical structures have great influences on material properties. 4 Thus, designing materials with specific micro–nanostructures has become a critical part in improving their diverse function and application. As an important multimaterials and multifunctional fabrication technology, direct ink writing (DIW) plays an important role in structure design. 5 However, it is difficult to endow DIW 3D printing of materials with specific micro/nanostructure, due to the complexity and resolution limitation of printing nozzle and controlled system of printer instruments. 6, 7 Large bone defect is still a thorny challenge in clinic because of the limited self‐repairing ability of bone tissues. 8 Recently, implantation of 3D‐printed scaffolds is considered as an effective approach for stimulating bone regeneration. 9 Human bones possess hierarchical structures from micrometer scale to nanometer scale, 7, 10 which plays important roles in the transportation and storage of the nutrients and tissue cells. 11 However, the conventional 3D‐printed scaffolds are composed of the stacked solid struts, but lack of hierarchical structures, limiting the delivery of nutrition and tissue cells, and further affecting the tissue formation in the inner of large bone defects. 12 Hot dog, a common food constituted by bread and sausage which supply the energy and nutrition, respectively. In this study, inspired by the structure and function of hot dog, hot dog‐like scaffolds with hierarchical structure are successfully fabricated by combining DIW 3D printing and bidirectional freezing. The scaffolds comprise macroporous hollow tubes (bread) embedded by bioceramic rods (sausage) with uniformly aligned lamellar microstructures. The hot dog‐like structure could significantly enhance the specific surface areas of scaffolds and efficiently promote the cells adhesion inside; In addition, the sausage of hot dog can supply the nutrition source. Meantime, the interior hierarchical rods in this study can help the delivery of osteogenic drugs, which further contributes to the cell differentiation and bone formation. Therefore, it is reasonable to prepare hot dog‐like scaffolds from both structure and function points of view. One of most interesting results of the study is that we successfully prepared hot dog‐like bioceramic scaffolds. Figure 1 a illustrated the typical fabrication process of hot dog‐like scaffolds by using the strategies of combined 3D printing and bidirectional freezing. Hollow tube bioceramic scaffolds were first prepared by 3D printing with modified nozzle (Figure S1a, Supporting Information). Then, the prepared scaffolds were put into the well‐dispersed bioceramic slurry for bidirectional freezing. During the bidirectional freezing process, lamellar ice crystals grew perpendicular to the steel plate due to the dual temperature gradients, while the bioceramic particles were squeezed to the position between two adjacent lamellar ice crystals (Figure S1b, Supporting Information). Subsequently, the ice crystals in the scaffolds were sublimated out through freeze‐drying, and the hot dog‐like scaffolds were finally achieved after being sintered. Figure 1 Fabrication and morphology of hot dog‐like scaffolds (HD‐AKT). a) The schemata of preparation of HD‐AKT, combining 3D printing and bidirectional freezing to realize 3D printed scaffolds containing rods with aligned lamellar microstructure. b–e) 3D micro‐CT images and f, g) 2D micro‐CT images of HD‐AKT from different views showing the scaffold structure, which is hollow tube macrospores embedded by bioceramic rods with the uniformly aligned lamellar structure of rods. The hierarchical structure of hot dog‐like scaffolds was observed by the scanning electron microscopy (SEM). As shown in Figure 1 a, the bioceramic rods with lamellar microstructure were embedded in the inner of the macroporous hollow tube of scaffolds. Micro computed tomography (Micro‐CT) images from different views showed that the structure of bioceramic rods inside the scaffolds was composed of large‐scale micro‐aligned lamellar framework (Figure 1 b–g). Therefore, the hot dog‐like scaffolds possessed macroporous hollow tubes (bread) and microporous rods (sausage) with aligned lamellar framework. Through traditional freeze‐casting, the microstructures and properties of materials could be controlled by regulating freezing temperature or slurry concentration. 13 To well control the lamellar microstructure of the bioceramic rods in the scaffolds, we fabricated several kinds of scaffolds with different slurry concentration. In this study, akermanite (AKT, Ca 2 MgSi 2 O 7 ), a representative silicate‐based bioceramic, was selected to fabricate the hot dog‐like scaffolds due to its excellent cytocompatibility. 14 Hot dog‐like AKT(HD‐AKT) scaffolds with slurry concentrations: 20%, 30%, 40%, and 50% were fabricated and named as HD‐20AKT, HD‐30AKT, HD‐40AKT, and HD‐50AKT, respectively. To compare with hot dog‐like scaffolds, traditional solid struts AKT (S‐AKT) scaffolds and hollow tube AKT (H‐AKT) scaffolds without hot dog structure were fabricated for the controls. Figure 2 a showed the cross‐sectional morphologies of HD‐AKT scaffolds. The diameter of rods in HD‐AKT scaffolds increased with the increment of the concentrations of AKT slurry during the bidirectional freezing process. Meanwhile, the lamellar distance in the bioceramic rods decreased from 34 ± 1. 0 to 18. 5 ± 0. 9 µm (Figure 2 b), while the thickness of ceramic layers increased from 17. 1 ± 1. 3 to 66. 8 ± 1. 5 µm (Figure 2 c). The porosities of the scaffolds obviously decreased from 43 ± 3. 06% to 27. 6 ± 1. 32% with the increase of slurry concentrations (Figure 2 d). By using the two‐step strategy, the hot dog‐like scaffolds with various materials including AKT (tubes)‐AKT (rods), Nagel (tubes)‐Nagel (rods), TCP (tubes)‐AKT (rods), and AKT (tubes)‐GO (rods) could be well prepared, suggesting the universality of the strategy (Figure 2 e). Due to the advantage of asynchronous sintering, the sintering shrinkage of rods occurred in the second step, leaving the tube (bread) and rod (sausage) partly separated. It was observed from the micro‐CT images that some parts of the rod are sintered together with the inner of tube, showing the stability of the structure (Figure S2a, b, Supporting Information). Figure 2 The hot dog‐like scaffolds are composed of hollow bioceramic tubes (mimicking the “bread” in hot dog) embedded by bioceramic rods (mimicking the “sausage” in hot dog) and the sausage‐like bioceramic rods possess uniformly aligned lamellar micropores. a) Morphology control of AKT scaffolds. Different structures of hot dog‐like scaffolds are obtained by changing the ratio of AKT in the freezing casting slurry from 20% to 50%, named for HD‐20AKT, HD‐30AKT, HD‐40AKT, and HD‐50AKT, respectively. Traditional 3D printed‐solid struts AKT scaffolds and 3D printed‐hollow tube scaffolds are named as S‐AKT and H‐AKT. The optical images (first line) and the SEM images show rods' lamellar structure of HD‐AKT. b) The characterizations of scaffolds: the distance of layers, c) the thickness of layers, and d) the scaffolds' porosity. e) Hot dog‐liked 3D printed scaffolds with different materials: AKT (tube)‐AKT (rods), Nagel (tube)‐Nagel (rods), TCP (tube)‐AKT (rods), and AKT (tube)‐GO (rods). The second interesting result of this study is that the prepared hog dog‐like bioceramic scaffolds could be used for drug/protein delivery due to their hierarchical microstructure. Previous studies showed that the microstructure characteristics of biomaterials play critical roles for providing absorption sites of drug and protein, and further contribute to modulating their delivery. 15 By mimicking the function of nutrition supply for sausages in hot dog, the potential of the scaffolds for loading icariin (Ica, a model osteogenic drug 16 ), was investigated in this study. As shown in Figure 3 a, b, the loading efficiency and capacity of the Ica could be well controlled through changing the lamellar microstructure of bioceramic rods of HD‐AKT. HD‐30AKT and HD‐40AKT possessed the highest Ica loading efficiency up to 7. 5% and 8 mg g −1 (mass ratio for drug/scaffold). However, the Ica loading efficiency and capacity of S‐AKT were much lower than that of HD‐AKT, indicating the excellent loading capacity of the hierarchical hot dog‐like scaffolds. The thermogravimetric analysis further verified the significantly improved loading efficiency of HD‐AKT scaffolds as compared to those without hot dog microstructure (Figure 3 c). Figure 3 Hot dog‐like scaffolds are an excellent carrier for drug and protein. The drug Ica and protein BSA loading and release properties of the HD‐AKT. a) The Ica loading efficiency, and b) the loading capacity of traditional solid struts scaffolds(S‐AKT), H‐AKT, and different kinds of HD‐AKT. c) Thermogravimetric analysis of the scaffolds after loading the Ica. d) The Ica release of different kinds of scaffolds after loading Ica (S‐AKT/Ica, H‐AKT/Ica, HD‐20AKT/Ica, HD‐30AKT/Ica, HD‐40AKT/Ica, and HD‐50AKT/Ica). e) The BSA loading efficiency. f) The BSA release of the scaffolds. g–i) SEM images of the hot dog rod of scaffolds g) after 90 d Ica release. h) Surface images and i) the cross‐section image show lots of mineralized hierarchical structures on the surface of hot dog rod of scaffolds. Compared with the S‐AKT and H‐AKT, the HD‐AKT have higher loading efficiency and capacity. Meantime, HD‐AKT possess longer Ica/BSA release time. Furthermore, the HD‐AKT scaffolds could maintain a sustained release of Ica for over 90 d, while Ica in S‐AKT and H‐AKT scaffolds completely released no more than 20 d, indicating that the hierarchical structure of HD‐AKT scaffolds plays a key role for maintaining the sustained release of drug. As compared with the S‐AKT and H‐AKT, the HD‐AKT scaffolds could provide more drug absorption sites due to the existence of lamellar structures which increase the specific surface area of HD‐AKT scaffolds (Table S1, Supporting Information). In addition, it was found that plenty of mineralized calcium phosphate microcrystals formed on the surface of the rods, which could contribute to the slow release of the Ica from scaffolds (Figure 3 g–i and Figure S4, Supporting Information). 17 Not only the small molecule drug Ica but also the large molecule protein bull serum albumin (BSA) could be well delivered by HD‐AKT scaffolds. Similar with Ica loading, HD‐AKT exhibited higher loading efficiency of BSA than H‐AKT and S‐AKT (Figure 3 e). In addition, HD‐AKT scaffolds displayed longer period of BSA release time than other scaffolds (Figure 3 f). The third interesting result of the study is that the prepared hot‐dog like scaffolds possess excellent bioactivity both in vitro and in vivo. To explore the effect of hot dog structure of HD‐AKT scaffolds on the cell delivery and osteogenic differentiation of rabbit bone mesenchymal stem cells (rBMSCs), the cells were cultured on the S‐AKT, H‐AKT, HD‐30AKT, HD‐30AKT/Ica, respectively. As shown in Figure 4 a–d, rBMSCs were well attached to the scaffolds after 3 d of culture. The rBMSCs could be delivered into the inside of the hierarchical struts of HD‐30AKT and HD‐30AKT/Ica; however, the cells were only attached on the outside surface of solid struts of S‐AKT (Figure 4 a 2 –d 2 ). Interestingly, there were lots of cells attached on the surface of lamellar rods in the HD‐30AKT from longitudinal section view (Figure 4 e). Cell proliferation assays revealed that rBMSCs in HD‐AKT and HD‐30AKT/Ica scaffolds with hot dog structure exhibited better activity than those in S‐AKT and H‐AKT indicating that the hierarchical structures of hot dog‐like scaffolds could be beneficial for the proliferation of rBMSCs, due to the higher specific surface area of lamellar rods which surfaces were conductive to the adhesion of cells (Figure 4 f). Moreover, HD‐30AKT and HD‐30AKT/Ica significantly enhanced the bone‐related gene expressions of rBMSCs (Figure 4 g–j). Figure 4 g showed that HD‐30AKT scaffolds significantly promoted the expression of runt‐related transcription factor 2 (Runx2) compared with H‐AKT scaffolds, which might be explained that more cells delivered on the rods might accelerate the gap junctions and regulated the Runx2 furthermore. 18 Due to the effective release of Ica from the lamellar rods in the scaffolds, bone‐related gene expressions osteocalcin (OCN), osteopontin (OPN), alkaline phosphatase (ALP) were essentially increased as compared to other groups without Ica delivery (Figure 4 h–j), suggesting that the prepared scaffolds are quite useful platform for drug delivery with significantly improved cell function toward to osteogenic differentiation for bone tissue regeneration. Figure 4 Hot dog‐like scaffolds are an excellent platform for cell delivery and differentiation. The proliferation, morphology, and relative genes expression of rBMSCs cultured on different scaffolds. a‐d) Confocal and SEM images of rBMSCs cultured on a) S‐AKT, b) H‐AKT, c) HD‐30AKT, and d) HD‐30AKT/Ica. e) The rBMSCs (red color) adhesion on the rod of HD‐AKT. f) The proliferation of rBMSCs after seeding in different kinds of scaffolds, showing hot dog‐like scaffolds are beneficial for cell proliferation. g) The relative osteogenic genes expression (OCN, Runx2, OPN, and ALP) of rBMSCs in scaffolds, indicating that the Ica release from the scaffolds promotes the relative osteogenic genes expression of rBMSCs ( n = 6, * P < 0. 05, and ** P < 0. 01. ). To further investigate in vivo bone‐forming bioactivity, S‐AKT, H‐AKT, HD‐30AKT, and HD‐30AKT/Ica were implanted into rabbit femoral defect for 8 weeks. Then, the overall photographs and micro‐CT analysis of femoral defect samples were displayed in Figure 5 a. The results demonstrated no inflammatory reaction in the defects (Figure 5 a 1 –d 1 ). Meantime, the newly formed bone tissues had grown into the scaffold tubes of H‐AKT, HD‐AKT, and HD‐AKT/Ica from the micro‐CT images of transverse view as indicated in Figure 5 a 2 –d 2 and a 3 –d 3. The sagittal view of micro‐CT showed that only the parts near the tissue had formed new bones for H‐AKT (Figure 5 b 4 ). However, considerable amount new bones were formed in the induction of HD‐AKT and HD‐AKT/Ica scaffolds (Figure 5 c 4, d 4 ). Micro‐CT reconstruction analysis showed that HD‐30AKT and HD‐30AKT/Ica scaffolds had higher the volume ratio of the new bone to the original defect regions (BV/TV) as compared to Blank, indicating superior bone repair ability of HD‐30AKT and HD‐30AKT/Ica (Figure 5 e). Interestingly, HD‐AKT/Ica were conductive to better repair bone defects as compared with HD‐AKT, verifying that the hierarchical hot dog‐like scaffolds with Ica delivery effectively improve bone‐forming bioactivity. As exhibited in Figure 5 g of bone staining images, there were few new bones in the Control group without the scaffolds. However, plenty of new bones (red) emerged and grew along the materials with the inducement of scaffolds (black) in Figure 5 h–j. In addition, bone tissues were found near the rods of HD‐AKT scaffolds (blue circle) in Figure 5 i, j compared with H‐AKT of Figure 5 h. The results illustrated that the hot dog‐like scaffolds could significantly promote the osteogenesis by inducing the new bone to grow into the hierarchical structure rods of the scaffolds. Moreover, HD‐AKT/Ica (Figure 5 j) exhibited more new bone than HD‐AKT (Figure 5 i). Furthermore, the newly formed bones were observed to grow in the interior of the lamellar microstructures of rod (green arrows) from the stained images (Figure 5 f), due to more nutrition and cell absorption sites provided from the hierarchical rods. Therefore, both hierarchical structure and the Ica delivery of the hot dog‐like scaffolds play a significant role for the regeneration of bone defects. Figure 5 Hot dog‐like scaffolds possess excellent bone‐forming bioactivity after implanted in the femoral defects of rabbits. The characterizations of hot dog‐like scaffolds for osteogenesis in vivo. a 1 –d 1 ) Digital, a 2 –d 2 ) 2D micro‐CT, and 3D micro‐CT images (a 3 –d 3 : transverse view, and a 4 –d 4 : sagittal view) of the defects at week 8. In 3D micro‐CT images, green, red and white represents new bone, scaffold, and primary bone, respectively. e) Micro‐CT reconstruction analysis exhibits the volume ratio of the new bone to the original defect regions (BV/TV) at week 8. HD‐30AKT and HD‐30AKT/Ica indicate significantly. f) The newly formed bones (red) grow into the lamellar microstructures (green arrows) of hot dog‐like scaffold. g–j) Hard histological sections stained with Van Gieson's picrofuchsin of g) Blank, h) H‐AKT, i) HD‐AKT, and j) HD‐AKT/Ica, red color stands newly formed bone and black color represents scaffolds. Newly formed bone can grow into the hierarchical rods of the scaffolds (blue circle point to the new bone in the hierarchical rods). improvement in new bone regeneration as compared to Blank control. In addition, Ica can also promote osteogenesis, suggesting that both hierarchical structure and the Ica delivery of the hot dog‐like scaffolds contribute to the bone regeneration ( n = 6, * P < 0. 05, ** P < 0. 01, and *** P < 0. 001. ). In summary, inspired by the constitution, structure and function of hot dog, we successfully fabricated the hierarchical hot dog‐like scaffolds which consisted of hollow tube embedded by bioceramic rods with uniformly aligned lamellar microstructures by combining DIW 3D printing with bidirectional freezing. The prepared hot dog‐like scaffolds exhibited hierarchical microstructure with improved specific surface area, which significantly enhanced the drug/protein delivery, cell proliferation, and osteogenic differentiation of rBMSCs. Due to the hierarchical lamellar microstructures and Ica delivery in scaffolds, the hot dog‐like scaffolds could significantly promote the formation of new bone tissue. Our study suggests that the hot dog‐like scaffolds can be used for the multifunctional biomaterials for drug delivery, tissue engineering, and regenerative medicine. The combined strategy of DIW 3D printing with bidirectional freezing is a promising method to prepare biomimetic and hierarchical biomaterials. Experimental Section Preparation of Hollow Tube Scaffolds by 3D Printing : The printing ink was prepared by mingling 2. 0 wt% sodium alginate (Alfa Aesar), 20 wt% Pluronic F‐127 (Sigma–Aldrich) polymer solution, bioceramic powder AKT (Ca 2 MgSi 2 O 7 ), which was synthesized according to our previous publication. 19 Modified printing nozzle was designed with a needle embedded in the nozzle (Figure S1a, Supporting Information) to prepare the hollow tube scaffolds by 3D printing based on our previous work. 20 After drying in the room temperature overnight, the hollow tube AKT scaffolds were sintered at 1350 °C for 3 h. The traditional solid struts AKT scaffolds without hollow tubes were printed by conventional nozzle as the control group. Other bioceramic powders with different composition (e. g. , β‐TCP, Nagel) were synthesized by the published methods. 21 Preparation of Hot Dog‐Like Scaffolds : Hot dog‐like scaffolds were successfully fabricated by ice crystals growth into the hollow tube scaffolds in the bidirectional temperature gradients. The slurry was first prepared through mixing the AKT powders (20, 30, 40, and 50 vol%) with 2 wt% of polyvinyl alcohol, assisting 1 wt% of sodium polyacrylate as the dispersant, and then put into the vacuum oven for 15 min to remove the gas. Subsequently, the hollow tube scaffolds were placed into the mold in the direction of vertical to the steel as shown in (Figure S1b, Supporting Information) before adding the slurry into the mold. Dual temperature gradients, along and perpendicular to the steel plate, could be generated after the liquid nitrogen was poured into the box, so that the ice crystals would grow into the tubes. Further, the samples were freezing dried for 48 h to sublimate out the ice crystals and then the outside of the scaffolds can be removed easily. Ultimately, the hot dog‐like scaffolds were obtained after sintered at 1350 °C for 3 h. Characterization of the Hot Dog‐Like Scaffolds : The scaffolds macroscopic morphologies were taken by optical microscopy (S6D, Leica, Germany). The microstructures were observed by micro‐CT (SKYSCAN1172, SKYSCAN, Belgium) and SEM (JSM‐6700F, Japan). The thickness of the layers and the distance between the layers were manually measured on basis of the SEM images. Ten measurements were performed at least for each parameter. The porosity was measured through Archimede method. In brief, the scaffolds were dried at 120 °C for 10 h first, weighed, and marked as M 1. After that, the scaffolds were immersed in water and placed under vacuum for more than 2 min until no bubbles came up. The scaffolds with water‐filled pores were weighed and marked as M 2. Finally, the buoyant weight of scaffolds was marked as M 3. The porosity (P) was calculated by following formula (1). (1) P = M 2 − M 1 / M 2 − M 3 × 100 % Characterization on the Ica and BSA Loading and Release of the Hot Dog‐Like Scaffolds : For the Ica and BSA loading, the Ica absorption wavelength of 365 nm was acquired with UV/vis (Figure S2a, Supporting Information). In addition, the absorption standard curve was measured through Microplate Reader (Epoch, BIO‐TEK, USA, Figure S2b, Supporting Information). The scaffolds were soaked in Ica solution (50 mg mL −1 ) and BSA solution (20 mg mL −1 ) respectively for 3 d at 37 °C. Then the scaffolds were taken out and concentrations of the remaining solution were measured. To further test the amount of the release of Ica and BSA in the scaffolds, the samples were soaked in phosphate buffered saline (PBS) again and the medium was collected for concentration analysis using microplate reader and replaced with fresh PBS solution immediately at certain time interval. The loading efficiency, loading capability and released ability of Ica and BSA were calculated as following Equations (2) and (3). (2) Loading efficiency % = C 0 − C 1 / C 0 × 100 % (3) Loading capability % = m 1 / m 2 × 100 % where C 0 is the concentration of Ica and BSA before loading and C 1 refers to the concentration of Ica and BSA after loading, and m 1 is the loading mass of Ica and BSA, in addition, m 2 refers to the mass of the scaffolds. In Vitro Bioactivity of the Hot Dog‐Like Scaffolds : The rabbit bone marrow stem cells (rBMSCs) were cultured in Dulbecco's Modified Eagle's Medium (DMEM, HyClone, China) supplemented with 10% fetal bovine serum (Invitrogen), penicillin‐streptomycin solution (Invitrogen). For the research of cell adhesion, rBMSCs were seeded on the scaffolds with the cell amount of 3 × 10 4 in 48‐well culture plates. After incubation for 3 d, cells were fixed on the scaffolds by using 2. 5% glutaraldehyde. Cell dehydration was accomplished through soaking in graded ethanol (30, 50, 70, 90, 95, and 100 v/v%). Before SEM analysis, the samples were dried by Hexamethyldisilazane. For better observation of the cell distribution, cells were fixed with 4% paraformaldehyde solution and stained with 4, 6‐diamino‐2‐phenyl indole (DAPI) and rhodamine phalloidin. Then, confocal images were obtained through fluorescence confocal microscopy (TCS SP8, Leica, Germany). To evaluate cell proliferation, the rBMSCs were cultured on different scaffolds for 1, 3, and 7 d for MTT assay at 490 nm by a microplate reader. To explore the expression of bone relative gene, the rBMSCs were cultured in differentiation medium (DMEM supplemented with 10 × 10 −3 M β‐glycerol phosphate, 0. 2 × 10 −3 M ascorbic acid, and 10% fetal bovine serum). After incubation for 7 d, the RNA was extracted using Trizol Reagent (Invitrogen Pty Ltd, Australia). The expression of relative bone genes was measured by RT‐qPCR: Runx2, OCN, OPN, and ALP. 22 In Vivo Bioactivity of the Hot Dog‐Like Scaffolds : All the animal experiments were carried out in compliance with the relevant laws. And all procedures were approved by The Ethics Committee of Nanjing First Hospital, Nanjing Medical University. 12 New Zealand white rabbits (2. 5–3 kg) were selected to evaluate the osteogenesis processes through implanting three kinds of scaffolds (H‐AKT, HD‐30AKT, and HD‐30AKT) into the critical‐sized femoral defect (diameter: 6 mm, and height: 8 mm), and the samples without the scaffold (Blank) were used as the control. After implanting for 8 weeks, the rabbits were sacrificed and the samples were gleaned. To estimate the osteogenesis, the samples were observed by micro‐CT (SKYSCAN1172, andSKYSCAN). Furthermore, histological images were acquired to evaluate the newly formed bone tissues after dehydrated, embedded in PMMA, sliced, and Van Gieson's picrofuchsin stained. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file.
|
10. 1002/advs. 201901173
| 2,019
|
Advanced Science
|
A Short Peptide Hydrogel with High Stiffness Induced by 3
|
Abstract Biological gels generally require polymeric chains that produce long‐lived physical entanglements. Low molecular weight colloids offer an alternative to macromolecular gels, but often require ad‐hoc synthetic procedures. Here, a short biomimetic peptide composed of eight amino acid residues derived from squid sucker ring teeth proteins is demonstrated to form hydrogel in water without any cross‐linking agent or chemical modification and exhibits a stiffness on par with the stiffest peptide hydrogels. Combining solution and solid‐state NMR, circular dichroism, infrared spectroscopy, and X‐ray scattering, the peptide is shown to form a supramolecular, semiflexible gel assembled from unusual right‐handed 3 10 ‐helices stabilized in solution by π–π stacking. During gelation, the 3 10 ‐helices undergo conformational transition into antiparallel β‐sheets with formation of new interpeptide hydrophobic interactions, and molecular dynamic simulations corroborate stabilization by cross β‐sheet oligomerization. The current study broadens the range of secondary structures available to create supramolecular hydrogels, and introduces 3 10 ‐helices as transient building blocks for gelation via a 3 10 ‐to‐β‐sheet conformational transition.
|
1 Introduction Peptide hydrogels are increasingly explored for biomedical applications such as wound healing patches, 1, 2 cell culture scaffolds for tissue engineering, 3 drug delivery vehicles, 4 or as substrates to study stem cell differentiation. 5 Peptides are particularly attractive as building blocks for hydrogels because: (i) their chemical structure and polydispersity is fully controlled, (ii) they exhibit high biocompatibility, 6 and (iii) their degradation products (amino acids) are readily cleared or reabsorbed by the metabolism. 6 In addition, bioactive functionality can be achieved: for example RGD peptides can be incorporated into the peptide sequence 7 to promote cell recognition or the peptide can be chemically modified with fluorescent probes and dye reporters 8 or with functional groups to promote subsequent crosslinking reactions. 9 In recent years, the ability to tune the gels' mechanical properties has become an increasingly important factor in the consideration of gel design. 10 While some hydrogels employ cross‐linking to vary the elastic properties, 11 others can be altered by varying the amount of salt in the gelation buffer or by adjusting the peptide concentration. 12 In many cases, short peptide‐based hydrogels are assembled from β‐sheets, β‐hairpins, or coiled‐coil α‐helices. 13, 14 Some peptides employ organic solvents to trigger gelation or toxic chemicals for cross‐linking, which is not ideal from a biocompatibility perspective. 11, 15 Hence peptide hydrogels assembled from uncommon secondary structural constructs may expand the existing peptide hydrogel libraries and have the potential to provide new characteristics, such as a broader range of moduli and water‐based gelation. Here, we present a short 8‐residue peptide (Ac‐ GLYGGYGV ‐NH 2 hereafter called GV8 ) that gels in water. GV8 exhibits a tunable, concentration‐dependent mechanical response with ≈25‐fold variation in storage modulus ( G′ ) and a maximum value reaching 35. 5 kPa that places it among the stiffest protein‐based hydrogels. The peptide sequence originates from suckerin proteins discovered in the sucker ring teeth (SRT) of the jumbo squid. 16, 17, 18 Suckerins are a protein family with a characteristic modular primary structure consisting of long Gly‐rich modules previously assumed to form mostly unordered domains, which are intervened by smaller Ala‐ and His‐rich modules that self‐assemble into stiffer β‐sheet elements. 17, 18 However, a recent NMR study indicated that the Gly‐rich domain can also form β‐sheets stabilized by aromatic side‐chain interactions. 19 GX8 peptides (where X = Val, Leu, and Phe) are specifically located in the Gly‐rich modules of suckerin‐19 with a high occurrence of 13 copies. Combining circular dichroism (CD), Fourier Transform Infrared Spectroscopy (FTIR), solution and solid‐state NMR, wide‐angle X‐ray scattering (WAXS), and molecular dynamics (MD) simulations, we find that GV8 peptide forms 3 10 helices in solution and undergoes a conformation change into antiparallel β‐sheets during gelation. 2 Results and Discussion 2. 1 Peptide Gelation We obtained the GV8 peptide hydrogel by simple incubation of the peptide in deionized (DI) water, with gelation occurring at peptide concentrations ranging from 10 to 20 × 10 −3 m and a concentration‐dependent gelation time between 5 and 9 h. The minimal critical gelation concentration ( C gc ) in water was 10 × 10 −3 m, below which we did not observe gelation. We monitored the gelation kinetics by measuring the absorbance (OD 550nm ) of the peptide solutions at 550 nm (Figure S1, Supporting Information), whereby OD 550nm increased during the gelation process and plateaued once gelation was complete. 20 We also attempted to mutate the C‐terminus Val residue with Leu ( GL8 ), Ala ( GA8 ), Phe ( GF8 ), Ser ( GS8 ), Lys ( GK8 ), or Ile ( GI8 ), but these peptides were not able to form gels in water, illustrating the key role of terminal Val in gelation as corroborated by NMR studies. GF8 and GI8 peptides remained in solution with some aggregates observed over time, whereas GL8 self‐assembled into large (mm‐size) and stiff beads (Figure S2, Supporting Information). 2. 2 Macro‐ and Microgel Structure We then examined the morphology and topology of GV8 hydrogel by Cryo‐Electron Microscopy (Cryo‐EM), atomic force microscopy (AFM), and scanning electron microscopy (SEM). 20 × 10 −3 m GV8 peptide solution was incubated for 3 h prior to blotting and vitrification to preserve the natural nanostructure of the sample in hydrated conditions 21 for Cryo‐EM imaging. Long fibers less than 10 nm wide were observed ( Figure 1 a) with consistent twisted morphologies and average periods of ≈80 nm along the fibers. AFM imaging was performed on a thin layer of dried gel, revealing a surface topology of a network of fibers (Figure 1 b), and the height profile (Figure S3, Supporting Information) revealed fibers of ≈5–10 nm in height thereby matching the Cryo‐EM observations. Since drying and conventional lyophilization causes the hydrogel structure to collapse, samples for SEM were prepared by snap‐freezing GV8 hydrogel in liquid N 2 for at least 5 min followed by cryo‐fracture and immediate lyophilization to obtain representative cross‐sections. SEM imaging revealed a porous structure (Figure 1 c, left) constructed by sheet‐like structures and closer examination indicated that the sheets were formed by a fibrous network of peptides (Figure 1 c, right). Figure 1 Structural features and physico‐chemical properties of GV8 peptide hydrogel observed with time‐series spectroscopy measurements during gelation. a) Cryo‐EM images of GV8 hydrogel fibrils and their twisted morphology (right image) with average periodicity of ≈80 nm. b) AFM amplitude profile of dried GV8 hydrogel with fibers of ≈6–10 nm height. c) SEM images of GV8 hydrogel cross‐section revealing sheet‐like morphology (left image) made of fibers (right image). A photograph of the hydrogel is shown in the inset. d) Scaling law plot of plateau G ′ versus peptide concentration. e) CD and f) FTIR spectra recorded over 50 h indicate significant increase in β‐sheet content (intensity increase at λ 218 nm and ν∼ 1634 cm −1, respectively). 2. 3 Rheology Characterization The hydrogel exhibited robust mechanical properties and could readily be manipulated and sectioned into thin slices (Movie S1, Supporting Information). In order to characterize the gel's mechanical properties, we prepared GV8 hydrogels with peptide concentrations C gc ranging from 10 to 20 × 10 −3 m and conducted oscillation frequency sweeps at 0. 25% shear strain (Figure S4, Supporting Information). The shear storage modulus ( G′ ) exhibited a scaling power law as a function of peptide concentration ( G′ vs C gc ) with a power law index of 3. 2 (Figure 1 d), allowing us to tune the storage modulus ≈25‐fold over a narrow range of peptide concentration. We note that this power law index is significantly higher than the “universal” scaling law determined for protein‐based semiflexible networks 22, 23, 24 where G′ scales as C 11/5 or for cross‐linked gels that exhibit a G′ ∝ C 2. 5 scaling law. 25 Instead, this behavior is well captured by the fractal gel model, G ′ ∝ C gc 3 + d b / 3 − d f, where d b is the fractal dimension of the connecting chain and d f is the dimensionality of the repeating fractal cluster. Taking d f = 1. 6 for the fractal dimension as obtained by small angle X‐ray scattering (SAXS, Figure 2 a) yields d b = 1. 5, which is a typical value of heat set protein gels. 26 This behavior suggests that gelation does not proceed by entanglement of long fibrils, but rather by growth of shorter fibrillar clusters and conformational transition as evidenced by NMR and WAXS measurements described later. In line with this picture, GV8 did not strain‐stiffen, likely because the chain length is shorter than in conventional biological gels. In quantitative terms, the maximum shear modulus of GV8 of 35. 5 kPa is on par with the stiffest, non‐crosslinked short peptide‐based hydrogel containing only natural amino acid residues. 12 A wide range of moduli have been reported for short peptide hydrogels 10, 27, 28, 29 with stiffest gels reached in gels containing modified amino acids, synthetic functional groups, or which have been crosslinked. 30, 31 The ability to tune the stiffness from 1. 3 to 35. 5 kPa is particularly appealing for stem cell differentiation studies since gel stiffness has been well‐documented to govern cell adhesion and regulation based on the substrate's mechanical feedback, 5, 32, 33, 34 with stiffness values in the range 0. 1–1, 8–17, and 25–40 kPa sought after for neurogenic, myogenic, and osteogenic differentiation, respectively. The advantage of our GV8 hydrogel for such applications is that its stiffness can be modulated solely by varying the peptide concentration without any additional chemical modifications. Figure 2 SAXS and WAXS patterns for the GV8 peptide extruded gel together with the fitting curve (red) and peaks assignment. a) SAXS profile for dried extruded GV8 gel. WAXS patterns of GV8 gel in the b) hydrated state and in the c) dried state. 2. 4 Circular Dichroism and FTIR Spectroscopy Next, we conducted time‐dependent CD and attenuated total reflection Fourier transform infrared spectroscopy (ATR‐FTIR) measurements on 20 × 10 −3 m GV8 from its initial solution state until its postgelation state in order to reveal secondary structural changes during self‐assembly (Figure 1 e, f). At time 0 h, the CD spectrum consisted of a minimum at 215 nm and two maxima at 200 and 228 nm (Figure 1 e). The bands at 215 and 228 nm are attributed to the resultant exciton couplet of Tyr–Tyr exciton interaction of their π→π* transitions, also known as CD Cotton effect 35, 36 indicating interaction between the Tyr aromatic chromophores, 37, 38 as also corroborated by our 2D NMR data. Over the course of gelation (1–5 h), the minimum shifted to 218 nm with a significant increase in intensity, and the maximum at 228 nm diminished, thereby transitioning to β‐sheet secondary structure. The CD spectra remained constant after 5 h in agreement with our OD 550nm measurements of 20 × 10 −3 m GV8 peptide (Figure S1, Supporting Information), whereby the absorbance plateau onset at 5 h indicated that gelation was complete without any significant structural changes after 5 h. Upon further incubation postgelation (25 and 50 h), a new maximum appeared at 237 nm and can be assigned to aromatic transitions of the Tyr residues, 39 which we postulate is related to the conformational transition of GV8 peptide during gelation. ATR‐FTIR was performed on dried 20 × 10 −3 m GV8 hydrogels incubated over the same time points as in CD studies. The samples were snap‐freezed in liquid N 2 to arrest the structural assembly of the peptides at stipulated time points. Amide I bands were deconvoluted and peaks assigned to β‐sheets, unordered regions, helices and turns, or 3 10 ‐helices (Figure S5, Supporting Information). 40, 41 β‐turns and 3 10 ‐helices were grouped together in our assignments as they are structurally alike 42, 43 with similar hydrogen bond strengths, and hence close frequency positions within the Amide I band. Over the course of gelation, the Amide I band maximum at ν˜ max = 1652 cm −1 shifted to 1634 cm −1 (Figure 1 f), confirming secondary structural change towards β‐structures. Semiquantitative analysis by deconvolution of Amide I band (Table S1, Supporting Information) indicated that the initial dominating secondary structures of GV8 peptide were turns and/or 3 10, whereas in the gel state β‐sheet structures were the most abundant (≈65 % at 50 h). 2. 5 Solution NMR In order to obtain the molecular level structure of GV8, we analyzed the 3D structure of the peptide in solution using NMR. To maintain the peptide in soluble form, its concentration was kept at 0. 5 × 10 −3 m. 2D 1 H– 1 H TOCSY (TOtal Correlation SpectroscopY) and 1 H– 1 H NOESY (Nuclear Overhauser Effect SpectroscopY) spectra showed well‐resolved cross‐peaks assigned to the individual amino acid residues of GV8 ( Figure 3 a, c). The 1 H α chemical shift deviations (CSD) 44 exhibited negative chemical shifts suggestive of a predominantly helical structure (Figure S6a, Supporting Information). However, precise examination of the 1 H– 1 H NOESY spectrum revealed the absence of ( i, i+4 ) medium range (H α –H N ) NOE connectivities typically observed in α‐helix (Figure 3 b). Instead, we only detected ( i, i+3 ) H α –H N NOEs in addition to the strong ( i, i+1 ) H N –H N NOEs, suggesting the presence of 3 10 helix. 45 Further analysis also revealed the presence of ring proton NOEs between Y3 and Y6 stabilizing the 3 10 helical structure (Figure 3 c). Collectively, these data indicated that the aromatic side chain interactions between Y3 and Y6 may lead the GV8 peptide to adopt a well‐defined 3 10 helix. This was also supported by NOEs between aliphatic side chains of L2 and V8 along the helical axis (Figure 3 c). The 3D structure of the GV8 monomeric 3 10 helix calculated using a total of 39 NOE constraints ( Table 1 ) is shown in Figure 3 d, e. When Val8 was mutated to Leu ( GL8 ) and Ala ( GA8 ), the aromatic interactions between Y3 and Y6 disappeared and both mutated peptides remained in extended conformations (Figure S6c, d, Supporting Information). Figure 3 Solution NMR characterization of GV8 at monomeric concentration (0. 5 × 10 −3 m ). a) 2D 1 H– 1 H TOCSY spectrum of 0. 5 × 10 −3 m GV8 peptide delineating the individual spins of GV8 amino acid residues. b) Bar diagram representation of NOE connectivities detected for GV8 peptide. c) 2D 1 H– 1 H NOESY spectra displaying weak ring interaction of Y3 and Y6. d) Superimposition of ten lowest energy structures of GV8 peptide. e) Representative structure of monomeric 3 10 ‐helix showing side chain stacking of Y3 and Y6. Table 1 Structural statistics summary of ten lowest energy structures of GV8 monomer and oligomer obtained from solution state NMR, and GV8 hydrogel by ssNMR GV8 monomer (solution NMR) GV8 oligomer (solution NMR) GV8 hydrogel (ssNMR) Distance restraints Intraresidue (| i−j | = 0) 9 18 22 Sequential (| i−j | = 1) 14 32 27 Medium range (2≤ | i−j |≤4) 16 56 0 Long range (| i−j |≤ 5) 0 32 10 Total NOE constraints (solution NMR)/dipolar contacts (ssNMR) 39 138 59 Distance restraints violations Number of violations 9 51 33 Maximum violation ≤0. 5 ≤0. 5 ≤0. 5 Average target function value 4. 49 34. 36 17. 37 Deviation from mean structure Backbone atoms [Å] 0. 52 0. 65 1. 49 Heavy atoms [Å] 0. 92 0. 70 1. 75 Ramachandran plot for the mean structure % residues in the most favorable and additionally allowed regions 100 100 85 % residues in the generously allowed region 0 0 15 % residues in the disallowed region 0 0 0 John Wiley & Sons, Ltd. To monitor gel formation, both 1D proton and 2D 1 H– 1 H NOESY spectra of 20 × 10 −3 m GV8 were recorded during a 4 h period. The peak intensities of the amide protons arising from residual peptides in solution decreased with time (Figure S6b, Supporting Information), implying that an increasing amount of peptide underwent structural rearrangement and were incorporated in the hydrogel. Analysis of 2D 1 H– 1 H TOCSY spectra acquired after 20 h demonstrated well resolved cross‐peaks corresponding to individual spins of GV8 peptide ( Figure 4 a). 2D 1 H– 1 H NOESY spectra displayed the ( i, i+3) NOEs that are fingerprints of a 3 10 ‐helix 45 (Figure 4 c). Strikingly, residues at the C‐terminal (G7 and V8) were involved in displaying long range NOEs with residues at N‐terminal (Y3 and L2) (Figure 4 c). The H α of G7 interacted with L2 and Y3 residues (blue arrows and dotted lines, Figure 4 c) and side chain β protons of V8 were also found to interact with L2 protons. These long‐range NOEs are attributed to cross‐strand NOEs resulting from oligomerization of the GV8 peptide after 20 h (Figure 4 b). The aromatic packing interactions between Y3 and Y6 were also clearly detected owing to the high peptide concentration (Figure 4 c). Figure 4 Solution NMR characterization of GV8 at oligomeric concentration (20 × 10 −3 m ). a) 2D 1 H– 1 H TOCSY spectrum of 20 × 10 −3 m GV8 peptide delineating the individual spins of GV8 amino acid residues. b) Bar diagram representation of residues that display sequential, medium range, and long‐range NOEs detected at 20 × 10 −3 m concentration. c) 2D 1 H– 1 H NOESY spectra displaying long range cross‐strand NOEs and ring proton interactions of Y3 and Y6 (marked with blue arrows and dotted lines). d) Superimposition of ten lowest energy structures of GV8 peptide at oligomeric 20 × 10 −3 m concentration. e) Representative structure of dimeric 3 10 ‐helix of GV8 oligomer showing side chain arrangement. Aliphatic residues (L2, V8, L2*, and V8*) are shown in pink color and aromatic residues (Y3, Y6, Y3*, and Y6*) in green. Using a total of 138 NOE constraints (Table 1 ), we calculated an ensemble of ten structures for GV8 composed of dimeric 3 10‐ helical building blocks (adding 5 Gly as a linker between the two monomers). The aromatic residues Y3 and Y6 delineated a higher number of medium and long‐range NOEs (Figure 4 b). Superposition of ten lower energy conformers led to root mean square deviations (RMSD) of backbone and heavy chains of 0. 65 and 0. 70 Å, respectively (Table 1 and Figure 4 d). 3D structure calculation revealed that the hydrophobic face of the dimeric 3 10 ‐helix is composed of π‐stacking interactions between Y3 and Y6, while the exposed side of the dimeric helix is made up of aliphatic side chains L2 and V8 (Figure 4 d, e). Procheck analysis of the 3D structure revealed that all residues resided in the sterically allowed regions (Table 1 ). 2. 6 Amide Temperature Coefficient and H/D Exchange NMR Studies The role of hydrogen bonds in stabilizing the 3 10 ‐helix was studied by calculating the protection factors from H/D exchange as well as the amide proton temperature coefficients (Δδ NH /Δ T ) at various temperatures. A series of 2D 1 H– 1 H TOCSY spectra were recorded every 30 min for the 0. 5 and 20 × 10 −3 m GV8 peptide dissolved in D 2 O. All Gly residues for both the monomer (0. 5 × 10 −3 m ) and the oligomer (20 × 10 −3 m ) concentrations displayed protection factor of 60–80, supporting a significant H/D exchange protection inside the core of the 3 10 ‐helical structure (Figure S7a, Supporting Information). The protection factor of Y3 and Y6 increased with the peptide concentration, indicating enhanced aromatic interactions for the oligomeric form (Figure S7a, Supporting Information). Comparably, the amide proton temperature coefficients of all GV8 residues at both monomer and oligomer concentrations exhibited values more positive than −4. 6 ppb/K. 46 The Gly residues also exhibited more positive values in line with their higher protection factor values (Figure S7b, Supporting Information). 2. 7 Solid State NMR NMR characterizations of the gel state were conducted by ssNMR under magic angle spinning (MAS) conditions. All amino acids of GV8 hydrogel prepared with uniformly labeled 13 C and 15 N peptide were unambiguously assigned using the sequential walking method of 3D NCACX, NCOCX, and CANcoCX spectra ( Figure 5 a). Analysis of the 2D 13 C– 13 C DARR spectra acquired at 50 ms contact time revealed long range dipolar contacts between L2 and V8 side chains (Figure 5 a, b). The Y3/Y6 ring packing interactions that were detected in oligomeric solutions of GV8 were no longer present in the hydrogel state. Figure 5 Solid NMR characterization of 13 C– 15 N labeled GV8 hydrogel. a) Strip plots of L2 and V8, 3D NCACX spectra of 13 C– 15 N labeled GV8 peptide hydrogel displaying long‐range contact of residues. b) 2D 13 C– 13 C DARR spectra with contact time of 50 ms showing long‐range dipolar contacts between L2 and V8 side chains (β, δ, γ). c) Representative structure of dimeric extended conformation of GV8 hydrogel. d) Side chain disposition representation of antiparallel β‐sheets of GV8 hydrogel displaying interchain connectivity between L2 and V8 residues (L2/V8* and L2*/V8). We then calculated the dimeric conformation of GV8 in the hydrogel state using intraresidue and sequential dipolar constraints (Experimental Section). An overview of ten lowest energy structures resulted in RMSD value of 1. 49 Å for backbone atoms and 1. 75 Å for heavy side chain atoms (Table 1 ). The lowest energy structure revealed that the GV8 hydrogel comprised of extended antiparallel β‐sheets (Figure 5 c). The absence of Tyr ring packing interactions strongly suggests that during gelation, Tyr side chains (Y3 and Y6) rearranged to be exposed to solvent, and that at the same time stronger hydrophobic interactions between L2, V8, L2*, and V8* stabilized the antiparallel β‐strand conformation of the GV8 hydrogel (Figure 5 d). Extended 3 10 ‐helices have been shown to act as intermediate seeds for the formation of amyloid (β‐sheet rich) aggregates, 47 but to the best of our knowledge it has previously not been reported to induce hydrogel formation. Furthermore, none of the Val8‐mutated peptides (Figure S1, Supporting Information) gelled under the same conditions and neither control peptides GL8 nor GA8 formed 3 10 ‐helices in solution (Figure S6c, d, Supporting Information) hence corroborating that 3 10 ‐helix is a critical transient conformation leading to gelation of GV8. Val8 at the C‐terminus therefore plays a crucial role in stabilizing the 3 10 ‐helix structural intermediate through an intrachain hydrophobic interaction, which is not achieved with Ala or Leu residues. 2. 8 WAXS of Extruded GV8 Hydrogel and MD Simulations Confirmation of β‐sheet presence in the gel was gained by performing WAXS measurements in both the wet and dry states. In the wet state (Figure 2 b) a peak at q = 5. 44 nm −1 and a shoulder at 13. 3 nm −1 were observed, corresponding to distances of 1. 16 nm and 4. 7 Å, respectively. These features are the hallmark of β‐sheet rich amyloid fibrils, with 1. 16 nm corresponding to the inter‐β‐sheet spacing and 4. 7 Å to the interstrand distance of β‐sheets. Interestingly, both features were greatly enhanced upon drying of the gel: the peak at 1. 16 nm shifted to 1. 09 nm due to dehydration, whereas the 4. 7 Å peak now became the dominant scattering peak of the dried gel. The emergence of additional correlations is highlighted by a deconvolution of the intensity profile with Laurentian curves as shown in Figure 2 c. To further assess the conformation propensity of GV8, we conducted MD simulations on both oligomeric and 40‐mer constructs. These simulations predicted that oligomers of GV8 prefer the antiparallel β‐sheet conformation, especially the Leu2, Tyr3, and Tyr6 residues ( Figure 6 a), whereas 3 10 ‐helices were not stable thus resulting in a very low concentration of 3 10 ‐helices after 200 ns simulations. These results indicate that antiparallel β‐sheets constituted the most stable structure at equilibrium. It is important to emphasize that the peptide concentration during MD simulations is much higher than in the soluble form and more representative of the gel state. Therefore, the very low concentration of 3 10 ‐helices at equilibrium is in line with the conformational transition detected in the gel by NMR. Based on solid‐state NMR and WAXS data in the gel state, we then conducted simulations on a 40‐mer antiparallel β‐sheet construct. The 100 ns simulation indicated a very high stability of antiparallel β‐sheets, in agreement with the WAXS measurements. Furthermore, the simulations indicated that β‐strands were stabilized by intersheet π–π stacking of Tyr residues (Figure 6 b). This result corroborates the molecular‐level structure of GV8 by solid‐state NMR, which indicated that Tyr side‐chains in the gel state pointed out perpendicular to the strand direction, making them available to engage in intersheet interactions as predicted by the simulations. In addition, in‐register antiparallel β‐sheets were observed for both ssNMR and MD simulations, which we postulate is due to the arrangement of the Tyr side‐chains that leads to the least sterically‐hindered conformation. Figure 6 MD simulations of GV8 conformation and oligomeric self‐assembly. a) Secondary structure distribution of each residue in dimer, tetramer, and octamer of GV8 at 300K for 200 ns. b) Initial and final structures of a 40‐mer model of the GV8 peptide with β‐sheets structures shown in blue. The inset in the bottom panel shows the representative structure in the model, whereby the π–π stacking and hydrophobic interactions mainly contribute to intersheet association. 3 Conclusions GV8 is an eight amino acid long peptide repeat from suckerin‐19—the most abundant protein forming the load‐bearing squid sucker ring teeth—that forms stiff hydrogels in water with tunable elastic modulus. Using CD, FTIR, and solution NMR spectroscopy, we have determined that GV8 self‐assembles into unusual 3 10 monomeric helices at low peptide concentration, which are intramolecularly stabilized by π–π stacking aromatic interactions between Y3 and Y6 residues, as well by the aliphatic side chains L2 and V8. As the concentration increases, GV8 dimerizes into antiparallel 3 10 helices driven by π‐stacking interactions between Tyr residues Y3, Y6, Y3*, and Y6*. In the gel state, ssNMR and WAXS measurements indicate that GV8 is made of antiparallel β‐sheets, inferring that gelation proceeds by a 3 10 ‐helix to β‐sheet conformational rearrangement. This mechanism is starkly different from previous reports on fibrous peptide‐based hydrogels. During this conformational transition, Tyr side‐chains reorient perpendicular to the chain direction according to both ssNMR and MD simulations, allowing to mediate intersheet interactions. Peptide‐based hydrogels with water gelation and the ability to tune the stiffness 25‐fold simply by increasing the peptide concentration may find notable opportunities for biomedical applications, such as tissue engineering, encapsulation of therapeutics, soft tissue adhesives, or matrix for stem cell differentiation. 4 Experimental Section Materials : Ac‐GLYGGYGV‐NH 2 peptide and Ac‐GLYGGYGX‐NH 2 peptides (where X = V, L, A, F, S, K, and I) were purchased from GL Biochem (Shanghai) Ltd. Peptides were checked to be >98% purity via trace HPLC and LC/MS prior to use. All of the peptides were acetylated at the N‐terminal and amidated at the C‐terminal to prevent end‐to‐end charge interactions. 13 C– 15 N uniformly labeled GV8 crude peptide purchased from Cambridge Isotopes was purified to >95% purity via HPLC and checked with LC/MS prior to use. UV–vis Spectroscopy : The peptides were dissolved in DI water at the respective concentrations and 100 µL was aliquoted into each well of a 96‐well microtiter plate, with a minimum of three wells per condition. UV–vis absorbance measurements at 550 nm were recorded on a Tecan infinite M200 Pro microplate reader at intervals of 30 min for the first 16 h and subsequently at increased time intervals. Peptide Hydrogel : GV8 peptide was dissolved in DI water at the desired concentration (between 10 and 20 × 10 −3 m ) and incubated at ambient temperature for at least 12 h. CD Spectroscopy : GV8 peptide was dissolved at 20 × 10 −3 m concentration in DI water and spectra were collected using a 0. 2 mm path length quartz cuvette. Data acquisition was performed using AVIV 420 Circular Dichroism (New Jersey, USA) spectrometer. A quartz sandwich cuvette with optical path length of 0. 2 mm was used for all data collection and the edges of the cuvette were sealed with parafilm to prevent loss of liquid. Data were acquired over a wavelength range of 190–260 nm and acquisition parameters were 0. 5 nm wavelength steps with an averaging time of 0. 1 s, 1. 00 nm bandwidth, and readings were averaged over three scans. Obtained spectra were smoothed at 12 pts via adjacent‐averaging method (ensuring that no visible existing peaks were removed or artefacts introduced) and plotted via OriginPro 9. 1. FTIR Spectroscopy : ATR‐FTIR spectroscopy of lyophilized GV8 samples were performed on a Bruker Vertex 70 (Massachusetts, USA) equipped with a PIKE Technologies MIRacle attenuated total reflection (ATR) ZnSe‐Diamond 3‐reflection accessory and a LN 2 cooled MCT detector. Scans were obtained at ambient temperature over the range of 4000–750 cm −1 with a resolution of 2 cm −1, averaged over 128 scans. GV8 peptide solutions were prepared at 20 × 10 −3 m concentration in separate vials of 20 µL and snap freezed by dipping the vials in liquid N 2 for 5 min at the stipulated time points and lyophilized immediately. All spectra processing were performed on OPUS 6. 5, and processed in the sequence of water vapor subtraction, baseline correction, then normalized using amide I band. Amide I band was deconvoluted by secondary derivation, with peak fitting performed using 100% Gaussian curves with individual FWHM kept relatively consistent. The deconvoluted peaks were assigned to β‐sheet, unordered, helix and turns or 3 10 structures. 40, 41, 48, 49 Cryo‐EM : GV8 peptide was dissolved at a concentration of 20 × 10 −3 m and incubated for 3 h. Copper grids with Ultrathin C Film on Lacey Carbon support film was plasma‐treated with JEOL DATUM HDT‐400 for 300 s to increase hydrophilicity of grid surfaces to allow aqueous samples to adhere and spread. Vitrified samples were prepared using Gatan Cryoplunge TM 3 (Cp3). 4 µL of sample was pipetted onto each plasma‐treated copper grid and blotted for 5 s followed by vitrification in liquid ethane at −180 °C. All images were taken in bright‐field mode with objective aperture inserted. Imaging was carried out with energy filtered Carl Zeiss TEM, LIBRA 120 with in‐column Omega spectrometer and operated an acceleration voltage of 120 kV and the sample temperature was maintained below −180 °C during imaging. SEM : Peptide hydrogel was snap freezed by dipping into liquid N 2 for at least 5 min. The frozen hydrogel was then cryo‐fractured with tweezers to expose the porous cross‐section and the fractured surfaces were placed face‐up on carbon tape and lyophilized immediately. Samples were Platinum‐coated below 5 Pa, at 20 mA for 30 s and imaging was performed using JEOL JSM‐FESEM 7600F (Massachusetts, USA), at SEI‐mode, 5 kV, and 92 µA emission current. AFM : 10 µL of GV8 peptide at 20 × 10 −3 m concentration was deposited onto freshly cleaved mica and air‐dried overnight. AFM images were obtained on Asylum Cypher S AFM (Oxfordshire, UK) in tapping mode using Nanoworld NCSTR silicon nitride soft‐tip cantilevers ( R f = 160 kHz, k = 7. 4 N m −1 ). All images were flattened to remove background curvature using Igor Pro software and no further image processing was carried out. Rheology : Rheological measurements were performed at ambient temperature on Anton Paar MCR501 rheometer with a parallel plate PP10 geometry. Five different concentrations (10, 12, 15, 18, and 20 × 10 −3 m ) of GV8 hydrogels were prepared in DI water, each pipetted into 1 mL syringes with nozzles removed, then left overnight to gelate. The hydrogel was extruded from the syringes and cut to 1–2 mm thick slices with a sterile blade and placed on the rheometer plate for measurements. Strain sweeps were first conducted at constant frequency of 1 Hz, from 0. 1% to 10 % strain to identify the linear viscoelastic region and 0. 25% strain was selected for subsequent frequency sweeps that were conducted from 0. 01 to 100 Hz. All measurements were triplicated with repeated measurements performed on fresh samples. NMR : All solution state NMR experiments were carried out on a Bruker 700 MHz spectrometer equipped with a cryoprobe. NMR data were processed with TOPSPIN (Bruker), then analyzed using Sparky 50 programs. 0. 5 or 20 × 10 −3 m GV8 peptide was dissolved in water, pH 6. 8 with 10% D 2 O for deuterium lock and DSS for signal reference. 2D 1 H– 1 H TOCSY and NOESY spectra were acquired with 80 and 200 ms mixing times, respectively. In order to monitor hydrogel formation, 1D 1 H spectrum and a series of 2D 1 H– 1 H NOESY spectra were recorded every 4 h for 20 h using the 20 × 10 −3 m peptide solution. For H/D exchange experiments, 0. 5 and 20 × 10 −3 m peptide samples were dissolved in 100% D 2 O and 2D 1 H– 1 H TOCSY spectra were recorded at 30 min intervals. The extrinsic exchange rates were obtained by fitting the peak intensity versus time to a single‐exponential decay equation. The protection factor were calculated as the ratio of intrinsic exchange rates (calculated from SPHERE 51 ) to the extrinsic exchange rates. A protection factor above 30 is indicative of stable hydrogen bonds, while values between 10 and 30 indicate an intermediate range of hydrogen bond strength. 52 Amide temperature coefficients were also determined by recording 1D 1 H spectra of 0. 5 and 20 × 10 −3 m GV8 peptide at 298, 303, 308, and 313 K. Amide proton chemical shift deviations were fitted linearly against temperature and the temperature coefficients were calculated as σδ HN/Δ T (ppb K −1 ). Solid state NMR experiments were carried out on a Bruker 600 MHz spectrometer equipped with a 1. 7 mm MAS probe. The MAS spinning frequency was 13333 Hz. 20 × 10 −3 m of 13 C– 15 N labeled GV8 peptide was dissolved in water, pH 6. 8 and allowed to incubate overnight for hydrogel formation. Sample was loaded in a 1. 7 mm thin wall zirconia rotor (Bruker) manually and the rotor was spun at 70 000 rpm for 30 min by ultracentrifugation (Beckman Proteomelab XL, IN, USA). 2D 13 C– 13 C DARR spectra were recorded over contact times ranging from 50 to 400 ms. 3D NCACX, NCOCX, and CANcoCX experiments were also recorded with 50 ms contact time. NMR Structure Calculation : The structure calculations were carried out using the CYANA 2. 1 program. The monomeric conformation of GV8 was calculated using the intensities of 1 H– 1 H NOE cross peaks that were classified as strong, medium, and weak and translated to upper bound distance limits of 2. 5, 3. 5, and 5. 0 Å. The dihedral Φ and Ψ angles were constrained between −120° to −30° and −120° to 120° as suggested in the CYANA program files. Out of the 100 structures generated, the ten lowest energy structures were used for more analysis. The dimer structures of GV8 were also calculated using the same constraints as monomeric structure calculation. The two monomeric units were linked by five glycine linkers. The structure of hydrogel was calculated using 13 C– 13 C dipolar contacts derived from solid state NMR spectra. All of the conformations were validated using PROCHECK. 53 For structure calculation from ssNMR, a total of 47 intraresidue and sequential dipolar constraints were used. The long‐range dipolar contacts included in the structure calculation were cross‐strand contacts used to generate a dimeric conformation. SAXS and WAXS : SAXS and WAXS experiments were performed using Rigaku MicroMax‐002+ equipped with a microfocused beam (40 W, 45 kV, 0. 88 mA) with the λ Cu Kα = 0. 15418 nm radiation collimated by three pinhole collimators (0. 4, 0. 3, and 0. 8 mm). The SAXS and WAXS intensities were collected by a two‐dimensional Triton‐200 gas‐filled X‐ray detector (20 cm diameter, 200 µm resolution) and a 2D Fujifilm BAS‐MS 2025 imaging plate system (15. 2 × 15. 2 cm 2, 50 µm resolution), respectively. An effective scattering vector range of 0. 05 nm −1 < q < 25 nm −1 was obtained, where q is the scattering wave vector defined as q = 4π sin θ/λ Cu Kα with a scattering angle of 2θ. Hamiltonian‐Replica Exchange Molecular Dynamics (H‐REMD) Simulations : H‐REMD simulations 54 were performed for the dimer, tetramer, and octamer of the Ac‐GLYGGYGV‐NH 2 peptide for 200 ns each. The CHARMM 36 mm force field parameters 55 were applied to peptides, and the dimer, tetramer, and octamer were put in a cubic box with TIP3P waters 56 and 0. 15 m NaCl. The minimum distance between the peptides and the box edge was larger than 1. 5 nm. The dimer, tetramer, and octamer systems have 8, 12, and 16 replicas from 300 to 600 K, respectively, and each was simulated for 200 ns. The trajectories were saved every 2 ps. Conventional MD Simulations : Conventional MD simulations were performed for the 40‐mer two‐layer antiparallel β‐sheet model for 100 ns using the AMBER 16 software 57 together with the AMBER14SB force field. SHAKE algorithm 58 was used to constrain all bonds involving hydrogens and electrostatic interactions were treated by the particle mesh Ewald sum method 59 with a 8 Å cutoff for nonbonded interactions in direct space. The model was solvated in a rectangular box filled with TIP3P waters, 56 with an at least 1. 0 nm distance between the peptides and the box edge. The whole system was first energy‐minimized, with a series of position restraints on the solute (all heavy atoms, backbone atoms, and Cα atoms). The simulation was continued for 100 ns at 1 bar and 298. 15 K. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201901198
| 2,020
|
Advanced Science
|
Adult Tissue Extracellular Matrix Determines Tissue Specification of Human iPSC‐Derived Embryonic Stage Mesodermal Precursor Cells
|
Abstract The selection of pluripotent stem cell (PSC)‐derived cells for tissue modeling and cell therapy will be influenced by their response to the tissue environment, including the extracellular matrix (ECM). Whether and how instructive memory is imprinted in adult ECM and able to impact on the tissue specific determination of human PSC‐derived developmentally fetal mesodermal precursor (P‐meso) cells is investigated. Decellularized ECM (dECM) is generated from human heart, kidney, and lung tissues and recellularized with P‐meso cells in a medium not containing any differentiation inducing components. While P‐meso cells on kidney dECM differentiate exclusively into nephronal cells, only beating clusters containing mature and immature cardiac cells form on heart dECM. No tissue‐specific differentiation of P‐meso cells is observed on endoderm‐derived lung dECM. P‐meso‐derived endothelial cells, however, are found on all dECM preparations independent of tissue origin. Clearance of heparan‐sulfate proteoglycans (HSPG) from dECM abolishes induction of tissue‐specific differentiation. It is concluded that HSPG‐bound factors on adult tissue‐derived ECM are essential and sufficient to induce tissue‐specific specification of uncommitted fetal stage precursor cells.
|
1 Introduction The control of interactions between cells and their environment is essential for tissue and disease modeling, and regenerative medicine. The extracellular matrix (ECM) serves as a key messenger between cells and the tissue environment by providing a dynamic scaffold. ECM‐embedded cues regulate organ homeostasis, meditate cellular responses, and promote tissue repair. 1 These cues are not only organ‐specific but also dynamically changing during tissue morphogenesis and maturation. Isolated ECM may not only provide a structural scaffold of the source organ, but also a developmental stage and tissue‐specific functional imprint that could be used as a basis for generating organ models. 2 For example, decellularized ECM (dECM) from heart, kidney, lung, and other organs has been used as scaffold for recellularization with tissue‐specific cells, non‐committed stem and progenitor cells to reestablish cell functionality and to study cell–ECM interaction with the vision to ultimately engineer organs for transplantation. 3, 4, 5, 6 The use of pluripotent stem cells (PSC) as a cell source for dECM recellularization was based on the assumption, that the ECM memorizes its tissue origin and provides cues, which drive organotypic differentiation processes. 6, 7, 8, 9, 10 Indeed, expression of organ‐specific cell markers was detected in renal, pancreatic, or cardiac dECM recellularized with undifferentiated PSC, however, reproducibility of obtaining structurally or functionally mature cells was usually poor. 6, 7, 8, 9, 10 More efficient and reproducible differentiation was shown with a tissue specifically committed progenitor cell population for recellularization of dECM from different organs. 8, 11 These data indicate that the developmental stage at which cells respond to dECM may be of importance. This is relevant as most PSC‐derived cells are immature. 12 Furthermore, for cell therapy, the most efficient tissue repair and regeneration may be achieved with proliferative active and plastic progenitor cells instead of terminally differentiated cells. However, these cells have to be able to respond to the adult, aged, or compromised environment, including the ECM. As reviewed, previous results from human and mouse lung 2, 14, 15 indicate that the adult ECM contains memory factors and cues that are able to promote tissue‐specific differentiation of tissue‐committed stem cells, very little is known about the nature and identity of the functional cues driving this process. 13 However, whether these ECM‐associated cues are conserved between species is unclear and using cells and ECM from different species may confound results. Strong evidence was provided that heparan sulfate proteoglycans (HSPG) are required to drive epithelial differentiation of endoderm cells on lung dECM as their removal diminishes differentiation. 14, 15 Moreover, in most studies to elucidate dECM effects on stem cells, differentiation promoting factors were provided with the cell cultivation medium. The presence of these factors makes it difficult to assess the specific role of ECM and ECM‐bound molecules for the observed cell specification effects. In fact, externally supplemented inducing factors also work efficiently with undefined matrices such as mouse tumor‐derived matrigel to promote differentiation of human stem cells. 16 Whether target tissue‐derived ECM may have any additional differentiation, promoting and specifying effects is uncertain and was not directly compared to matrigel. Moreover, the property of autonomous self‐organizing morphogenesis of more committed precursors, for example, of metanephric mesenchyme precursor cells of the kidney, 8 may additionally blur the role and need of exogenous ECM as a promoter of tissue‐specific differentiation. 17, 18 Finally, the use of both dECM and cells from human will eliminate variability introduced by cross species approaches when investigating specification of early human precursor cells in a human environment. Here, we aimed at elucidating whether organ‐specific imprinting of mature adult ECM provides essential and sufficient cues for tissue‐specific differentiation of uncommitted embryonic stage human mesodermal precursor cells. To minimize experimental variables such as species differences, medium components, decellularization methodologies, cell source, and differentiation stage, we established a humanized system, where these variables are controlled and standardized. In addition, we aimed at elucidating the nature of the ECM cues responsible for inducing organotypic cell specification and whether structural factors may also trigger differentiation. The results show that dECM retains distinct tissue‐specific memory, which is imprinted in adult ECM‐decorating heparin‐binding growth factors and act on immature embryonic stage precursor cells. 2 Results and Discussion 2. 1 Differentiation Specificity of Human Induced PSC‐Derived Mesodermal Precursor Cells (P‐meso) Depends on Tissue Origin of Decellularized ECM The use of ECM as a bioactive matrix for cells in vitro and in vivo is based on its interactive properties and tissue‐specific imprints, which are recognized as cues by cells to adopt tissue‐specific phenotypes. For example, decellularized kidney, lung, heart, or skeletal muscle ECM, have been used for recellularization with pluripotent stem cells, tissue committed precursor and mature cell types. Assessment of the resulting phenotypic patterns revealed in most cases variable proportions of tissue‐specific cell types. 3, 4, 13, 17, 19, 20, 21, 22, 23 However, the addition of differentiation‐inducing factors in the culture media, and the use of non‐human and non‐standardized dECM preparations made it difficult to distinct the role of the ECM in the observed organotypic determination. 24, 25 Here, we eliminated confounding variables to determine the potency of intrinsic imprinting of ECM. We omitted inductive factors within cultivation media, applied a standardized decellularization method for all tissues used, and used only human ECM and cells. Moreover, we tested early fetal stage not yet tissue‐committed mesoderm precursor cells on ECM from mature adult tissues most likely encountered by cells in regenerative medicine and in vitro modeling. Finally, we directly compared the effects of human kidney, heart, and lung ECM on these cells. Human kidney, heart, and lung tissues were decellularized and analyzed for preservation of structure and tissue architecture. For characterization, the 800 µm thick dECM slices were attached to glass slides for stabilization, improved handling, and reproducible analysis. Maintained collagen and laminin structures and absence of nuclear stain indicated complete cell removal and maintenance of major structural properties (Figure S1, Supporting Information). To investigate the effects of the tissue origin of the dECM on cell differentiation, human induced pluripotent stem cells (hiPSC) were differentiated into P‐meso, characterized by the uniform expression of T (Brachyury), HAND1, and goosecoid 26 and absence of endodermal and pluripotency marker expression (Figure S2, Supporting Information). The P‐meso cells were seeded on dECM from healthy adult kidney, heart, and lung tissue in an air‐liquid interphase (ALI) cultivation system in growth factor‐free medium for 14 days. We expected that mesoderm precursor cells may recognize marks on adult ECM‐derived from tissues of mesoderm origin such as kidney and heart, but not on dECM from endoderm‐derived tissues such as the lung. On kidney dECM, the cells were initially scattered, uniformly distributed, and some organization patterns were observed by day 3 ( Figure 1 a, e). By day 7, cells organized and formed tubular structures reminiscent of those found in nephron‐forming elements during renal development (Figure 1 b, f). Patterning continued until day 14, when the cells became more arranged, tubular elements were more prominent, rounded structures were increasing in number, and more elongated tubular structures appeared. Cells organized at the border of tubule‐ and blood vessel‐like matrix structures and densely grouped within glomerular areas (Figure 1 c, g). The renal cells started to express typical markers of differentiated nephronal cells, including epithelial cells of the proximal and distal tubules, loop of Henle, collecting duct, glomerular podocytes, and endothelial cells by day 7 (Figure S3a–j, Supporting Information). Expression of these markers was maintained until day 14 with increasing cell numbers and structural patterning ( Figure 2 a–j). Figure 1 Histological characterization of kidney, heart and lung dECM repopulated with P‐meso cells. Representative hematoxylin and eosin (H&E) staining of kidney, heart, and lung scaffold sections on days 3, 7, and 14 post seeding with P‐meso cells, and of respective native tissue sections. Overview of: a–c, e–g) recellularized kidney, i–k) heart, and m–o) lung dECM. b) Circles indicate circular and longitudinal tubule/vessels‐like structures. c) Tubule/vessel‐like structures (circles) appear more compact compared to day 14. d) Native human kidney. Circles indicate blood vessel and tubular structures. e–g) Higher magnification of a renal glomerulus in dECM (circles). g) Cells arrange similar to native glomerulus seen in (h). j) P‐meso cells on heart dECM clustered at day 7, when beating cells were first observed. k) Clusters increased in size and number by day 14 with continues beating. l) Native human heart. m–o) Cells scattered on lung dECM with no structural changes over time. p) Native human lung. Seven independent experiments were performed for each tissue with three different hiPSC‐lines. Scale bar: 75 and 20 µm. Figure 2 Immunohistochemical characterization of hiPSCs‐derived P‐meso cells on kidney, heart, and lung dECM. a–j) Kidney, heart, and lung cell markers were selected to identify different renal cell types, k–n) immature and mature cardiac cells, and o–r) different lung epithelial cell types: a, b) AQP1 and Na, K‐ATPase for proximal tubules, c) Calbindin for distal tubules, d) CK19 for loop of Henle, e, f) podocin and synaptopodin for glomerulus, g) CD144 for endothelial cells, h) AQP2 for collecting duct, i, j) CK18 and E‐Cadherin for nephron epithelia. k) MEF2C as cardiac progenitor marker, l–n) α‐actinin, c‐Troponin and MF20 as more mature cardiomyocyte markers, o, p) lung epithelia, q) alveolar type‐II cells, r) multiple lung cell types. Images are from different depths of the 800 µm thick ECM showing uniform cellular penetration. Scale bar: 20 µm. n = 7. s, t) Percentage of cells expressing kidney (s) and heart (t) markers at day 14 post differentiation induction. ± SEM, n > 2. To determine whether the P‐meso‐derived renal proximal tubular cells on kidney dECM have the capability of electrolyte reabsorption, we performed sodium uptake analysis. Exposure of the cells to ouabain enhanced sodium uptake by inhibiting Na, K‐ATPase in most of the cells ( Figure 3 ). Figure 3 Electrolyte reabsorption hiPSC‐meso‐derived cells on kidney dECM (day 14). a–c) Sodium‐green fluorescence demonstrates sodium uptake as observed by the intracellular fluorescence signal within tubular‐like structures (circles). d–f) Ouabain inhibition of Na, K‐ATPase increased intracellular sodium levels. g–i) No fluorescence was detected when sodium‐green was omitted. j) Percentage of cells absorbing electrolytes. Scale bar: 75 µm mean ± SEM, n = 2. Spreading and organization of the P‐meso cells on heart dECM was distinctly different from the pattern observed on kidney dECM (Figure 1 i–k). On day 3, in heart, dECM the cells were evenly scattered and accumulated into cell condensates by day 7, which started to beat (Video S1, Supporting Information) and to express typical markers of cardiomyocytes from day 7 (Figure S3k–n, Supporting Information) until at least day 14 (Figure 2 k–n), including the cardiac progenitor marker Myocyte enhancer factor 2C (MEF2C) and markers of more mature cardiac cells c‐troponin, α‐actinin, and myosin. Cell condensates were maintained by day 14 with increasing numbers of beating cell clusters (Figure 1 k), which were stable at least until day 30 when the experiment was terminated (not shown). In contrast, the P‐meso cells on lung dECM spread uniformly over the matrix and proliferated but did not show any differentiation pattern (Figure 1 m–o) or expression of the lung epithelial cell markers Prosurfactant Protein C (proSp‐C), Pan‐cytokeratin, Epithelial membrane protein 2 (EMP2), and Caveolin1 until day 14 (Figure 2 o–r; Figure S3o–r, Supporting Information). This corroborates our assumption that ECM of endoderm‐derived lung tissue is unable to support and promote mesoderm‐lineage specification and is unable to transdifferentiate iPSC‐derived mesoderm precursors into lung epithelial cells. However, CD144 positive endothelial cells, which are of mesoderm origin, were induced from the P‐meso cells on all three matrices ( Figure 4 h). The percentage expressions of different renal and cardiac markers at day 14 were quantified (Figure 2 s–t). Figure 4 Transcription of renal and cardiac markers in P‐meso‐derived cells on kidney, heart dECM, and single matrix proteins collagen IV, laminin, fibronectin, and geltrex. RNA expression analysis by qPCR reveals increased expression from day 7 to day 14 of tissue‐specific renal transcripts AQP1, Na, K‐ATPase, NCCT, CK19, AQP2, E‐Cadherin, and podocin only in cells on kidney dECM (a–h), and of cardiac transcripts NKX2. 5, MEF2C, GATA 4, MHC, MLC2, and troponin only in cells on cardiac dECM (i–n). h) The endothelial marker CD144 was expressed in cells on kidney, heart, and lung dECM. f) On geltrex, only E‐Cadherin was induced in P‐meso cells and detected at days 7 and 14. i) Similarly, the endothelial marker CD144 was induced by geltrex. j–n) The cardiac markers MEF2C and GATA4 were induced on geltrex on days 7 and 14 post seeding. Gene expression was normalized to the native human tissue, mean ± SEM, * p < 0. 005, n = 7. To elucidate whether cells integrate into the full depth of the 800 µm thick dECM slices, analysis of an average of ninety 5 µm sections taken from various tissue depths from recellularized kidney, heart, and lung slices demonstrated cell penetration throughout the full matrix thickness (Figures 1 and 2 ; Figures S3 and S4, Supporting Information). To confirm selective differentiation of P‐meso cells on cardiac and renal dECM, mRNA expression of renal, cardiac, and endothelial markers was determined by quantitative real‐time polymerase chain reaction (qPCR) on days 7 and 14 post seeding (Figure 4 a–n). Interestingly, expression of immature cardiac marker genes homeobox gene NKX2. 5, Myocyte enhancer factor 2 (MEF2C), and the zinc‐finger transcription factor GATA4 were expressed higher on day 7, with declining expression by day 14 (Figure 4 i–k), while expression levels of the more mature cardiac markers myosin light and heavy chain and cardiac troponin increased between days 7 and 14 (Figure 4 l–n). Only CD144 positive endothelial cells, which are of mesoderm origin, were induced from the P‐meso cells on dECM of all three tissues (Figure 4 h). To assess whether the matrix proteins Col IV, Laminin, Fibronectin, or the murine tumor‐derived matrigel (geltrex) are able to promote differentiation of P‐meso cells, we seeded the cells on these individual proteins and on geltrex. As for the dECM, the cells were cultured under ALI conditions without supplementing the media with differentiation‐inducing growth factors. Expression of renal, cardiac, and lung‐specific marker genes at days 7 and 14 was determined by qPCR. Low expression of the selected genes was detectable on any of the single matrix proteins. On geltrex, the cardiac markers GATA4 and MEF2C, the epithelial marker E‐Cadherin and the endothelial marker CD144 were expressed at days 7 and 14 (Figure 4 ). Interestingly, the distinct ability of kidney and heart dECM to differentiate early mesoderm into nephronal and cardiac cells was completely lacking on isolated matrix proteins such as Fibronectin, Laminin, and ColIV, heavily diminished on the murine tumor‐derived geltrex and absent on lung dECM. On lung dECM, the mesoderm cells failed to differentiate into lung cells, which derive from the endoderm lineage. However, mesoderm‐derived endothelial cells were readily detected in lung dECM. Interestingly, the use of porcine kidney dECM also induced differentiation of human P‐meso cells into cells of the renal lineage, however, expression of markers was reduced and not all renal cell types were detected (data not shown). These data indicate that the specific dECM decoration rather than common structural matrix proteins are responsible to germ‐line sensitive cell differentiation. To further confirm tissue specificity of differentiation, we tested the expression of kidney markers on heart dECM and of cardiac markers on kidney dECM on days 7 and 14 post seeding with P‐meso cells. No expression of cardiac cell markers on kidney dECM or of renal cell markers on heart dECM was observed. Similarly, no renal or cardiac cell markers were expressed on lung scaffolds (Figure S4, Supporting Information). Differentiation and specification of P‐meso cells into nephronal and cardiac cells using human kidney and heart dECM were reproduced with three different human iPSC‐lines: WISCi004‐A (IMR90), BCRTi005‐A, and BIHi004‐A (Figure S2, Supporting Information). The mesodermal markers‐expressing hiPSC‐derived P‐meso cells arise early in human embryogenesis and are plastic mesendodermal precursor cells committed to develop into a wide range of mesodermal cell types, including muscle, kidney, endothelia, and connective tissues. We posited that these cells are better suited than pluripotent cells or further tissue committed renal, cardiac, or endothelial precursor cells to elucidate the influence of adult ECM on cellular plasticity. Indeed, kidney dECM alone directed differentiation of these P‐meso cells into a variety of nephronal cells of the main structural components, including glomerulus, proximal and distal tubules, loop of Henle, collecting duct, and endothelial cells. Other studies using pluripotent hiPSC or embryonic stem cells (hESC) or kidney‐committed metanephric mesenchyme cells also detected the expression of some renal markers when seeded on kidney dECM. While hPSC differentiate also into renal cells, the efficacy and yield is variable and not quantitatively and qualitatively reproducible. The rather tissue restricted metanephric mesenchyme cells, in connection with inductive media components differentiated preferentially into renal tubular cells. 20, 27, 28, 29 A direct comparison on dECM from other tissues was not performed. We showed that heart dECM directed the P‐meso cells to cardiac progenitor and mature cells as early day 7 of culture in the absence of specific inducers with increasing expression of mature markers and reduction of early maturation markers over time. This was achieved with notable morphological and functional patterning in the used 3D in vitro dECM platform, where the P‐meso cells penetrated the full layer of the 800 µm thick dECM sections. Our defined and standardized 3D‐ECM model is thus generally suitable to investigate cell‐matrix signaling and offers an approach to efficiently generate renal and cardiac cells from hiPSC. 2. 2 Removal of HSPG from dECM Abolishes dECM Induced Tissue‐Specific Differentiation of P‐meso Cells HSPGs are ECM associated glycoproteins with the common characteristic of containing one or more covalently attached heparan sulfate (HS) chains. 30 HSPGs non‐covalently bind a number of chemokines, cytokines, enzymes, growth factors, or other bioactive molecules. 30 We hypothesize that HSPG‐binding heparin‐binding growth factors (HBGF) may be responsible for the observed tissue‐specific memory of dECM. When we assessed the presence of HBGFs vascular endothelial growth factor (VEGF), fibroblast growth factor 2 (FGF‐2), bone morphogenetic protein 2 (BMP‐2), hepatocyte growth factor (HGF), epidermal growth factor (EGF), platelet‐derived growth factor‐beta (PDGF‐BB), and transforming growth factor‐beta (TGF‐ß), these were all detectable in ECM even after decellularization, albeit at reduced levels compared to native tissues ( Figure 5 a–g). Heparitinase treatment to eliminate HSPGs removed these dECM‐bound HBGFs as shown exemplary for VEGF. Assessment of VEGF retention by immunostaining confirmed typical distribution patterns with highest VEGF concentrations in glomerular renal structures, which was partially maintained after decellularization and eliminated after heparitinase treatment, while the structural dECM proteins were preserved (Figure 5 h–j). Figure 5 HBGF concentrations in native kidney and heart tissues, dECM, and dECM‐HSPG. a–g) Detection of indicated HBGFs in native kidney, kidney dECM, and dECM‐HSPG by ELISA. h–j) VEGF detection in kidney dECM, native kidney, and dECM‐HSPG sections by immunohistochemistry. Arrows indicate glomeruli, with strongest VEGF expression. Masson trichrome staining shows collagen (blue). Mean ± SEM, * p < 0. 005, n = 3. Recellularization of heparitinase‐treated dECM (dECM‐HSPG) with P‐meso cells resulted in a complete loss of condensation or structural organization of the cells until day14 on both heart and kidney dECM. The cells were scattered in a disorganized manner on kidney dECM and no vesicle or tubule‐like structures formed. Similarly, on heart dECM, the cells were randomly distributed, did only loosely assemble (Figure S5, Supporting Information) and no beating cells were observed. Immunostaining for markers of differentiated renal or cardiac cells did not show expression on the respective kidney and heart dECM‐HSPG on day 7 or day 14 ( Figure 6 ; Figure S6, Supporting Information). This revealed that ECM‐inductive cues essential for renal and cardiac lineage differentiation and organizational patterning are dependent on HSPG and factors bound to HS. Figure 6 Removal of HSPGs abolishes induction of tissue‐specific cell types from P‐meso cells on kidney and heart dECM‐HSPG. Kidney and heart cell markers were used to identify different renal structures and immature and mature cardiac cell types at day 14 post seeding of P‐meso cells on dECM‐HSPG. a–h) No expression of kidney markers was detected for Na, K‐ATPase (proximal tubules), Calbindin (distal tubules), CK19 (loop of Henle), podocin and synaptopodin (glomerular podocytes), AQP2 (collecting duct), CK18, and E‐Cadherin (nephron epithelia). i–l) No expression was detected of cardiac markers MEF2C (cardiac progenitors), α‐Actinin, c‐Troponin, and MF20 (mature cardiomyocytes). Scale bar: 20 µm. Our results provide a direct comparison of dECM from different tissues and conclusively show that adult mature ECM harbors tissue‐specific imprints, which support organotypic differentiation even of early fetal‐stage uncommitted precursor cells. As HSPGs themselves may alone not drive differentiation, these imprints are very likely based on heparin‐binding factors decorating the ECM in a tissue‐specific pattern. The matrix‐associated HSPGs are secreted by the tissue cells to support their own homeostasis. 1, 29, 30, 31, 32, 33 We show that these HSPG‐binding factors remain to some degree in dECM preparations. When removed, no tissue‐specific differentiation was observed. The remaining ECM molecules such as collagen and laminin and the preserved physical ECM properties and architecture are alone not sufficient to induce tissue‐specific P‐meso differentiation. It is thus likely, that the degree to which the applied decellularization method preserves these HSPG‐binding factors influences the potency for specification by ECM preparations. Their comparative analysis may thus lead to the design of compositions that aid enhancement of ECM‐driven cell specification. For example, selective removal of HBGF by electrostatic intervention, which leaves HSPG intact, will allow the decoration of ECM with inductive HSPG‐binding proteins. It has been established that the patterns of ECM markings by HSPG, HBGFs, and ECM‐cleavage fragments are dynamic and change in tissues compromised by inflammation, fibrosis, and cell death. 34 Ex vivo or in situ ECM‐modulation and cell stage selection could thus improve stem cell integration and stem cell based repair in such compromised tissues, and our model provides new means for in vitro testing of such applications. 3 Conclusion Human dECM from aged, adult tissues carries factors, which directs organ‐specific differentiation of uncommitted iPSC‐derived mesoderm cells. This tissue memory is lost when HSPGs embedded in the dECM are removed. The direct comparative analysis of dECM from three different organs was possible by strict standardization and reduction of confounding variables, including the use of only human dECM, of iPSC‐derived mesoderm cells only committed to mesoderm, but not to a specific tissue fate, and of medium without inductive factors that could conceal ECM effects. The data indicate that uncommitted hiPSC‐derived cells will recognize cues in adult tissue dECM and react by tissue specification. The method and findings will allow pinpointing HSPG‐associated molecules essential for cell specification and utilization of dECM and its modification to generate functional niches for tissue engineering. 4. Experimental Section Tissues and Cell Lines Normal human kidney tissue was obtained after nephrectomy because of the presence of renal tumors. Normal cardiac and lung tissues were collected from explanted hearts and lung patients who underwent heart and lung transplantation. The tissues were evaluated by a pathologist and normal tissue areas were cut from the healthy region of the organs, to exclude diseased tissue. Tissues were obtained after informed consent based on approvals from the Ethics Commission of Charité with approval numbers for kidney (EA1/134/12), heart (EA4/028/12), and lung (EA2/079/13). Tissues from a total of 14 different kidneys, 5 different hearts, and 4 different lungs were obtained (Table S3, Supporting Information), cut into ≈1 cm 3 cuboid pieces and stored at −80 °C before further processing. The human iPSC lines WISCi004‐A (IMR90), obtained from WiCell, BIHi004‐A, and BCRTi005‐A, generated at Charité, were used (information available at https://hpscreg. eu ). Decellularization Tissue pieces were sectioned into 800 µm thick slices on a cryostat (Leica CM 1950) and one slice per well put into a 6‐well culture plate in distilled water. For decellularization, slices were washed with ice cold water for 2 h at 4 °C before being subjected to 0. 1% w/v sodium dodecyl sulfate (SDS) (Sigma‐Aldrich), pH 7. 5 at room temperature for 3 h. The detergent was changed every 30 min. Slices were washed with water at 4 °C for 30 min followed by incubation in 350 IU mL −1 DNase1 (Roche) in phosphate buffered saline (PBS) pH 7. 5 for about 2 h at 4 °C. The slices were again washed in water at 4 °C for 30 min and incubated at 4 °C in PBS supplemented with 100 U mL −1 penicillin and 100 µg mL −1 streptomycin for 2 h. All incubation and wash steps were performed with agitation. To remove heparin sulfate proteoglycans together with the bound growth factors, dECM was treated with heparitinase‐1 solution (0. 1 m sodium acetate, 10 m m calcium acetate, and 10 mU heparitinase‐1) (amsbio) for 3 h at 37 °C and washed three times each for 5 min with sterile PBS as described in ref. 14. Histological Analysis For histological and immunohistochemical analysis, tissue and dECM slices were fixed in 4% phosphate‐buffered formaldehyde solution (PFA) (Carl Roth) for 60 min at room temperature, embedded in paraffin and cut into 5 µm sections. Before staining, sections were deparaffinized and subjected to H&E staining (Sigma‐Aldrich, Carl Roth) according to established protocols. 35 Imaging was performed using an inverse microscope (Axio Observer Z1). Quantification of dECM‐Associated Heparin‐Binding Growth Factors Native tissue, dECM, and dECM treated with heparitinase‐1 (dECM‐HSPG) were grinded into powder in liquid N2. The powder was lyophilized and the dry weight determined and dissolved in 1 mL of RIPA buffer (150 m m NaCl, 50 m m Tris, 1% TX‐100, 0. 5% SDC, 0. 1% SDS, pH: 7. 4). Lysates were sonicated for 20 s, incubated for 24 h at 4 °C on a shaker and centrifuged at 13000 × g for 10 min. The following heparin‐binding growth factor concentrations were determined by enzyme‐linked immunosorbent assay (ELISA): FGF‐2, VEGF, HGF, EGF, PDGF‐BB, TGF‐ß using the respective Quantikine (R&D systems) and BMP2 (abcam) kits. All assays were performed according to the manufacturer's instructions. Absorbance was measured at 450 and 650 nm (Spectra Max 340C). Cytokine concentrations were normalized to the tissue dry weight. Mesodermal and Endodermal Differentiation of Human iPSC Human iPSC‐lines were kept in culture in TeSR‐E8 medium (Stem Cell Technologies) on geltrex (Life Technologies) coated dishes. Cells were fed daily and passaged every 4–6 days with gentle cell dissociation reagent Trypsin‐EDTA (Biochrome AG) for 5 min at 37 °C and then manually detached from the dish using a cell scraper. The resulting clumps of cells were plated in a ratio of 1:6. For hiPSC‐differentiation to mesodermal cells, the protocol established by Orlova et al. was used. 26 Briefly, hiPSC were seeded on geltrex coated dishes and incubated in APEL‐2 (Stem Cell Technologies) and protein‐free hybridoma medium (PFHMII) (Life Technologies) and differentiation induced by the addition of CHIR99021 (1. 5 µ m ) (Tocris), BMP4 (30 ng mL −1 ), activin A (25 ng mL −1 ), and VEGF (50 ng mL −1 ) (all from Peprotech). On day 3, the factors were removed, and cultivation continued with VEGF (50 ng mL −1 ) and the TGFβ pathway inhibitor SB431542 (10 µ m ) (abcam). On day 4, hiPSCs‐derived mesodermal cells (P‐meso) were harvested and used for recellularization. To induce endoderm, hiPSC were incubated for 24 h at 37 °C in differentiation medium (RPMI1640/L‐Glu, B27, 100 ng mL −1 activin A, 3 µ m CHIR99021, 10 µ m ROCK‐Inhibitor), washed twice in 2 KnockOut‐DMEM‐F12, passaged using Accutase. Single cells (2. 0 × 10 5 cells cm −2 ) were subsequently incubated in differentiation medium for 24 h at 37 °C/5% CO 2 Incubator. Medium was subsequently supplemented with 0. 5 m m sodium butyrate, and cells further incubated for 2–6 days at 37 °C/5% CO 2 with daily medium change. Recellularization of dECM with P‐meso Cells Approximately 800 µm thick slices of kidney, heart, and lung dECM and heparitinase‐1 treated dECM, were placed on hydrophobic floating membranes (Whatman) in a six‐well plate to provide an ALI condition (Figure S2f, Supporting Information). To analyze single matrix proteins and geltrax, membranes were coated with the individual matrix proteins Collagen IV (Sigma‐Aldrich), laminin (Biolamina), fibronectin (Corning), and geltrex (Life Technologies). 500 000 hiPSC‐meso cells were placed on the different matrix preparations in APEL‐2 + PFHMII. Medium was supplemented with Rock Inhibitor Y‐27632 (WAKO) for the initial 24 h. Cultivation continued for up to 14 days in APEL‐2 + 5% PFHMII and medium was changed every 48 h. For histology and immunohistochemistry, cells were fixed at days 3, 7, and 14 in 4% PFA at room temperature for 1 h and embedded in paraffin for further analysis. Immunohistochemistry For immunostaining, 5 µm paraffin sections cut from fixed dECM or recellularized dECM were deparaffinized and subjected to antigen retrieval solution (DAKO). The sections were permeabilized with 0. 1% TX‐100 in PBS pH 7. 4 (T‐PBS) three times for 5 min, blocked for 10 min with 1% bovine serum albumin (BSA) (Sigma‐Aldrich) in T‐PBS and for 60 min with 5% donkey serum (Merck Millipore), and 1% BSA in PBS before immunostaining. Primary antibodies were applied overnight at 4 °C; all antibodies were diluted in 5% donkey serum and 1% BSA in PBS. The sections were then washed with T‐PBS three times for 5 min each, and incubated with secondary antibodies for 1 h at room temperature. Finally, after washing in TPBS three times for 5 min each, sections were mounted with immunoselect‐antifading mounting medium including 4′, 6‐diamidino‐2‐phenylindole (DAPI) (Dianova). The same protocol was used for negative control staining, except that the primary antibody was omitted. For a list of the antibodies used, see Table S1, Supporting Information. To detect VEGF, anti VEGF (Santa Cruz Biotechnology Inc. ) and secondary antibody IgG H&L (horse‐radish peroxidase, HRP) preadsorbed (abcam) were used, followed by 3, 3′‐diaminobenzidine (DAB)/plus (abcam) chromogen detection. To visualize structures, the sections were counterstained with Masson trichrome (Sigma‐Aldrich) using a standard protocol. 35 Imaging was performed using either an inverse microscope (Axio Observer Z1, Carl Zeiss) or the Operetta high content imager and Columbus image analysis server (both PerkinElmer). Quantitative Real‐Time Polymerase Chain Reaction (qPCR) RNA was extracted from recellularized dECM cultures at days 7 and 14 using picopure RNA isolation kit (ThermoFisher Scientific) and cDNA was prepared by TaqMan Reverse Transcription Reagents (ThermoFisher Scientific). The components of SensiFAST SYBR Hi‐ROX (Bioline), were mixed with the yielded cDNA. Specific genes were amplified by the application of primers listed in Table S2, Supporting Information. For each primer an additional negative control was applied (without cDNA). The real‐time PCR‐QuantStudio 6Flex (Applied Biosystems, Life Technologies) was used and run with 40 cycles. Functional Assay of dECM‐Induced Proximal Tubular Epithelial Cells Electrolyte reabsorption assays were performed using NaCl as electrolytes to examine functional properties of cells expressing renal proximal tubule epithelial cell markers. Cellular sodium green (ThermoFisher Scientific) uptake was evaluated as described. 36 Recellularized kidney dECM at day 14 of cultivation was incubated with 10 m m sodium green in 90 m m NaCl, 60 m m N ‐methyl‐ d ‐glucamine, 2 m m NaH 2 PO 4, 5 m m KCl, 1 m m CaCl 2, 1. 2 m m MgSO 4, 32 m m 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES), 10 m m glucose at pH 7. 4 for 60 min at room temperature, and washed with PBS. To assess specificity of cellular sodium uptake, cells were incubated with 50 µ m ouabain (Sigma‐Aldrich) for 1 h to restrain Na/K ATPase. The incubated samples were washed with PBS and fixed in 4% PFA, nuclei were stained with DAPI and uptake of sodium was visualized by fluorescence microscopy (Axio Observer Z1, Carl Zeiss). Statistical Analysis Quantitative results are reported as mean ± standard error of the mean (SEM). Statistical comparisons were performed using unpaired t ‐tests, unless specified otherwise. For multiple comparisons of more than two groups, one‐way ANOVA was used with Tukey's multiple comparison post hoc tests for significance. GraphPad Prism 5 (GraphPad Software, La Jolla, USA) was used for statistical analysis. p < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest. Author Contributions I. U. and A. K. conceptualized the study; I. U. , J. F. B. , A. R. , B. E. , C. S. , C. K. , S. H. , P. R. and A. K. provided the methodology; I. U. and A. K. performed the analysis and investigation; I. U. , J. F. B. , A. R. , B. E. , C. S. , C. K. , S. H. and A. K. provided resources and materials; I. U. and A. K. drafted the manuscript; I. U. , J. F. B. , A. R. , C. K. , S. H. , P. R. , A. K. reviewed and edited the draft, with input from all authors; J. F. B. , P. R. and A. K. provided supervision to the experiments; A. K. managed funding acquisition. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file.
|
10. 1002/advs. 201901240
| 2,019
|
Advanced Science
|
Tissue‐Engineered Neural Network Graft Relays Excitatory Signal in the Completely Transected Canine Spinal Cord
|
Abstract Tissue engineering produces constructs with defined functions for the targeted treatment of damaged tissue. A complete spinal cord injury (SCI) model is generated in canines to test whether in vitro constructed neural network (NN) tissues can relay the excitatory signal across the lesion gap to the caudal spinal cord. Established protocols are used to construct neural stem cell (NSC)‐derived NN tissue characterized by a predominantly neuronal population with robust trans‐synaptic activities and myelination. The NN tissue is implanted into the gap immediately following complete transection SCI of canines at the T10 spinal cord segment. The data show significant motor recovery of paralyzed pelvic limbs, as evaluated by Olby scoring and cortical motor evoked potential (CMEP) detection. The NN tissue survives in the lesion area with neuronal phenotype maintenance, improves descending and ascending nerve fiber regeneration, and synaptic integration with host neural circuits that allow it to serve as a neuronal relay to transmit excitatory electrical signal across the injured area to the caudal spinal cord. These results suggest that tissue‐engineered NN grafts can relay the excitatory signal in the completely transected canine spinal cord, providing a promising strategy for SCI treatment in large animals, including humans.
|
1 Introduction Spinal cord injury (SCI) causes irreversible tissue loss, including neurons in the gray matter and nerve tracts and oligodendrocytes in the white matter. 1 Spontaneous regeneration after SCI is limited, and there are no satisfactory interventions to enhance endogenous regeneration to replenish lost tissue, resulting in functional deficits below the level of injury. 2, 3 Reconstruction of neural tissue in the injured area using exogenous neural grafts may supplement the injury gap with an appropriate neural population to facilitate motor, sensory, and autonomic functional recovery. 4, 5 The rapid development of stem cell therapy and tissue engineering technology offers a promising therapeutic strategy for SCI. 6 For example, neurons and glia derived from grafted human neural stem cells (NSCs) were able to replace lost neurons and glia excised by hemisection surgery, and they functioned as interneurons to reconnect severed neural circuits in a rodent SCI model. 7 A prodifferentiation regimen is required to increase neuronal yield from NSCs, either endogenous or exogenous, because they are prone to differentiate into astrocytes in the post‐SCI milieu. 8 When NSCs were delivered in a fibrin–thrombin matrix containing trophic factors including brain‐derived neurotrophic factor (BDNF), neurotrophin‐3 (NT‐3), platelet‐derived growth factor (PDGF‐AA), insulin‐like growth factor 1 (IGF‐1), epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), acidic fibroblast growth factor (aFGF), glial cell line‐derived neurotrophic factor (GDNF), hepatocyte growth factor (HGF), and calpain inhibitor, they mainly differentiated into neurons in the spinal cord. 9, 10 However, concerns were raised regarding the risk of long‐distance migration and ectopic neurogenesis by donor NSCs or neural progenitor cells that were transplanted immediately after SCI modeling in a trophic factor‐enriched microenvironment. 11 Thus, an alternative approach would be to augment in situ neuronal differentiation at the injury site through application of functionally defined single reagents such as the histone deacetylase inhibitor valproic acid (VPA), 12 the EGF receptor signaling antagonist cetuximab, 13 the microtubule‐stabilizing agent paclitaxel, 14 and others. To drive neuronal differentiation of NSCs with minimal adverse effects on host homeostasis, the doses of these adjuvant compounds should be fine‐tuned and their release should be spatiotemporally controlled, which is a substantial engineering challenge. Our team has established a tissue engineering methodology to construct neural network (NN) tissue in vitro using biocompatible bioscaffolds, NSCs, and lentiviral‐based trophic factor delivery. 15 Our previous results showed that NSCs expressing a genetically modified version of NT‐3 receptor TrkC can be induced to differentiate into neurons with mature neuronal properties including firing action potentials and trans‐synaptic electrochemical activities when they were cocultured with Schwann cells (SCs) with genetically modified NT‐3 in a 3D gelatin sponge scaffold. Together with deposited extracellular matrix (ECM), SCs and NSC‐derived neurons formed a homeostatic system allowing for intercellular interactions between cells and the ECM. This tissue engineering construct functions as NN tissue. 16 Previous studies showed that prebuilt NN tissue survived the hostile post‐SCI microenvironment for up to 2 months and acted as a relay to repair neural circuits in a rat‐transected SCI model without ectopic colony formation. 17, 18 Results from rat SCI models support our hypothesis that transplantation of a tissue engineering NN tissue is a promising therapeutic approach to repair SCI. However, given significant interspecies differences in terms of anatomy, pathology and pathophysiology, findings derived from rodent models may not generalize to humans. 19, 20 Therefore, large animal models should be used before clinical studies to bridge the gap between rodent and human studies. The canine model has several innate advantages for studying therapeutic SCI interventions. 21, 22 First, canines with SCI underwent similar pathological processes and repair mechanisms as observed in patients. 23 Compared with other large animal SCI models, postoperative care is easier to administer in canines with SCIs. Moreover, domesticated canines are more compliant with multimodal behavioral tests than other large animals, allowing for more reliable evaluation of therapeutic effects. 21, 24 We recently established a complete SCI canine model and tested the therapeutic effects of canine bone marrow mesenchymal stem cell (MSC)‐derived NN tissue on motor function recovery. 16 Although MSC‐derived NN tissues exhibit potent therapeutic efficacy and translational value, the mixed population of neuron‐like cells and other MSC‐derived cell populations made analysis of functional recovery mechanisms difficult. How exogenous neurons participate in repairing damaged neural circuits in large animals with considerable spinal cord tissue loss remains to be determined. Therefore, we adopted an established complete SCI canine model to assess the effects of NSC‐derived NN tissue on SCIs with large tissue deficits. Our aim was to evaluate whether and how tissue engineered NSC‐derived NN tissues could integrate with host neural circuits and serve as neuronal relays to help cortex excitatory signal cross the lesion gap to the caudal spinal cord. 2 Results 2. 1 Canine NSCs and SCs Culture and Gene Modification NSCs were isolated from the hippocampus, and nestin‐positive neurospheres were transfected with a lentivirus carrying a TrkC coding sequence (pLent‐EF1a‐TrkC‐Flag‐CMV‐GFP‐P2A‐Puro). After transfection, nearly all cells within neurospheres were GFP positive, with most expressing the NSC marker nestin (Figure S1A, Supporting Information). SCs were bipolar in shape after two passages. S100 immunofluorescence staining showed that ≈90% of the bipolar cells were SCs (Figure S1B, Supporting Information). SCs were genetically modified with an NT‐3 lentivirus (pLent‐EF1a‐NT‐3‐Flag‐CMV‐P2A‐Puro). After 14 d of coculture in collagen sponge (CS) scaffolds, > 90% of NSCs maintained TrkC and GFP expression (T‐NSCs, Figure S1C, Supporting Information). Most SCs (GFP negative) maintained NT‐3 expression (N‐SCs, Figure S1D, Supporting Information). The experiment groups were as follows: NSCs, NSCs+SCs, T‐NSCs+SCs, NSCs+N‐SCs, and T‐NSCs+N‐SCs. 2. 2 Phenotypic and Functional Identification of NSC‐Derived NN Tissue In Vitro To construct NSC‐derived NN tissues, T‐NSCs and N‐SCs were cocultured in a 3D CS scaffold for 14 d (the T‐NSCs+N‐SCs group, Figure 1 A). Immunofluorescence staining of slices from the T‐NSCs+N‐SCs group showed that NSCs differentiated into neurons expressing microtubule‐associated protein 2 (Map2, a neuronal marker) and postsynaptic density protein 95 (PSD95, a postsynaptic marker, Figure S1E, Supporting Information). These neurons also showed immunoreactivity for choline acetyl transferase (ChAT, a cholinergic marker, Figure S1F, M, Supporting Information), glutamate decarboxylase 67 (GAD67, a GABAergic marker, Figure S1G, M, Supporting Information), and Ca 2+ /calmodulin‐dependent protein kinase 2α (CaMK2α, an excitatory marker, Figure S1H, M, Supporting Information). SCs within the NN tissue were GFP negative and expressed myelin basic protein (MBP, a myelin marker, Figure S1I, Supporting Information) or glial fibrillary acidic protein (GFAP, an astrocyte marker, Figure S1J, Supporting Information). Figure 1 Construction of canine NSC‐derived NN tissue. A) Schematic diagram illustrating construction of canine NSC‐derived NN tissue in a 3D CS scaffold. NSCs with genetically modified TrkC (TrkC‐NSCs) were cocultured with SCs with genetically modified NT‐3 (NT‐3‐SCs) in the 3D CS scaffold for 14 d. The H&E staining revealed the cross section of the 3D collagen sponge containing cells. B) SEM showing neuron‐like cells with long processes (green), oligodendrocyte‐like cells with multiple short processes wrapping other processes (red), and SC‐like cells (blue) in NSC‐derived NN tissue. C–E) TEM images showing features of synapses (presynaptic element, green in (C) and postsynaptic element, red in (C)) between processes of two neurons. These were characterized by the presence of vesicular profiles (arrows in (D) and (E)) in one process and a layer of electron‐dense structures (arrowheads in (D) and (E)) in the other. F) Multilamellar myelin sheaths (arrows) wrapping an axonal profile (asterisk) in the NN tissue. G1, G2) Neurons unloaded prelabeled FM1‐43 dye (red, white arrows) following membrane depolarization triggered by high [K + ] stimulation, as shown by H) the steep drop of fluorescence intensity after stimulation. I) Whole cell patch‐clamp recording showed action potentials of neurons following J) 21 d of culture. K) High‐frequency miniature excitatory postsynaptic currents (mEPSCs) and L) miniature inhibitory postsynaptic currents (mIPSCs) were detected. Scale bars = 500 µm in (A); 10 µm in (B); 0. 5 µm in (C); 100 nm in (D) and (E); 200 nm in (F); 20 µm in (G1), (G2), and (I). Western blot analysis showed that protein levels of neurofilament 200 (NF, a neuron marker), PSD95, synaptophysin (SYP, a presynaptic marker), ChAT, and GAD67 were highest in the T‐NSCs+N‐SCs group (Figure S1N, O, Supporting Information), suggesting that T‐NSCs had the greatest levels of neurogenesis, synapse formation, and neurotransmitter synthesis when cocultured with N‐SCs in vitro. MBP expression levels indicated that the T‐NSCs+N‐SCs group had the highest potential for myelination in vitro (Figure S1N, O, Supporting Information). Scanning electron microscope (SEM) analysis showed extensive contacts between long processes of SC‐like, oligodendrocyte‐like, and neuron‐like cells (Figure 1 B). Synaptic features, such as synaptic vesicles at axonal terminals, and PSD formation were detected in neurons within the NN tissue using transmission electron microscope (TEM, Figure 1 C–E). High magnification of the boxed areas revealed the details of the relatively mature synapse (Figure 1 D), and relatively immature symmetrical synapses (Figure 1 E). The results suggest dynamic changes in the formation synapses in NN tissue. Furthermore, multiple lamellar structures appeared to form myelin sheaths around axons in the NN tissue (Figure 1 F). When a high [K + ] solution containing N ‐3‐triethylammonmpropyl‐4‐4‐(dibutylamino) styryl (FM1‐43) dye was added to the culture dish containing the T‐NSCs+N‐SCs scaffold, the cells endocytosed the FM1‐43 dye, rendering the membrane red (Figure 1 G1) when visualized using a fluorescence microscope. When these cells were restimulated with a high [K + ] solution without FM1‐43 dye, fluorescence intensity rapidly decreased (Figure 1 G2, H), indicating that the cells had the ability to quickly exocytose FM1‐43 dye in a manner similar to synaptic vesicle release. 25 Based on these findings, whole‐cell patch clamp was used to evaluate the electrophysiological properties of NSC‐derived neuron‐like cells (Figure 1 I, J). Evoked action potentials were recorded in 8/10 neuron‐like cells in the T‐NSCs+N‐SCs group. Postsynaptic currents including miniature excitatory postsynaptic currents (mEPSCs) and miniature inhibitory postsynaptic currents (mIPSCs) were detected (Figure 1 K, L), suggesting trans‐synaptic communication. These findings suggest that canine NSCs expressing genetically modified TrkC differentiated into neurons with neuronal phenotypes, synaptic structure, and electrophysiological function when cocultured with SCs with genetically modified NT‐3 in the 3D CS scaffold for up to 14 d. The scaffold provided a matrix for cell adherence and growth, and an anchoring surface for ECM deposition (Figure S1K, L, Supporting Information). This allowed differentiating cells, their surrounding ECM, and the CS to develop into NN tissue after 14 d of culture. 2. 3 Transplantation of NSC‐Derived NN Tissue Improved Hindlimb Motor Function After cord transection at the T10 spinal cord segment, canines completely lost sensory and motor function below the injury level. Motor function of the pelvic limbs in the NN group (transplantation of NN tissue into the 4 mm injury gap of spinal cord, n = 11), CS group (implantation of CS scaffold into the 4 mm injury gap, n = 10), and SCI group (removal of 4 mm spinal cord tissue without any implantation, n = 4) was assessed using the Olby scoring scale ( Figure 2 A, B). At 4 weeks after SCI, deep pain sensation in response to pinch was recovered in canines in all three groups (Olby score for the NN, CS, and SCI groups: 1. 32 ± 0. 47, 1. 15 ± 0. 36, and 1. 00 ± 0. 00). At 8 weeks after SCI, gradual recovery of pelvic limb motor function was observed in three groups. Two canines (2/11) in the NN group showed occasional multijoint swings. Pelvic limb protractions were also observed in the CS and SCI groups but with much lower frequency (Olby score for NN: 3. 95 ± 0. 98, CS: 2. 80 ± 0. 60, and SCI: 2. 75 ± 0. 43). Twelve weeks after SCI, pelvic limbs swings of canines in the NN group were more frequent. Two (2/11) in the NN group showed occasional weight‐bearing protractions. Six (6/10) canines in the CS group showed more than two pelvic limb joint movements but without weight‐bearing protraction. Canines in the SCI group showed mild motor function improvement in the pelvic limbs (Olby score for NN: 4. 7 ± 0. 95, CS: 3. 45 ± 0. 50, SCI: 3. 00 ± 0. 50). At 24 weeks after SCI, nine (9/11) canines in the NN group presented with frequent weight‐bearing gaits of the pelvic limbs for short distances (1–5 m each time, Video Clip 1 for the best performance and Video Clip 2 for average performance). In contrast, no weight‐bearing movement was observed in the CS or SCI group (Olby score for NN: 7. 08 ± 1. 50, CS: 4. 70 ± 0. 64, SCI: 3. 88 ± 0. 78; Video Clip 3 for the CS group and Video Clip 4 for the SCI group), consistent with observations in chronic SCI patients who have recovered mild motor function. This result suggests that transplantation of NSC‐derived NN tissue has therapeutic efficacy for restoring motor function following SCI. Figure 2 Behavior, imaging, and electrophysiological observations following NN tissue implantation. A) Gradual recovery of pelvic limb motor function was observed from the fourth week after SCI in all canines in all three groups. Olby scores in the NN group were higher than those in the CS or SCI groups 8–24 weeks after SCI. B) At 24 weeks after surgery, weight‐bearing gaits were frequently encountered in the NN group but not in the CS or SCI groups. Nor, Normal group. C) At 24 weeks after surgery, MRI confirmed complete transection injury at the T10 level in the NN, CS, and SCI groups. Canines in the NN group displayed a narrower gap between the two ends of the transected spinal cord as determined by MRI and more significant nerve tract regeneration as determined by DTI, compared with the CS and SCI groups. D) FA values in the areas rostral or caudal to the I/G site in the NN group were higher than those in the CS and SCI groups. E) ADC values in the I/G site and areas rostral or caudal in the NN group were lower than those in the CS and SCI groups. F) Schematic diagram showing the stimulation and recording sites for cortical motor evoked potentials (CMEPs). G) CMEPs were detected at the L1–L2 segments of the spinal cord in the NN group ( n = 11), relative to negligible CMEP waves in the CS ( n = 10) and SCI groups ( n = 4). H) Representative spinal somatosensory evoked potentials (SSEPs) were detected in 3/11 canines in the NN group, but not in the CS and SCI groups. Bar charts for I) CMEP latency and J) amplitude showing significantly shorter latency and higher amplitude in the NN group relative to the CS and SCI groups (* p < 0. 05). Nor, Normal group. To better evaluate coordinated gaits, an underwater treadmill test was used to observe locomotion in a microgravity environment (Figure S2, Supporting Information). Canines in all groups were able to regain single joint movement on the underwater treadmill 8 weeks after SCI. Locomotion recovery was not significant in the CS and SCI groups; however, pelvic limb movement in the NN group showed continuous improvement. For example, coordinated inter‐pelvic limb stepping was first observed at 12 weeks after SCI. Up to 24 weeks, frequent front‐pelvic limb coordinated stepping was observed, suggesting that NSC‐derived NN tissue may help repair the neural circuits controlling the front and pelvic limbs. Canines in the CS and SCI groups did not regain coordinated stepping even 24 weeks after SCI (Video Clip 5). 2. 4 Imaging and Electrophysiological Assessments Quality control of transection modeling and nerve tract regeneration was evaluated using neuroimaging. Magnetic resonance imaging (MRI) findings showed loss of spinal cord continuity at the injury site after SCI. At 24 weeks after surgery, the gap between the rostral and caudal stumps of the transected spinal cord was seen in all groups. It appeared to be narrower in the NN group compared with the CS and SCI groups (Figure 2 C, top panels). Diffusion tensor imaging (DTI) was used to show tractography of the severed descending and ascending neural tracts (Figure 2 C, bottom panels). At 24 weeks after SCI, DTI showed that a number of nerve bundles extended into the injury/graft (I/G) site from both the rostral and the caudal stumps in the NN group. This was in sharp contrast to the nerve fiber bundles reconstructed in the CS and SCI groups that exhibited mild regeneration into the injury site. Fiber tractography and measurement of fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values are algorithms derived from MRI data to quantify nerve fiber bundle continuity. FA values in the NN group were higher in the areas rostral and caudal to the injury site compared with the CS and SCI groups (Figure 2 D). In contrast, ADC values in the NN group were lower in the injury site and the areas rostral and caudal to the injury site compared with the CS and the SCI groups (Figure 2 E). These results suggest that transplantation of NSC‐derived NN tissue facilitated host nerve fiber regeneration after SCI. Electrophysiological results (Figure 2 F–J) showed that canines in the NN group regained cortex motor evoked potentials (CMEPs) transmission across the I/G site to the lumbar 2 (L2) segment of the spinal cord (Figure 2 F) when the motor cortex was electrically evoked. CMEP latency was significantly shorter in the NN group (25. 26 ± 3. 92 ms) compared with the CS (46. 20 ± 7. 21 ms) and SCI (46. 20 ± 7. 21 ms) groups (Figure 2 G, I), suggesting faster electrical transmission spreading across the injury site. In addition, CMEP amplitude was evidently higher in the NN group compared with the CS and SCI groups (Figure 2 G, J), suggesting that more motor pathway neurons were electrically activated. There were 3 canines (3/11) in the NN group for which spinal somatosensory evoked potentials (SSEPs) were recorded (Figure 2 H). The CS and SCI groups failed to exhibit typical SSEP waves (Figure 2 H). These results indicated that implanted NSC‐derived NN tissue may contribute to improved nerve conductivity in the I/G site. 2. 5 Integration of NSC‐Derived Neurons in the Injured Spinal Cord Eight weeks after transplantation of NN tissue, numerous GFP‐positive cells were found within the I/G site ( Figure 3 A). A portion of the donor cells expressed growth‐related protein 43 (GAP43, Figure 3 B), a marker for the growth cone of nerve fibers, suggesting dynamic axon growth. Long processes extending from donor cells in the I/G site (4 mm long) were observed in the areas rostral or caudal to the I/G site. Some processes were immunoreactive for NF (a marker for nerve fibers) or Map2, and created contacts with host neurons or nerve fibers (Figure 3 C–F). Some of these contacts were immunoreactive for either PSD95 or SYP (Figure 3 C–E), suggesting the presence of synaptic connections between the donor and host neurons. In addition, some donor nerve fibers extending to the area caudal to the I/G site made close contacts with host neurons that were positive for vesicular glutamate transporter 1 (VGluT1, a marker of glutamatergic neurons, Figure 3 F), suggesting that excitatory synapses were established between the donor and host neurons. TEM verified the formation of synapses between neurons in the I/G site (Figure 3 G, H). Asymmetrical membrane features and the presence of spherical vesicles in the presynaptic components may indicate the presence of excitatory synapses in the I/G site of spinal cord. Figure 3 Survival and integration of NSC‐derived neurons in the injured spinal cord. A) Representative image of surviving GFP‐positive cells in the I/G site in the NN group 8 weeks after transplantation. B) GFP‐positive donor cells expressed both GAP43 and NF (arrows) in the I/G site. C–E) GFP‐positive cells extended long NF‐positive nerve fibers into the rostral and caudal areas of the I/G site, making contacts with host nerve fibers expressing the postsynaptic marker PSD95 (arrows in (C) and (D)) or expressing the presynaptic marker SYP (arrows in (E)). F) GFP‐positive nerve fibers either expressed VGluT1 or made close contact with host VGluT1‐positive fiber terminals in the area caudal to the I/G site (arrows), suggesting the presence of excitatory synapses established between donor and host neurons. Hoe, Hochest33342. G, H) TEM showed that neuronal connections in the center of the I/G site exhibited asymmetrical synapse features, as characterized by aggregation of synaptic vesicles (arrow), presynaptic components (H), and focal condensation (arrowheads) of postsynaptic membranes (H). Scale bars = 1 mm in (A); 20 µm in (B)–(F); 0. 5 µm in (G); 200 nm in (H). 2. 6 Phenotypic Characterization of NSC‐Derived Cells and Reinnervation of the I/G Site At 24 weeks after SCI, immunofluorescence staining was used to determine the types of cells in the NSC‐derived NN tissue in vivo. A portion of GFP‐positive cells coexpressed Map2 and ChAT or GAD67 or glutamate (Glu, a marker of glutamatergic neuron), suggesting the formation of a mixed population of excitatory and inhibitory neurons ( Figure 4 A–C, K). In addition, a small portion of the donor cells expressed GFAP (9. 35 ± 1. 76%, n = 5). Expression of nestin (an NSC marker) or MBP was rarely detected in GFP donor cells (Figure 4 D, E). In addition, there were numerous GFP‐positive cells expressing TrkC, surrounded by GFP‐negative cells expressing NT‐3 (Figure 4 F). Figure 4 Donor cells in grafted NN tissue maintain a neuronal phenotype and contribute to enhanced innervation in the I/G site at 24 weeks. A–C) Most GFP‐positive cells were Map2‐positive neurons that expressed the neurotransmitter marker ChAT (arrow in (A)), GAD67 (arrow in (B)), or Glu (arrow in (C)). D, E) A small population of donor cells showed GFAP immunoreactivity (9. 35 ± 1. 76%, n = 5). The NSC population, as shown by nestin immunostaining (arrowhead in (D)), and oligodendrocyte population, as shown by MBP immunostaining (arrows in (E)) were negligible among GFP donor cells. F) Expression of TrkC (arrow) in GFP‐positive cells and expression of NT‐3 (arrowhead) in adjacent cells. G) Immunofluorescence staining for NF shows innervations of the I/G site and adjacent rostral and caudal areas. Hoe, Hochest33342. H) Nerve fiber outgrowths for GFP‐positive neurons (arrows) contributed to innervation in the I/G site. Host NF‐positive nerve fibers traveled longitudinally through the I/G site (arrowheads). I) Histogram showing that the NF‐positive axon density was highest in the NN group in the I/G site and areas rostral or caudal, relative to the corresponding areas in the CS or SCI groups (* and # indicate p < 0. 05 for NN versus the CS or SCI group, respectively, n = 5 in the NN and CS groups, n = 4 in the SCI group). J) Bar chart showing the quantification of the numbers and areas of grafted GFP‐positive cells (* and # indicate p < 0. 05 comparing 24 weeks with the 8 or 0 week data, respectively, n = 5 at 0 weeks, n = 2 at 8 weeks, n = 5 at 24 weeks). K) Bar chart showing the percentages of ChAT‐, GAD67‐, and Glu‐positive cells among all GFP‐positive cells in vivo at 24 weeks after transplantation (* and # indicate p < 0. 05 when ChAT was compared with GAD67 or Glu, respectively, n = 5). Scale bars = 20 µm in (A)–(F) and (H), 1 mm in (G). At 24 weeks after SCI, immunofluorescence staining for NF showed that the positive area was significantly larger in the NN group (Figure 4 G, H) compared with that observed in the CS and SCI groups in the I/G site and in the areas rostral or caudal to the I/G site (Figure 4 I, Figure S3, Supporting Information). The nerve fibers growing out from the transplanted neurons (GFP positive) and host neurons (GFP negative), assisted in increasing innervation of the I/G site. At 8 weeks after transplantation, 60% of GFP‐positive donor cells had survived; however, at 24 weeks, this rate decreased to 35% (Figure 4 J). The surviving cells at 24 weeks after transplantation were dominantly excitatory neurons (ChAT or Glu positive) rather than inhibitory neurons (GAD67 positive, Figure 4 K). 2. 7 Remyelination after Transplantation of NSC‐Derived NN Tissue MBP immunostaining showed tube‐like immunoactive profiles wrapping the donor nerve fibers (NF and GFP positive) or host neurite in the I/G site and in rostral and caudal areas in the NN group (Figure S4A–D, Supporting Information). GFP‐positive SCs (GFP‐NT‐3‐SCs) were used to examine their potential contribution to remyelination of nerve fibers. Immunostaining showed that a large portion of GFP donor SCs were both S100 and MBP positive. SCs wrapped NF‐positive axons, suggesting that they contributed to the remyelination of nerve fibers in the I/G site (Figure S4E, E1–E3, Supporting Information). The donor SCs maintained NT‐3 expression, which may help construct a favorable microenvironment for cell survival, axon regeneration, and myelination in the I/G site of spinal cord. 2. 8 Modification of the Microenvironment after Transplantation of NSC‐Derived NN Tissue After transplantation of NSC‐derived NN tissue into the SCI site, the area of deposited collagenous fibers was significantly decreased as revealed by Masson's trichrome staining. In contrast, substantial deposition of collagenous fiber at the injury site was observed in the CS and SCI groups (Figure S5A, Supporting Information). The results suggested that transplantation of NSC‐derived NN tissue may help reduce the formation of fibrotic scars in the I/G site. The expression of fibronectin (FN) and laminin (LN) was not significantly different in the I/G site in the CS or SCI groups relative to the NN group (Figure S5B, C, Supporting Information). In addition, microglia/macrophages at the injury site were quantified after immunofluorescence staining of ionized calcium binding adaptor molecule 1 (IBA‐1). IBA‐1‐positive cell areas in the I/G sites and rostral or caudal to the I/G site were not significantly different among the three groups (Figure S5D, Supporting Information). 2. 9 NSC‐Derived NN Tissue Integrated into Host Neural Circuits At 24 weeks after transplantation, 5‐hydroxytryptamine (5‐HT)‐positive nerve fibers were observed rostral to the I/G site of the spinal cord in the NN group ( Figure 5 A–C). Some 5‐HT‐positive nerve fibers grew into the I/G site and made contacts with GFP‐positive donor cells (Figure 5 D). However, no 5‐HT‐positive nerve fibers were found in the area caudal to the I/G site (Figure 5 E). Semi‐quantitative analysis of 5‐HT‐positive areas in the three groups suggest that the number of spared 5‐HT‐positive nerve fibers in the area rostral to the I/G site was significantly greater in the NN group compared to the other groups (Figure 5 B, Figure S3, Supporting Information). Immune electron microscopy (IEM) showed synaptic connections between 5‐HT‐positive nerve fibers and GFP‐positive cells rostral to the I/G site (Figure 5 F–H). Tyrosine hydroxylase (TH)‐positive nerve fibers showed more significant regeneration capability; they were present both rostral and caudal to the I/G site ( Figure 6 A–D). TH‐positive nerve fibers also made close contacts with GFP‐positive donor cells in the I/G site and caudally (Figure 6 C, D). Similarly, IEM showed that TH‐immunoreactive nerve fibers formed synaptic connections with GFP‐immunoreactive donor neurons (Figure 6 E, F). Figure 5 Donor neurons formed synaptic connections with descending 5‐HT‐positive nerve fibers 24 weeks after SCI. A) Overview of a longitudinal section of the spinal cord segment containing the I/G site in the NN group. Numerous donor cell processes (green) extended to the areas rostral and caudal to the I/G site, some of which were closely adjacent to 5‐HT‐positive nerve fibers (red). B) Comparison of 5‐HT‐positive areas in the I/G sites and the areas rostral or caudal in the NN, CS, and SCI groups (** p < 0. 001). C) Higher magnification of the boxed area in (A) showing that 5‐HT‐positive nerve fibers were present among the dense GFP‐positive donor cell processes rostral to the I/G site. D) Some 5‐HT‐positive nerve fibers traversed into the I/G site and formed close contacts with the grafted neurons. E) 5‐HT‐positive nerve fibers were scarce in the area caudal to the I/G site. Hoe, Hochest33342. F, H) IEM showing that 5‐HT‐positive nerve fibers (labeled by nanogold particles, superimposed in light purple in (F), white arrows in (G) and (H)) formed synaptic connections with GFP‐positive donor cells (labeled by diaminobenzidine, DAB, asterisks in (F)) and resembled presynaptic components containing the vesicles as shown by nanogold particle labeling (red arrows in (G) and (H)). PSDs (arrowheads in (G) and (H)) of the donor cells were shown by DAB labeling. Scale bars = 1 mm in (A); 20 µm in (C)–(E); 1 µm in (F); 200 nm in (G) and (H). Figure 6 Donor neurons formed synaptic connections with descending TH‐positive nerve fibers 24 weeks after SCI. A) A horizontal section of the spinal cord containing GFP‐positive donor cells in the I/G site in the NN group. TH‐positive nerve fibers were observed B) rostral to the I/G site, C) in the I/G site, and D) in the area caudal to the I/G site. Arrows in (B)–(D) indicate where GFP‐positive donor neurons formed close contacts with TH‐positive nerve fibers. Hoe, Hochest33342. E, F) IEM showing a TH‐positive nerve fiber (labeled by nanogold particle, white arrows in (F)) formed synaptic connections with the donor cell (labeled by diaminobenzidine, DAB, asterisks). The synaptic connection had presynaptic components containing the vesicles (red arrows in (F)) and PSD (arrowheads in (F)). Scale bars = 1 mm in (A); 20 µm in (B)–(D); 1 µm in (E); 200 nm in (F). 2. 10 Structural and Functional Repair of Neural Circuits after NN Tissue Transplantation To evaluate whether disrupted neuronal circuits were repaired following NN tissue transplantation, trans‐multisynaptic viruses were used for anterograde and retrograde tracing. Vesicular stomatitis virus (VSV), an anterograde tracer, 26 transmits across multiple synapses via the presynaptic component to the postsynaptic component. The VSV used in this study encoded blue fluorescent protein (BFP) to visualize labeled nerve tracts. The viruses were injected into the motor cortex of canines in the NN and CS groups 24 weeks after SCI. After 2 weeks, histological analysis showed no BFP fluorescence in the I/G site or caudally in the CS group (Figure S6A–C, Supporting Information). In contrast, VSV‐labeled neurons in the NN group were clearly present in the areas rostral and caudal to or in the I/G site of spinal cord ( Figure 7 A–E). BFP fluorescence signal was observed in a large portion of GFP‐positive donor cells in the NN group (30. 12% of 249 GFP‐positive cells were VSV‐positive, Figure 7 C). In addition, BFP fluorescence appeared in some large neurons located in the area caudal to the I/G site (Figure 7 E), suggesting that VSV transmitted across the I/G site. Figure 7 VSV anterograde and PRV retrograde tracing of transplanted NSC‐derived neurons in the I/G site. A) Low magnification of a longitudinal section from the NN group. B–D) Higher magnification of the boxed areas in (A) showing that VSV was trans‐synaptically transported from the motor cortex to neurons rostral to the I/G site (arrows in (B)) or caudal to the I/G site ((D) and arrows in (E)). VSV was clearly present in GFP and Map2 double‐positive donor neurons in the I/G site (30. 12% of 249 counted GFP‐positive cells were VSV positive, arrows in (C)). F) Low magnification of horizontal spinal cord sections. G–I) Representative images showing NSC‐derived neurons (38. 32% of 214 counted GFP‐positive cells were PRV positive, arrows in (H) and (h)) in the I/G site or host neurons in the areas rostral (arrows in (G) and (g)) and caudal (arrows in (I) and (i)) to the I/G site in the NN group, as shown by RFP‐PRV retrograde labeling. Hoe, Hochest33342. Scale bars = 2 mm in (A) and (F); 20 µm in (B), (H), (g), and (i); 10 µm in (C), (E), and (h); 50 µm in (D), 200 µm in (G); 500 µm in (I). Next, pseudorabies virus (PRV) was used as a retrograde trans‐multisynaptic tracer. 27 to test whether transplanted neurons were synaptically connected with host neurons in the caudal spinal cord. In the CS group, the red fluorescent protein (RFP) signal encoded by PRV revealed host neurons several millimeters caudal to the I/G site (Figure S6D, Supporting Information), but not in the I/G site or in the area rostral to the I/G site (Figure S6E, Supporting Information). In contrast, RFP fluorescence was observed in transplanted neurons in the I/G site and in host neurons in the area rostral to the I/G site in the NN group (Figure 7 F–I, g–i). Notably, 38. 32% of transplanted neurons (82/214 counted GFP‐positive neurons) in the I/G site contained PRV‐RFP (Figure 7 H, h). These neurons may transmit PRV to neurons in the area rostral to the I/G site of the spinal cord. To determine whether donor neurons were functionally integrated with host neural circuits, c‐fos immunofluorescence staining was performed after electrical stimulation of the motor cortex. Immunoreactivity was detected in host neurons rostral and caudal to the I/G site and in the transplanted neurons in the I/G site in the NN group ( Figure 8 A–D), indicating that descending electrical signal were transmitted across multiple synapses including those established in the I/G site. Overall, 12. 89% (29/225 GFP‐positive cells) of transplanted neurons expressed c‐fos (Figure 8 C), suggesting that they were functionally integrated into the motor pathway. However, no c‐fos‐labeled cells were present in the I/G site or caudal to the I/G site in the CS group (Figure S6F, G, Supporting Information). These results indicate that the transplanted neurons play a pivotal role in relaying motor cortex signal across the I/G site after complete SCI (Figure 8 E). To investigate long‐term survival and fate of the donor cells in the spinal cord, a canine survived up to 72 weeks after NN tissue transplantation. Histological samples show that a portion of GFP‐positive donor cells survived in the I/G site (Figure S7, Supporting Information). A subset of the donor cells maintained the phenotype of neurons (NF immunopositive, Figure S7C, Supporting Information). Figure 8 c‐fos expression in the NN group. A) Overview of a longitudinal section from the NN group. After electrical stimulation of the cerebral motor cortex, c‐fos expression was detected in the B) T7–T8 spinal cord segments, C) I/G site, and D) caudal to the I/G site. A portion of the grafted neurons were c‐fos positive (12. 89% of 225 counted GFP‐positive cells, arrows in C). Hoe, Hochest33342. E) Schematic diagram depicting the putative role of the grafted NSC‐derived NN tissue in relaying motor signal as determined by anterograde transport of BFP‐VSV from the motor cortex, retrograde transport of RFP‐PRV from the sciatic nerve, and electrical stimulation of the motor cortex. Scale bars = 2 mm in (A); 20 µm in (B) and (D); 10 µm in (C). 3 Discussion Findings from canine studies are considered to have greater translational value than those from rodent studies. 28 Recent reports highlight the importance of validating translational programs prior to human implementation, 24 and proof of principle in a canine model is key to this endeavor. 28 Buoyed by success in canine SCI modeling, postoperative care, and subsequent behavioral and histological assessments, we made several important observations. First, canine NT‐3‐SCs effectively induced neuronal differentiation of TrkC‐NSCs and formation of NN tissue with functional synaptic connections. This in vitro tissue engineering approach eliminates the uncertainty of stem cell differentiation and the possibility of ectopic colony formation as seen when cells are grafted in vivo. Second, a multimodal assessment battery including locomotion scoring, MRI, DTI, and evoked potential recording showed that SCI canines with NSC‐derived NN tissue transplantation showed continuous motor and sensory improvement and were eventually able to regain coordinated weight‐bearing locomotion. Third, histological analysis showed that the transplanted NSC‐derived NN tissue created a pro‐regenerative microenvironment for donor survival and axonal regeneration. Donor neurons survived up to 24 weeks and formed synaptic connections with host neurons through long processes. Meanwhile, a significant number of host nerve fibers also regenerated into the I/G site and formed synapses with grafted neurons. Fourth, antegrade and retrograde neural tracing techniques confirmed that the transplanted neurons were successfully integrated into host neural circuits. More importantly, c‐fos in donor neurons after cortical stimulation indicated the establishment of functional synaptic connections between host cortical descending nerve fibers and donor neurons. Therefore, the NSC‐derived NN tissue integrated into existing neural circuits and functioned as a neuronal relay to restore severed connections (Figure 8 E). Unlike transplantation of embryonic spinal cord tissue 29 or embryonic neural progenitor cells (NPCs), 30 we preconstructed transplantable NSC‐derived NN tissue to replace lost spinal cord tissue. Our findings reinforce those from previous rat studies that NSC‐derived NN tissue was effective for structurally and functionally repairing complete SCI. 17, 18 Indeed, canine NSCs and SCs, along with trophic factors and secreted ECM, formed a homeostatic microenvironment following 14 d of coculture in a cytocompatible 3D CS scaffold. Within the preconstructed NN tissue, NSCs continuously matured to achieve terminal differentiation toward multiple types of neurons, including excitatory and inhibitory neurons. These neurons were capable of firing action potentials and, more importantly, of communicating with each other through synaptic transmission, as evidenced by TEM and detection of excitatory and inhibitory postsynaptic currents in the NN tissue. SCs contributed to the formation of myelin sheathes to encapsulate neurites of NSC‐derived neurons. The NSC‐derived NN tissue demonstrated neuronal functions but was distinct from embryonic tissues, undifferentiated NSCs, or neural precursor cells (NPCs). We propose that transplantation of an NSC‐derived NN tissue may be superior to transplantation of freshly isolated NSCs or NPCs 8, 11 for three reasons. 1) The homeostatic microenvironment formed during long‐duration 3D culturing conferred greater cellular resilience in the harsh post‐SCI milieu. Donor cells survived longer, allowing them to integrate with the host. 2) Transplantation of differentiated NSCs eliminated undesirable effects including excessive astrocytic differentiation and ectopic colony formation. 3) Tissue engineering in vitro allows for real‐time monitoring of the quality of each batch, thus improving the safety of stem cell therapy. Following NN tissue transplantation in a rat model, animals regained partial motor function of the paralyzed hindlimbs. The plateau of motor recovery typically began 6 weeks after transplantation and remained stable. However, weight‐bearing hindlimb locomotion (i. e. , Basso, Beattie and Bresnahan score > 9 points) was rarely observed following complete SCI in rats, 17, 31 consistent with other studies using the same model. In contrast, pelvic limb motor function recovery in the complete SCI canine model following NSC‐derived NN transplantation was significantly delayed relative to rat studies. The canines started to show distinguishable motor recovery 8 weeks after NN transplantation, as assessed by Olby scoring. Significant motor recovery in the treatment group, relative to that in the controls (CS and SCI groups), was observed 12 weeks after NN transplantation. Motor function recovery in the treatment group slowed at 20 weeks and plateaued 24 weeks after transplantation. Remarkably, most canines that received NN transplantation regained weight‐bearing locomotion of the affected pelvic limbs and front‐pelvic limb coordination by the end of the study. Combined use of MRI, DTI, and electrophysiological evaluation, which is recommended as routine practice for comprehensive evaluation of SCI severity and recovery progress, 23 also suggest that canines that received NN tissue treatment had better structural and functional improvement than those in the control groups. In contrast, canines in the control groups showed mild recovery of pelvic limb motor function with no weight‐bearing locomotion or coordinated stepping. This is similar to observations in severe SCI patients for whom spontaneous recovery of neural function is rare. Because of the biological similarities in recovery following SCI between canines and human, 24 we hypothesize that transplantation of NSC‐derived NN tissue may also benefit SCI patients. Given that large animals have more complex immune systems and stronger rejection to grafts than rodents, 32 donor cell survival in vitro remains the primary challenge to stem cell‐based SCI repair. 24 This requires immunosuppressants after cell transplantation. 3, 17, 31 To acclimate to the drastic milieu shift, donor cells can be grafted in a trophic factor‐enriched microenvironment. For example, NSCs/NPCs encapsulated in a growth factor cocktail‐fibrin matrix had a satisfactory survival rate when transplanted into the injured monkey spinal cord. 33 Our previous studies showed that NT‐3 overexpressing SCs promoted neuronal differentiation of NSCs in vitro 17 and improved the microenvironment, allowing for increased graft cell survival in a rat SCI model. 18, 34 In the present study, we inferred that NT‐3 overexpressing SCs may be key contributors to donor survival in the injured canine spinal cord, as we observed that NT‐3 exerted anti‐inflammatory effects when delivered to the parenchyma of the injured canine spinal cord. 35 In addition, higher cerebrospinal fluid (CSF) pressure in the spinal cords of large animal may flush away grafted cells, preventing them from remaining and functioning at the I/G site unless intraoperative draining of CSF was performed prior to cell transplantation. 33 However, the optimal amount of CSF draining requires further study. Moreover, draining CSF may not be a clinically acceptable strategy in humans due to numerous side effects. Alternatively, a tissue engineering approach to construct organoid or tissue‐like implants can eliminate the need to drain CSF prior to transplantation. In our study, NN tissue composed of cells, ECM, and a collagen matrix, formed a homeostatic entity following 14 d of culture. In particular, abundant ECM components including FN and LN were deposited in the scaffold. These ECM molecules enhanced attachment of seeded cells onto the CS scaffold through specific adhesion molecules such as integrins. This made resident cells more resistant to CSF pressure during implantation. Although combined strategies have great potential, donor population loss after transplantation might be inevitable. For example, when human NPCs were grafted into nonhuman primate spinal cord for 9 months, only ≈25% of the original population remained, despite the use of a triple immunosuppressant regimen. 33 We also observed a significant decrease in donor cell number from 8 to 24 weeks after transplantation; however, motor function continued to show improvement during this time. We speculate that this paradox may be caused by gradual functional maturation of the transplanted NN tissue and refined integration into host neural circuits. To initiate front‐pelvic limbs coordinated stepping, excitatory inputs must transmit across the grafted neurons and relay to the motor neurons controlling the pelvic limbs. It is possible that the donor neurons not functionally involved in this signaling might have been suppressed, or even eliminated. Indeed, elimination of redundant synaptic connections is essential for functional neural circuit formation. 36 Moreover, nonintegrated neurons may have been eliminated due to a lack of survival signals from other neurons. 37 Future work is needed to distinguish whether neuronal death following transplant is a result of neural circuit maturation or lack of transplanted tissue viability. 38 Nonetheless, there were still GFP positive donor cells survived in the I/G site of spinal cord up to 72 weeks after NN transplantation. It would be worthy to systematically analysis the biological basis supporting the long‐term survival of the donor neurons in the future. A pro‐regenerative microenvironment during transplantation of NSC‐derived NN tissue may also promote endogenous neural regeneration. For example, the fibrotic scarring represented by collagen deposition was significantly reduced in the NN group. Conversely, the expression of FN and LN was upregulated in the I/G site of the NN group. The shift between the inhibitory and pro‐regenerative ECM molecules may also promote regeneration of host nerve fibers and endogenous neurogenesis in the I/G site. The canines in the NN group showed continuous motor function recovery; therefore, endogenous neural regeneration may compensate for the gradual loss of donor neurons. However, further research is warranted to determine the delicate interactions between endogenous neural regeneration and exogenous neurons. Histological analysis was performed to show how NSC‐derived NN tissue contributed to structural repair of the damaged canine spinal cord. At 8 weeks after transplantation, most GFP‐positive donor neurons located at the I/G site had long processes extending rostral or caudal to the I/G site. Synaptic connections between the donor and host neurons were verified by immunocytochemistry and IEM. It was inferred that NT‐3 activated TrkC signaling to promote synapse formation 18, 39 and that TrkC‐overexpressing neurons more readily formed synaptic connections with host‐regenerated nerve fibers expressing protein tyrosine phosphatase σ (PTPσ). 40 However, we did not observe long axons that extended to remote spinal cord tissue more than two spinal cord segments rostral or caudal to the I/G site. In other words, there were no direct synaptic connections between the donor neurons and lumbar motor neurons controlling pelvic limb muscles. However, trans‐synaptic virus tracing demonstrated establishment of multisynaptic connections in the motor neural pathway. VSV, a type of anterograde trans‐multiple synaptic virus, 26 transmitted from the motor cortex (injection site) to the transplanted neurons in the I/G site, as evidenced by BFP fluorescence encoded by VSV in GFP/Map2 double‐immunopositive donor neurons. BFP was detected in host neurons in the area caudal to the I/G site, meaning that VSV was transmitted from the donor neurons in the I/G site to the host neurons caudal to the I/G site through established synapses. No BFP was detected in lumbar neurons, which might be due to insufficient time after injection to allow the virus to transmit. However the retrograde trans‐multiple synaptic virus PRV transmitted from the lumbar motor neurons to neurons in the area caudal to the I/G site, where they retrogradely transmitted to the implanted neurons in the I/G site. In contrast, no VSV or PRV transmission was detected in the I/G site of canines in the CS group. These results indicated that the transplanted NN tissue contributed to repair of the severed spinal cord with a large tissue deficit. Although our previous publication suggested that MSC‐derived NN tissue may also contribute to structural repair of canine injured spinal cord, 16 this study included de facto synaptic connection mapping of the key elements in the motor pathway starting from the cortex to lumbar motor neurons. It is plausible that transplanted NSC‐derived NN tissue may relay excitatory signals across the I/G site, contributing to recovery of paralyzed motor function after SCI. This is supported by the observations from this study. 1) Canines transplanted with NSC‐derived NN tissue regained partial motor function of the pelvic limbs, as evidenced by recovery of weight‐bearing locomotion and coordinated stepping on an underwater treadmill, relative to mild motor recovery in canines in the control groups. 2) CMEP recordings indicated that electrical signals initiated from the motor cortex could pass through the I/G site to reach the lumbar spinal cord. 3) After electrical stimulation of the cerebral cortex, c‐fos expression was found in neurons rostral to the I/G site, in grafted neurons, and in neurons caudal to the I/G site. This suggests that action potentials were conducted through functional synapses connecting the grafted and host neurons. 41 4) Histological analysis further verified that host descending 5‐HT‐ and TH‐positive nerve fibers, both of which conduct excitatory signals from the brain, were able to regenerate significantly to the rostral and central areas of the I/G site to form synaptic connections with the transplanted neurons. 5) More than half of the donor neurons were capable of synthesizing excitatory neurotransmitters. Axons of donor neurons extended to the adjacent caudal spinal segment of the host tissue and made synaptic connections with the host neurons via VGluT1‐positive presynaptic boutons. The PSD was immunopositive for PSD95, a marker for excitatory postsynaptic components. Asymmetric synapse features observed by IEM further suggest that excitatory synapses were established between donor and host neurons. Taken together, these results indicate that the implanted NN tissue played an essential role in re‐establishing functional connections in the motor pathway, allowing for excitatory transmission from the brain to the muscle, which may explain how weight‐bearing locomotion and coordinated stepping were regained. Indeed, reactivating the spared motor pathway is believed to directly control motor function. For example, epidural stimulation of lumbar motor circuits improves motor recovery in SCI patients. 42, 43 In contrast, suppressing overinhibition caused by inhibitory interneuron activity following SCI was shown to promote functional recovery in a mouse SCI model. 44 Sensory recovery is considered fundamental for motor recovery. Front‐pelvic limb coordinated stepping on the underwater treadmill, as observed in this study, may indicate that canines with NN tissue transplantation had at least partial proprioception. Validation of sensory recovery in animals requires special equipment (e. g. , functional MRI) or sophisticated training and compliance to the test, which were not performed in this study. However, SSEP detection indicated that the implanted NN tissue may have helped relay the ascending sensory signals. This is supported by the fact that the ascending sensory pathway helps the cortex adjust the accuracy of the descending motor pathway, which is essential for recovery of coordinated movements. 45 Furthermore, SC‐derived myelin may contribute to functional recovery after SCI as remyelination enables more stable and efficient neuronal relay by NN tissue. 46 4 Conclusions Functional NN tissue was successfully constructed in a 3D culture system by coculturing canine NT‐3‐SCs and canine TrkC‐NSCs in a CS scaffold. These NN tissues survived up to 72 weeks at the SCI site, served as interneurons to relay descending excitatory signals from the brain to host neurons caudal to the I/G site, and helped canines regain weight‐bearing locomotion and coordinated stepping. Adverse effects such as autotomy, hyperalgesia, and tetanic spasm were not encountered following transplantation. Histological analysis ruled out ectopic migration or tumorigenesis by donor cells. These findings support the safety and efficacy of transplantation of NN tissue to treat SCI in large animals and provide a framework for future clinical translation using tissue engineering construction of NSC‐derived NN tissue. Using induced pluripotent stem cell‐derived NPCs 6, 7 or SCs to construct NN tissue may be a future strategy to mitigate cell source concerns. Combination with physiotherapy or rehabilitation 47, 48 may enhance integration of the transplanted NN tissue with donor neural circuits, which could further increase neural function restoration following SCI. 5 Experimental Section Ethics : All animal experiments were approved by the Institutional Animal Care and Use Committee of Sun Yat‐sen University (Approval number: 20181000244), and the Laboratory Animal Regulations of Guangdong Province (2010 No. 41). Animal welfare was in compliance with Laboratory Animal Guidelines for Ethical Review of Animal Welfare, General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China/Standardization Administration of China (GB/T 35892‐2018). Cultivation of NSCs and SCs : To acquire NSCs and SCs, newborn male beagle canines (1–3 d, ChaiMen Biological Inc. , Nanjing, China) were anesthetized and sacrificed by cervical dislocation after inhalation of overdosed isoflurane. NSCs were isolated from the hippocampus similar to a previously described procedure. 49 Briefly, the whole hippocampus was dissected and dissociated into single cell suspension. NSC medium composed of Dulbecco's minimum essential medium (DMEM)/F12 (1:1, Life Technologies, USA) supplemented with 1 × B27 and 20 ng mL −1 bFGF. Cells were grown as neurospheres in suspension and passaged by mechanical dissociation every 5 d. Neurosphere progenitor content was assessed by nestin immunostaining (Figure S1A, Supporting Information). To obtain SCs, the sciatic nerves and brachial plexus were dissected and placed in ice‐cold D‐Hank's solution. The epineurium and connective tissue were removed under a dissecting microscope. All nerves were cut into small pieces (< 2 mm) and dissociated with 0. 16% collagenase (Sigma‐Aldrich, USA) at 37 °C for 15 min and centrifuged at 1000 rpm min −1 for 5 min. Pellets were resuspended in 1 mL culture medium containing DMEM/F12, 10% fetal bovine serum (2 mmol L −1 forskolin (Sigma‐Aldrich, USA)) and 20 mg mL −1 bovine pituitary extract (Sigma‐Aldrich, USA), then seeded in 75 mL culture flasks precoated with polylysine at 37 °C with 5% CO 2. After 30 min, 4 mL of culture medium was added to each flask. Cells were maintained in an incubator at 37 °C in 5% CO 2. Culture medium was changed every 2 d. The cells were passaged at 90% confluence and purified by differential adhesion and differential digestion techniques. 16 SC purity was assessed by immunochemical staining with S100. In Vitro Induction of NN Tissue : Recombinant lentiviruses were used to modify NSCs and SCs. Neurospheres were transfected with a lentivirus carrying a TrkC coding sequence (pLent‐EF1a‐TrkC‐Flag‐CMV‐GFP‐P2A‐Puro) (Vigenebio Biosciences Inc. , China); SCs were infected with a lentivirus vector carrying an NT‐3 sequence (pLent‐EF1a‐NT‐3‐Flag‐CMV‐P2A‐Puro or pLent‐EF1a‐NT‐3‐Flag‐CMV‐GFP‐P2A‐Puro) (Vigenebio Biosciences Inc. ). After the lentivirus was added into the culture medium at a multiplicity of infection of 50 for 48 h, the supernatant was removed and replaced with fresh culture medium containing 2 µg mL −1 puromycin for cell purification. The screened TrkC gene‐modified NSCs (TrkC‐NSCs) and NT‐3 gene‐modified SCs (NT‐3‐SCs) were mixed in a 1:1 ratio with 1 × 10 6 cells seeded into a cylindrical CS scaffold (5 mm diameter and 4 mm long) using a micropipette. The scaffolds were incubated for 14 d, and the culture medium was changed every day. The experimental groups included NSCs, NSCs+SCs, T‐NSCs+SCs, NSCs+N‐SCs, and T‐NSCs+N‐SCs, with a total of 1 × 10 6 cells (for NSCs alone or 1:1 for NSCs and SCs coculturing) in 30 µL of culture medium seeded into each CS. Western Blotting : Scaffolds in each group ( n = 5) were chopped into pieces on a cold stage, added to radioimmunoprecipitation assay buffer, and sonicated to extract total protein. Equal amounts of protein in each group were loaded onto a 10% polyacrylamide gels for electrophoresis. After transfer to polyvinylidene fluoride membranes, the membranes were treated with the following primary antibodies: NF, TrkC, PSD95, ChAT, GAD67, GFAP, SYP, MBP, NT‐3, and glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) (all antibodies are from rabbit host species) and incubated overnight at 4 °C. Membranes were then incubated with antirabbit horseradish peroxidase‐conjugated IgG. Bands were detected with an enhanced chemiluminescence western blot kit (Cwbiotech, China) using a chemiluminescence imaging system (ChemiDoc, Bio‐Rad, USA). GAPDH was used as a loading control. The antibodies used in this study are listed in Table S1 (Supporting Information). Immunofluorescence Staining : Expression of specific proteins was detected by immunofluorescence. Briefly, spinal cord tissue was cut into 25 µm thick longitudinal sections using a cryostat microtome, then rinsed with 0. 01 m phosphate‐buffered saline (PBS) three times, blocked with 10% goat serum for 30 min, and incubated with primary antibody containing 0. 3% Triton X‐100 to increase penetration at 4 °C overnight. The sections were washed three times with PBS and then incubated with a secondary antibody at 37 °C for 1 h, Hoechst33342 (Hoe) was used to stain the nuclei. The sections were observed with a fluorescence microscope (Leica, Germany) or a laser confocal microscope (LSM780/LSM800, Zeiss, Germany) to produce a z ‐axis scan. The list of antibodies used is shown in Table S1 (Supporting Information). Live‐Cell FM1‐43 and Whole Cell Patch‐Clamp Detection : FM1‐43 [ N ‐3‐triethylammonmpropyl)‐4‐(4‐(dibutylamino) styryl] dye (Life Technologies) was used to determine whether induced neurons had synaptic vesicle releasing features. 50 After 14 d of coculture, the NN tissue was rinsed with low [K + ] 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES). The first dose of high [K + ] solution stimulated recycling of FM1‐43‐containing endocytic synaptic vesicles. After rinsing three times (15–20 min each) with culture medium in the absence of FM1‐43, endocytic/exocytotic activities decreased to basal levels, and nonspecific labeling of cytoplasmic membranes was also eliminated, while synaptic vesicles maintained FM1‐43 labeling. After rinsing, cells were excited with a second dose of high [K + ] solution without FM1‐43, resulting in depolarization. FM1‐43‐labeled synaptic vesicle release was visualized using a confocal laser scanning microscope (LSM780/LSM800, Zeiss, Germany). A control experiment to assess nonspecific bleaching of fluorescence was performed simultaneously on NN tissue without high [K + ] solution stimulation. To investigate excitability of NSC‐derived neurons, whole‐cell patch clamp was performed with a HEKA EPC amplifier 10 (HEKA Inc. , Germany) after culturing for 21 d in vitro. Results were analyzed using Patchmaster software (HEKA Inc. ). Signals were filtered at 1 kHz and sampled at 5 kHz. The external solution contained 140 × 10 −3 m NaCl, 5 × 10 −3 m KCl, 2 × 10 −3 m CaCl 2, 1 × 10 −3 m MgCl 2, 10 × 10 −3 m HEPES, and 10 × 10 −3 m glucose (320 mOsm, pH set to 7. 3 with Tris base). The patch electrodes had a resistance of 3–5 MΩ when filled with pipette solution containing 140 × 10 −3 m CsCl, 2 × 10 −3 m MgCl 2, 4 × 10 −3 m ethylene glycol‐bis(β‐aminoethyl ether)‐ N, N, N ′, N ′‐tetraacetic acid (EGTA), 0. 4 × 10 −3 m CaCl 2, 10 × 10 −3 m HEPES, 2 M magnesium adenosine triphosphate (Mg‐ATP), and 0. 1 × 10 −3 m guanosine triphosphate (GTP). The pH was adjusted to 7. 2 with Tris base, and the osmolarity was adjusted to 280–300 mOsm with sucrose. Briefly, when the micropipettes were at the appropriate distance from the cell membrane, brief and gentle suction was applied to create tight contact with resistance up to 1 GΩ. Then, brief and strong suction was used to form a whole‐cell configuration with tip resistance of 3–5 MΩ. Electrophysiological recordings were performed at room temperature (22–24 °C). Finally, the membrane potential of the cells was clamped at −70 mV using a voltage clamp. mPSCs were counted and analyzed using Fitmaster (HEKA Inc. ). Spinal Cord Transection Modeling and Transplantation : Twenty‐seven healthy female beagles (6–8 months old, 8–9 kg, supplied by Frontier Biotechnology Inc. , China) were randomly divided into three groups: 1) the NN group ( n = 11 for 24 weeks, n = 2 for 8 weeks, n = 1 for 72 weeks); 2) the CS group ( n = 10 for 24 weeks); and 3) the SCI group ( n = 4 for 24 weeks). To ensure adherence to inclusion criteria, all animals were screened for signs of spinal cord disorder prior to the experiment. Spinal cord transection was performed as previously described. 16 Briefly, canines were anesthetized, and the spine was cut at thoracic vertebrate T10. Laminectomy was performed to provide a window for a longitudinal dura incision of about 1 cm. After the T9‐T10 spinal segments were exposed, a sharp blade was used to transect the rostral and caudal spinal cord to make a 4 mm long gap. The severed spinal cord was removed until the ventral dura was exposed to ensure that no residual spinal cord tissue remained in the gap. The corresponding paired spinal roots were also completely removed. The NN or CS scaffold was implanted after sufficient hemostasis in the NN and CS animals, respectively. No scaffold was implanted in the SCI group, but the area was washed three times with saline. Low‐tension dural sutures were placed to prevent scaffold mobility. Intensive postoperative care was administered during the first week, including daily intravenous rehydration with 50 mL lactated Ringer's solution. Penicillin was administered to prevent infection (800 000 units, intramuscular daily). Pain was managed by oral administration of meloxicam (0. 1 mg kg −1 /24 h for 5 d). Defecation was stimulated manually three times per day until bladder function returned. Cyclosporin A (20 mg kg −1 ) was administered once daily on a strict 24 h cycle until the end of the experiment. Assessment of Locomotor Performance : The behavioral assessment battery included an open field locomotion test platform equipped with a high‐speed videotaping device and an underwater treadmill locomotion evaluation system. Motor function recovery assessment was performed by a double‐blind protocol using Olby scoring. 51, 52 During evaluation, canines were free to walk in the open field for more than 10 min. Scores were obtained based on limb movement functions including deep pain reflex, joint movement, weight‐bearing capability, muscle strength, and gait on a 15‐point scale. To further assess pelvic limb locomotion under non‐weight‐bearing conditions, canines were placed on an underwater treadmill for visual observation of minor joint movement and coordinated movements between the front and contralateral pelvic limbs, and between the pelvic limbs. 16 Canines were placed on a treadmill submerged in warm water (35–39 °C). Track belt speed was adjusted according to individual maximum values (5–20) cm s −1 and the duration for each canine on the treadmill was 5 min for each recording cycle. Two investigators blinded to the experimental groups scored the videos captured at 0, 1, 2, 4, 8, 16, and 24 weeks postoperatively. Electrophysiology Assessment : CMEPs and SSEPs were assessed at 24 weeks after SCI using NeuroExam M‐800 (Medocon Technology, China). Canines were anesthetized with pentobarbital sodium (3% dissolved in saline, 45 mg/kg, intraperitoneal [i. p. ]) and ketamine (10 mg kg −1, intramuscular injection every 20 min during surgery), and placed in a prone position during the operation. For CMEPs, one stimulus electrode was inserted into the subcutaneous tissue 3 cm lateral to the intersection of the cranial midline and eyebrow, and another electrode was inserted into the contralateral deep tissue approaching the bone surface between the rear ear and inion. The recording electrode was placed on the spinal cord surface at L1–L2. The grounding electrode was inserted subcutaneously near the T1 spinal segment. Multiple pulse stimulation was used to elicit a CMEP with the following parameters: 250 times gain, 150 µs time constant, 100 mA pulse width, and 1000 µs interval repeated four times. Each test was repeated 40 times to ensure waveform stability. For SSEPs, the recording electrodes were placed in the same location as the stimulus electrode for CMEPs, and the stimulus electrode was placed at the gluteus maximus. A series of stimulations with 7 mV, 1 Hz, and 300 repeats was performed for SSEP waveform stacking. Induction of Spinal Cord c‐fos Expression : At 24 weeks after SCI, canines in the NN group ( n = 2) and the CS group ( n = 2) received unilateral electrical stimulation of the motor cortex to induce c‐fos expression in the spinal cord. Canines were anesthetized with 1% pentobarbital sodium (40 mg kg −1, i. p. ), and the motor cortex zone was exposed by craniotomy. Stimulation was delivered between an electrode (0. 7 mm diameter) on one side of the hindlimb motor cortex (1. 0–1. 2 cm posterior to the bregma, 1. 0 cm lateral to the midline) and a needle electrode on the hard palate. A 300 ms train current stimulation was delivered at 10 mA with a pulse duration of 0. 5 ms at a frequency of 200 Hz continuously for 1 h. 16, 41 Then, the canines remained deeply anesthetized for 2 h until sacrifice. Trans‐Multisynaptic Virus Labeling : At 24 weeks after SCI, two canines per group were randomly selected from the NN and CS groups and received VSV (BrainVTA Technology Co. Ltd. , China) to label the motor cortex for anterograde virus mapping. The labeling procedure was performed as previously described. 16, 53 Briefly, after anesthetization, the head of the canine was fixed in a custom‐made stereotactic frame. A longitudinal incision (6–7 cm) was made at the midline above the parietal bone. To expose the motor cortex, two 1. 5 × 3. 0 cm 2 oval windows were made in the parietal bone using a drill and a pair of rongeurs with the following coordinates: 1. 0–1. 2 cm posterior to the bregma, 1. 0 cm lateral to the midline. VSV‐encoding BFP (2. 00E+09 PFU mL −1 ) was loaded into a Hamilton microsyringe (Hamilton Co. , USA) attached to the stereotactic frame. A total of ten injections (3 µL each) were made in both cerebral hemispheres, delivering a total of 30 µL VSV. For retrograde virus mapping, two canines per group were randomly selected from the NN and CS groups and underwent retrograde PRV (BrainVTA Technology Co. Ltd. ) labeling at the sciatic nerve. Briefly, after anesthetization, a 5 cm longitudinal incision was made along the biceps femoris and semitendinosus muscles of both pelvic limbs to expose the sciatic nerve as much as possible. With the aid of a dissecting stereomicroscope (Leica), the needle tip of a 30 G Hamilton syringe was inserted into sciatic nerve along its longitude axis for 10 mm, and then withdrawn 2 mm to make space for injection. PSV (20 µL, 2. 00E+09 PFU mL −1 ) which encoded the RFP reporter gene (RFP‐PRV) was slowly injected into the bilateral sciatic nerves. After injection, the sciatic nerves were clamped with hemostatic forceps 2 cm above the injection point for 30 s to maximize PRV particle uptake by the nervous tissue, and the wound was sutured layer by layer. After virus mapping, the canines received extensive postoperative care including intramuscular injection of penicillin (80000 U/kg/d) for 3 d to prevent infection. The animals were sacrificed 14 d after virus injection. MRI and DTI : The canines were examined by MRI and DTI 24 weeks after surgery to observe the morphology of the spinal cord lesion site. Fiber tractography and measurement of FA and ADC values were acquired to assess the continuity of nerve fiber bundles in the spinal cord, which is not detectable by MRI. 54, 55 The average FA and ADC values were measured and represent the rostral, caudal, and I/G site of the spinal cord for each animal. Ultrastructural Observation : Tissue engineering NN tissue was visualized by SEM. Samples were washed three times with PBS, then fixed with 2. 5% glutaraldehyde for 90 min before dehydration using an alcohol gradient. After freeze‐drying for 2 d, the samples were coated with gold, and then observed by SEM (Philips XL30 FEG, Philips, Netherlands). TEM was used to observe the NN tissue or spinal cord tissue. The samples were fixed with 2. 5% glutaraldehyde and 15% picric acid for 2 h at 4 °C, followed by 1% osmic acid for 1 h at room temperature, dehydrated using an alcohol and acetone gradient, then embedded in Epon and polymerized for 48 h at 60 °C. The embedded tissue was sliced into semithin sections (2 µm thickness, Leica RM2065 microtome). Sections were mounted on slides, stained with toluidine blue (5% in a borax solution), and mounted using neutral balsam prior to observation. The remaining contents were cut into ultrathin sections (100 nm thickness), double‐stained with lead citrate and uranyl acetate, and examined using an electron microscope (Philips CM 10, Philips, Netherlands). For IEM, the injured spinal cord segment was removed and immediately immersed in 4% paraformaldehyde containing 0. 15% glutaraldehyde and 15% saturated picric acid at 4 °C for 2 h. After rinsing in cold 0. 1 M PBS, the tissues were fixed without glutaraldehyde at 4 °C for 4 h, then sagittally sectioned into 50 µm thick slices using a vibratome. Sections were transferred to 0. 1 m PBS containing 25% sucrose and 10% glycerol overnight at 4 °C, then freeze‐thawed three times using liquid nitrogen. For double‐labeling, the tissue slices were treated with 0. 3% H 2 O 2 to scavenge endogenous peroxidase prior to adding blocking serum. After washing with PBS, sections were treated with 20% goat serum (Tris buffer, pH 7. 4) for 40 min prior to incubation with primary antibodies. The sections were incubated with primary antibodies in 2% goat serum for 24 h at 4 °C, then incubated with secondary antibody overnight at 4 °C. The tissues were postfixed with 1% glutaraldehyde for 10 min and further treated using an SABC‐DAB Kit and gold enhanced using a GoldEnhance EM Plus Kit (NanoProbe 2114, USA). The sections were then subjected to osmiumization, gradient dehydration, then embedded with Epon. Finally, ultrathin sections were observed using the electron microscope. Morphological Quantification : For the in vitro quantification of immunopositive cells, one in every five of the whole series of horizontal sections (five sections) from each NN tissue was selected ( n = 5 per group). After immunostaining with the respective markers, five areas (0. 7 mm × 0. 5 mm including four corners and one center) were chosen for each of the sections. The percentage of immunopositive cells was calculated by counting the total number of immunopositive cells and dividing it by the total number of GFP positive cells. For the quantification of surviving cells, prior to transplantation (0 weeks), one in every 10 of the whole series of horizontal sections (five sections) from each in vitro cultured NN tissue was selected ( n = 5). Five areas (0. 7 mm × 0. 5 mm including four corners and one center) were chosen for each of the sections, and the total numbers of GFP‐positive cells was counted. The numerical value obtained was subsequently divided by the total area. For the quantification of surviving cells at 8 weeks ( n = 2) and at 24 weeks ( n = 5), one in every ten slices of the whole series of horizontal sections (five sections) from each animal was selected. The transected area with implantation was defined as the I/G site. Five areas (0. 7 mm × 0. 5 mm including four corners and one center) for each section were chosen in the I/G site. The total number of GFP‐positive cells was counted and divided by the total area. For the in vivo quantification of immunopositive cells, areas in the I/G site of each of the horizontal sections were scrutinized. One in every ten sections from each animal was processed, and a total of five sections per animal were analyzed ( n = 5). Five areas (0. 7 mm × 0. 5 mm including four corners and one center) for each of the sections cut through the I/G site were chosen. The percentage of immunopositive cells was calculated by counting the total number of immunopositive cells and dividing by the total number of GFP‐positive cells. To investigate nerve fiber density, one in every five slices of the whole series of sections from each animal was selected, and the entire areas of each slice containing T8–T10 spinal cord segments were visualized using a laser confocal fluorescence microscope after immunofluorescence staining. A total of 10 sections per animal were analyzed ( n = 5 for the NN and CS groups respectively, n = 4 for the SCI group). The range of rostral or caudal areas was defined as a 0. 5 mm wide area rostral or caudal to the transection margin, respectively. The transected area with or without implantation was defined as the I/G site. Three areas (325 µm × 325 µm) were selected as regions of interest (ROIs) from each photographed slice. These ROIs were processed using ImageJ software (version 1. 48, NIH, USA) and the positive area ratio was calculated. Images cavity areas > 50% were excluded. This method was utilized for the semi‐quantitative analysis of NF‐, 5‐HT‐, FN‐, LN‐, and IBA‐1‐positive areas. Statistical Analysis : All statistical analyses were performed using SPSS 17 (SPSS Inc. , USA). Data were expressed as the means ± standard deviations (mean ± SD). Data were analyzed using one‐way analysis of variance (ANOVA). If equal variances were found, the least‐significant difference (LSD) test was applied. If variances were not equal, Kruskal‐Wallis test and Dunnett's T3 were applied. Student's t ‐test was used to compare two groups. p < 0. 05 was considered statistically significant. Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file. Supplementary Click here for additional data file.
|
10. 1002/advs. 201901388
| 2,019
|
Advanced Science
|
Self‐Healing Hydrogels and Cryogels from Biodegradable Polyurethane Nanoparticle Crosslinked Chitosan
|
Abstract Hydrogels are widely used in tissue engineering owing to their high water retention and soft characteristics. It remains a challenge to develop hydrogels with tunable degradation rates, proper environmental responsiveness, and injectability. In this study, biodegradable difunctional polyurethane (DFPU) nanoparticle dispersions are synthesized from an eco‐friendly waterborne process involving the use of glyoxal. Such DFPU is used to crosslink chitosan (CS). Schiff base linkages between DFPU and CS successfully produce self‐healing hydrogels at room temperature. Moreover, cryogels are generated after being frozen at −20 °C. These gels are found to be sensitive to low pH and amine‐containing molecules owing to the property of Schiff bases. Furthermore, the degradation rates can be adjusted by the type of the component oligodiols in DFPU. Rheological evaluation verifies the excellent self‐healing properties (≈100% recovery after damage). Both the self‐healing gels and cryogels are injectable (through 26‐gauge and 18‐gauge needles, respectively) and biocompatible. Rat implantation at 14 d shows the low immune responses of cryogels. The functionalized biodegradable polyurethane nanoparticles represent a new platform of crosslinkers for biomacromolecules such as chitosan through the dynamic Schiff reaction that may give rise to a wide variety of self‐healing gels and cryogels for biomedical applications.
|
1 Introduction Hydrogels are polymers with high water retention, soft characteristics, and good biocompatibility, which gain wide applications in biomedical fields. Recent attention has been paid to environmentally responsive hydrogels, injectable hydrogels, self‐healing hydrogels, and cryogels as potential biomedical materials owing to their excellent functionality. 1 Environmentally responsive hydrogels are also known as smart hydrogels, which can change their chemical properties or physical structures in response to environmental stimuli and are often used in drug release and sensors. 2 Injectable hydrogels have in situ formability which allows surgical operation in a minimally invasive way. Therefore, they hold promises in drug delivery, cell encapsulation, and tissue engineering. 3 Meanwhile, self‐healing hydrogels are inspired by the self‐healing ability in biology. 4 Materials with the ability of self‐healing have potential biomedical applications, such as drug release, capillary network construction, and hemostasis. 5 Cryogels have attractive features of interconnected macropores' structure that allow the transport of nutrients and wastes through the cryogels. 6 Cryogels are often used in chromatographic materials, efficient carriers for the immobilization of biomolecules and cells, and matrices for cell separation, cell culture, and tissue padding. 7 Chitosan is a polycationic biopolymer that is naturally derived and is one of the important environmentally friendly renewable materials obtained by deacetylation of chitin. 8 When the degree of deacetylation of chitin reaches about 50%, it becomes soluble in acidic condition and is named chitosan. [qv: 8b] Chitosan is generally nontoxic, biocompatible, biodegradable, and bacteriostatic, and has been employed for many pharmaceutical and medical applications including orthopedic/periodontal devices, tissue engineering, wound healing, and drug/gene delivery. 9 However, chitosan can only be dissolved in water under acidic conditions, limiting its use as a living cell and tissue matrix. [qv: 5a] Therefore, the water‐soluble derivatives of chitosan such as glycol chitosan and N‐carboxyethyl chitosan have been developed that can be prepared as hydrogels. [qv: 5a, 10] Self‐healing hydrogels based on chitosan derivatives have attracted much recent attention. [qv: 5a, 11] For the purpose, various crosslinkers have been synthesized and used. Among them, functionalized polyethylene glycol (PEG) crosslinkers are the most popular choices, [qv: 5a] for example, difunctional polyethylene glycol (DFPEG). DFPEG has good solubility, biocompatibility, and reacts with chitosan to form Schiff base (also known as imine, —N=CH—). The dynamic covalent bond in DFPEG crosslinked chitosan has the ability to deform and reform due to the Schiff base instability, and such structurally dynamic hydrogels can respond to many chemical and biological stimuli by liquefaction of the hydrogels. 12 However, the hydrogel formed by DFPEG or multifunctionalized PEG may be dissolved rather rapidly in vivo. 13 Besides, the soluble PEG cannot be decomposed in vivo. 14 Therefore, biodegradable crosslinkers which have turnable degradation rates are demanded in the field. Meanwhile, Schiff base crosslinkers such as glutaraldehyde and dextran oxide are also used to prepare chitosan cryogels. 15 As a crosslinker, DFPEG is not reported for cryogels, and dextran oxide is not reported for self‐healing hydrogels. Issues regarding if self‐healing hydrogels and cryogel can be made of the same main chain and crosslinker are rarely discussed. Recent attention has been paid to nanoparticles as crosslinker because of their potential benefits. Incorporation of inorganic nanoparticles has been demonstrated to reinforce the hydrogel matrix, resulting in stronger nanocomposite gels. 16 Through specific interactions with the hydrogel polymer chains, nanoparticles can effectively contribute to the polymer network elasticity and thereby increase the mechanical strength of the hydrogels. 17 Furthermore, dynamic hydrogels with intrinsic self‐healing capabilities can result if the polymer–particle interfacial crosslinks are reversible. 18 Moreover, nanoparticles can introduce a variety functionalities of inorganic materials to the hydrogel, such as electronic conductivity and magnetic response. 19 Despite the great achievements of inorganic nanoparticle crosslinkers, the poor biodegradability and biocompatibility may limit their biomedical applications. Based on the literature, biodegradable crosslinkers for chitosan self‐healing hydrogels or cryogels are still limited, and thus highly demanded. Here, we report an effective and novel crosslinker type of Schiff base from eco‐friendly waterborne polyurethane in nanoparticle (NP) form. By adjusting the reaction temperature, self‐healing hydrogels or cryogels with multiresponsiveness and injectability can be easily prepared. Novel self‐healing hydrogels and cryogels are injectable through 26‐gauge and 18‐gauge needles, respectively. Moreover, the soft segments of the developed crosslinker can be altered to obtain biodegradable hydrogels and cryogels with tunable degradation rates. 2 Results 2. 1 Synthesis and Characterization of Difunctional Polyurethane (DFPU) Crosslinkers New difunctional polyurethane crosslinkers with tunable degradation rates were synthesized, as shown in Figure 1 A. The synthesis was based on a green, waterborne procedure. Two types of difunctional polyurethane crosslinkers were synthesized in this study. The major one was abbreviated as DFPU. The oligodiol employed to synthesize DFPU was polycaprolactone diol (PCL diol, Mn 2000 Da). The second difunctional polyurethane crosslinker with a different soft segment and faster degradation rate was abbreviated as DFPU'. The oligodiol employed to synthesize DFPU' was a combination of PCL diol and poly(1, 4‐butylene adipate) diol (PEBA diol, Mn 2000 Da) in 2:3 mass ratio. 10 g of oligodiol and 3 g of isophorone diisocyanate (IPDI) were added in a glass vessel with the catalyst Sn(Oct)2 and reacted for 3 h under a nitrogen atmosphere at 75 °C. After that, 4. 5 g of methyl ethyl ketone (MEK) and 0. 6699 g of 2, 2‐bis(hydroxymethyl) propionic acid (DMPA) were added and reacted for 1 h. 0. 505 g of triethylamine (TEA) was added at 50 °C for 30 min. 0. 21 g of ethylenediamine (EDA) and 36 mL of distilled water were added to the vessel under vigorous stirring for 1 h. Finally, 0. 5075 g of glyoxal was added to the vessel and reacted for 30 min. The stoichiometric ratio of oligodiols/IPDI/DMPA/TEA/EDA/glyoxal was 1:2. 7:1:1:0. 7:0. 7. As displayed in Figure 1 B, the mixture of DFPU and glycol chitosan (CS) formed self‐healing hydrogel at room temperature (abbreviated as CS‐PU hydrogel) and formed cryogel after being frozen at −20 °C (abbreviated as CS‐PU cryogel). Figure 1 A) The synthetic route of novel DFPU crosslinkers. The oligodiol employed was PCL diol alone, or a combination of PCL diol with another oligodiol (such as PEBA diol). B) Schematic representation of the simple process to form self‐healing hydrogel or cryogel. The functional groups of DFPU were confirmed by attenuated total reflectance Fourier transform infrared (ATR‐FTIR) spectroscopy. A peak representing the presence of aldehyde group was observed at 1380 cm −1 (C—H bending of aldehyde) (Figure S1A, Supporting Information). The X‐ray diffractometer (XRD) patterns for nonfunctionalized polyurethane (abbreviated as PU) and DFPU are displayed in Figure S1B (Supporting Information). The data indicated that nonfunctionalized PU was amorphous. After functionalization, the crystalline PCL peaks at 2θ = 21. 21° and 23. 51° showed up. The average molecular weight, the average hydrodynamic diameter, and the zeta potential of PU and DFPU are shown in Table S1 (Supporting Information). The zeta potentials of the PU NPs and the DFPU NPs were −57. 2 ± 0. 4 and −51. 4 ± 0. 9 mV, respectively. These values indicated good stability of the prepared dispersions. The average hydrodynamic diameter of PU NPs was 36. 0 ± 0. 6 nm and that of DFPU NPs was 39. 5 ± 9. 6 nm, respectively. The data for DFPU' are also demonstrated in Table S1 (Supporting Information). The images from transmission electron microscopy, displayed in Figure 2, showed that the DFPU NPs had a spherical shape with a diameter of about 40 nm (Figure 2 A), and after mixing with CS, transformed into irregular oval shape (Figure 2 B). The small‐angle X‐ray scattering (SAXS) profiles for DFPU dispersions in different solid contents are shown in Figure 2 C. When the solid content increased, the flat area (black arrow) in the q range 0. 01 was more apparent, an influence from the structure factor of DFPU NPs. In Figure 2 D, for the homogeneous mixture of DFPU and CS, the hump peak (black arrow) in the q range 0. 01 became obvious after the sol–gel transition, indicating that DFPU NP deformed after crosslinking CS. Figure 2 Characterization of DFPU by TEM and SAXS. A) TEM images for DFPU NPs alone. The size was in the range of 30–50 nm. B) TEM images for DFPU NPs mixed with glycol chitosan. The shape of DFPU nanoparticles changes when mixing with CS. C) SAXS profiles (measured at 37 °C) for DFPU dispersions of various solid contents (5–30 wt%). D) SAXS profiles for DFPU (5 wt%) mixed with twice the volume of glycol chitosan (3 wt%). The composition of the resulted CS‐PU hydrogel is DFPU 1. 7 wt%, glycol chitosan 2 wt%, and 96. 3 wt% water. 2. 2 Optimization of the Composition for the CS‐PU Hydrogel (Contents of Main Chain and Crosslinker) The CS‐PU hydrogel was conveniently achieved by mixing DFPU and CS. To optimize and select the proper composition to form CS‐PU hydrogel with better properties, the amounts (concentrations) of DFPU and CS were adjusted as listed in Table S2 (Supporting Information). For the compositions DFPU 5 wt%/CS 1. 5 wt%, DFPU 2. 5 wt%/CS 1. 5 wt%, and DFPU 2. 3 wt%/CS 2 wt%, the hydrogel shrank and dehydrated after storage. For the composition DFPU 1. 7 wt%/CS 2 wt%, the hydrogel remained stable ( Figure 3 ). For the composition DFPU 1 wt%/CS 2 wt%, a hydrogel could not form ever after 3 d. A saturated water content was achieved when the molar ratio of amine group (CS)/aldehyde group (DFPU) was 1:0. 005. A higher proportion of DFPU (0. 007) let to dehydration and a lower proportion of DFPU (0. 003) failed to form network. These observations indicated that the ratio of the main chain CS and the crosslinker DFPU had a significant influence on the stability of the CS‐PU hydrogel. The composition DFPU 1. 7 wt%/CS 2 wt% was selected to perform subsequent experiments. Figure 3 Optimization of the ratio of DFPU nanoparticulate crosslinker and glycol chitosan (DFPU 1. 7 wt%/CS 2 wt%) to form stable CS‐PU hydrogel without deswelling. A higher ratio of DFPU (DFPU 5 wt%/CS 1. 5 wt%) resulted in shrinkage and deswelling (dehydration) of the hydrogel in 3 d. Details of optimization are supplemented in Table S2 (Supporting Information). 2. 3 Characteristics of the CS‐PU Self‐Healing Hydrogel In Figure 4 A, two hydrogels with different colors were prepared and then cut into two semicircular pieces, and put together for observation of self‐healing at room temperature. One hour later, the scar at the damage site disappeared. The healed gels could support their own weight after healing and endure stretching without breaking at the cut/healed position. Figure 4 Characteristics of the CS‐PU self‐healing hydrogel (from DFPU 1. 7 wt% and CS 2 wt%). A) Macroscopic hydrogel recovery process. B–D) Rheological properties of the hydrogel. E) Needle injectability. In (A), two circular samples were cut into half and then cross placed together for 5 h. After that, the healed sample was stretched by a pair of tweezers. In (B), the strain for the deconstruction was evaluated by the strain sweep (1–500% strain) experiment at 37 °C and 1 Hz. The gel to sol transition occurred when the strain was ≧ 80%. In (C), the damage‐healing properties of hydrogels were demonstrated by measurements under three cycles of the strain change (1% strain → 130% strain → 1% strain → …) at 37 °C and 1 Hz, and the CS‐PU hydrogel could restore its structure after high strain‐induced structural damage, i. e. , with the self‐healing capability. In (D), the static shear viscosities of CS‐PU self‐healing hydrogel versus the shear rate at 37 °C. In (E), the CS‐PU self‐healing hydrogel could be injected through a 26‐gauge needle (260 µm internal diameter). The time‐dependent and frequency‐dependent viscoelasticity of the CS‐PU hydrogel at 37 °C are shown in Figure S2 (Supporting Information). The stabilized storage modulus (G′) of the hydrogel was ≈700 Pa at 37 °C and 1 Hz (Figure S2A, Supporting Information). There was no significant difference in the stiffness of the hydrogel in the frequency range of 1–100 rad s −1 (Figure S2B, Supporting Information). The strain‐dependent viscoelasticity of the hydrogel at 1 Hz 37 °C is shown in Figure 4 B. The G′ values decreased as the dynamic strain increased over the range of 1 to 500%. The gel‐to‐sol transition occurred when the strain exceeded 80%. Finally, the damage‐healing cycles evaluated by successive step changes in the dynamic strain between 130% and 1% as demonstrated in Figure 4 C. Upon the change to the higher strain (130%), the G′ of the hydrogel dropped from ≈700 Pa to ≈100 Pa and was lower than G″. Repeated damage‐healing experiments showed that the self‐healing hydrogel could fully restore their structure after multiple cycles. In Figure 4 D, the static experiment revealed that the steady shear viscosity of the hydrogel decreased as the shear rate increased, which indicated that the CS‐PU hydrogel exhibited shear‐thinning property and suggested good injectability. Indeed, the CS‐PU hydrogel could be injected through a 26‐gauge (260 µm internal diameter) needle, as shown in Figure 4 E. 2. 4 Characteristics of the CS‐PU Cryogel CS solution was mixed with DFPU dispersion at −20 °C for overnight to obtain CS‐PU cryogel. The bulk composition of CS‐PU cryogel was the same as that of CS‐PU hydrogel (DFPU 1. 7 wt%/CS 2 wt%). As displayed in Figure 5 A, the CS‐PU cryogel was compressed (1 mm height) and then recovered their initial shape (8 mm height) by rehydration. The compressibility was about 87. 5%. The mechanical strength, swelling degree, and porosity of the cryogel are listed in Table S3 (Supporting Information). In addition to the high swelling ratio of 2730%, the CS‐PU cryogel absorbed water very fast, demonstrated in Movie S1 (Supporting Information). In Figure 5 B, the image from scanning electron microscopy (SEM) showed interconnected macroporous (250 µm) network in the cross‐section of CS‐PU cryogel. Meanwhile, the freeze‐dried self‐healing hydrogel showed a different porous structure, i. e. , the holes were smaller (100 µm) and fewer holes than those of the CS‐PU cryogel (Figure S3, Supporting Information). The profiles of water swelling for the lyophilized hydrogels and cryogels against the immersion time at 37 °C are shown in Figure 5 C. It was apparent that cryogels were swollen to around 27‐fold of the dried mass in less than 1 min but CS‐PU hydrogel could only be swollen to around 16‐fold of the dried mass in 1 min (and 19‐fold after 48 h). The thermal stability of CS‐PU hydrogel and CS‐PU cryogel is shown in Figure S4A, B (Supporting Information). Derivative thermogravimetry (DTG) is a type of thermal analysis in which the rate of material weight changes upon heating is plotted against temperature. The DTG curves indicated that both CS‐PU hydrogel and CS‐PU cryogel were stable at high temperatures up to about 200 °C. Besides, the structure of CS‐PU cryogel was more heterogeneous than that of CS‐PU hydrogel. Figure 5 Characteristics of the CS‐PU cryogel (from DFPU 1. 7 wt% and CS 2 wt%). A) The compressed cryogel could return to the original shape after rehydration. B) The SEM cross‐sectional image of CS‐PU cryogel showing interconnected macroporous network. C) Water swelling of the lyophilized hydrogels and cryogels, against the immersion time. D) Cryogels distorted by external force could return to the original shape in 1 s after immersion in water. E) The cryogel (length 4 mm, thickness 1 mm) could be injected by a conventional 18‐gauge needle (838 µm internal diameter) and recover the original shape after injection in water without being distorted. Deformed CS‐PU cryogels had the ability to recover to the original shape within 1 s, as illustrated in Figure 5 D and Movie S2 (Supporting Information). Moreover, the triangle‐shaped CS‐PU cryogels (5 mm side length, 1 mm thickness) could be squeezed through an 18‐gauge needle (838 µm inner diameter) and immediately return to their original geometry after injection, as illustrated in Figure 5 E and Movie S3 (Supporting Information). 2. 5 The Responsive Ability and In Vitro Degradation of CS‐PU Hydrogel and Cryogel The pH and aniline responsiveness of CS‐PU hydrogel and cryogel is summarized in Table 1 and Figure 6 A, B. CS‐PU hydrogel and cryogel were fully liquefied after soaking in acetic acid for 5 and 7 min, respectively (Figure 6 A). In contrast, CS‐PU hydrogel and cryogel were fully liquefied after soaking in aniline for 48 and 72 h, respectively (Figure 6 B). In vitro degradation profiles in 37 °C phosphate buffered saline (PBS) are demonstrated in Figure 6 C, D. Gels prepared from CS with DFPU' crosslinker were named as CS‐PU' gels. After 28 d CS‐PU hydrogel remained 87. 8% by weight, while CS‐PU' hydrogel remained 60. 0% by weight (Figure 6 C). Meanwhile, CS‐PU cryogel remained 91. 5% by weight, and CS‐PU' cryogel remained 77. 1% by weight after 28 d (Figure 6 D). The comparative degradation profiles of PU films are also shown in Figure S5 (Supporting Information). In addition, the gel fraction for PU films, CS‐PU hydrogels, and cryogels is shown in Figure S6 (Supporting Information). No obvious difference in the gel fraction was observed among CS‐PU hydrogel, CS‐PU' hydrogel, CS‐PU cryogel, and CS‐PU' cryogel. Taken together, these data suggested that by changing the soft segment used in the synthesis of the crosslinker, the degradation rates of the resulted hydrogels and cryogels could be adjusted. Table 1 Responsive ability of CS‐PU hydrogel and CS‐PU cryogel Materials Exposed to Outcome CS‐PU hydrogel 1 mL 97% acetic acid (pH < 1) Liquefied in 5 min 1 mL aniline (pH ≈ 10) Liquefied after 48 h CS‐PU cryogel 1 mL 97% acetic acid (pH < 1) Liquefied in 7 min 1 mL aniline (pH ≈ 10) Liquefied after 72 h John Wiley & Sons, Ltd. Figure 6 The responsive ability and in vitro degradation of cryogels and hydrogels (from DFPU 1. 7 wt% and CS 2 wt%). The CS‐PU cryogel and CS‐PU hydrogel were soaked in A) acetic acid and B) aniline, where the remaining volume of the cryogel and hydrogel were measured and compared. The remaining volume 0% indicates dissolution. Meanwhile, in vitro degradation was conducted by immersion in PBS at 37 °C. C) The degradation of CS‐PU hydrogel was compared with that of CS‐PU' hydrogel, and D) the degradation of CS‐PU cryogel was compared with that of CS‐PU' cryogel to demonstrate the tunable degradation rate. 2. 6 Cell Survival and Proliferation in the CS‐PU Hydrogel and CS‐PU Cryogel The immediate viability of neural stem cells (NSCs) in CS‐PU hydrogel evaluated by the VB‐48 assay is shown in Figure 7 A. The cell viabilities in the groups of medium, CS, DFPU, and DFPU+CS were 88. 2%, 81. 5%, 90. 5%, and 85. 1%, respectively. There was no significant difference in the cell viability among the groups. Meanwhile, cell survival in longer term was assessed by the cell growth over 7 d of culture. The data are shown in Figure 7 B. After 3 d, the amounts of cells were ≈312% and ≈360% in CS‐PU hydrogels and CS‐PU cryogels, respectively, compared to the initial values. After 7 d, the amounts of cells were ≈497% and ≈571% in CS‐PU hydrogels and CS‐PU cryogels, respectively. Figure 7 Cell survival and proliferation in the CS‐PU hydrogel and the CS‐PU cryogel. A) The vitality of NSCs determined by the VB‐48 assay. The data of control group were obtained from cells in the culture medium. B) The viability and proliferation of NSCs embedded in the CS‐PU hydrogel and CS‐PU cryogel determined by the CCK‐8 assay. The cell viability value (%) was calculated from optical density after deduction from the blank control (i. e. , the hydrogel or the cryogel without cells) and normalized to that of initial cells. *** p < 0. 001 and **** p < 0. 0001 among the indicated group. 2. 7 Biocompatibility by Rat Subcutaneous Implantation The foreign body reaction was evaluated by histological staining of the explanted samples and the result is shown in Figure 8 A. Mild inflammation at the border of CS‐PU cryogel was observed after two weeks with the presence of inflammatory cells. In Figure 8 B, PU (nonfunctionalized) was used as the control, which showed a fibrous capsule of 57. 3 um thickness. CS‐PU cryogel did not show any fibrous capsule. In addition, immunofluorescence staining was performed to obtain the population ratio of M1 macrophages to M2 macrophages, as shown in Figure 8 C, D. There was no significant difference in the M2/M1 ratio between CS‐PU cryogel and PU film. Both groups had an M2/M1 ratio of about 3, higher than that (about 0. 5) reported for polylactide. 20 Figure 8 Foreign body reaction of CS‐PU cryogels after rat subcutaneous implantation. A) Histology of H&E‐stained sections after implantation for 14 d. The scale bar represents 500 µm. B) The extent of foreign body reaction could be revealed by the thickness of the fibrous capsule (white arrows) based on the histology. C) Immunofluorescent images (marker protein expression) of macrophages, stained by the mouse monoclonal anti‐CD86 antibody for M1 macrophages (red), or mouse monoclonal anti‐CD163 antibody for M2 macrophages (green). D) Quantification of M1 macrophage and M2 macrophage populations. Results are expressed as mean ± SD, N = 3. ** p < 0. 01, and **** p < 0. 0001 among the indicated groups. PU (nonfunctionalized) films were used as the control. 3 Discussion Biodegradable crosslinkers in the form of polyurethane nanoparticles were successfully synthesized by a green water‐based process. This type of crosslinker for producing dynamic Schiff bonding is rarely reported. According to the dynamic light scattering (DLS) measurement, the DFPU NPs were stably suspended in water. Furthermore, ATR‐FTIR results revealed that PU was successfully modified by glyoxal with aldehyde groups. Meanwhile, XRD patterns of PU films demonstrated that the modification induced crystallinity of the PCL segment was induced after modification. In addition, the morphology of DFPU was investigated by SAXS and observed by transmission electron microscopy (TEM). The TEM image of DFPU NPs showed the NPs in spherical shape. After mixing with CS, DFPU NPs gradually transformed into irregular oval shape gradually. The deformation of DFPU NPs was probably caused by the different reaction rates of aldehyde groups and amine groups during the crosslinking process. Optimal synthetic conditions exert a significant influence on the properties of the product. 21 The formation of the hydrogel was optimized with the procedures described below. First of all, the undiluted DFPU (28 wt%) was mixed with CS (3 wt%). In observations of the properties of CS‐PU hydrogel, the hydrogels underwent syneresis after 24 h. According to the literature, the properties of swelling and syneresis were correlated to the effective crosslinking density described by the polymer network theory. 22 Therefore, the crosslinker DFPU dispersion was prepared in various concentrations to optimize the composition of the hydrogel. After the formation of hydrogel, the crosslinking reaction kept going, which increased the crosslinking density. When the crosslinking density was too high, the structure of the hydrogel started to shrink, causing dehydration. The shape of the hydrogel could be maintained for a longer period of time when the proportion of the main chain (CS) in the hydrogel increased. The hypothetical mechanism for the formation of deswelling hydrogels and stable hydrogels is proposed in Figure 3. In addition, it was unable to form a hydrogel when the concentration of DFPU was too low. These results indicated that the hydrogel was stable and water‐saturated when the ratio of crosslinker to main chain was optimized. Considering the stability of CS‐PU hydrogel, the composition DFPU 1. 7 wt%/CS 2 wt% was selected for major experiments. The CS‐PU hydrogel was prepared by Schiff reaction from the aldehyde group of DFPU and the amine group of CS. As mentioned previously, dynamic Schiff base could lead to the self‐healing ability of CS‐PU hydrogel. The healed gel could endure stretching without breaking at the cut/healed position and the rheometer was utilized for the further analysis of self‐healing properties of hydrogel. We found that the mechanical strength of hydrogel was close to the initial state after multiple cycles of damage‐healing processes, indicating the quick recovery of the inner network of CS‐PU hydrogel. CS‐PU hydrogel not only showed favorable self‐healing ability but revealed strong shear thinning behavior. To be more specific, CS‐PU hydrogel could be thinned into liquid form under the high shear rates and recovered back into gel rapidly after removal of shear stress. With such characteristics, the CS‐PU hydrogel was not only suitable for subcutaneous injection through a 26‐gauge needle, but may also have potential for future applications in 3D printing. We unexpectedly discovered in our study that the mixture of DFPU and CS could form cryogel after freezing at −20 °C in our study. Self‐healing hydrogels and cryogels made of the same main chain and the crosslinker has not been reported so far. According to the literature, the formation rate of crosslinking network and ice crystal exerted a significant influence on the generation of a cryogel. 23 Therefore, we assumed that the DFPU crosslinker played an important role on forming CS‐PU cryogels. The spherical crosslinker resulted in the formation of unique network possibly because each aldehyde group on the NPs exhibited different steric hindrance on the main chain. When the mixture of CS and DFPU was frozen, the ice crystals in the mixture grew until they met the facets of other crystals. DFPU crosslinked the chains of CS tightly around the ice crystals. After thawing the ice crystals, interconnected pores were generated inside the gel. Because of the interconnected macropores (250 µm) in the cryogel, the CS‐PU cryogel had higher swelling ratio and swelling rate than that of freeze‐dried CS‐PU hydrogel. In addition, CS‐PU cryogel (5 mm side length, 1 mm thickness) could be squeezed through an 18‐gauge needle (838 µm inner diameter) without shape loss or gross damage. In our study, the compressibility of target cryogel was consistent with its high degree of porosity. The result indicated that the CS‐PU cryogel could be compressed to almost its minimum volume and injected through a needle. The CS‐PU cryogel regained its geometry and architecture after being injected. According to the literature, cryogels were suitable as carriers for minimally invasive delivery of adherent cells with effective protection from compression forces during injection. 24 Compared with traditional procedures, CS‐PU cryogels were able to be injected to a specific location without invasive surgery, which could decrease scarring, lessen the risk of infection, and reduce recovery process. CS‐PU hydrogel and cryogel developed here were sensitive to acetic acid and aniline. On the basis of the literature, Schiff bases were often sensitive to hydrolysis under acidic conditions, 25 which accounted for the liquefaction of both CS‐PU hydrogels and cryogels under acidic conditions. In addition, when treated with aniline, CS‐PU hydrogels and cryogels were also liquefied. Aniline with monoamine group could compete with CS to crosslink DFPU, leading the destruction of 3D network. 12 Besides chitosan, many biomacromolecules containing amine groups are expected to form gels through the crosslinking of DFPU based on Schiff base. Meanwhile, the gels with multiresponsive properties may have interests in drug release. Degradation of the hydrogels after implantation is important in tissue engineering. The degradation rates should be adjusted in order to suit for various tissues. Gels with tunable degradation rates are highly demanded. Hsieh et al. prepared a gel slowly degraded to provide better support for vasculogenesis in vivo. [qv: 13b] Hsu et al. developed green/water‐based, biodegradable polyurethanes of which the degradation rate could be altered by the types of the component oligodiols. 26 In the present study, two of the water‐based polyurethanes (DFPU and DFPU') were functionalized as crosslinkers. It is reasonable to assume that the degradation rates of CS‐PU hydrogels and cryogels prepared with different types of crosslinkers remained to be tunable. In fact, CS‐PU degraded slower than CS‐PU'. We speculated that the soft segments of the NP crosslinker developed here may be further changed to obtain hydrogels and cryogels with desired properties. Analyses of cell viability and cell growth were performed on CS‐PU hydrogels and cryogels in vitro to evaluate the tissue engineering applications. The results indicated that no significant difference was observed in the immediate cell viability among the control group (culture medium), DFPU, CS, and DFPU+CS. Meanwhile, significant cell proliferation was confirmed after 7 d of cell culture in CS‐PU hydrogel and cryogel. These data supported the cytocompatibility of DFPU crosslinker, CS‐PU hydrogel, and CS‐PU cryogel. The hydrogel as negative control group was prepared by mixing glyoxal and CS (Figure S7, Supporting Information). After 3 and 7 d, the amounts of cells in negative control groups were lower than those in the experimental CS‐PU hydrogel groups, indicating that the crosslinker DFPU was superior to the conventional crosslinker glyoxal in the biocompatibility. In terms of the mechanical properties, the literature indicated that the matrix of 0. 1–10 kPa stiffness could offer a favorable environment for neural cells. 27 During the long‐term culture, we found that the degradation rates of CS‐PU hydrogel and cryogel both increased, probably because of the pH change in the microenvironment around the samples. In addition, amine‐containing molecules from the metabolism of cells could also influence the degradation rates of the gel. These conditions may contribute to the degradation of CS‐PU hydrogel and cryogel in a biological environment. Rat implantation showed that the CS‐PU cryogel completely degraded after 28 d (Figure S8, Supporting Information), possibly because the multiple responsiveness of CS‐PU cryogel accelerated the biodegradation rates of the gel. Moreover, rat implantation at 14 d revealed mild inflammation at the border of CS‐PU cryogel. According to the literature, the M2 macrophages play an essential role in tissue remodeling and suppression of inflammatory immune reactions. 28 In the present study, CS‐PU cryogel as well as PU film (control group) recruited less M1 macrophages and more M2 macrophages than PLA films. These data indicated that CS‐PU cryogel was superior to the conventional PLA in biocompatibility. Polyurethane nanoparticles and films were reported to inhibit the macrophage polarization toward the M1 phenotype. 20 The good individual biocompatibility of PU and CS may explain the low immune reaction of the CS‐PU cryogel. In summary, we developed a novel biodegradable NP crosslinker which could be used to fabricate CS‐PU self‐healing hydrogels and cryogels with good cytocompatibility, biocompatibility, injectability, and multiresponsiveness. Moreover, the degradation rates of target gels could be further fine‐tuned by changing the soft segment of the crosslinker. The CS‐PU gels are promising biomedical materials and may have good potential in tissue engineering applications. 4 Conclusion Novel waterborne and biodegradable PU NP crosslinker dispersions with aldehyde groups were synthesized and characterized in this study. CS‐PU self‐healing hydrogels and cryogels were formed from the same ingredients by Schiff reaction between DFPU and CS under room temperature and −20 °C, respectively. These gels were found to be sensitive to pH values and amine‐containing molecules owing to the property of Schiff bases. The degradation rates of gels could be adjusted by the types of the component oligodiols in DFPU. Both CS‐PU self‐healing hydrogels and cryogels were injectable (26‐gauge and 18‐gauge needles, respectively) and demonstrated good cell proliferation (497% and 571% after 7 d, respectively). CS‐PU self‐healing hydrogels had excellent self‐healing properties that fully (≈100%) recovered after damage. CS‐PU cryogels exhibited great water absorption (≈2730%), compressibility (eightfold), and the ability of shape recovery. These materials showed low immune responses in rat 14‐d implantation. Therefore, DFPU crosslinkers and CS‐PU gels are promising new materials for biomedical applications. 5 Experimental Section Synthesis of DFPU Crosslinkers : DFPU was synthesized as described earlier. PCL diol was purchased from Sigma (USA). PEBA diol was supplied from Greco (Taiwan). IPDI, DMPA, and EDA were acquired from Acros (USA), Sigma (USA), and Tedia (USA), respectively. MEK and TEA were both obtained from J. T. Baker (USA) and used as received. Glyoxal was purchased from Alfa Aesar (USA). Nonfunctionalized PU, to contrast DFPU, was synthesized from a waterborne process previously developed. 29 The process was similar to that of DFPU except that no glyoxal was added. Characterization of DFPU Crosslinkers : The waterborne DFPU synthesized was in the nanoparticle form. The hydrodynamic diameter ( D h ) and the zeta potential of the NPs in dispersion (3000 ppm) were measured by a nanoparticle analyzer (Delsa Nano, Beckman Counter) involving the principle of dynamic/electrophoretic light scattering. The morphology of the DFPU NPs in dispersion was examined by TEM (Hitachi H‐7100, Japan). The DFPU dispersion was diluted with distilled water to a concentration of 5000 ppm. The diluted DFPU dispersion was dropped on the copper grid for 2 min and removed of excess DFPU using a filter paper. After that, the phosphotungstic acid was dropped on the copper grid for 30 s before observation. The DFPU dispersions in various solid contents (5–30 wt%) were investigated by SAXS at the beamline 23A of the National Synchrotron Radiation Research Center (NSRRC) at Hsinchu, Taiwan. DFPU dispersions were cast into films and characterized using an ATR‐FTIR spectrometer (Spectrum 100 model, Perkin Elmer). PU films cast from nonfunctionalized PU dispersion were also analyzed for comparison. The size distribution (polydispersity index, PDI) of DFPU and the molecular weight (Mw and Mn) were obtained by the gel permeation chromatography (GPC, JASCO, Japan) coupled with an RI (RI‐930) detector using N, N ‐dimethylacetamide (DMAc, Tedia) as the eluent. The diffraction peaks of DFPU were investigated using an XRD (X'Pert, PANalytical, Netherland). The pyrolytic temperature of DFPU was obtained using the thermogravimetric analyzer (TGA, Q50, TA, USA) at a heating rate of 10 °C min −1 under N 2. Preparation of CS‐PU Self‐Healing Hydrogels and CS‐PU Cryogels : The water‐soluble CS (Wako, Japan) was crosslinked by DFPU to prepare CS‐PU self‐healing hydrogels and CS‐PU cryogels. Glycol chitosan (CS) powder was dissolved in deionized water and vortexed in 3 wt% concentration. The crosslinker DFPU dispersion was prepared in various concentrations (10, 7, 5, and 3 wt%) to optimize the composition of the hydrogel. CS solution was mixed with DFPU dispersion by vortex under room temperature to form a homogenous mixture, which were CS‐PU self‐healing hydrogels. The cryogel composition was selected with the ratio of optimized ingredients. To prepare the cryogel, CS solution was mixed with DFPU dispersion by vortex under room temperature and then immediately placed in a freezer set at −20 °C for overnight. Characterization and Rheological Evaluation of CS‐PU Self‐Healing Hydrogels : The morphology of the DFPU NP crosslinked CS was examined by a TEM (Hitachi H‐7100, Japan). The DFPU dispersion mixed with CS solution was diluted with distilled water to a concentration of 5000 ppm. Sample preparation followed that described previously. The microstructure of DFPU mixed with CS was also examined by the SAXS at the beamline 23A of the NSRRC. For visualization of the self‐healing property of the CS‐PU hydrogel, two hydrogel disks were separately prepared. Trypan blue was added in one hydrogel disk to make the color different. Each of the white and blue hydrogels were cut into two pieces and put two semicircles of different colors together to test if they formed a united disc. Rheological properties of the CS‐PU self‐healing hydrogel was examined by a rheometer (Rheometric RS5, TA) at 37 °C. The storage modulus and loss modulus (G' and G'') were determined against time at a constant frequency of 1 Hz (6. 28 rad s −1 ) and 1% strain. After that, the frequency sweep was obtained at a constant strain of 1% in the angular frequency range (1−100 rad s −1 ). The dynamic strain sweep was evaluated at a frequency of 1 Hz from 0. 1% to 500% strain. The shear thinning property was analyzed by the steady shear experiment, where the viscosity was measured against the shear rate. For quantitative evaluation of the self‐healing process, G' and G'' at the constant frequency of 1 Hz were measured by damage‐healing cycles at the high strain (130%) for damage and at the low strain (1%) for healing. Characterization and Compression Property of CS‐PU Cryogels : The porous structure of the CS‐PU cryogel (cross‐section) was examined by SEM (Hitachi TM3000, Japan) operated in 3 kV. The porosity of the CS‐PU cryogel was measured by immersion in ethanol. The percent porosity was calculated by the equation( W w − W d )/ ρV × 100%, where W w was the wet weight of CS‐PU cryogel after immersion in ethanol, W d was the dry weight of CS‐PU cryogel before immersion, ρ was the density of ethanol, and V was the volume of the CS‐PU cryogel. The water swelling ratio of the cryogel was measured against time at 37 °C. Pre‐weighed dry cryogels ( W d ) were immersed in deionized water. The swollen weight ( W w ) of the cryogels was surveyed after removal of excess water from the surface. The swelling ratio of the cryogel was calculated by the equation ( W w − W d )/ W d × 100%. The dynamic compression modulus of the cryogel was evaluated with a dynamic mechanical analyzer (DMA, Q800, TA, USA) at a constant frequency of 1 Hz and 0. 1% compression strain at 37 ο C. The ability of CS‐PU cryogel to flow through a conventional‐gauge needle and to regain the original shape after injection was examined. Triangle‐shaped CS‐PU cryogels were suspended in 1 mL of water and injected by an 18‐gauge needle. Responsiveness and In Vitro Degradation of CS‐PU Hydrogel and CS‐PU Cryogel : The environmental responsiveness of the CS‐PU hydrogel and cryogel was tested by immersion of the sample (0. 9 mL) in 1 mL acetic acid and aniline, respectively. The volume of samples remained in each solvent at each time point ( V ) was measured. In vitro degradation of CS‐PU hydrogel and cryogel was evaluated by immersion in PBS under 37 °C. Degradation of the hydrogel and cryogel prepared from CS using DFPU' instead of DFPU was also evaluated for comparison. Before the experiment, CS‐PU gels and CS‐PU' gels were lyophilized for 48 h and weighed ( W i ). Samples were washed with deionized water, freeze‐dried, and weighed ( W f ) after 7, 14, 21, and 28 d. The remaining weight (%) of samples was obtained by the equation W f / W i × 100%. The degree of chemical crosslinking for the DFPU films, CS‐PU gels, and CS‐PU' gels was estimated by gel fraction. The samples were lyophilized for 24 h and weighed ( W i ). Samples were immersed in MEK for 24 h. They were freeze‐dried for 24 h and weight ( W f ). The gel fraction (%) for each sample was obtained from the equation ( W f / W i ) × 100%. Cell Culture and Cell Viability Analysis : NSCs were obtained from adult mouse brain. NSCs were cultured in the medium containing 1:1 Ham's F12 (Gibco) and HG‐DMEM supplemented with 10% fetal bovine serum (FBS) and 400 µg mL −1 G418 (Invitrogen). Cell viability after exposure to various ingredients (medium, DFPU, CS and DFPU+CS) for 10 min was analyzed by the VitaBright‐48 (VB‐48) assay. The viability of cells was determined by staining with VitaBright‐48 (VB‐48, blue) and propidium iodide (PI, red) solution. VB‐48‐positive and PI‐positive cells each represented viable and dead cells, and the respective population was quantified by the Nucleo‐Counter NC‐3000. Long‐term cell proliferation in the CS‐PU hydrogel or CS‐PU cryogel up to 7 d was obtained by the cell counting kit‐8 (CCK‐8, Sigma‐Aldrich) assay. The cell viability value (%) was calculated from optical density after deduction from the blank control (i. e. , the hydrogel or the cryogel without cells) and normalized to that of initial cells. Rat Subcutaneous Implantation : The CS‐PU cryogel and PU film (as the control) of ≈1 cm × 1 cm were inserted into the subcutaneous sites of the adult Sprague‐Dawley rats. The samples were removed after implantation for two weeks. The samples with surrounding tissues were fixed in formalin. The thickness of the fibrous capsule was obtained by examination under the optical microscope after hematoxylin and eosin (H&E) staining. The repair and the inflammatory response in vivo were investigated by immunofluorescence staining of tissue sections. The tissue sections were treated with mouse anti‐CD86 antibody (Bio‐Rad, USA) for staining of M1 macrophages, or mouse anti‐CD163 antibody (Bio‐Rad, USA) for staining of M2 macrophages at 4 °C overnight. After that, the tissue sections were incubated with secondary antibodies at room temperature for 1 h. The number of fluorescent cells was counted under a fluorescence microscope (Nikon, eclipse 80i, Japan). All procedures were approved by the Animal Care and Use Committee (NTU105‐EL‐00029). Statistical Analysis : All the experimental data were independently confirmed for three times. Statistical analysis was represented by the Student t ‐test. Results were considered statistically significant when p values were < 0. 05. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file. Supplemental Video 2 Click here for additional data file. Supplemental Video 3 Click here for additional data file.
|
10. 1002/advs. 201901412
| 2,019
|
Advanced Science
|
The miR‐193a‐3p‐MAP3k3 Signaling Axis Regulates Substrate Topography‐Induced Osteogenesis of Bone Marrow Stem Cells
|
Abstract Substrate topographical features induce osteogenic differentiation of bone marrow stem cells (BMSCs), but the underlying mechanisms are unclear. As microRNAs (miRNAs) play key roles in osteogenesis and bone regeneration, it would be meaningful to elucidate the roles of miRNAs in the intracellular signaling cascade of topographical cue‐induced osteogenic differentiation. In this study, the miRNA expression profile of the topographical feature‐induced osteogenic differentiation group is different from that of the chemical‐factors‐induced osteogenic differentiation group. miR‐193a‐3p is sensitive to substrate topographical features and its downregulation enhances osteogenic differentiation only in the absence of osteogenesis−inducing medium. Also, substrate topographical features specifically activate a nonclassical osteogenetic pathway—the mitogen‐activated protein kinase (MAPK) pathway. Loss‐ and gain‐of‐function experiments demonstrate that miR‐193a‐3p regulates the MAPK pathway by targeting the MAP3k3 gene. In conclusion, this data indicates that different osteogenic‐lineage‐related intracellular signaling cascades are triggered in BMSCs subjected to biophysical or chemical stimulation. Moreover, the miR‐193a‐3p‐MAP3k3 signaling axis plays a pivotal role in the transduction of biophysical cues from the substrate to regulate the osteogenic lineage specification of BMSCs, and hence may be a promising molecular target for bone regenerative therapies.
|
1 Introduction Stem cells reside in a complex microenvironment in which extracellular biophysical cues, such as the topography and stiffness of the extracellular matrix (ECM), as well as various biochemical factors such as growth factors and hormones, play pivotal roles in effecting cell survival, self‐renewal, and differentiation. 1, 2 An understanding of how biophysical cues regulate stem cell function and lineage fate specification would facilitate the design of novel biomaterials to regulate stem cell differentiation and provide guidance for the development of new tissue engineering and regenerative medicine strategies. Stem cells cultured on biomaterials can perceive and respond to topographical features by altering cellular adhesion, cytoskeletal organization, and ultimately cell fate. 3 However, at present our knowledge of such extracellular signal‐to‐intracellular signal transductions is limited. microRNAs (miRNAs) are a class of small noncoding RNAs that regulate gene expression at the post‐transcriptional stage by binding to the 3′ untranslated region (UTR) of target mRNAs, which either changes mRNA stability or inhibits protein translation. 4 miRNAs often simultaneously modulate the activity of several target genes and signaling/regulatory networks, resulting in profound biological effects. 5 miRNAs are important regulators of gene expression in various organisms and biological/pathological 6 processes, including the proliferation, migration, and differentiation of bone marrow stem cells (BMSCs). However, the roles of miRNAs in regulating the osteogenesis of BMSCS in response to topographical substrate cues are largely unknown. 7 In this study, electrospun poly‐ l ‐lactide (PLLA) membranes with a random nanofibrous arrangement (random group) were utilized to explore the mechanism by which topographical cues in the niche regulate stem cell function and fate. The random arrangement of nanofibers in the scaffold mimics the extracellular collagen distribution and promotes the osteogenic differentiation of human bone marrow‐derived stem cells (hBMSCs). An miRNA chip assay and a bioinformatics/functional analysis were carried out to explore the role of miRNAs in topographical cue‐induced osteogenic differentiation. Substrate stiffness is another important microenvironmental biophysical cue that determines the fate of BMSCs. Therefore, in this study, Gel‐MA hydrogel were utilized as another material model in a preliminary investigation to evaluate whether the miRNAs‐related mechanism elucidated in the PLLA model is of universal significance in biophysical cue‐induced osteogenesis. 2 Results Electrospun PLLA nanofibrous substrates promoted the spreading of hBMSCs and enhanced their osteogenic differentiation. We utilized electrospun PLLA membranes with a random fiber arrangement (random group) ( Figure 1 b) that mimicked the topology of collagen in the ECM as an artificial substrate, to explore the mechanisms by which topographical cues in the niche regulate stem cell function and lineage fate specification. Cast, flat PLLA membranes (flat group) (Figure 1 a) were used as the negative control. Figure 1 Morphology and osteogenic differentiation of human bone marrow‐derived stem cells (hBMSCs) on poly‐ l ‐lactide (PLLA) fibrous scaffolds and cast PLLA films. a) Representative scanning electron microscopy images showing the smooth surface in the flat group. b) Representative scanning electron microscopy image of randomly arranged nanofibers (scale bar, 5 µm). c, d) Confocal immunofluorescence staining of nuclei with 4′, 6‐diamidino‐2‐phenylindole (DAPI) in hBMSCs from the c) flat group and d) the random fiber group after 4 h in culture. e, f) Confocal immunofluorescence staining of F‐actin with rhodamine‐labeled phalloidin in hBMSCs from the e) flat group and f) the random fiber group after 4 h in culture. g) Merged confocal images of (c) and (e), with bright‐field microscopy images of the flat group. h) Merged confocal microscopy images of (d) and (f), with bright‐field microscopy images of the random group. Scale bar, 25 µm. Quantification of i) the nuclear size and j) cell spreading area of hBMSCs in the flat group and random group ( n = 30). j) Quantification of hBMSCs in the flat and random groups ( n = 30). k) Runx2 and l) Ocn mRNA levels in hBMSCs in the flat and random groups at 7, 14, and 21 days, respectively. Results are means ± SEM ( n = 3). * p < 0. 05, * by two‐sample t ‐test. The hBMSCs in the random group had cytoplasmic projections that were aligned along individual nanofibers (Figure 1 h, white arrows), resulting in an elongated, branched morphology (Figure 1 d, f, h). By contrast, the hBMSCs in the flat group displayed a spindle‐like morphology (Figure 1 c, e, g). The area of cellular spread and the nuclei were significantly larger in the random compared to the flat group (Figure 1 i, j). The expression levels of the osteogenic marker genes Runx2 and Ocn were upregulated in the random group, particularly at 14 days (Figure 1 k, l). Therefore, the topological structure of the nanofiber scaffold can induce an osteogenic morphology of hBMSCs, driving their osteogenic differentiation. Randomly arranged PLLA fibrous substrates downregulated miR‐193a‐3p expression. Next, miRNA microarrays were used to investigate the role of miRNAs in topographical cue‐induced osteogenic differentiation. hBMSCs in the flat group cultured in chemical osteogenic medium served as the positive control (flat OS+ group). A cluster heatmap analysis showed significantly differentially expressed miRNAs in the random, flat, and flat (OS+) groups ( Figure 2 a). We next performed a 3D principal component analysis (PCA) to evaluate the spatial distribution of the nine samples from the three groups (Figure 2 b). The results showed that hBMSCs in the flat OS+ group and random group have markedly different miRNA profiles. Figure 2 Differential microRNA (miRNA) levels determined using miRNA microarrays and according to component distribution. a) Heatmap of hierarchical clustering and b) 3D PCA of miRNA levels in hBMSCs cultured on the flat, flat (OS+), and random substrates for 14 days. c) The 2D PCA of significantly downregulated miRNAs in the random and flat (OS+) group. d) miR‐193a‐3p levels in hBMSCs from the random, flat (OS+), and random (OS+) groups after 14 days in culture. Results are means ± SEM ( n = 3). Samples were subjected to one‐way analysis of variance (ANOVA) with Tukey's post hoc test (* p < 0. 05, ** p < 0. 01, *** p < 0. 001). To identify the miRNAs involved in topographical cue‐activated osteogenic differentiation, we screened out 9 significantly downregulated miRNAs in the flat OS+ group, together with 19 downregulated miRNAs in the random group (2. 5‐fold or greater difference in expression level compared to the flat group). In the 2D PCA of significantly downregulated miRNAs, the component distribution that represented the characteristic downregulated miRNAs in the random group (blue circle) differed from that of the flat (OS+) group (green circle) (Figure 2 c). This suggests that topographically activated osteogenic differentiation differs from that induced biochemically. The 2D PCA showed that miR‐193a‐3p was significantly downregulated in the random group, which is the farthest from the center of the flat (OS+) group. Thus, we selected miR‐193a‐3p as a representative miRNA for further analysis. Next, quantitative polymerase chain reaction (qPCR) showed that miR‐193a‐3p expression was downregulated in the random group irrespective of use of chemical osteogenic medium, while there was no significant difference between the flat (OS+) group and flat group after 14 days in culture (Figure 2 d). These results indicate that miR‐193a‐3p is downregulated by topological activation and is involved in the cellular sensing of changes in surface topography. Downregulation of miR‐193a‐3p promoted the osteogenic differentiation of hBMSCs. Next, we investigated the role of miR‐193a‐3p in the function of hBMSCs by treating them with agomir‐193a‐3p (a miR‐193a‐3p agonist) and antagomir‐193a‐3p (a miR‐193a‐3p inhibitor), together with their scrambled controls. Intracellular miR‐193a‐3p was markedly upregulated by agomir‐193a‐3p, and markedly downregulated by antagomir‐193a‐3p (Figure S1, Supporting Information). qPCR analysis showed that the Runx2 and Ocn mRNA levels were markedly upregulated by antagomir‐193a‐3p, but downregulated by agomir‐193a‐3p, when compared to the corresponding scrambled controls ( Figure 3 a, b). Functionally, antagomir increased the protein expression levels of RUNX2 and OCN, while agomir decreased the protein expression levels of RUNX2 (Figure 3 c; Figure S2, Supporting Information). Consistently, overexpression of miR‐193a‐3p weakened ALP staining, while knockdown of miR‐193a‐3p enhanced it (Figure 3 e, f). Interestingly, agomir‐193a‐3p did not decrease the protein expression levels of OCN (Figure 3 c; Figure S2b, Supporting Information). These results thus revealed that downregulation of miR‐193a‐3p expression promotes the osteogenic differentiation of hBMSCs. Figure 3 Downregulation of miR‐193a‐3p correlates with enhancement of the osteogenic differentiation of hBMSCs. a–c) Effects of agomir‐193a‐3p, antagomir‐193a‐3p, and their scrambled controls on the mRNA levels of a) Runx2 and b) Ocn, and their protein levels c) in hBMSCs. d) Gross and e) magnified images of alkaline phosphatase (ALP) staining of hBMSCs treated with agomir‐193a‐3p, antagomir, or their scrambled controls for 5 days. Results are means ± SEM ( n = 3). Samples were subjected to one‐way ANOVA with Tukey's post hoc test (* p < 0. 05, ** p < 0. 01, *** p < 0. 001). miR‐193a‐3p antagomir‐loaded nanofiber scaffolds enhanced the healing of critical‐sized bone defects in rat cranium. To confirm the role of miR‐193a‐3p in bone defect healing, we covered critical‐sized bone defects in rat cranium with four membranes composed of PLLA nanofiber scaffolds loaded by lyophilization with i) agomir‐193a‐3p, ii) antagomir‐193a‐3p, iii) scrambled control, and iv) water. After 4 weeks, microcomputed tomography (micro‐CT) image reconstruction showed that in the group with PLLA membranes loaded with antagomir‐193a‐3p, newly formed high‐density bone almost filled the defect, while there was scant newly formed bone in the groups loaded with agomir‐193a‐3p, scrambled control, and blank ( Figure 4 a). At 8 weeks after implantation, the bone defect in the antagomir‐193a‐3p‐loaded group had almost completely healed, while markedly less newly formed low‐density bone was found in the groups loaded with scrambled control and blank. Only minimal new bone formation occurred in the group loaded with agomir‐193a‐3p (Figure 4 a). Quantitative analysis showed that both the regenerated bone volume (BV) and the bone mineral density (BMD) were highest in the antagomir‐193a‐3p group (Figure 4 b, c). The results of the histological analysis corroborate the micro‐CT data (Figure 4 d). At 8 weeks postimplantation, the bone defect in the antagomir‐193a‐3p‐loaded group had almost completely healed with obvious bone‐structure formation, while less newly‐formed bone was observed in the groups loaded with scrambled control and blank. Only minimal new bone formation was observed in the group loaded with agomir‐193a‐3p, with obvious unrepaired defect areas. Therefore, these results showed that downregulation of miR‐193a‐3p markedly enhanced the healing of critical‐sized bone defects. Figure 4 Nanofiber membranes loaded with miR‐193a‐3p antagomir enhanced the healing of critical‐sized bone defects. a) Representative micro‐CT images and sagittal views of rat cranial critical‐sized full‐thickness defects at 4 and 8 weeks after surgery (scale bar, 5 mm). Yellow circles and white arrows indicate the bone defect area. b, c) Quantitative analysis of BV and BMD of the newly formed bone. Data are means ± SE) ( n = 6) and all p ‐values are based on one‐way ANOVA with a post hoc test (* p < 0. 05). d) Histological results of 8‐weeks H&E staining (Top row) and Masson staining (Bottom row). Blue arrows denote the newly‐formed bone. (bar = 200 µm) miR‐193a‐3p regulates the mitogen‐activated protein kinase (MAPK) signaling pathway. The miRNA–mRNA integrated assay was used to analyze the pathways regulated by miRNAs during the osteogenic differentiation of BMSCs activated by topological cues. Eighteen miRNAs were more than twofold significantly downregulated in BMSCs in the flat (OS+) group versus the flat group, and thirty‐six miRNAs were significantly downregulated more than twofold in the random group versus the flat group. We next compared the predicted 6843 target genes of the 18 downregulated miRNAs in the flat (OS+) group with the 1378 upregulated (≥1. 5‐fold) genes identified by microarray analysis in our previous study. 8 We found that 650 upregulated genes ( Figure 5 a) overlapped. We also compared the predicted 9243 target genes of the 36 downregulated miRNAs in the random group with the 451 upregulated (≥1. 5‐fold) genes identified by microarray analysis in our previous study. 8 We found that 257 upregulated genes overlapped (Figure 5 b). To explore the intracellular signaling pathways to which these genes belong, we performed an ontology analysis using KOBAS 3. 0 software ( http://kobas. cbi. pku. edu. cn/anno_iden. php ) (Figure 5 c, d). A pathway enrichment analysis revealed substantial differences between the two groups. Comparison of the top 10 pathways between the two groups (Tables S1 and S2, Supporting Information) revealed that 6 overlapped (Figure S3a, Supporting Information). Among the top 20 pathways in the two groups (Tables S1 and S2, Supporting Information), only 9 overlapped (Figure S3b, Supporting Information) and 14 were found to be consistently enriched (Figure S3c, Supporting Information). Overall, there was only 50% similarity between the two groups, indicating marked differences between the physical and chemical activation of osteogenesis. A gene enrichment analysis showed that the MAPK pathway was the top pathway in the random group but not in the flat (OS+) group. Hence, MAPK signaling is likely involved in activation of the osteogenic differentiation of BMSCs by topographical cues. As stated above, miR‐193a‐3p was involved in topological cue‐activated osteogenic differentiation of BMSCs, and we hypothesized that activation of the MAPK signaling pathway and miR‐193a‐3p expression are closely correlated. Interestingly, some key factors in the MAPK pathway, such as those encoded by Erk1 and Jnk, are predicted by bioinformatics to be target genes of miR‐193a‐3p. We found that antagomir‐193a‐3p significantly increased the mRNA levels of Erk1 and Jnk, while agomir‐193a‐3p significantly decreased the mRNA levels of Erk1 and Erk5 (Figure 5 e–g). The protein expression levels of ERK1, JNK, p‐ERK1, and p‐ERK5 were increased by antagomir, while the expression levels of ERK1, JNK, and p‐ERK1 were decreased by agomir‐193a‐3p (Figure 5 h). These results verified that miR‐193a‐3p suppresses the MAPK pathway. Figure 5 PLLA nanofibrous substrates activated the mitogen‐activated protein kinase (MAPK) signaling pathway by downregulating miR‐193a‐3p. a, b) miRNA–mRNA integrated analysis of the flat (OS+) (a) random groups (b). c) Significantly enriched pathways for the 650 putatively upregulated genes in (a) ( p < 0. 001). d) Significantly enriched pathways for the 257 putatively upregulated genes in (b) ( p < 0. 05). e–g) mRNA and h) protein levels of MAPK kinases in bone marrow stem cells (BMSCs) treated with agomir, antagomir, or their scrambled controls. Results are means ± SEM ( n = 3). Samples were subjected to one‐way ANOVA with Tukey's post hoc test (* p < 0. 05, ** p < 0. 01). miR‐193a‐3p directly targeted MAP3K3. To gain insight into the mechanisms by which miR‐193a‐3p regulates the MAPK pathway activity and affects the osteogenic differentiation of BMSCs, we used miRWalk 2. 0 ( http://zmf. umm. uni-heidelberg. de/apps/zmf/mirwalk2/index. html ) to predict its potential targets. Among the candidate target genes, Map3k3 has a miR‐193a‐3p–binding site at its 3′ UTR. Next, we verified that the Map3k3 mRNA level in hBMSCs was upregulated in the random group ( Figure 6 a). We found that the Map3k3 mRNA level (assessed by qPCR) was downregulated by agomir‐193a‐3p but upregulated by antagomir‐193a‐3p, as compared to the corresponding scrambled controls (Figure 6 b). Also, agomir decreased, and antagomir increased, the MAP3K3 protein level (Figure 6 c). Figure 6 miR‐193a‐3p directly targeted MAP3K3. a) Map3k3 mRNA expression levels of hBMSCs in the flat and random groups. b) mRNA and c) protein levels of MAP3K3 in hBMSCs treated with agomir‐193a‐3p, antagomir‐193a‐3p, and their corresponding scrambled controls. d) Schematic diagram of the design of luciferase reporters with the wild‐type Map3k3 3′UTR (WT MAP3K3 3′UTR) or the site‐directed mutant Map3k3 3′UTR (MUT MAP3K3 3′UTR). e) The effects of agomir‐193a‐3p, antagomir‐193a‐3p, or their corresponding scrambled controls on the luciferase activity of WT MAP3K3 3′UTR and MUT MAP3K3 3′UTR reporter in hBMSCs. f) Transfection efficiency of Map3k3‐small‐hairpin RNA (shRNA). g–i) mRNA levels of osteogenic genes in hBMSCs treated with MAP3K3‐shRNA. j) Transfection efficiency of the Map3k3‐overexpression vector. k–m) mRNA levels of osteogenic genes in MAP3K3 overexpressing hBMSCs. n). Protein levels of osteogenic factors in hBMSCs treated with MAP3K3‐shRNA. o) Protein levels of osteogenic factors in MAP3K3 overexpressing hBMSCs. p) Alizarin red staining of MAP3K3 overexpressing hBMSCs after 21 days of culture. Data are means ± SD ( n = 3). * p < 0. 05, ** p < 0. 01, *** p < 0. 001. To further confirm the miR‐193a‐3p target region in the Map3k3 mRNA, we constructed MAP3K3 3′UTR luciferase reporters that contained wild‐type (WT MAP3K3 3′UTR reporter) and mutant (MUT MAP3K3 3′UTR reporter) sequences of the miR‐193a‐3p binding sites (Figure 6 d), and co‐transfected hBMSCs with these together with miR‐193a‐3p oligonucleotides. We found that agomir‐193a‐3p decreased, while antagomir‐193a‐3p increased, the luciferase reporter activity of WT MAP3K3 3′UTR, but not that of MUT MAP3K3 3′UTR (Figure 6 e). Therefore, miR‐193a‐3p directly targets Map3k3 and binds to its 3′UTR. Next, we knocked down Map3k3 in hBMSCs using Map3k3‐targeting small‐hairpin RNAs (shRNAs) to disrupt the expression of the Map3k3 gene (Figure 6 f), and used a lentivirus to overexpress the Map3k3 gene (Figure 6 j). We found that the mRNA levels of Ocn and Bsp were significantly decreased by Map3k3 knockdown (Figure 6 g, h) and that the expression levels of these genes were markedly increased by Map3k3 overexpression (Figure 6 k, l). But the mRNA levels of Runx2 remained unchanged with either Map3k3 knockdown (Figure 6 i) or overexpression (Figure 6 m). The protein levels of OCN and RUNX2 were significantly decreased by MAP3K3 knockdown but markedly increased by its overexpression (Figure 6 n, o). Increased Alizarin red staining in the MAP3K3 overexpression group was observed after 21 days in the absence of chemical osteogenic medium (Figure 6 p). Therefore, miR‐193a‐3p regulated topographical feature‐activated osteogenic differentiation of hBMSCs in the random group by directly binding to Map3k3. miR‐193a‐3p activated the MAPK signaling pathway through the miR‐193a‐3p‐MAP3K3 axis. We have shown that miR‐193a‐3p was downregulated in the random group (Figure 2 ), while the miRNA–mRNA integrated analysis indicated that the MAPK signaling pathway was enriched in the random group and that miR‐193a‐3p expression was negatively correlated with activation of the MAPK signaling pathway (Figure 5 ). Map3K3 is a target gene of miR‐193a‐3p (Figure 6 ). Next, we investigated the effect of loss‐ and gain‐of‐function of MAP3K3 on the MAPK pathway. We used a lentivirus to knock down and overexpress MAP3K3 in hBMSCs. The qPCR results showed that the mRNA level of only Erk5 was significantly decreased after Map3K3 knockdown ( Figure 7 a). Also, the mRNA level of only Erk1 was significantly increased by Map3K3 overexpression (Figure 7 b). The results of Western blotting showed that the JNK, ERK5, p‐ERK1, p‐JNK, and p‐ERK5 levels were markedly decreased by MAP3K3 knockdown (Figure 7 c). Conversely, the levels of JNK, ERK5, p‐ERK1, p‐JNK, and p‐ERK5 were significantly increased by MAP3K3 overexpression (Figure 7 d). The protein level of ERK1 was unaffected by knock down or overexpression of MAP3K3 in hBMSCs (Figure 7 c, d). These results suggest that miR‐193a‐3p negatively regulates, and MAP3K3 positively regulates, the MAPK pathway, and that both regulate the activation (or phosphorylation) of ERK1, JNK, and ERK5, which are key signaling molecules in the MAPK pathway. Next, we co‐transfected MAP3K3‐shRNA lentivirus and agomir‐193a‐3p or antagomir‐193a‐3p into hBMSCs to verify the relationships of miR‐193a‐3p, Map3k3, and Ocn. The qPCR results showed that antagomir‐193a‐3p upregulated the Map3k3 and Ocn mRNA levels, and that MAP3K3‐shRNA interfered with this effect (Figure S4, Supporting Information). Therefore, miR‐193a‐3p activates the MAPK pathway through the miR‐193a‐3p‐MAP3K3 signaling axis. Figure 7 MAP3K3 activates the MAPK signaling pathway. a) mRNA levels of MAPK kinases upon MAP3K3 downregulation by lentivirus‐shRNA. b) mRNA levels of MAPK kinases upon MAP3K3 overexpression by lentivirus‐MAP3K3. c) Protein levels of MAPK kinases upon MAP3K3 downregulation by lentivirus‐shRNA. d) Protein levels of MAPK kinases upon MAP3K3 overexpression by lentivirus‐MAP3K3. Data are means ± SD ( n = 3). All p ‐values by Student's t ‐test. * p < 0. 05. The miR‐193a‐3p‐MAP3K3 axis modulates the osteogenic differentiation of hBMSCs in response to substrate rigidity. Substrate stiffness is a microenvironmental biophysical cue that determines the fate of BMSCs and has been the focus of recent research. 9 In this study, Gel‐MA hydrogel 10 was utilized to investigate the role of miR‐193a‐3p in substrate stiffness‐induced osteogenic differentiation of hBMSCs. We found that the numbers of pseudopod protrusions and the activation of cytoskeletal components of hBMSCs increased as matrix stiffness increased. hBMSCs displayed a globular‐like phenotype on the 3% (w/v) Gel‐MA hydrogel with a poorly developed cytoskeleton, while hBMSCs cultured on the 20% (w/v) Gel‐MA hydrogel were polygonal and had an osteoblast‐like morphology with a well‐developed cytoskeleton ( Figure 8 a–d). Also, the spread of the hBMSCs increased with increasing matrix stiffness (Figure 8 e, f). The mRNA levels of the osteogenic marker genes Runx2 and Ocn also increased with increasing matrix stiffness (Figure 8 g, h). These phenotypes indicated that biophysical cues, such as matrix stiffness, determine the fate of BMSCs by altering their adhesion morphology. Next, we investigated the role of miR‐193a‐3p and its target gene Map3k3 in the substrate stiffness‐induced osteogenic differentiation of hBMSCs. We found that miR‐193a‐3p was downregulated, but its target gene Map3k3 was upregulated, in the groups with higher Gel‐MA stiffness (Figure 8 i, j). Therefore, the miR‐193a‐3p‐MAP3K3 signaling axis may be involved in the substrate stiffness‐induced osteogenic differentiation of BMSCs. Figure 8 miR‐193a‐3p mediated hBMSC osteogenic differentiation in response to substrate stiffness on gelatin methacrylate (Gel‐MA). a–d) Confocal immunofluorescence of cytoskeletal actin and nuclear DNA of hBMSCs cultured on Gel‐MA with different substrate stiffnesses at 24 h postseeding (scale bar, 50 µm). Actin, red; nuclear DNA, blue. e) The stiffness of the Gel‐MA hydrogel could be modified by altering its concentration; the higher the gel concentration, the higher the Young's modulus. f) Cell spreading areas on substrates with different stiffnesses ( n = 10). g–j) mRNA levels of Runx2, Ocn, miR‐193a‐3p, and Map3k3 in cells cultured on substrates with different stiffnesses ( n = 3). Results are means ± SEM. Samples were subjected to one‐way ANOVA with Tukey's post hoc test. Significant differences are denoted by * p < 0. 05. 3 Discussion Biomaterials can be tailored to provide a microenvironment that directs the differentiation of stem cells into the desired lineages. 1, 11 The physical properties of biomaterials can also modulate stem cell differentiation and lineage fate by means of cellular mechanosensing and/or other mechanisms. 12 The data presented in this study revealed an miRNA‐related mechanism for topographical feature‐induced osteogenic differentiation of BMSCs. We found that downregulation of miR‐193a‐3p expression was specifically involved in substrate topography‐induced osteogenic differentiation of hBMSCs. miR‐193a‐3p can directly bind to Map3k3 mRNA, which results in downregulation of Map3k3 expression. On substrates with topographical features, miR‐193a‐3p expression was downregulated, leading to upregulation of Map3k3 and subsequent activation of the MAPK signaling pathway, which promoted the osteogenic differentiation of hBMSCs. We further found that the miR‐193a‐3p‐MAP3K3 signaling axis plays a role in modulating substrate‐rigidity‐induced osteogenic differentiation. Hence, the miR‐193a‐3p‐MAP3K3 signaling axis is needed for transduction of biophysical stimuli from the substrate to induce the osteogenic differentiation of hBMSCs. Osteogenic differentiation trigged by biophysical and chemical cues are different processes. During biophysical cue‐stimulated osteogenic differentiation, a change in focal adhesion behavior and cytoskeleton re‐arrangement occur before any alteration of gene expression, 13, 14 while in chemical cue‐induced osteogenic differentiation, gene expression changes before cytoskeleton re‐arrangement and a shift in cell morphology. 15 The different miRNA expression profiles of the random and flat (OS+) groups further suggest that biophysical and chemical induction of osteogenic differentiation are mechanistically different (Figure 2 a, b). We found that some miRNAs are sensitive to biophysical, but insensitive to chemical osteogenic induction (Figure 2 c, d). Also, the topographical features in the random group downregulated the pri‐ and pre‐miRNA levels, suggesting that the different miRNA expression profile in the random group is due to the effect of topographical features on the biogenesis of miRNA (Figure S5a, b, Supporting Information). We found that the miR‐193a‐3p‐MAP3K3‐MAPK axis mediated topographical feature‐induced osteogenic differentiation. This axis also plays a role in the osteogenesis induced by other biophysical cues, as similar results were obtained using a second biophysical material (Figure 8 ). BMP/Smad, 16, 17 Wnt, 18 and TGF‐β 16, 19 are the classical signaling pathways involved in chemical‐induced osteogenic differentiation, and drive the osteogenic differentiation of stem cells. 16, 17 The MAPK signaling pathway is activated by various extracellular stimuli, including radiation, osmotic pressure, temperature, and mechanical force, and regulates cell proliferation, differentiation, and survival. 20 It is not a typical signaling pathway driving the osteogenic differentiation of stem cells, in agreement with the finding that biophysical cues induce only moderate osteogenic differentiation. 21 In this study, we found that the role of miR‐193a‐3p in chemical‐induced osteogenic differentiation is likely obscured (Figure S6, Supporting Information). Thus, the results of this study deepen our understanding of how biophysical cues determine the lineage into which stem cells differentiate. In this study, the downregulation of miR‐193a‐3p enhanced the healing of the critical‐sized bone defects (Figure 4 ). This data showed that miR‐193a‐3p could be used as a molecular switch to regulate mechanosensing of biophysical cues in vivo. Treatment with miR‐193a‐3p antagomir mimicked the mechanosensing of biophysical cues such as topographical features and matrix stiffness. As a result, the healing of critical‐sized bone defect was enhanced. This phenomenon also suggested that when designing implantable biomedical devices or orthopedic substitutes, more attention should be focused on their biophysical properties, including their topographical features and elastic modulus. In summary, the miRNA expression profiles differed between biophysical‐ and chemical‐induced differentiation. Also, some miRNAs were sensitive to biophysical, but insensitive to chemical, induction of osteogenesis. We found that the miR‐193a‐3p‐MAP3K3‐MAPK axis mediated topographical feature‐induced osteogenic differentiation, and that this signaling axis is common to osteogenesis induced by other biophysical cues. The results of this study deepen our understanding of how biophysical cues determine the lineage into which stem cells will differentiate. Conflict of Interest The authors declare no conflict of interest. Author Contributions Y. L. and Y. H. contributed equally to this work. Y. L. , Y. H. , X. Z. , Y. W. , and X. D. conceived the idea for the study. Y. L. , Y. H. , M. X. , C. Y. , and C. C. collected the data. Y. L. , Y. H. , Z. H. , X. C. , M. G. , and W. L. discussed, and contributed to, the final design of the study. Y. L. and Y. H. wrote the first draft of the manuscript with significant assistance from B. C. H. and X. D. All the authors contributed to the completion and revision of the manuscript. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201901511
| 2,019
|
Advanced Science
|
2D MXene‐Integrated 3D‐Printing Scaffolds for Augmented Osteosarcoma Phototherapy and Accelerated Tissue Reconstruction
|
Abstract The residual of malignant tumor cells and lack of bone‐tissue integration are the two critical concerns of bone‐tumor recurrence and surgical failure. In this work, the rational integration of 2D Ti 3 C 2 MXene is reported with 3D‐printing bioactive glass (BG) scaffolds for achieving concurrent bone‐tumor killing by photonic hyperthermia and bone‐tissue regeneration by bioactive scaffolds. The designed composite scaffolds take the unique feature of high photothermal conversion of integrated 2D Ti 3 C 2 MXene for inducing bone‐tumor ablation by near infrared‐triggered photothermal hyperthermia, which has achieved the complete tumor eradication on in vivo bone‐tumor xenografts. Importantly, the rational integration of 2D Ti 3 C 2 MXene is demonstrated to efficiently accelerate the in vivo growth of newborn bone tissue of the composite BG scaffolds. The dual functionality of bone‐tumor killing and bone‐tissue regeneration makes these Ti 3 C 2 MXene‐integrated composite scaffolds highly promising for the treatment of bone tumors, which also substantially broadens the biomedical applications of 2D MXenes in tissue engineering, especially on the treatment of bone tumors.
|
1 Introduction Bone cancer is a general term for malignant bone tumors such as osteosarcoma, chondrosarcoma, and fibrosarcoma. 1 It is usually divided into autologous skeletal system cancer and bone‐metastases cancer (primary tumors in breast, lung, and kidney). 2 At present, the treatment of bone cancer usually combines destructive surgery (amputation or comprehensive limb salvage surgery) with multidrug chemotherapy, which has significantly improved the survival rate of patients. 3 Unfortunately, the invasiveness of cancer as well as anatomical complexity determines an unattainable radical resection followed by inevitable local recurrence. Moreover, massive bone defects caused by surgery have surpassed the self‐healing ability of bone tissue, bringing long‐term pain to patients and even causing the failure of surgery. 4 Consequently, it is highly urgent and necessary to construct multifunctional tissue‐engineering biomaterials with simultaneous bone‐tumor killing and bone‐tissue remodeling capacity. Recently, ultrathin MXene nanosheets, as a new class of early transition metal carbides/nitrides/carbonitrides, have significantly enriched the 2D material families, 5 which are featured with unique structural characteristics including large specific surface area and adjustable physiochemical property such as excellent electroconductibility. In 2D MXene, “M” denotes transition metal atoms, “X” means carbon or nitrogen, and “ene” suffix originating from “graphene” represents the materials with ultrathin 2D structure. 6 For 2D MXenes' peculiarities, they have been broadly explored in versatile applications such as energy storage, 7 catalysis, 8 electromagnetic shielding, 9 water purification, 10 etc. The fast development of theranostic nanomedicine has promoted the extensive biomedical applications of these 2D MXenes in biosensing, 11 intracellular fluorescent imaging, 12 antibacterial, 13 and photothermal therapy (PTT). 14 Their fascinating biomedical performances promote the further extensive exploring of the unique and specific applications in versatile biomedical fields such as tissue engineering, which has not been achieved so far. 2D Ti 3 C 2 MXenes possess high biocompatibility and desirable photothermal‐conversion efficiency in near‐infrared (NIR) biowindow. 15 Especially, by the interaction of water and oxygen, they would degrade to release Ti‐based species, which is expected to promote the growth of new bones. 16 Therefore, it is highly feasible to utilize the photothermal‐conversion property of 2D Ti 3 C 2 MXene nanosheets (NSs) for ablating bone‐tumor cells, and then employ their biodegradable performance and biodegradation products for accelerating the bone reconstruction. In general, the regrowth and regeneration of large bone defects still require some biomaterials to bridge the tissue gap and afford structural support to sustain the physiological activities and cellular behaviors during the new bone formation, such as nutrient transport, cell adhesion, proliferation, migration, differentiation, and maturation. 17 Bioactive glass (BG) is a typical biomaterial for bone‐tissue regeneration, 18 which has been demonstrated to be featured with high biocompatibility, osteoconductivity, osteoinductivity, and degradability. 19 Therefore, the BG scaffolds (designated as BGSs) with 3D interconnected macropores, precisely controlled appearance and internal structures, as fabricated by the intriguing 3D‐printing technique, are generally regarded as the desirable candidate bridge biomaterial for hard‐tissue regeneration. 20 In this work, we report, for the first time, on the rational integration of 2D Ti 3 C 2 MXenes with 3D‐printing BG scaffold (designated as Ti 3 C 2 ‐BG scaffold or TBGS) for the construction of multifunctional biomaterial scaffold for bone‐cancer treatment with simultaneous bone‐tumor killing and bone‐tissue regeneration functionalities ( Scheme 1 ). On one hand, the integrated 2D Ti 3 C 2 MXene NSs kill the bone cancer cells based on their specific photothermal‐conversion property. On the other hand, the implanted 3D BG component assists the differentiation of human bone marrow mesenchymal stem cells (hBMSCs) into osteoblasts by its bridging functionality. 21 In the process of bone‐tumor treatment and bone‐tissue reconstruction, the titanium‐based species, as the biodegradation product from Ti 3 C 2 MXene NSs, accelerate the formation of new bone. 16, 22 During these activities, the BG component also gradually degrades to provide the necessary minerals and space for the newly formed bone tissue. Therefore, this rationally designed multifunctional 3D composite scaffold represents the novel therapeutic biomaterial for bone‐tumor therapy with concurrent cancer cell‐killing and tissue‐engineering performances. Scheme 1 Schematic illustration of the fabrication of TBGS, ablation of bone cancer, and regeneration of bone tissue. I) Fabrication procedure of TBGS, including 3D printing of pure BGS, integration of Ti 3 C 2 MXene, and degradation of Ti 3 C 2 MXene on BGS. II, III) TBGS used for osteosarcoma cell elimination by photothermal ablation both in vitro (II) and in vivo (III). IV) Bone‐tissue reconstruction and the therapeutic results after the implantation of BGS and TBGS. 2 Results and Discussion 2. 1 Synthesis and Characterization of 2D Ti 3 C 2 MXenes 2D Ti 3 C 2 MXenes, as a new photothermal nanoagent with excellent photothermal‐conversion property and high biocompatibility, were integrated with 3D‐printed BG scaffolds for killing bone cancer cells and regenerating massive bone defects. These 2D ultrathin Ti 3 C 2 NSs were fabricated by hydrofluoric acid (HF) etching and subsequent tetrapropylammonium hydroxide (TPAOH) exfoliation of the original bulk MAX‐phase Ti 3 AlC 2 ceramics ( Figure 1 a). 23, 42 The as‐prepared 2D Ti 3 C 2 NSs could be well dispersed in aqueous solution as demonstrated by the obvious Tyndall effect (Figure 1 b), enabling the further facile integration with 3D‐printing scaffolds. Figure 1 Synthesis and characterization of 2D Ti 3 C 2 MXene NSs. a) Schematic illustration of the fabrication process of 2D Ti 3 C 2 NSs, including HF etching and TPAOH intercalation of the original bulk Ti 3 AlC 2 ceramic. b) Digital photographs of Ti 3 C 2 NSs dispersed in aqueous solution. c) SEM image of bulk Ti 3 AlC 2 ceramic (scale bar: 5 µm). d) Magnified SEM image of the selected area in (c) (scale bar: 200 nm). e) SEM image of multilayer Ti 3 C 2 MXene after HF treatment (scale bar: 2 µm). f) TEM image of Ti 3 C 2 NSs (scale bar: 200 nm). g) HRTEM image of Ti 3 C 2 NSs (scale bar: 3 nm). h) SAED pattern of single‐layer Ti 3 C 2 NSs (scale bar, 5 1/nm). The bulk Ti 3 AlC 2 ceramic (MAX phase, Figure 1 c, d) was initially etched by HF aqueous solution for 3 days to selectively remove the middle Al layer, which could fabricate multilayer Ti 3 C 2 MXene with accordion‐like microstructure (Figure 1 e). 9 The elemental mapping of Ti 3 AlC 2 ceramic reveal the co‐existence of Ti, Al, and C elements (Figure S1, Supporting Information), and the Al content of multilayer Ti 3 C 2 was substantially decreased (Figure S2, Supporting Information) after HF etching. Subsequently, the etched Ti 3 C 2 powder was intercalated with TPAOH solution for another 3 days to fabricate delaminated 2D ultrathin Ti 3 C 2 NSs (Figure 1 f). Both high‐resolution transmission electron microscopy (HRTEM, Figure 1 g) image and the corresponding selected area electron diffraction pattern (SAED, Figure 1 h) exhibit that the prepared 2D Ti 3 C 2 NSs were featured with planar topology and hexagonal crystallize structure with well crystallinity, demonstrating the successful fabrication of 2D ultrathin 2D Ti 3 C 2 MXene NSs. 2. 2 Design, Fabrication, and Characterization of BG/Ti 3 C 2 ‐BG Scaffolds Figure 2 a schematically depicts the formation of pure BGS by 3D‐printing technology. 24 In this work, we applied a facile and efficient strategy, i. e. , the direct solution‐soaking method, to prepare TBGS. 25 To achieve a suitably modified amount for photothermal ablation, BGSs were integrated with Ti 3 C 2 NSs at elevated initial concentrations (1. 0, 1. 5, and 2. 0 mg mL −1 ). BGS integrated with 1. 0 mg mL −1 Ti 3 C 2 NSs was termed as 1. 0 TBGS, and other TBGSs were renamed by this analogy. Digital photographs (Figure 2 b‐a1–d1) reveal that the scaffolds fabricated by the 3D‐printing technique were featured with well‐designed microstructure, and the printed pure BGSs and TBGSs exhibited white color and black color, respectively. Scanning electron microscope (SEM) images (Figure 2 b‐b2–d2) show that the Ti 3 C 2 NSs modified the whole surface of BGSs and the corresponding structure was not compromised as compared to BGS (Figure 2 b‐a2). Pure BGSs showed rough and loosened surfaces (Figure 2 b‐a3, a4) while TBGSs had relatively smooth surfaces after the adsorption of nanosheets (Figure 2 b‐b3–d3, b4–d4), which might be attributed to the planar structure of MXene covered onto the surface of BGSs. Figure 2 Fabrication and characterization of BGS/TBGS. a) Schematic illustration of the formation of pure BGS and TBGS, including 3D printing and further Ti 3 C 2 MXene integration. b) Digital photographs and SEM images of pure BGS and TBGS: digital photographs of: a1) pure BGS, b1) 1. 0 TBGS, c1) 1. 5 TBGS, and d1) 2. 0 TBGS; SEM images of: a2–a4) pure BGS, b2–b4) 1. 0 TBGS, c2–c4) 1. 5 TBGS, and d2–d4) 2. 0 TBGS. From top to bottom, scale bars are 3 mm, 500 µm, 5 µm, and 1 µm, respectively. Each row of images shares the same scale bar. c, d) The fracture morphologies of 1. 0 TBGS. Scale bar is 100 and 1 µm, respectively. e–g) Element mappings of 1. 0 TBGS (all scale bars: 1 µm; e: merged image, f: Ti element, and g: Si element). h) SEM image and EDS (inset image) of 1. 0 TBGS (scale bar: 3 µm). i) XRD patterns of Ti 3 C 2 NSs, pure BGS, and TBGS. j) Ti 2p XPS spectra of TBGS. Furthermore, the detailed fracture morphologies of 1. 0 TBGS showed an obvious core–shell structure consisting of an ≈577 nm thick Ti 3 C 2 shell and a BG core (Figure 2 c, d). The element‐mapping analysis demonstrated the desirable element distribution, which exhibited that the Si, Ca, and P signals tended to increase from the surface to the interior, whereas the Ti and C signals tended to decrease (Figure 2 e–g; Figure S3, Supporting Information). In addition, this tendency was also confirmed by the energy‐dispersive spectrometer (EDS) analysis (Figure 2 h; Figure S4, Supporting Information), which provided the solid proof of coating a Ti 3 C 2 MXene layer onto the surface microstructure of 3D BGS. X‐ray diffraction (XRD) patterns of freeze‐dried Ti 3 C 2 powder, BGS, and TBGS powders (Figure 2 i) revealed the successful fabrication of Ti 3 C 2 NSs and the effective integration of Ti 3 C 2 NSs with BGSs. The distinct (002) peak in Ti 3 C 2 powder could be ascribed to the 2D Ti 3 C 2 MXene. 26 The peak at 2θ ≈ 26. 65° of BGS particles was indexed to the (011) planes of hexagonal SiO 2 phase as the uppermost composition of BG, which was in accordance with the standard card of PDF#78‐1253. Both the (002) peak of Ti 3 C 2 MXene and the (011) peak of BGS could be found in the diffraction peaks of TBGS, demonstrating the successful integration of Ti 3 C 2 MXene and BGS. In addition, Raman spectra of Ti 3 AlC 2, Ti 3 C 2 NSs, BGS, and TBGSs powder are shown in Figure S5 in the Supporting Information. All of these consequences provided solid evidences that Ti 3 C 2 NSs have been successfully adhered to BGSs from different aspects. The surface status and chemical composition of BGSs and TBGSs were detected by X‐ray photoelectron spectroscopy (XPS). The characteristic peaks of BGSs were assigned to Si (Si 2p 103. 3 eV), Ca (Ca 2p 347. 5 eV), P (P 2p 133. 5 eV), and O (O 1s 532. 4 eV). The peak at the binding energy of 458. 3 eV was assigned to Ti 2p of TBGSs, which indicates the existence of Ti 3 C 2 NSs on the surface of modified BGSs (Figure S6, Supporting Information). Furthermore, the Ti 2p peak of TBGSs can be deconvoluted into six subpeaks (Figure 2 j) at 453. 9, 454. 9, 456. 9, 458. 3, 460. 2, and 464. 2 eV corresponding to 2p 3/2 (Ti), 2p 3/2 (Ti–C), 2p 3/2 (Ti 2 O 3 ), 2p 3/2 (TiO 2 ), 2p 1/2 (TiO), and 2p 1/2 (TiO 2 ), respectively. In addition to the inherent Ti—C bond of Ti 3 C 2 NSs, XPS spectra demonstrated the presence of Ti and Ti x O y, revealing the Ti 3 C 2 NSs were partially oxidized during the modification. 2. 3 In Vitro Photothermal Performance, Cytotoxicity Assay, and Cell Ablation of Ti 3 C 2 ‐BG Scaffolds The essential characteristics of photothermal nanoagents for photonic tumor hyperthermia are the efficient optical absorption and high photothermal‐conversion efficiency in NIR biowindow. 27 As illustrated in Figure S7a in the Supporting Information, the optical absorbance spectrum of Ti 3 C 2 NSs in water presented a pronounced absorption in the range of 750–850 nm. 28 Subsequently, we systematically assessed the effect of Ti 3 C 2 concentrations, power density of laser irradiation, and the environment (dry and wet) on photothermal properties of the composite scaffolds after exposure to NIR laser irradiation. Pure BGS and TBGSs at varied initial integrating concentrations (1. 0, 1. 5, and 2. 0 mg mL −1 ) were exposed to an 808 nm laser irradiation for 10 min at a power density of 1. 0 W cm −2 in the air (Figure S7b, Supporting Information). From 1. 0 TBGS to 2. 0 TBGS, the final equilibrium temperature increased from 55 to 65 °C within 10 min, while the temperature of pure BGS did not increase significantly. It indicated that the final equilibrium temperature and the heating rate of scaffolds were positively correlated with the integrated Ti 3 C 2 amount. Furthermore, with the elevation of power density of irradiation laser from 0. 5 to 1. 0 W cm −2 (808 nm, 10 min), the equilibrium temperature of 1. 0 TBGS increased from 40 to 65 °C in the air ( Figure 3 a), and from 42 to 58 °C in phosphate buffer solution (PBS, Figure 3 b), exhibiting that the photothermal effect of the composite TBGS was dependent on the power density of laser irradiation. The elevated temperature in PBS was slightly lower than the temperature in air because of the endothermic property of the liquid. Subsequently, the BGS and 1. 0 TBGS were irradiated by 808 nm laser for 10 min at a power density of 1. 0 W cm −2 (Figure S7c, Supporting Information). The temperature of TBGS increased by ≈20 °C in 10 min, but the pure BGS showed no obvious temperature variation. Figure 3 In vitro photothermal performance, cytotoxicity assay, and cell ablation of BGS/TBGSs. a) Photothermal‐heating curves of 1. 0 TBGS under the irradiation with 808 nm laser at varied power densities (0. 5, 0. 75, and 1. 0 W cm −2 ) in dry environment. b) Photothermal‐heating curves of 1. 0 TBGS under the irradiation with 808 nm laser at varied power densities (0. 5, 0. 75, and 1. 0 W cm −2 ) in wet environment. c) Heating curves of a TBGS for five laser on–off cycles (1. 0 W cm −2 ) under irradiation with 808 nm laser. d) Schematic illustration for cancer‐cell ablation by Ti 3 C 2 NSs possibly shedding from TBGS. e) Relative cell viability of bone‐tumor cells (Saos‐2) after the treatment with different conditions as described in the figure. The “200 µL” and “Dry” groups respectively mean TBGS +NIR group in 200 µL DMEM and dry environment (discarding all DMEM). The other groups were performed in 400 µL DMEM. The cell viabilities in “TBGS + NIR, ” “200 µL, ” and “Dry” groups were about 36%, 21%, and 13%, respectively. This result demonstrated that the increase of fluid could reduce the tumor‐killing effect of PTT. n = 4. f) Relative cell viability of Saos‐2 cells treated with 1. 0 TBGSs for different irradiation durations. n = 4. g) CLSM images of live (green) and dead Saos‐2 cells (red) on BGSs or TBGSs with different treatments as indicated in the figure. Scale bar is 200 µm. * p < 0. 05, ** p < 0. 01, *** p < 0. 001. To further investigate the photothermal stability of these MXene‐modified composite scaffolds, 1. 0 TBGS was irradiated by 808 nm laser for ≈3 min (laser on) and then naturally cooled down to room temperature (laser off). After five “on–off” cycles of laser, the laser‐induced temperature increase showed no obvious deterioration (Figure 3 c), indicating the high photothermal stability of these MXene‐integrated composite TBGSs for the potential continuous photothermal hyperthermia of bone tumors. It has been demonstrated that the temperature around 45 °C could induce the tumor‐cell death. 29 Based on the aforementioned photothermal evaluation results, 1. 0 TBGS was chosen as the implanting composite scaffold for photothermal bone‐tumor hyperthermia. The in vitro cytotoxicity assay and cancer‐cell ablation of scaffolds were quantitatively evaluated by a standard Cell Counting Kit‐8 (CCK‐8) assay. Saos‐2 cells (osteosarcoma cells) were incubated with pure BGSs and TBGSs, which were further irradiated by 808 nm laser for triggering photothermal ablation (Figure 3 d). As shown in Figure 3 e, compared to the control group (blank), the percentages of viable cells in the BGS, BGS + laser, laser only, and TBGS groups ranged from about 90–100% with no significant difference, which manifested high biocompatibility of the fabricated TBGSs. Comparatively, less than 40% of Saos‐2 cells survived in the TBGS + laser group, revealing the capability of TBGS for efficiently killing cancer cells by photothermal ablation. Especially and importantly, with the prolonging of the laser‐irradiation duration (Figure 3 f), times, and power density (Figure S7e, f, Supporting Information), there were distinctly much fewer living cells survived in the TBGS + laser group. For instance, the percentages of viable cells after laser irradiation for 15 min (1. 0 W cm −2, one time), irradiation for three times (1. 0 W cm −2, 10 min), and irradiation at 1. 0 W cm −2 (10 min, one time) were about 25%, 25%, and 38%, respectively, demonstrating the high controllability of TBGS‐assisted photothermal ablation of cancer cells. In addition, the Saos‐2 cell apoptosis after photothermal ablation was further intuitively confirmed by confocal laser scanning microscopy (CLSM) observations (Figure 3 g). After laser irradiation (10 min, 1. 0 W cm −2 ), the dead and live cells in scaffolds were specifically stained by propidium iodide (PI; red) and calcein‐AM (green), respectively. It is found that the experimental group (TBGS + laser) and control groups (BGS, BGS + laser, TBGS) exhibited a sharp contrast in fluorescence color where the experimental group showed the significant red fluorescence, indicating the effective cell apoptosis as induced by photothermal ablation. Furthermore, the number of apoptotic cells (Q1: dead cells + Q2: late apoptotic cells + Q4: early apoptotic cells) of flow cytometric analysis (Figure S8, Supporting Information) in group BGS, TBGS, BGS + NIR, and TBGS + NIR were determined to be 8%, 7%, 9. 3%, and 43%, respectively. These results strongly demonstrated that TBGSs possessed powerful tumor‐cell ablation capacity under the irradiation of NIR laser in vitro. Especially, as could be seen from all bright‐field images (Figure 3 g), each group of scaffolds still maintained a well‐ordered hierarchical 3D geometric structure after laser irradiation and further immersion scouring. 2. 4 In Vivo Photothermal Tumor Ablation by MXene‐BG Scaffold under NIR Irradiation Encouraged by in vitro excellent photothermal performance of TBGSs, a localized in vivo photothermal tumor ablation was scheduled. However, there are some critical challenges to establish an orthotopic osteosarcoma model for this multilevel research. First and foremost, leakage is a severe complication during intrafemoral/intratibial injection of tumor cells to establish the orthotopic model. 30 Therefore, intramedullary injection might induce direct local pollution or indirect pulmonary seeding via circulation. 31 This orthotopic model is potentially a failure for the biological research of osteosarcoma. Second, the orthotopic osteosarcoma model is not suitable for the surgical removal and corresponding investigation of bone repair. Given the osteosarcoma model was successfully established in the distal femur or proximal tibia of rats, in order to simulate the scenario clinically, we would excise the tumor first, then implant the scaffold, and conduct photothermal treatment. However, a limb‐sparing strategy is impossible for the osteosarcoma‐bearing rat. Third, although the orthotopic model of osteosarcoma has several distinct advantages including the osseous microenvironment and anatomical similarity, 32 for immunological, pharmaceutical, and therapeutic studies, an ectopic model of osteosarcoma is sufficient due to its straightforward biological performance, 33 which also achieved popularity in the past 30 years. 34 In this study, we designed two independent but correlated animal models to reflect the unique property of TBGS, namely, photothermal tumor ablation in ectopic osteosarcoma‐bearing nude mice and newborn bone regeneration in rat with critical cranial defect. The data collectively supported the potential clinical value of multifunctional TBGS in the treatment of osteosarcoma. To evaluate the photothermal ability of TBGS, a localized in vivo photothermal tumor ablation was further assessed by employing female BALB/c nude mice bearing Saos‐2 bone tumor ( Figure 4 a). These mice bearing Saos‐2 xenograft (subcutaneous tumor) were randomly divided into four groups ( n = 6 for each group) for diverse treatments including BGS, BGS + NIR, 1. 0 TBGS, and 1. 0 TBGS + NIR, which was set based on the in vitro PTT experiments when the tumor volume reached around 120 mm 3. Photonic tumor hyperthermia (808 nm, 1. 0 W cm −2, 10 min) was conducted in the BGS + NIR and 1. 0 TBGS + NIR groups 1 day after implanting the scaffolds into the tumor. The corresponding IR thermal images were shown at tumor sites in groups of BGS + laser and TBGS + laser (Figure 4 b). As clearly shown in Figure 4 c, the surface temperature of tumors implanted with 1. 0 TBGSs was rapidly elevated to the equilibrium temperature of as high as 63 °C under NIR laser irradiation only within 2 min. In striking contrast, the temperature of tumors implanted with BGSs without the integrated Ti 3 C 2 MXene only showed a slight increase to about 37 °C. As shown from the corresponding tumor photographs (2 weeks after treatment; Figure 4 d), the tumors in treated groups (TBGS + NIR) were completely removed by photonic tumor hyperthermia without reoccurrence. Comparatively, the tumors in other treatment groups grew continuously without any therapeutic effect. Figure 4 In vivo photothermal‐performance evaluation of TBGSs. a) Schematic illustration of TBGSs for in vivo photothermal cancer ablation. b) The corresponding IR thermal images at tumor sites of Saos‐2 tumor‐bearing mice in groups of BGS + laser (upper) and TBGS + laser (bottom). c) Temperature elevations at tumor sites of Saos‐2 tumor‐bearing mice in groups of BGS + laser and TBGS + laser. d) Photographs of Saos‐2 tumor‐bearing mice on 14th day after different treatments, and the tumor tissues stained by H&E, TUNEL (apoptosis), and Ki‐67 (proliferation) in 1 day after different treatments (scale bars: 10 µm). e) Time‐dependent tumor‐growth curves ( n = 5, mean ± s. d. ) after different treatments. f) Time‐dependent body‐weight curves of mice after different treatments. Inset: tumor weights of mice on the 14th day after varied treatments. n = 5. g) H&E staining of major organs (heart, liver, spleen, lung, and kidney) of Saos‐2 tumor‐bearing mice on the 14th day after different treatments (scale bars: 100 µm). To further reveal the corresponding mechanism of high photothermal‐ablation efficacy, in 24 h after photothermal ablation, the necrosis of tumor tissues was qualitatively measured by hematoxylin and eosin (H&E) and terminal deoxynucleotidyl transferase‐mediated dUTP‐biotin nick end labeling (TUNEL) staining and the in vivo cellular proliferation was evaluated by Ki‐67 antibody staining (Figure 4 d; Figure S9, Supporting Information). In H&E images, TBGS + NIR group was colored less blue/purple (nuclei of cells) than the control groups, namely, the number of apoptotic osteosarcoma cells in TBGS + NIR group was larger than the control groups after photothermal ablation. According to TUNEL images, more apoptotic cells (brown colors) were detected in TBGS + NIR group than that in control groups, which presented that TBGS + NIR group possessed the best therapeutic efficacy on ablating Saos‐2 tumor cells. The Ki‐67 images also manifested that the TBGS + NIR group owned the least proliferative cancer cells (dark brown), which indicated that the proliferation of Saos‐2 cancer cells was dramatically suppressed in TBGS + NIR group among the four groups and matched with the H&E and TUNEL results. The tumor volume and mice weight of the four groups were acquired every other day. Apparently, the tumor volumes of the therapeutic group represented conspicuous suppression with the final complete eradication while the tumor volume of the control groups increased rapidly (Figure 4 e). Meanwhile, the body weight (Figure 4 f) of all groups revealed no significant difference, implying that no obvious toxicity was induced by either pure BGSs or TBGSs. NIR laser has limited tissue‐penetration depth. Therefore, in practical osteosarcoma removal surgery, the tumor lesion is exposed after removing the bone tumor and part of the surrounding tissues by surgery, and then TBGS are implanted. The photothermal therapy is conducted subsequently. After the photothermal therapy, the muscularis layer, subcutaneous tissue, and the skin will be successively closed. Therefore, after surgical resection, it can be considered that the bone tumor has changed from a deep tumor to a superficial tumor. In this way, the photothermal therapy can achieve the excellent therapeutic effect without the shield of soft tissues. Subsequently, to reveal the potential acute toxicity and long‐term toxicity of composite scaffolds, we further evaluated the histocompatibility of the composite scaffolds by H&E staining of the major organs (heart, liver, spleen, lung, and kidney) of mice on the 1st, 14th, and 28th days after photothermal ablation (Figures S10 and S11, Supporting Information; Figure 4 g). The H&E staining of these organ sections displayed that there was no obvious histomorphology and pathology change in these organs among the treatment group and control groups, indicating that the fabricated composite scaffolds with the integrated 2D Ti 3 C 2 MXene have no significant acute and chronic pathological toxicity to the major organs, i. e. , they are featured with high histocompatibility. 2. 5 Ti 3 C 2 ‐BG Composite Scaffolds for Stimulating Proliferation and Differentiation of hBMSCs In Vitro Bone mesenchymal stem cells (BMSCs) are able to differentiate into osteoblasts in a specific environment, 35 therefore investigations on the in vitro adhesion and differentiation of BMSCs on TBGS have been conducted to evaluate the effect of the material on the osteogenic potential of hBMSCs. It has been found that TBGS provided hBMSCs with favorable growth environment and space, recruiting hBMSCs to adhere to its surface. In addition, hBMSCs exhibited well‐spread morphology and extended abundant pseudopods after seeding for 1 day ( Figure 5 a; Figure S12, Supporting Information). CLSM images (Figure 5 b) revealed the proliferation of hBMSCs on BGS or TBGS, which adhered to the surface of scaffolds. As compared to the BGS group, TBGS group markedly increased the proliferation of hBMSCs at day 7. Especially, hBMSCs on TBGSs exhibited abundant filopodia while the cells on BGSs had much fewer filopodia. Furthermore, the typical CCK‐8 assay also quantitatively demonstrated that TBGSs were highly biocompatible and capable of promoting cell proliferation (Figure 5 c). In order to further demonstrate the bioactivity of Ti 3 C 2 on BMSCs without bioglass, we also performed relevant experiments (Figure S13, Supporting Information). The experimental data clearly exhibited that Ti 3 C 2 NSs with different concentrations (from 6 to 200 ppm) had no obvious toxicity to BMSCs during the evaluation period of 7 days, and the Ti 3 C 2 NSs at low concentrations (6 ppm) even promoted the proliferation of BMSCs. Figure 5 In vitro evaluation on the proliferation and osteogenic differentiation as assisted by BGS/TBGSs for bone regeneration. a) SEM image of hBMSCs after seeding on 1. 0 TBGS for 1 day (scale bar: 5 µm). b) CLSM images of hBMSCs stained with DAPI (cell nuclei, blue fluorescence) and rhodamine phalloidin (cytoskeleton, red fluorescence) on BGSs/TBGSs at days 1 and 7 (scale bar: 100 µm). c) Cell proliferation as measured by a standard CCK‐8 assay at days 1, 3, 5, and 7. n = 3. d) Alizarin red S staining of control, BGSs, and TBGSs at day 21. e) Osteogenic gene expression (COL1, RUNX2, OPN, and OCN) of hBMSCs in control, BGS, and TBGS groups on day 7. n = 3. * p < 0. 05, ** p < 0. 01, *** p < 0. 001. During the osteogenic differentiation of hBMSCs, calcium deposited, mineralized, and specifically formed red precipitates with Alizarin red S dye. 36 Extracellular matrix (ECM) mineralization of hBMSCs on the control, BGSs, and TBGSs groups were estimated by Alizarin red assay. The result revealed that the number of calcium nodules was distinctly enhanced in the TBGS group at day 21 as compared to the other two groups, indicating that TBGS improved the osteogenic capability of hBMSCs in vitro (Figure 5 d). Furthermore, to evaluate the differentiation of hBMSCs in various groups (the control, BGSs, and TBGSs groups), osteoblast‐related gene expression was analyzed, 37 including collagen type I (COL I), Runt‐related transcription factor 2 (RUNX2), osteocalcin (OCN) and osteopontin (OPN) genes. The expression of osteogenic‐specific genes in TBGSs groups was significantly enhanced at day 7 as compared to BGSs group (Figure 5 e), which demonstrated that TBGS could act as the bioactive material for promoting the osteogenic differentiation of hBMSCs in vitro. All the above results further confirmed that TBGSs distinctly improved the osteogenesis of hBMSCs in vitro potentially by some titanium‐based species originating from the biodegradation productions of integrated Ti 3 C 2 MXene, providing a promising biomaterial platform for the restoration of defective bone tissue. 2. 6 Ti 3 C 2 ‐BG Composite Scaffolds for Stimulating Osteogenic Activity In Vivo To explore the conceivable clinical application of 3D‐printed TBGSs, the in vivo efficacy of bone regeneration of TBGSs was further assessed on Sprague–Dawley rats (SD rats) with critical cranial defect. Photothermal therapy is a potent technique for cancer therapy with minimal invasiveness and high selectivity. 38 Meanwhile, previous results have demonstrated that short‐time NIR‐induced photothermal therapy did not impair the long‐term bone‐regeneration process. 25 Two possible reasons are clarified as follows. On one hand, the beginning stage of the bone healing is an inflammation phase, to recruit the MSCs to the injury site. 39 Similarly, a local inflammatory reaction will occur after photothermal treatment. On the other hand, circulating MSCs are present in the peripheral blood in minimal concentrations under normal conditions. However, their numbers significantly increase in the blood of patients with bone fracture, bone sarcomas, osteoporosis, etc. It is believed that these increased MSCs may be released from the bone marrow. In addition, previous research has demonstrated that some BMSCs involved in bone regeneration are systemically mobilized and recruited to the defective site from remote bone marrow. 40 Therefore, we did not investigate the toxicity of NIR to local normal tissue for the bone defect repair. The gross observation (3D reconstruction) and micro computed tomography (micro‐CT) analysis were conducted on samples collected at week 24 after the implantation of the composite scaffolds. 3D reconstruction of harvested craniums showed that much more calcified tissues were present in the defect implanted with TBGS, which confirmed the fact that TBGSs featured better regeneration outcome for bone defects than pure BGS without MXene integration ( Figure 6 a, b). The micro‐CT images directly presented this result by displaying both front and back surface of a cranium that TBGS (Figure 6 d, e) was more effective than BGS (Figure 6 c, f) in bone regeneration at week 24. Quantitative analysis of fundamental parameters was conducted based on the histomorphometric micro‐CT analysis, such as the relative bone volume/tissue volume (BV/TV), bone mineral density (BMD), and porosity (TOT). The BV/TV, representing the percentage of newborn osseous tissue volume accounting for the entire defect space, was higher in TBGS than in BGS (Figure 6 g). The BMD and TOT revealed the average bone density of circular defect areas from two perspectives (Figure 6 h, i). These data collectively revealed the excellent osteogenic performance of TBGSs as compared to BGS. Figure 6 In vivo osteogenesis performance of BGS and TBGS. a, b) 3D reconstruction of circular defects at 24 weeks after scaffolds implantation. c–f) Micro‐CT images of cranial defect areas with a diameter of 5 mm at 24 weeks postoperation. g) Value of BV/TV in newborn osseous tissue ( n = 6). h) Value of BMD in newborn osseous tissue ( n = 6). i) Value of TOT in newborn osseous tissue ( n = 6). * p < 0. 05, ** p < 0. 01, *** p < 0. 001. Newborn osseous tissue was also further assessed by CLSM through scanning in circular defect regions. The samples of both groups were marked with tetracycline hydrochloride (HCL) (blue), Alizarin red (red), and calcein‐AM (green), and different colors represented newborn bone tissue at different stages of osteogenesis (blue fluorescence: weeks 2–4; green fluorescence: weeks 4–6; red fluorescence: weeks 6–8). The fluorescence intensity in left (BGS group) was significantly weaker than that in the right (TBGS group), which suggested that the TBGS stimulated more efficient osteogenic activity as compared with BGS ( Figure 7 a–d). Even though both control and treatment groups were significantly stained with the three colors, the newborn osseous tissue around TBGS group (Figure 7 g, h) showed better osteogenic performance compared with the BGS group (Figure 7 e, f). Green and red fluorescence in the TBGS group were obviously more than that in the BGS group, indicating that more newborn osseous tissues were formed in the TBGS group than that in BGS group in the latest 4 weeks. The quantitative analysis of CLSM images (Figure S14, Supporting Information) made the osteogenesis capacities of BGS and TBGS more clearly presented, which showed that the osteogenesis rate of BGS and TBGS were about 20% and 50%, respectively. The above practices further confirmed the powerful bone reconstruct capability of TBGS in animal levels. Figure 7 Confocal fluorescence images for superficial analysis of newborn osseous tissue of BGS and TBGS groups at week 8. a) Tetracycline HCL (blue fluorescence) injected intramuscularly into calvarial defect model rats at week 2. b) Alizarin red (red fluorescence) injected intramuscularly into calvarial defect model rats at week 4. c) Calcein‐AM (green fluorescence) injected intramuscularly into calvarial defect model rats at week 6. d) Merged image of three fluorochromes. These three fluorochromes represent newborn osseous tissue in different therapeutic duration. Scale bar in (a)–(d) is 1 mm. e, f) Magnified images represented newborn bone around BGS. g, h) Higher‐magnification images indicating the hierarchical architecture of bone around TBGS and its corresponding material‐guided regeneration process. Scale bar in images (e)–(h) is 125 µm. To further evaluate the efficacy of TBGSs for bone‐defect regeneration in other aspects, the H&E staining ( Figure 8 a–c) and Goldner staining (Figure 8 d–i) were conducted. H&E staining showed that there was no inflammatory cell in either BGS or TBGS group. A large number of mineralized bone tissues (yellow triangles) was found in the bone defect implanted with TBGSs (Figure 8 c). Meanwhile, there was no obviously visible residual scaffold (black asterisks) in the experiment group as compared to the BGS group (Figure 8 b). Goldner staining exhibited that the defect region in the BGS group displayed a mixture of new osteoid tissue (red tissue) around the residual materials (black asterisks) (Figure 8 d, e). At the same time, there was quite a lot of mineralized bone tissues (emerald green tissue) filled in the defect region of TBGS group, indicating a better newborn bone formation in TBGS group (Figure 8 f). In addition, Figure 8 g–i displays newborn bone‐tissue formation during different periods (weeks 8, 16, and 24) of TBGSs. Images at week 8 (Figure 8 g) revealed a large amount of fibroblast and macrophage crawled in and through the pores of scaffolds. Red osteoid tissue was generated around the materials while the old scaffolds were degraded, which demonstrated the desirable simultaneous process of the degradation of old scaffolds and the formation of new osseous tissue. A lot of mineralized bone tissues were around the residual old scaffolds (Figure 8 h), collectively revealing the excellent regeneration performance of TBGS. There were no obvious scaffolds left in the bone defect of TBGS group at week 24 (Figure 8 i). The defect region was covered with mineralized bone, without a visible difference with the old bone tissue around the defect region. This desirable therapeutic outcome, which was attributed to effects of TBGSs, manifested the material‐guided bone regeneration process that osteoblast adhered and proliferated on the TBGSs with both osteoconduction and osteoinduction, accompanied with the formation of new osseous tissue on the vanishing scaffolds substrates. Furthermore, to evaluate the degradation of scaffolds, the BGSs and TBGSs were soaked in simulated body fluid (SBF) for 14 days at 37 °C, and the degradation rates of TBGS and BGS are ≈5% and 3%, respectively (Figure S15, Supporting Information). In addition, according to the Goldner staining images (Figure 8 g–i), with the prolonging of the in vivo experiment, the residual amount of scaffold was gradually degraded. Therefore, this result proved that TBGSs owned biodegradability, high biocompatibility, and the desirable performance of accelerating tissue reconstruction. Figure 8 Histology staining of harvested craniums of Sprague–Dawley rats implanted with BGS/TBGS at week 24. a–c) H&E staining of harvested craniums obtained from SD rats at week 24 after operation. d–f) Goldner staining of harvested craniums of SD rats at week 24 after implanting with BGS and TBGS. g–i) Goldner staining of TBGS group at different period of weeks 8, 16, and 24. The defect areas were implanted with BGS and TBGS. Black asterisks mark implanted scaffolds that were not biodegraded completely. Yellow triangles indicate newborn osseous tissue. Scale bar in (a) and (d) is 2 mm. Scale bar in (b), (c), (e), and (f) is 500 µm. Scale bar in (g)–(i) is 200 µm. To further investigate the in vivo long‐term toxicity (24 weeks) of BGSs/TBGSs, venous blood was collected and the major organs of rats (heart, liver, spleen, lung, and kidney) were dissected, which were fixed in a 10% formalin and stained with H&E for histological analysis after all the SD rats were executed. The hematology parameters including leucocyte, erythrocyte, hemoglobin (HGB), the percentage of neutrophil, albumin/globulin (A/G), albumin (ALB), blood urea nitrogen (BUN), cholinesterase (CHE), uric acid (URCA), K +, Na +, and Ca 2+ were tested (Figure S16a, Supporting Information). It has been found that there were no meaningful changes in the TBGSs group in comparison to the control group. In addition, the corresponding histological sections of major organs (Figure S16b, Supporting Information) also exhibited no significant abnormalities between the control and treatment groups. Based on the above results, there were no obvious toxicity, inflammation, and infection as observed in the treated SD rats during a long therapeutic period. It also demonstrated that TBGSs were highly biocompatible for the further safe in vivo osteogenic surgery. 3 Conclusions In summary, we have successfully integrated 2D Ti 3 C 2 MXene with 3D‐pringting scaffolds for achieving simultaneous photonic bone‐tumor killing and bone‐tissue regeneration, which has been respectively demonstrated by the subcutaneous osteosarcomas model in nude mice and the bone defect model in SD rats. This composite scaffold takes the unique photothermal‐conversion performance of 2D Ti 3 C 2 MXene and bone‐regeneration capability of BG scaffolds. The TBGS developed in this work is expected to be used for the postoperative treatment of osteosarcoma, that is, TBGSs would be implanted into the bone defect site formed by the surgical resection of bone tumor. Then, the high photothermal‐conversion performance in NIR region of TBGS would be initially used to kill the potentially residual bone‐tumor cells, and the excellent bone conduction and induction characteristics of TBGS would be then employed to repair the bone defects. As demonstrated in the experimental results, both in vitro and in vivo systematic assessments have demonstrated that these Ti 3 C 2 MXene‐integrated composite scaffolds efficiently induced the death of bone cancer cells and eradicated the tumor on bone‐tumor xenograft by NIR irradiation. Especially and importantly, the integration of 2D Ti 3 C 2 MXene has been demonstrated to efficiently accelerate the growth of newborn bone tissue of the composite BG scaffolds. The dual functionality of bone‐tumor killing and bone‐tissue regeneration makes these Ti 3 C 2 MXene‐integrated composite scaffolds highly promising for the treatment of bone tumor. This first report on introducing MXene‐based nanoplatforms into tissue‐engineering biomedical field not only broadens the applications of 2D MXenes in biomedicine, but also provides an intriguing biomaterial system for initiating the related tissue engineering‐related researches. 4 Experimental Section Synthesis of Raw Ti 3 C 2 Nanosheets : First, the Ti 3 AlC 2 ceramic powder was fabricated by uniformly mixing titanium powder (Alfa Aesar, Ward Hill, USA, 99. 5 wt% purity; −325 mesh), aluminum powder (Alfa Aesar, Ward Hill, USA, 99. 5 wt% purity; −325 mesh), and graphite powder (Alfa Aesar, Ward Hill, USA, 99 wt% purity; particle size <48 µm, −300 mesh). The powder mixture (Ti/Al/C molar ratio: 2/1/1) was ground in a planetary ball mill for 10 h and then sintered in Ar atmosphere (1500 °C, 2 h). Then, Ti 3 C 2 NSs were synthesized by selectively removing the Al layer from the Ti 3 AlC 2 ceramic with HF etching at room temperature according to previous report. 41 Typically, the Ti 3 AlC 2 powder (10 g) was immersed into HF aqueous solution (40%, 50 mL; Sinopharm Chemical Reagents Co. , Ltd. , Shanghai, China) in a polytetrafluoroethylene (PTFE) container, and the mixture was stirred for 48 h at room temperature. After centrifugation and washing, the precipitations were dispersed into TPAOH (50 mL, 25 wt% aqueous solution; J&K Scientific Co. , Ltd. , Beijing, China) under stirring for 72 h at room temperature. Finally, the resulting suspension was centrifugated and washed three times with deionized water for removing the remnant TPAOH. By this method, the raw 2D Ti 3 C 2 NSs were obtained. 42 Synthesis of Raw BG Powders : Briefly, raw BG powders were prepared via an evaporation‐induced self‐assembly (EISA) method. 43 Typically, 80S15C BG powders (Si/Ca/P molar ratio: 80/15/5) were synthesized by dissolving tetraethoxysilane (TEOS, 53. 6 g), Ca(NO 3 ) 2 ⋅4H 2 O (11. 2 g), triethyl phosphate (TEP, 5. 84 g), and HCl (0. 5 m, 8 g) into ethanol (480 g). Then, the mixture was stirred at room temperature. After 24 h, the resulting sol was transferred into a petri dish for EISA process at room temperature for 7 days in a fume cupboard and then dried at 60 °C for 48 h. After being further ground, the raw BG powders were passed through 400 mesh sieve for eventually forming homogeneous size less than 37 µm. 3D Printing of BG Scaffolds : All printed scaffolds were fabricated by a 4th generation 3D Bioplotter (Envision GmbH, Germany). The printing ink was introduced into a polyethylene syringe tube which was fixed onto the 3D Bioplotter. A tapered nozzle (inner diameter: 400 µm) was attached to the syringe tube. Then, scaffolds ( Φ 10 × 2 mm, pore size: 350 µm) were plotted layer by layer by extruding the paste as a fiber. The architecture was changed by plotting fibers with 0 and 60 angle steps between two successive layers. The dosing pressure to the syringe pump was 2. 8–4. 4 bar. The printing speed was 8–18 mm s −1 and the layer thickness was about 0. 32 mm. Nozzle temperature was set at 30 °C and build plate temperature was consistent with the room temperature. The printed scaffolds were dried (37 °C, 12 h) and sintered (1060 °C, 3 h) to obtain pure BGSs. Ti 3 C 2 MXene Integration into BG Scaffolds : Ti 3 C 2 NSs were suspended in distilled water by ultrasonic treatment to obtain the homogeneous Ti 3 C 2 aqueous suspension. To prepare TBGSs, the BGSs were soaked in Ti 3 C 2 aqueous solution at different concentrations (1. 0, 1. 5, and 2. 0 mg mL −1 ) for 10 min and dried at 60 °C for 4 h. This operation was repeated three times and finally TBGSs were obtained. BGS integrated with 1. 0 mg mL −1 Ti 3 C 2 NSs was termed as 1. 0 TBGS, and other TBGSs were renamed by this analogy. Characterization : SEM images, EDS, and element mapping were measured on a SU8220 microscope (Hitachi, Japan). Both TEM and HRTEM images were observed by a JEM‐2100F transmission electron microscope. XPS was recorded on ESCALAB250 (Thermal Scientific, US). XRD analysis was operated on a Rigaku D/MAX‐2200 PC XRD system. Raman spectra were recorded on a high‐resolution Raman microscope (HORIBA LabRAM HR800). The CLSM images were acquired by an Olympus BX53 fluorescence microscope. NIR laser was produced using an 808 nm high‐power multimode pump laser (Shanghai Connect Fiber Optics Company). The temperature detection and thermal‐image record were conducted on an infrared thermal imaging instrument (FLIR A325SC camera, USA). The element quantitation was analyzed by inductively coupled plasma‐optical emission spectrometry (ICP‐OES, Agilent 725, Agilent Technologies, USA). In Vitro Photothermal Performance of Ti 3 C 2 ‐BG Scaffolds : The surface temperature of scaffolds was monitored by an infrared thermal imaging instrument. To explore the photothermal performance of different scaffolds, BGS, 1. 0 TBGS, 1. 5 TBGS, and 2. 0 TBGS were exposed to an 808 nm laser irradiation at the power density of 1. 0 W cm −2. Then, the photothermal performance of 1. 0 TBGS at varied power densities (0. 5, 0. 75, and 1. 0 W cm −2 ) was also investigated to explore appropriate laser power density for ablating tumor. The above experiments were conducted in a dry environment (in air). Analogously, the photothermal performance of 1. 0 TBGSs at varied power densities (0. 5, 0. 75, and 1. 0 W cm −2 ) was also assessed under wet environment (in 400 µL PBS). Finally, the photothermal stability of TBGSs was acquired (five laser “off–on” cycles, 1. 0 W cm −2 ). In Vitro Cytotoxicity Assay and Cell Ablation : Osteosarcoma Saos‐2 line (noted as Saos‐2 cells, Cell Bank of Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences) was maintained in McCoy's 5A Medium (HyClone) and supplemented with 1% penicillin/streptomycin and 10% fetal bovine serum (FBS) in a humidified incubator (5% CO 2, 37 °C). To investigate in vitro toxicities and anticancer effects of TBGSs, Saos‐2 cells were seeded in 48‐well plates (Corning, USA) for 24 h (1. 0 × 10 5 per well, 800 µL medium), and then the 1. 0 TBGSs and BGs ( Φ 8 mm × 1. 5 mm) were gently placed on the plates to co‐incubate for additional 24 h. Afterward, the standard CCK‐8 assay was performed to quantify the cell viabilities after different treatments ( n = 4). Different irradiation durations (0, 5, 10, and 15 min), irradiation times (0, 1, 2, and 3 times), and power densities (0, 0. 5, 0. 75, and 1. 0 W cm −2 ) for photothermal ablation were also systematically evaluated. The OD value of wells without NIR irradiation (control, BGS, 1. 0 TBGS) indicated the in vitro toxicities of materials, and the other groups (NIR, BGS + NIR, 1. 0 TBGS + NIR) presented the ability of photothermal ablation against osteosarcoma cells. To visually evaluate the photothermal‐ablation effect of scaffolds on osteosarcoma cells, Saos‐2 cells were incubated in 48‐well plates with BGSs and TBGSs. After 24 h, the BGSs and TBGSs were irradiated by 808 nm laser (10 min, 1. 0 W cm −2 ). Subsequently, cells in BGSs and TBGSs with or without irradiation were stained with PI/calcein‐AM. Finally, the scaffolds were visualized by CLSM. Dead cells stained with PI showed the red fluorescence, and live cells stained with calcein‐AM exhibited the green fluorescence. All scaffolds were sterilized by UV radiation for 24 h before experimental evaluation. Cell Culture : Primary hBMSCs were obtained from ScienCell Research Laboratories (the United States, #7500) and cultured with α‐MEM (Gibco) supplemented with 10% FBS (Gibco) in 5% CO 2 at 37 °C. Every 3–4 days, the cells were detached (from the surface of the 75 cm 2 cell culture flask (Greiner Bio‐One) using 0. 25% trypsin), washed, centrifuged (1000 rpm × 5 min), resuspended (in 12 mL α‐MEM), and subcultured (in 1:3 volume ratio). Cells, from 4th to 9th generations, were used for the experiments. The cells were regularly examined under an optical microscope to monitor growth and possible contamination. Sample Preparation for SEM and CLSM Observation : In short, hBMSCs (1. 0 × 10 4 ) were seeded in 48‐well culture plates with BGSs and 1. 0 TBGSs. One day later, to observe the morphology and adhesion of hBMSCs on the scaffolds, BGSs and TBGSs were fixed with glutaraldehyde and then dehydrated with gradient concentrations of ethanol (30, 40, 50, 60, 70, 80, 90, and 100 v/v%). Then, the scaffolds with hBMSCs were observed by SEM. To further investigate the cytoskeletal change during the osteoblastic differentiation, the hBMSCs were co‐cultured with BGS or TBGS and stained by 4′, 6‐diamidino‐2‐phenylindole (DAPI, blue)/rhodamine phalloidin (red) on the 1st and 7th days. Then, the CLSM photographs were recorded. Finally, BGSs and TBGSs, seeded with hBMSCs, were stained by Alizarin red to specifically mark calcium salt which was generated during the mineralization. Quantitative Real‐Time Polymerase Chain Reaction (QPCR) Analysis : The effects of different scaffolds on the osteogenic differentiation of hBMSCs were assessed by measuring the mRNA expression of COL I, RUNX2, OCN, and OPN genes. The total cellular RNA was harvested with TRIzol (Invitrogen) after osteogenic induction at the 7th day. One microgram of purified RNA was then reversely transcribed into complementary DNA (cDNA) using the PrimeScript RT reagent kit (Takara, Shiga, Japan). The reverse transcription reaction was quantified by the ABI Prism 7900. Thermal Cycler used a real‐time PCR kit (SYBR Premix EX Taq, Takara, Japan). The product was quantified using a standard curve, and levels of gene expression were normalized to glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH). Relative gene expression was analyzed by the 2 −ΔΔCt method. 44 Cell Toxicity and Proliferation : To evaluate the cell toxicity and proliferation, hBMSCs were seeded on BGSs and TBGSs. Cell toxicity and proliferation were observed by CLSM after 7 days. The proliferation of cells was measured by CCK‐8 assay on the 1st, 3rd, 5th, and 7th days. Alizarin Red Staining : Cells were co‐cultured with BGSs and TBGSs in 24‐transwell plates in osteogenic medium for 3 weeks. The culture medium was renewed every 2 days. After 21 days, cells were washed with PBS and fixed with 4% paraformaldehyde for 30 min at 4 °C. After that, the cells were stained with Alizarin red S solution (40 × 10 −3 m, 2% aqueous, Sigma) for 15 min. Cells were rinsed again with PBS before being observed by microscopy. In Vivo Photothermal Therapy in NIR Biowindow : Female BALB/c nude mice (about 13 g) were subcutaneously injected with Saos‐2 cells (4 × 10 6 cell per site) to establish the ectopic osteosarcoma model. When the volume of the tumors reached about 120 mm 3, the mice were divided into four groups including BGS group, BGS + NIR laser group, TBGS group, and TBGS + NIR laser group ( n = 6 in each group). Then, a small incision was carefully made to expose the tumor, and the scaffolds (BGSs or TBGSs, 8 mm × 1. 5 mm × 1. 5 mm) were implanted into the center of the lesion, and the subsequent surgical sutures were used to close the wound. Twenty‐four hours later, the in vivo photothermal therapy was performed. Laser irradiation was carried out on the BGS + NIR laser group and TBGS + NIR laser group. Each mouse was anesthetized and exposed to the 808 nm laser for 10 min (1. 0 W cm −2 ). The tumor surface temperature and the thermal images of mice were recorded by an infrared thermal camera during the treatment. The NIR treatment time was set as day 0. From day 0, the tumor volume and the body weight of all mice were monitored every 2 days during half a month after the corresponding treatments. The tumor volume was calculated according to the following formula: tumor volume ( V ) = (tumor length) × (tumor width) 2 /2 − scaffold volume. The tumors were dissected and sectioned into slices to qualitatively measure the necrosis. The tumor slices were stained with H&E, TUNEL, and Ki‐67 antibody. To further investigate in vivo toxicity of pure BGSs and TBGSs, the major organs (heart, liver, spleen, lung, and kidney) of four groups of mice were obtained and stained with H&E on the 1st, 14th, and 28th days, respectively. Animal Surgical Procedures : All surgical procedures were performed on 8 week old male SD rats. Following anesthesia with intraperitoneal pentobarbital (5 mg/100 g; Sigma), two 5 mm defects in the frontal‐parietal bone were created using an electric trephine (Nouvag AG, Goldach, Switzerland). After that, the calvarial defects were filled with BGS and TBGS, respectively ( Φ 5 mm × 2 mm). Finally, the incision was closed by suturing the periosteum and skin separately. The HCL, Alizarin red, and calcein‐AM were injected intramuscularly at weeks 2, 4, and 6. Rats were successively killed by an overdose of anesthetic after 8, 16, and 24 weeks. Craniums were gathered and fixed in a 4% paraformaldehyde solution overnight before further analysis. Micro‐CT Analysis : All the harvested specimens were examined using the mCT‐80 system to evaluate new bone formation within the defect region. The undecalcified samples were scanned at a resolution of 18 µm. After 3D reconstruction, the relative BV/TV, BMD, and total porosity (TOT) in the defect regions were used to calculate new bone formation using the auxiliary software of the mCT‐80 system44. Histological Analysis of Newborn Osseous Tissue : After decalcification and paraffin embedding, specimens were cut into 5 µm thick sections and then incubated at 60 °C for 1. 5 h. To evaluate the newborn osseous tissues around the BGS/TBGS, hard tissue slices were stained with H&E and Goldner's trichrome method. For Goldner staining, sections were placed in Weigert's Hematoxylin for 30 min, washed in running tap water for 10 min, and then stained in Ponceau Acid Fuchsin, phosphomolybdic acid–Orange G solution, and Light Green stock solution. Photomicrographs were acquired using a LEICA DM 4000. Meanwhile, standard blood tests were also performed. Statistical Analysis : All data were reported as mean ± standard deviation. Statistical comparisons were conducted with Student's two‐sided t ‐test as * p < 0. 05 (statistically significant), ** p < 0. 01 (moderately significant), and *** p < 0. 001 (highly significant). Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201901719
| 2,019
|
Advanced Science
|
Formation of Polarized, Functional Artificial Cells from Compartmentalized Droplet Networks and Nanomaterials, Using One‐Step, Dual‐Material 3D‐Printed Microfluidics
|
Abstract The bottom‐up construction of synthetic cells with user‐defined chemical organization holds considerable promise in the creation of bioinspired materials. Complex emulsions, droplet networks, and nested vesicles all represent platforms for the engineering of segregated chemistries with controlled communication, analogous to biological cells. Microfluidic manufacture of such droplet‐based materials typically results in radial or axisymmetric structures. In contrast, biological cells frequently display chemical polarity or gradients, which enable the determination of directionality, and inform higher‐order interactions. Here, a dual‐material, 3D‐printing methodology to produce microfluidic architectures that enable the construction of functional, asymmetric, hierarchical, emulsion‐based artificial cellular chassis is developed. These materials incorporate droplet networks, lipid membranes, and nanoparticle components. Microfluidic 3D‐channel arrangements enable symmetry‐breaking and the spatial patterning of droplet hierarchies. This approach can produce internal gradients and hemispherically patterned, multilayered shells alongside chemical compartmentalization. Such organization enables incorporation of organic and inorganic components, including lipid bilayers, within the same entity. In this way, functional polarization, that imparts individual and collective directionality on the resulting artificial cells, is demonstrated. This approach enables exploitation of polarity and asymmetry, in conjunction with compartmentalized and networked chemistry, in single and higher‐order organized structures, thereby increasing the palette of functionality in artificial cellular materials.
|
1 Introduction The creation of artificial cells with functional properties, that are analogous to their biological counterparts, is envisaged to give rise to a wealth of opportunities, in a diversity of application areas, in both biotechnology and therapeutics, 1, 2, 3, 4, 5 for example, in tumor cell targeted drug delivery 6 and artificial kidneys for clinical treatments. 7 Imparting functionality to artificial cells is typically underpinned by organization of chemical reactions, with the aim to engineer specific dynamic or responsive behaviors. This itself is enabled by the spatiotemporal organization of chemical reactions, and their selective, controlled communication. 8, 9, 10, 11, 12 The aspiration to engineer such chemical models which harness biology, represents a conceptual shift from linear systems with fixed physical‐chemical properties, toward the exploitation of emergent, nonequilibrium, and dynamic properties. 13 Such approaches are both inspired by, and become increasingly realizable with, our understanding of biological complexity, and the cell's exploitation of emergent behaviors; for instance, those arising from (i) molecular self‐assembly, (ii) the barrier properties of membranes, (iii) the compartmentalization of multiple chemistries, and (iv) the process of symmetry‐breaking, to develop new properties. 14, 15, 16, 17, 18, 19, 20, 21, 22 In this regard, and of particular interest, is the ability to manufacture miniaturized, asymmetric entities with spatially organized chemistries to dictate functional polarity. 23 Such features have been demonstrated to possess emergent and collective properties in solid‐state materials. 24, 25 In biological systems, these properties are essential for dictating directional motility, 26 governing cellular migration, 27 enabling response to external chemical gradients 28 and defining directional growth. 29, 30 Consequently, there has been significant interest in incorporating such polarizable properties in artificial cells. 31, 32, 33, 34, 35, 36, 37, 38 However, to date, the practical establishment of this has remained limited. 39, 40 Despite the importance of discrete chemical organization in artificial cells, the majority of research in bottom‐up synthetic biology, has focused on the development of much simpler models. There has been significant interest in the construction of spontaneous, self‐assembled, artificial cells, in the form of vesicles, colloidosomes, or phase‐separated oil–water, or water–water (coacervate) systems. 41, 42, 43, 44, 45 The exploitation of such processes can shed important light on the emergence of early life, and resultant structures may represent minimally complex, artificial cell chassis, often termed “protocells. ” However, using these spontaneous self‐assembly processes alone, it is challenging to recapitulate the controlled subcompartmentalization and spatial organization of biological cells, and, therefore, the higher functionality of even simple biological systems. In this regard, it is notable that few biological structures are formed de novo, instead being spatially and temporally constructed, or formed, from existing structures. 46 Consequently, droplet microfluidics represents an interesting parallel, where channel architecture and flow provide a route for directing the assembly of compartmentalized, and spatially patterned, chemically organized, soft‐matter materials. 22, 47, 48, 49 Beyond these self‐assembled systems, increased structural complexity has been sought in droplet networks 50, 51, 52 and compartmentalized vesicles 53 to produce interconnected, membrane‐segregated compartments, as artificial cellular systems. Such materials have been constructed manually, 18 by the printing or assembly of individual droplet units 20 and by using microfluidics 54 and demonstrated as networked reactors possessing emergent properties. 55, 56 Microfluidic construction of such materials offers the opportunity of scalable production, using massively parallel microfluidics, 57, 58 and typically involves their formation from hierarchical water/oil emulsions. Ourselves, and others, have demonstrated the creation of droplet networks, segregated by lipid bilayers using microfluidic approaches. 54, 59 By extending this concept to a third phase and forming a triple emulsion, such droplet structures can be effectively encapsulated within a permeable, freestanding, hydrogel shell. This droplet assembly is able to communicate with its external environment and be functionally enhanced by the integration of transmembrane proteins. 59 It is notable that the microfluidic formation of such hierarchical droplets, or higher‐order emulsions, requires the integration of both hydrophilic and hydrophobic surfaces within the same microfluidic device. This determines the channel wetting properties and governs whether water or oil droplets form at a given junction. Consequently, the requirement to alternate between aqueous and nonaqueous wetting phases represents a fabrication challenge to achieve the integration of both hydrophilic and hydrophobic channels. This may entail the complex assembly of multimaterial, or surface‐functionalized, fluidic channels. Furthermore, traditional planar fabrication tools usually constrain droplet flows to a single plane, which creates 2D axisymmetric morphologies in double and triple emulsions. Consequently, in both lipid bilayer stabilized, and more traditional multiphase emulsions, a number of constraints are placed on the ability to create nonsymmetric droplet structures in a well‐controlled manner. In this regard, asymmetric droplets, or so‐called Janus droplets, have been explored as a means to break the morphological symmetry in droplets. 60, 61, 62 Janus droplets were first reported by Nisisako et al with microfluidic formation of such structures deterministically created in single emulsion laminar flows. 63 Janus particles have been used in applications that exploit the particle anisotropy, ranging from chemical sensors 64 to electrically activated, reconfigurable displays. 63 Double emulsions have also been used as a template to create Janus polymer structures, with the immiscible oil and water volumes templating the synthesis of two conjoined particle hemispheres. 65 Double emulsions with a Janus shell have also been reported, 66, 67 but the Janus patterning has remained limited to a single phase, and in a single plane. The ability to expand such approaches and combine with droplet interface bilayer networks could provide a means to pattern individual layers of hierarchical, membrane‐based, cellular structures. Further, the ability to create droplet structures with symmetry‐breaking in more than one plane would enable the discrete patterning of each phase, and its orientation, in the droplet hierarchy. As a result, this could produce chemical gradients or polarized phases, and position compartments spatially within the fluidic or soft matrices. This would, therefore, facilitate the construction of synthetic cell chassis, with compartmentalized chemistries, access to membrane biochemistry, with overall spatial control of assembly. Such order would enable control over chemical polarity, and increase the functional complexity of artificial cells. To realize this, we report here the development of one‐step, multimaterial microfluidic devices with 3D arranged channels, to produce hierarchical droplets with discrete patterning of component liquid phases. Using this approach, we demonstrate the production of symmetry‐breaking droplet structures, incorporating patterned organic and inorganic components, with biological lipid bilayers for the first time. This increases the palette of functionality and afforded assembly control in such chemically organized synthetic cells. We demonstrate the ability to impart polarization‐driven behaviors in the constructs, by creating magnetically polarized, encapsulated droplet networks that can orientate, rotate, and migrate, as well as collectively establish a common orientation, and undergo shared collective motions in larger populations. 2 Results 2. 1 Single‐Step 3D‐Printed Microfluidic Devices for the Production of Multiphase Emulsions The advent of 3D printing has enabled the construction of 3D fluidic architectures. 68, 69 In droplet microfluidics, hydrophobic fluidic channels govern the formation of water droplets in a continuous oil flow, and hydrophilic fluidic channels enable the dispersion of oil droplets in continuous water flows. Here, we developed a monolithically constructed, 3D fluidic device, consisting of both hydrophobic and hydrophilic materials, alternately deposited in a single, integrated build process, using a dual‐material printer (Ultimaker 3). This created both hydrophilic and hydrophobic ducts in a single fabrication step, without the requirement for surface modification, or post‐manufacture integration of different substrates. Using this approach, it was possible to produce water droplets in oil using a hydrophobic PLA channel ( Figure 1 a), and oil droplets in water, within a more hydrophilic polyvinyl alcohol (PVA) channel (Figure 1 b). In Figure 1 a, b, the aqueous wetting contact angle of PLA and PVA is depicted, together with the resultant droplet regimes. The increased hydrophilic properties of PVA enable inversion of the usual droplet formation regime to produce oil droplets in water. These material architectures were then combined in a single device to sequentially produce water droplets in oil, followed by oil droplets in water, and finally, the ejection of water droplets from the device, to a collection bath (Figure 1 c, and Table S1, Video S1, Supporting Information). This enabled the hierarchal assembly of complex, multiphase emulsions, using the monolithic microfluidic device, which was fabricated in a one‐step process. Figure 1 c‐iii demonstrates the production of aqueous droplets (red) encapsulated within an oil droplet (blue), itself encapsulated within an alginate hydrogel shell (colorless) using this approach. These triple emulsions are shown to be sequentially deposited from the fluidic device, in effect, “printing” individual complex emulsion droplets. Figure 1 3D‐printing of microfluidic devices in hydrophobic and hydrophilic polymers enables the production of both water droplets in oil and oil droplets in water. Dual‐material printed microfluidic devices allow the sequential application of these droplet formation regimes to produce triple emulsions of hierarchical water/oil/water droplet structures. a) 3D‐printed polylactic acid (PLA) substrate displays hydrophobic properties demonstrated by water contact angle measurement. 3D‐printed PLA T‐junction generates water droplets in oil (schematic and experimental image). b) 3D‐printed polyvinyl alcohol (PVA) substrate demonstrates more hydrophilic properties by water contact angle measurement. 3D‐printed PVA T‐junction generates oil droplets in water (schematic and experimental image of oil (blue) droplet ejection from PVA device into a water bath). c‐i) Dual head extrusion printer enables fabrication of integrated PLA and PVA, dual material, microfluidic devices. c‐ii) (Top) Schematic depiction of operational concept; sequential, droplet generating, flow‐focusing junctions in PLA and PVA, enable the formation of double emulsions (water‐in‐oil‐in‐hydrogel (ungelled) (w/o/w)). Monolithic device realized in lower image, comprised of PLA (pink) and PVA (colorless) materials. c‐iii) Triple emulsions formed as hydrogel droplets are ejected from the microfluidic device. The outer, liquid, alginate phase is gelled upon dripping into a CaCl 2 containing gelling bath, creating a triple water‐oil‐water emulsion with a gelled outermost shell. These complex emulsions (pink: aqueous internal phase, blue: oil midphase, colorless: outer alginate shell phase) are individually deposited from a dual material, microfluidic device. All scale bars: 1 mm. 2. 2 Characterization of Complex Emulsion Formation The formation of oil droplets in a continuous water phase, without surface modification, is usually challenging, owing to the native preferential hydrophobic wetting of many substrate materials, which favor the formation of water droplets in oil. Consequently, to date, only water droplet formation, in extrusion 3D‐printed microfluidic devices, has been demonstrated. 70 We observe that in the formation of complex emulsions, the comparatively hydrophilic PVA junction is capable of operating in two distinct droplet generating regimes for the formation of oil droplets in water. These regimes are essentially, either dripping or jetting ( Figure 2 a), both being determined by the combination of input fluid flow rates, and the specific composition of the respective fluid phases (Figure 2 b and Table S2, Supporting Information). In the formation of triple emulsions, this results in the formation of either single, or multiple, encapsulated middle phase oil droplets (designated m ), containing inner phase aqueous droplets (designated i ) in the larger emulsion constructs, under jetting and dripping regimes, respectively. Figure 2 b shows the manifestation of this with respect to the relative, and absolute, flow rates of the three input fluid phases, for both a fatty‐acid‐rich sunflower oil, and a hexadecane/silicone oil midphase. The complex interplay of fluid flow properties has been characterized in the determination of droplet formation, by dripping and jetting regimes and their transitions. 71 Fluid density, velocity, droplet and channel diameter, surface tension and viscosity, all play an important role, as drag and inertial forces compete with surface tension. 72, 73 It is unsurprising that this relationship becomes more complex in triple‐phase systems, where flow rates govern both upstream droplet formation, and also influence linear flow velocity downstream at subsequent encapsulation. The fluid dynamics of droplet formation are additionally perturbed by both internal and external immiscible interfaces. 74 New dimensionless numbers have been proposed to describe triple emulsion droplet formation, 75 albeit by a different microfluidic method with fluid phases unsuited to the creation of artificial cells. In this work, we take an empirical approach to control the number of inner ( i ) and middle droplets ( m ) that are encapsulated in the shell phase, by the tuning of fluid flow rates, to govern the droplet generation frequency of each of the three phases. The output triple emulsion droplets have uniform morphology (Video S1, Supporting Information), and relatively monodisperse size distributions (<4%) for each phase (Figure S1, Supporting Information). The size of the internal and midphase droplets is observed to change with the relative flow rates of the component fluid phases, corresponding to the internal and midphase droplet number. This is further detailed in Figure S2 (Supporting Information). Figure 2 Tuning fluid flows enables control of droplet morphology. a) The PVA junction responsible for oil‐in‐water droplet formation is capable of operating in two distinct droplet generating regimes in the formation of triple emulsions; either by dripping (1: left and top) or jetting (2: right and lower). These result in the formation of multiple (1: left and top) or single (2: right and lower), encapsulated middle phase oil droplets (blue), containing inner phase aqueous droplets (pink) within the larger emulsion construct (gray). Images (right) depict tipple emulsion formation at the device exit, under dripping (top) and jetting (lower) regimes. b) Ternary phase diagrams depicting flow ratio, and stacked column graphs detailing volumetric flow rates, of the three input phases and corresponding triple emulsion morphology. These detail the number of middle‐phase droplets ( m ) and their number of constituent internal aqueous droplets ( i ), for operation with two different oil phases. On each plot, yellow illustrates two or more midphase oil droplets as a result of operation in a dripping regime, and green indicates a single midphase oil droplet from operation in a jetting regime. Index labels (a–j) enable cross‐correlation between plots. Marker size on phase diagram denotes number of encapsulated inner aqueous droplets ( i ), with value quoted under stacked column graph ( n = 30 under each condition). 2. 3 3D Microfluidic Architectures Enable Formation of Asymmetric and Patterned Complex Emulsions, Creating Hierarchal Polarized Structures The use of 3D printing enables the creation of fully 3D microfluidic architectures, unlike more traditional fabrication approaches, which are usually constrained to machining in a single plane. Using the extra‐dimensionality afforded by 3D printing allowed us to create devices which control the phase orientations and encapsulations within the complex emulsions system, thereby creating patterned and asymmetric droplet structures around different geometric axes ( Figure 3 ). The dominance of low Reynolds number flows maintains the laminar streams of fluids, thereby minimizing mixing. Axial symmetry at the droplet generating junctions was used to enable asymmetric fluid additions with respect to the major channel axis. While also creating a symmetric shear force on the disperse (droplet) phase. This maintained the patterning or organization of emulsions, when the dispersed phase was pinched off by the continuous (shell) phase. Complementary computational fluid dynamic modeling finds that the organization within the resultant w/o/w emulsion flow is stabilized in the extended fluidic duct in our experimental conditions (Table S3, Video S2, Supporting Information). We found that, while in flow, the inner aqueous droplets reached their equilibrium positions within the middle oil droplet, without ejection into the continuous hydrogel phase. This process is dominated by the viscous stress, competing with interfacial tension 76 (see Discussion, Supporting Information). Similarly, we find internal ( i ) droplets are constrained within lateral domains of the encapsulating oil droplet, as are differential additions of middle phase ( m ) oils, thus enabling the maintenance of internal hemispherical patterning and gradients (see Table S4, Supporting Information and Discussion, Supporting Information). Figure 3 3D configurations of dual‐material fluidic channels enable the spatial patterning of complex emulsions. a) (i–iv) Four different fluidic channel geometries. The ability to manufacture channels and junctions in 3D space (ii–iv) provides new opportunities for the spatial patterning and organization of multiphase emulsions, compared to traditional planar (2. 5D) channel arrangements (a‐i). a‐i) With fluidic channels constrained in a single plane, Janus‐core and Janus‐shell (bi‐Janus) double emulsions can be formed, with the hemispherical divide occupying the same orientation, in both instances. a‐ii) Rotating the plane of one droplet generating flow focusing junction geometry by 90° results in the rotation of the plane of Janus droplet formation. Here, creating the depicted bi‐Janus droplet hierarchies, with independent and opposing Janus orientation (90° rotation) in the core (mineral oil with and without oil blue N) and shell phases (alginate shell with and without silica particles or graphene oxide). a‐iii) The addition of a fifth aqueous input ( v ) upstream of the consecutive hydrophobic and hydrophilic droplet generating geometries of (a‐ii), enables the addition of internal aqueous droplets (pink) as inner droplets in bi‐Janus triple emulsions with perpendicular Janus patterning. a‐iv) Extension of this principle to two independent upstream aqueous droplet generating geometries, combined to produce a parallel coflow, enables the addition of two types (differing size and contents) of aqueous internal droplets within the resultant triple emulsion. The lateral offset delivery of these droplets into the common channel, defines the position of entry in the encapsulating droplet. Small aqueous droplets (yellow) are retained to the right‐hand perimeter of the larger (pink) aqueous droplets within the Janus oil middle phase of the triple emulsion. Complementary CFD modeling of the emulsion stabilization within fluidic duct can be found in Table S4 (Supporting Information). b) Schematic illustration of parameter space, droplet number, and arrangement in triple emulsions (see also Figure 2 ). c) In combination, the spatial patterning enabled by 3D fluidic geometries demonstrated in (a) with the ability to control number of inner and middle phase droplets (b), a diverse range of complex and patterned triple emulsions can be made. ( i and m subscripts illustrate droplet numbers of construct). d) The encapsulating hydrogel shell provides mechanical stability, thereby enabling physical manipulation of resultant emulsion droplets. Scale bars: 1 mm. Using this experimental approach, we orientated sequential droplet generating geometries in 3D. This allowed us to spatially pattern the droplet structures with control over the designated patterned phase or phases, and the geometric orientation of patterning. Consequentially, this provided the ability to create both Janus shells, Janus and gradient mid‐droplets ( m ), and achieve the controlled hemispherical placement of inner droplets ( i ). These may differ in size and contents, and be located in different regions within the encapsulating droplet (Figure 3 ). Figure 3 a‐i illustrates conceptually how traditional 2. 5D channel architectures can be used to create Janus shells with inner droplets as double emulsions. Enabled by 3D fabrication, it becomes possible to rotate the fluid inlets (by 90° in this instance) providing the opportunity to create asymmetric coflows in different orientations with respect to the traditional single fabrication axis. In this way, it becomes possible to combine multiple coflow orientations, creating both Janus shells and Janus interiors, with independent and opposing orientations in both double (Figure 3 a‐ii) and triple (Figure 3 a‐iii) emulsions. In principle, this process can be extended to increasing hierarchies of droplet structures in complex emulsions. We also demonstrate that this approach can be extended to the production of differing inner droplets ( i ) in the triple emulsion, and dictate their location in the final droplet construct. This is achieved by joining two streams of water/oil emulsions at a Y‐shaped junction to form a Janus flow before its breakup into discrete encapsulated droplets (Tables S1 row 3 and S4, Supporting Information). Figure 3 a‐iv illustrates the production of inner droplets ( i ) of two different sizes and contents (red ≈430 µm and yellow ≈125 µm) within a Janus encapsulating oil droplet ( m ), itself within a hydrogel shell. The smaller (yellow) inner droplets are retained to the right‐hand perimeter of the larger red droplet population. With this approach to spatial patterning, enabled by 3D fabrication, in combination with the flow‐rate controlling size and number of inner and midphase droplets (Figure 3 b), we demonstrate that it is possible to produce a complex diversity of spatially compartmentalized and patterned, triple emulsion structures (Figure 3 c) from these new 3D‐printed, microfluidic devices. The outer shell phase is gelled on exit from the microfluidic device by the Ca 2+ quickly diffusing and crosslinking the alginate shell in the gelling bath. As such, the final droplet structures are freestanding and able to withstand mechanical manipulation (Figure 3 d). The alginate shell remains water permeable, thus enabling communicative access from the environment to the internal contents, where inner droplets can be segregated from the hydrogel shell by a biomimetic, lipid bilayer. 59 2. 4 Functionalized Encapsulated Droplet Interface Bilayers (eDIBs) In an immiscible, aqueous–oil system, phospholipids may serve as an effective amphiphilic surfactant, stabilizing aqueous droplets in oil through the self‐assembly of lipid monolayers at the water–oil interface. The contacting of two such interfaces leads to the spontaneous creation of a lipid bilayer, resembling the foundational structure of the cell membrane and membrane‐bound organelles. 50, 59 In this way, membrane compartmentalized systems can be constructed within hierarchal droplet architectures, where functional membrane proteins can be reconstituted to enable both the communication and transfer between different compartments and between compartments and the environment. 59 Expansion of the microfluidic patterning approach described earlier was implemented for the creation of encapsulated, droplet interface bilayer systems with layered hydrogel shells. Within these constructs functional organic and inorganic materials were incorporated alongside lipid bilayers ( Figure 4 a and Video S3, Supporting Information). In this way, we produce eDIBs (Figure 4 b‐i), with 25 µm chemically exfoliated graphene oxide (GO) sheets incorporated into the outer shell at concentrations of up to 1. 44% w/w (Figure 4 b‐ii) (GO characterization Figure S2, Supporting Information). GO is an atomically thin 2D, functional nanomaterial that has attractive mechanical, electrical, chemical, and photonic properties. 77 GO has been used to nonspecifically bind proteins and peptides, 78 thus it may be possible to serve as a protective shell layer to protect lipid bilayers from destabilizing peptides and proteins, while retaining the capacity for water and small molecular diffusion to the lipid bilayer. At higher concentrations, the presence of GO increased the viscosity of the hydrogel phase. While this did not hamper the microfluidic formation of the hierarchical droplet assemblies, the slower surface‐tension‐driven relaxation of the outer droplet to a spherical shape during gelation resulted in the gelation of tear‐drop, or ovoid‐shaped, outer shells. Figure 4 In the presence of phospholipids, droplet networks segregated by lipid bilayers may be formed within a triple emulsion. This creates encapsulated droplet interface bilayer networks (eDIBs) that can be patterned with functional materials using the fluidic patterning approach illustrated in Figure 3. a) The presence of amphiphilic phospholipids in the oil midphase serves as a surfactant, forming self‐assembled lipid monolayers, at each oil–water interface. b‐i) The contacting of two such aqueous interfaces results in the formation of a lipid bilayer between the aqueous volumes, creating lipid membranes between internal aqueous droplets, and the internal droplets and the hydrogel shell, where they make contact. Incorporation of two successive outer alginate coflows, in emulsion formation, can be used to form double‐layered hydrogel shells. These can be used to encapsulate droplet networks with Janus‐, or whole‐shell‐, patterning that incorporates functional materials, such as graphene oxide or silica magnetic particles. b‐ii) Bright‐field and fluorescent images of encapsulated droplet interface bilayers (eDIBs), with an outer hydrogel shell loaded with atomically thin graphene oxide (GO) sheets. b‐iii) Encapsulated droplet interface bilayers (eDIBs) with a double hydrogel shell, comprising an inert spacer shell, and an asymmetric Janus‐shell. The Janus shell incorporates silica magnetic microparticles to impart polarized magnetic properties on the membrane‐based artificial cell construct. We also produced eDIBs with a functionally inert inner shell of osmotically matched buffer, followed by a second Janus outer shell, containing embedded, 2 µm porous magnetic silica microparticles. In this way, a thin protective inner hydrogel region could serve as a spacer, in effect, isolating larger silica particles from direct contact with the lipid bilayers, but maintaining aqueous and diffusive continuity of the shell with the membrane. Silica particles have been reported to disrupt artificial bilayers 79 and we observed results consistent with this in the absence of the additional inner hydrogel layer. By the creation of two‐layered shells, we could produce stable particle‐laden encapsulated bilayer systems, with polarized properties, such as a Janus shell (Figure 4 b‐iii), or an asymmetric encapsulated droplet network (Figure S3, Supporting Information). In these systems with inner ( i ) aqueous droplets containing fluorophore sulforhodamine B, fluorescent imaging 3 d following manufacture demonstrated that the lipid bilayer networks remained intact (Figure S4, Supporting Information). The modulation of salt conditions of the alginate phase to avoid excessive osmotic stress on the assembled lipid bilayers was also observed to slow the alginate gelation processes, giving rise to some shape fluctuation in the shell of the final constructs, as they gel on entry to the gelling bath, usually giving rise to slight elongation of the overall structure. It should be possible to expand our reported microfluidic methodology to employ in‐channel gelation, 59 to facilitate gelation while maintaining sphericity within the in‐channel flow. The methodology reported here provides a route for the fabrication of artificial cell chassis containing both lipid membranes and functional nano‐ and microparticulate materials in spatially organized architectures. By using controlled spatial organization, these materials may, therefore, harness the functionality of both the biological and nonbiological components, while circumventing the physical incompatibilities that would otherwise hamper realization. In this way, designer hierarchical droplet structures could be created to control bio, chemical, and physical properties. 2. 5 Polarization‐Induced Properties of Artificial Cell Chassis The particle‐containing encapsulated bilayer systems can gain specific functionality from patterned and encapsulated components. Here, we demonstrate the polarization‐induced functional properties of the eDIBs complete with an outer Janus shell of embedded, paramagnetic, silica microparticles. The incorporation of paramagnetic particles enables the manipulation of the eDIBs, by an external magnetic field, enabling mobility and orientational control, in aqueous environments (Video S4, Supporting Information). These artificial cell chassis can be moved individually in an aqueous environment ( Figure 5 a). In a rotating magnetic field, individual constructs orientate with respect to the magnetic field and to the polarity of their Janus shell, thereby providing a directionality to the droplet structure. These behaviors could be combined in an aqueous pool, where Janus shell eDIBs could seek, orientate, move, and submerge to interface with a static magnetic target placed beneath the Petri dish. The preferred orientation of the construct is governed by the polarized patterning of the eDIB structure (Figure 5 c). Such polarization‐induced properties, created by the ability to spatially organize and pattern eDIBs, provide the opportunity to recapitulate biological and biological‐like behaviors. These may include the harnessing of polarization in directional control, polarized chemical organization, or the exploitation of anisotropy, for example, in the self‐organization of higher‐order assemblies or artificial tissues. Figure 5 d shows a population of core–shell Janus magnetic architectures in an aqueous environment under a rotating magnetic field. All constructs are observed to spontaneously orientate such that their hemispherical division is in the vertical plane. Individual constructs are observed to migrate clockwise, while simultaneously rotating about their own central axis. This is a consequence of both the overall, and locally polarized, magnetic properties of each construct. Colliding constructs are observed to temporarily comigrate with inhibited rotational velocity, before separation and recovery of rotational motion (Figure 5 d‐ii and Table S4, Supporting Information, correlation test). The orientation of individual constructs may all be unified with this type of patterning, and their motion within a larger population may influence others within the group. 25 Such mobility synchronization results in the emergence of simple, group level behaviors. Thus, these principles may serve to pave the way for the design and programming of more sophisticated, population level behaviors, governed by polarized or directional functionality, in membrane‐based artificial cell communities. Figure 5 Polarized encapsulated droplet interface bilayers (eDIBs), with an asymmetric Janus‐shell of silica magnetic microparticles, exhibit directional locomotion in an aqueous environment. a) A single magnetically polarized eDIB is corralled from a population with a noncontacting magnetic wand. b) Time sequence of images of a magnetically polarized eDIB, orientating with respect to a rotating magnetic field. The construct rotates clockwise about its central axis. The dotted line indicates the radial boundary of the magnetic and nonmagnetic hemispheres. c) Time‐sequence images of magnetically polarized eDIB orientating and migrating in three dimensions to a magnetic target beneath the Petri dish. c‐i) Orientation in the plane to face the magnet. c‐ii) Migration on the surface plane. c‐iii) Reorientation of the magnetic hemispherical face toward the magnet and submersion of the construct. c‐iv) Localization at lower surface of the Petri dish at magnetic target location. d) A population of core–shell, Janus constructs, containing magnetic microparticles in one hydrogel hemisphere, collectively migrate and spin in a rotating magnetic field. d‐i) Tracked trajectories (subset shown for clarity) highlight clockwise migration about a central point. All constructs adopt a common axial orientation, and also spin clockwise about this axis while migrating. d‐ii) The directionality of each construct is observed by the polarized hemispherical face. Colliding constructs experience a shared reduction in angular rotational velocity, temporarily before moving apart, and restoring velocity (see also Video S4, Supporting Information). Scale bars: 1 mm. 3 Discussion In summary, we have demonstrated the feasibility of a new, simple, one‐step process that uses 3D printing to fabricate monolithic, dual‐material, microfluidic devices, which are able to produce both water‐in‐oil and oil‐in‐water droplet emulsions without the need of surface modification. With this approach to device fabrication, we demonstrate the sequential operation of hydrophobic and hydrophilic droplet generating geometries to create hierarchal droplet structures, where flow rates determine the mode of operation, and, govern the number of encapsulated inner and midphase droplets. By exploiting the ability to completely fabricate manifolds in three dimensions, we are able to generate triple emulsions with 3D‐patterned morphologies. This provides the capability to produce not only Janus shell materials but also hierarchical double‐Janus (shell and midphase) droplets with independent control of the hemispherical orientations. These features can be combined with inner droplet populations of differing size, number, and hemispherical location, alongside the ability to create layered and patterned shells. These droplet templates are able to incorporate lipid bilayers, providing membranes between neighboring aqueous droplets, and between inner droplets and their contact of the midphase‐and‐hydrogel interface. This provides the opportunity to harness lipid membrane, and membrane protein biochemistry, together with biochemical compartmentalization and environmental communication. We are able to combine the presence of lipid membranes with the integration of functional organic and inorganic materials, patterned within the same hydrogel matrix shell. The hydrogel encapsulation provides mechanical rigidity and aqueous compatibility, while microfluidic manufacture provides a potential means to scale manufacture through parallelization of capillary structures. 57, 58 3D microfluidic architectures constructed from dual hydrophilic and hydrophobic materials enable the control of the landscape of hierarchal droplet construction and provides the ability to pattern the morphologies of the droplet hierarchies and break axial symmetry. This enables the creation of polarized droplet structures, which can be further used to impart additional functionality. This is demonstrated here for the first time, via the construction of hemispherically magnetic, encapsulated droplet interface bilayers, enabling controlled orientation, locomotion, and rotation. The rapid, dual‐material, fabrication process means that device designs can be rapidly iterated and customized for different desired droplet arrangements. Something not previously possible. This in effect creates the possibility of disposable devices for complex emulsion production, or even use as cartridges for droplet printing. Using 3D channel fabrication techniques, it is conceivable that nonhomogeneous, but spherically centrosymmetric, droplet hierarchies may be produced via a radial array of droplet inlet channels into the pre‐encapsulating flow. Current work is directed toward improved understanding of precise inner‐droplet arrangements to further increase design control. In the reported work, the new use hydrophilic PVA serves to demonstrate feasibility of the dual‐material approach, reporting the creation of complex emulsions using a fused‐filament 3D‐printed device for the first time. However, ultimately, the hydrophilic PVA material is nonoptimum, as it is partially soluble in aqueous media, in this case the hydrogel shell phase, over prolonged periods of use. In the immediate term, this time‐limited use may be offset by the rapidity of device manufacture. While the current commercial palette of printable polymeric materials is limited, it is rapidly growing. 80 Therefore, the prospect of further hydrophilic polymer substrate materials is promising, as 3D printing enjoys continued growth across ever widening application areas. Similarly, advances in printer resolution can be expected to facilitate the production of narrower channel architectures, thereby enabling the miniaturization of patterned emulsions. With the growing interest in 3D printing of soft materials, 81 the ability to deposit stable patterned, complex emulsions from a 3D‐printed device, as demonstrated here, may serve as the basis for the 3D printing of large‐scale materials, with complex emulsions or artificial cells as the constituent building blocks. This may find use in bioprinting applications, or in the construction of increasingly complex artificial tissues made from populations of synthetic cells, affording increased material complexity, in comparison to single‐phase droplet printing. 20 The reported spatial patterning of complex emulsions could be applied with incorporated living cells, enabling the concept of biohybrid materials, which could combine membrane‐bound droplets and cells, 82 or the integration of cells into hydrogels. 70, 83 The ability to impart polarization or directional preferences in cell‐laden constructs may find use in the self‐organization of 3D tissue engineering and repair systems. The engineering of orientation‐dictated, higher‐order assembly of artificial tissues from component cells holds significant promise. Likewise, polarizable artificial cells may take advantage of directional‐specific behaviors, analogous to phototropic or gravitropic behavior in plants, albeit by alternative mechanisms. The ability to create soft‐matter Janus materials with integrated biological components also paves the way for applied uses similar to their solid‐state counterparts, such as in sensing or displays, but with integrated biochemical capabilities. Droplets, lipid bilayers, soft‐matter compartmentalized structures and functional nano‐ and micromaterials are all the subject of great interest as building blocks of bioinspired engineered materials. The ability to create spatially organized and patterned droplet structures, comprising all these elements within the same entity, will likely serve to increase the sophistication of artificial cell models, as functional materials combining biological, organic, and inorganic material properties. The ability to impart asymmetric, polarized, or gradient patterning on such hierarchal droplets in any desired orientation, at their formation, through the use of 3D‐printed microfluidics, affords a new level of control in the design and functional utility of these materials. Biology routinely exploits the emergent functionality of the barrier properties of membranes, in combination with chemical gradients and polarization, to impart directional or symmetry‐breaking properties in cells, and enable subsequent individual and collective functionality. This serves as inspiration for the next generation of dynamic, functional, synthetic materials, built using these approaches. The data presented here demonstrate the capability to engineer chemical patterning and impart such properties in aqueous compatible, hydrogel‐encapsulated, membrane‐bound droplet networks. This holds significant promise for the advancement of behaviors and functionalities that more closely mimic the sophisticated functionality of biological cells. These patterned, hierarchal droplet materials are operational in aqueous environments, in which they remain freestanding, and retain diffusive communication between the environment and internal membrane‐bound architectures. This renders them promising candidates for applications across the life sciences, both within and outside the laboratory, such as for responsive drug delivery, self‐repairing and reconfigurable materials, and biochemical computation. 4 Experimental Section Chemicals and Components : Alginic acid sodium salt, calcium chloride, sodium chloride, silicone oil AR20, hexadecane, chloroform, Oil Blue N, PVA powder, tween 20, and span 80 were purchased from Sigma‐Aldrich. Sunflower oil (pure) was purchased from Sainsbury's. Hydrophilic, magnetic silica microparticle (SiMAG/MP‐DNA 2. 0um) was purchased from Chemicell. 1, 2‐Diphytanoyl‐sn‐glycero‐3‐phosphocholine (DPhPC) lipid was purchased from Avanti Polar Lipids. Sulforhodamine B was purchased from Sigma Aldrich. Neodymium magnets were purchased from RS Components. Large flake size of graphene oxide suspension was synthesized using a method previously reported by Rocha et al. elsewhere. 84 The concentration of GO in the water (1. 44 wt%) was calculated by weight difference after freezing and freeze drying (48 h) 4–6 g of suspension. Triple emulsion templates are shown in Table S1 (Supporting Information). The preparations of the precursor for the eDIBs formation were as in a previous work. 59 Calcium chloride solution (0. 5 m ) was prepared for the outer alginate matrix gelation. To prepare the microparticle containing alginate solutions, the original particle‐containing solution was diluted in deionized water (1% v/v for SiMAG/MP‐DNA, and up to 1. 44 wt% for GO slurry), and the alginate powder was added to prepare a 3 wt% alginate solution using a magnetic stirrer (IKA RCT basic safety control) agitated at 800 rpm and 50 °C for 4 h. The solution was stored at 4 °C. Triple Emulsion Construct Materials : Internal aqueous phase contained 0. 5 m NaCl with 100 × 10 −3 m sulforhodamine B (pink) for observation. The oil midphase for eDIB formation was formulated as hexadecane and silicone oil AR20 (1:2 v/v) with DPhPC 8. 33 mg mL −1 for droplet network production. The oil phase for non‐bilayer constructs was formulated with either sunflower oil, or mineral oil, with surfactant (see Table S2, Supporting Information). Oil Blue N was added at 0. 05 wt% for visualization, to demonstrate liquid Janus cores, visualize internal concentration gradients, and their persistence throughout and beyond the microfluidic production process. Shell phases were formulated with 3% alginate solution, containing additional 0. 5 m NaCl (for the eDIB inner alginate shell to osmotically match the internal aqueous phase droplet), or GO or silica nanoparticles (for the eDIB outer alginate shell patterning) as described in the text. The gelling bath contained 0. 5 m CaCl 2 solution. Microfluidics : Microfluidics devices were designed and modeled with COMSOL Multiphysics (the dimensions of tubular channels are in the range of 240–1200 µm diameter), and were printed using fused filament fabrication printers (Ultimaker 3) with PLA filaments and PVA filaments (Ultimaker), using 0. 4 mm AA and 0. 4 mm BB print‐cores. The print g‐codes were programmed using Cura (version 3. 3. 1) software with customized settings (Figure S1, Supporting Information). The layer height was controlled at 0. 06 mm, and the printing speed was tuned at 100 mm s −1 with default printing temperatures (200 °C for PLA filaments and 215 °C for PVA filaments). The printed parts were treated with chloroform vapor within an enclosed metal box for 8 min, and the treated pieces were left in a fume cupboard for 2 h before use. 70 Precursors were loaded in syringes (5 mL, gas‐tight, SGE Analytical Science), and were delivered to the microfluidic devices at a constant flowrate, through PEEK and PFA interconnects and FEP tubing, using syringe displacement pumps (KD Scientific, model 789200L). Details of the controlled sequential emulsification using multiphase microfluidics are given in a previous work. 22 Experiments and Measurements : General images and videos were taken using a 12MP digital camera with mounted zooming lenses on 3D‐printed stands. Epifluorescence and light microscopy images were captured using a modified Nikon Eclipse Ti‐U inverted microscope and Andor iXon ultra 897 EMCCD camera. For white light images, illumination was provided by the microscopes integrated 100 W halogen lamp, while a Shanghai Dream Lasers 532 nm DPSS laser with a power output of 100 mW was utilized for epifluorescence illumination. Laser coupling into the microscope was achieved via a custom‐built optical circuit utilizing components sourced from Thorlabs Chroma and Semrock followed by a single mode fiber‐optic launch. A low magnification 1× (Nikon Plan UW) objective was used in all acquisitions. Excitation and emission fluorescence wavelengths were separated using a 532 nm edge dichroic mirror combined with a 542 nm edge long‐pass filter and a 565–615 nm bandpass filter. Two‐color overlays were generated using the FIJI distribution of ImageJ. Particle coordinates data were collected and analyzed by Vernier Video Physics (version 3. 0. 5). Image processing was performed using ImageJ. Computational fluid dynamic modeling was done in COMSOL Multiphysics (version 5. 4), using moving mesh method. Data Analysis : See the Supporting Information for GO characterization data and methods. Droplet Rotation Analysis : The raw data of the moving droplets were evaluated and processed using IBM SPSS software. The linear relationships of the droplets' angular displacement profiles were analyzed using a Pearson product‐moment correlation test. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file. Supplemental Video 2 Click here for additional data file. Supplemental Video 3 Click here for additional data file. Supplemental Video 4 Click here for additional data file.
|
10. 1002/advs. 201901818
| 2,020
|
Advanced Science
|
Direct Conversion of Human Dermal Fibroblasts into Cardiomyocyte‐Like Cells Using CiCMC Nanogels Coupled with Cardiac Transcription Factors and a Nucleoside Drug
|
Abstract Using direct conversion technology, normal adult somatic cells can be routinely switched from their original cell type into specific differentiated cell types by inducing the expression of differentiation‐related transcription factors. In this study, normal human dermal fibroblasts (NHDFs) are directly converted into cardiomyocyte‐like cells by drug and gene delivery using carboxymethylcellulose (CMC) nanoparticles (CiCMC‐NPs). CMC‐based multifunctional nanogels containing specific cardiomyocyte‐related genes are designed and fabricated, including GATA4, MEF2C, and TBX5 (GMT). However, GMT alone is insufficient, at least in vitro, in human fibroblasts. Hence, to inhibit proliferation and to induce differentiation, 5‐azacytidine (5‐AZA) is conjugated to the hydroxyl group of CMC in CiCMC‐NPs containing GMT; in addition, the CMC is coated with polyethylenimine. It is confirmed that the CiCMC‐NPs have nanogel properties, and that they exhibit the characteristic effects of 5‐AZA and GMT. When CiCMC‐NPs‐containing 5‐AZA and GMT are introduced into NHDFs, cardiomyocyte differentiation is initiated. In the reprogrammed cells, the mature cardiac‐specific markers cardiac troponin I and α‐actinin are expressed at twofold to threefold higher levels than in NHDFs. Engineered cells transplanted into live hearts exhibit active pumping ability within 1 day. Histology and immunohistology of heart tissue confirm the presence of transplanted engineered NHDF cells at injection sites.
|
1 Introduction Heart failure, a major health problem associated with critical complications, is one of the most lethal diseases in the world. 1 When cardiomyocyte cells are damaged, they cannot be regenerated to restore the beating action of the heart. Therefore, complete heart transplantation is the only way to cure heart failure. However, this procedure is technically challenging, limited by the small number of available donors, and has severe side effects associated with the accompanying immunosuppressive therapy. To overcome these issues, cell therapy has been proposed as an alternative to transplantation. However, this approach does not yet meet the functional criteria for current clinical trials. 2 Therefore, research efforts have been devoted to differentiating somatic cells into functional cardiomyocyte cells by tissue engineering, followed by grafting of the engineered cells. 3 The direct conversion method can switch normal human dermal fibroblasts (NHDFs) to another mature cell type. Many research groups have based their efforts on a novel concept of differentiation, distinct from the induced pluripotent stem cells (iPSCs) method using the four factors identified by Yamanaka. Instead, these scientists have directly redifferentiated cells without taking them through an iPSC stage. 4 In this study, we directly reprogrammed NHDFs into cardiomyocytes using mature cells. In direct reprogramming, transcription factors and drugs are the mediators of differentiation. Therefore, suitable particles are needed to carry these agents into the target cell. 5 Nano‐sized materials have been investigated as carriers for simple and safe delivery of genes and drugs into cells. Due to their small size, these materials can easily penetrate cells without any obstructions. 6 Nanocarriers must be biocompatible with cells, tissues, and the human body as a whole. In addition, they must be safe, high in water content, easy to manufacture, and they must also have a high protein‐loading efficiency. 7 A material that satisfies all of these properties is carboxymethylcellulose (CMC), an environmentally friendly and non‐toxic saccharide. CMC is water‐soluble and has a temperature specificity, making it suitable for use as a carrier for drugs and proteins. 8 When carboxymethyl cellulose (CMC) is dissolved in water, the hydroxyl group of the glucose residues in cellulose can be substituted, making it possible to conjugate CMC to various types of drugs with cationic groups. 9 Similarly, the carboxyl groups in CMC can interact with the cationic groups of amines. In previous studies of gene delivery, our group formed complexes of a typical cationic polymer, polyethylenimine (PEI), to the carboxyl groups of CMC. 9 The “polyplexed” PEI/CMC easily complexes with genes and drugs, and can be easily transferred into human mesenchymal stem cells via endocytosis. 10 Transcription factors are directly used for the differentiation process as the research on the factors expressing in the process of growing from the embryonic state to the adult cell became active. 11 The factors that differentiate fibroblasts into cardiomyocytes include MyoD, which promotes differentiation into muscle. 12 Recently, the transcription factors GATA4, MEF2C, and TBX5 (collectively, GMT) were used to directly differentiate mouse fibroblasts into myocardial cells. 13 Initially, many studies have focused on GMT conduction alone. Induced cardiomyocytes (iCMs) are produced less efficiently by GMT, but upon addition of Hand2, the efficiency increases, and both GMT and GMTH decrease remodeling following myocardial infarction and cardiac dysfunction. 14 The efficiency of reprogramming can be further increased by adding Mesp1, Hand1, Hand2, Nkx2. 5, and Myocard (Myocd). 15 In this study, we differentiated NHDFs into iCMs using GMT, the main factors involved in inducing differentiation into cardiomyocytes. 16 In addition, due to the low efficiency of differentiation using transcription factors alone, we combined these factors with 5‐azacytidine (5‐AZA) to further promote cardiomyocyte production. 5‐AZA, an inhibitor of proliferation, demethylates DNA, thereby weakening the effects of gene‐silencing mechanisms mediated by methylation. 17 Demethylation decreases the stability of the silencing signal, resulting in gene activation. Thus, 5‐AZA is used not only to study tumor degeneration, but also to promote the differentiation of adult stem cells into cardiomyocytes. 18 In combination with GMT, 5‐AZA is injected intracellularly to induce cross‐differentiation directly into cardiomyocytes with greater efficiency. 19 We developed delivery vehicles capable of self‐assembly with DNA through formation of a complex with the cationic group of 5‐AZA and electrostatic interactions with strong cationic polymers. When these vehicles are used to introduce transcription factors and drug NHDFs, they activate differentiation into cardiomyocytes. Therefore, we transfected cardiomyocyte induced CMC‐nanoparticles (CiCMC‐NPs) into human fibroblasts and monitored the cells for evidence of direct conversion. In the process of cardiomyocyte differentiation using CiCMC‐NPs, cardiac‐like phenotypes are induced in human fibroblasts ( Scheme 1 ). For example, iCMs express multiple cardiac markers and exhibit spontaneous contractility for some period of time. According to various studies, the cardiac markers GMT are expressed when cells are differentiated from embryonic cardiomyocytes, and GMT also promotes expression of NKx2. 5. 20 In the late stage, cardiac troponin I (cTnI) and alpha‐actinin are expressed. Scheme 1 Schematic of direct reprogramming of normal human dermal fibroblasts (NHDFs) using nanogel‐type CMC complexed with DNA encoding three cardiogenic factors (GATA4, MEF2C, and TBX5) and a chemical drug (5‐azacytidine) to promote DNA demethylation and cell arrest. 2 Results and Discussion 2. 1 Synthesis of CiCMC‐NPs with the Temperature and pH Properties of Nanogels We fabricated novel nano‐sized gel‐type nanoparticles for use in direct conversion of somatic cells into myocardial cells. The newly generated nanoparticles were simultaneously loaded with a drug (5‐AZA) and complexed with genes (GMT plasmid DNA [pDNA]) that can induce differentiation into cardiomyocytes. The combination of CMC nanogels and both of these agents is important for efficient differentiation into cardiomyocyte cells, because particle size changes depending on CMC nanogel concentration and the CMC–polymer bond ratio. We confirmed that, when 0. 1% CMC nanogel and polymer were bonded at a ratio of 1:1, they formed nano‐sized gels (Figure S1, Supporting Information). New nanoparticles were formed by ionic bonding. All of the materials used in this approach were water‐soluble and had individual charges. 5‐AZA ionically bonded to the carboxyl groups of CMC nanogels via amine groups. Because GMT pDNA has a negative charge, it cannot complex directly with CMC nanogels. Therefore, we first complexed the nanogels with a cationic polymer, PEI, which has amine groups; this caused the nanogels to carry a positive charge, enabling electrostatic interactions with GMT pDNA. This interaction was confirmed by the sizes and morphologies of CiCMC‐NPs, which can be verified in detail in Figure 1 A. All bonds in the CiCMC‐NPs were ionic bonds (Figure 1 A, a). We confirmed that the CiCMC‐NPs were of nanometer scale through dynamic light scattering (DLS) and scanning electron microscopy (SEM) (Figure 1 A, b). Monitoring of size and morphology at each step during the production of CiCMC‐NPs indicated that the polymer had nanoscale dimensions starting from the time of complex formation (Figure S2, Supporting Information). When the amine groups of 5‐AZA bind to the ester groups of CMC by electrostatic attraction, the CMC is reduced in size to about 320 nm because the hydrophobic moieties are toward to the core of nanogels. In addition, when PEI with another strong amine group is complexed with residue ester groups of CMC that are not bound with 5‐AZA, it forms a nanogel about 190 nm smaller than that of CMC complexed with 5‐AZA due to its core entanglement. However, if the core‐concentrated nanogels form complexes with pDNAs, the size of the nanogels increases up to 266 nm. This means that the entangled cation groups and anion groups in the core portion were loosened and the loosened cation groups were headed outwards while they were bound to the phosphate groups of the pDNA. Figure 1 Characteristics of CiCMC‐NPs in terms of the properties of CMC. A) Schematic of CiCMC‐NPs. a) Detailed diagram of CiCMC‐NPs. b) Sizes of CiCMC‐NPs. B) Sizes and morphologies of CiCMC‐NPs as a function of temperature. a) Size, as determined by DLS and turbidity. b) Morphology, as determined by SEM. C) Sizes and morphologies of CiCMC‐NPs as a function of pH. a) Size, as determined by DLS and turbidity. b) Morphology, as determined by SEM. D) Viscosities of CiCMC‐NPs as a function of a) temperature and b) pH; c) images of the change in nanogel viscosity. Figure 1 B–D shows that CiCMC‐NPs were of nanometer scale, but the exact size changed depending on temperature and pH. Our CiCMC‐NPs were fabricated based on CMC, which is itself sensitive to both of these physical factors. 21 Therefore, we sought to determine whether CiCMC‐NPs were similar to CMC in terms of their pH and temperature dependence. To this end, we investigated how their sizes changed with temperatures and pH. CMC nanogels decreased in size as temperature increased, and CiCMC‐NPs exhibited similar trends. According to DLS measurements, CMC nanogels had dimensions of 400, 300, 300, and 250 nm at 4, 25, 37, and 48 °C, respectively, and CiCMC‐NPs had dimensions of 280, 230, 220, and 190 nm, respectively (Figure 1 B, a). Thus, CiCMC‐NPs were of optimal size for endocytosis. The morphology‐ and temperature‐dependent trends were confirmed in greater detail by SEM (Figure 1 B, b). Both CMC‐NPs and CiCMC‐NPs were sensitive to temperature and had sizes that were suitable for application to cells as the temperature increased. Other properties of nanogels include sensitivity to pH. Size changed less dramatically in response to pH than to temperature (Figure 1 C, a). CMC nanogels were 350, 300, 310, and 400 nm at pH of 4, 5. 5, 7. 4, and 8, respectively, and CiCMC‐NPs were 290, 210, 250, and 300 nm, respectively. These results show that the particles were significantly smaller at pH 5. 5, which is the pH in the cytoplasm, indicating that CiCMC‐NPs are suitable nanoparticles for in‐cell applications. In addition, we confirmed that the temperature is similar to that of CMC nanogels. Morphological analysis confirmed the size change (Figure 1 C, b). Another common feature of nanogels is their viscosity. Hence, we investigated how viscosity changed in response to temperature and pH. In both CMC nanogels and CiCMC‐NPs, the viscosity of the nanogel decreased with increasing temperature, but was not affected by pH (Figure 1 D, a, b). The reduction in viscosity at higher temperatures could be observed visually. As the temperature increased, gelation decreased and the gel sank (Figure 1 D, c). 2. 2 Identification of CiCMC‐NPs with 5‐AZA in the Cell Cycle Next, we tried to induce differentiation of NHDFs into cardiomyocytes by transfecting the cells with CiCMC‐NPs containing 5‐AZA and GMT pDNA, both of which promote differentiation into cardiomyocytes. 5‐AZA inhibits DNA methyltransferase of stem cells caused to cease the proliferation and differentiation at low concentrations. However, at high concentrations, it is known to cause cytotoxicity and cell death. Therefore, the useful concentration of 5‐AZA in the case of cardiomyogenesis has been reported to be about 10–20 µM. Based on this study, we prepared and analyzed CiCMC nanogels with 5‐AZA at various concentrations (10–200 µM). Since 10–20 µM 5‐AZA, which is a useful concentration in cardiomyocyte differentiation, forms nanoparticles of less than 200 nm, it is a suitable nano‐sized material for use as a delivery vehicle for drugs and genes to cells (Figure S2, Supporting Information). However, for CiCMC nanogel formations at different concentrations of 50, 100, and 200 µM of 5‐AZA and CMC, the results of the DLS analysis showed that the nanoparticles were widely formed over ranges from 400 to 650 nm (Figure S3, Supporting Information). Nanogels formed to a size of 200 nm or more are thought to be less effective than CMC‐nanogels formed by 10–20 µM of 5‐AZA as drug‐delivery vehicles. As a result of cell cycle arrest analysis using the CiCMC nanogels into cells, the fraction of G2/M phase formed by 20 µM 5‐AZA‐coupled nanogel was about 19. 2%, which is about a 10% increase compared to the control group. However, the 10 µM 5‐AZA‐coupled CMC‐nanogel showed an increase of about 5% over the 14. 5% fraction on G2/M. Thus, DLS and cell cycle arrest assay results show that the 20 µM 5‐AZA‐coupled CMC‐nanogels are optimal for nanoparticle size and potency of 5‐AZA (Figure S3, Supporting Information). When the CiCMC‐NPs were transfected into NHDF cells, 5‐AZA caused cell cycle arrest ( Figure 2 A). When cell cycle progression is blocked, levels of Cdk2, which regulates S phase, and cyclin D1, which regulates G1 phase, are reduced. Hence, the cells ceased proliferating, as confirmed by hematoxylin and eosin (H&E) staining. Figure 2 Characterization of 5‐AZA‐complexed CMC‐NPs. A) Schematic of the effect of 5‐AZA, including cell cycle arrest and downregulation of cell cycle proteins (cyclin D1 and Cdk2). After drug treatment, cell proliferation ceased. B, C) Cell cycle arrest. FACS revealed a high proportion of cells in G2. D) Western blots and expression levels of cell cycle proteins. E) Immunofluorescence staining. a) Confocal microscopy. b) Graph of expression levels. We then investigated in greater detail the effects of 5‐AZA in the cell. The presence of 5‐AZA in CiCMC‐NPs was determined by UV spectroscopy (Figure 2 B). At a wavelength of 270 nm, negative control (NC) and GMT pDNA with CMC‐NPs decreased the sharpness of the peak and 5‐AZA was presented. The sample group had weak absorbance, confirming that the absorbance properties varied depending on the presence of 5‐AZA. Thus, the absorbance of CiCMC‐NPs was similar to that of 5‐AZA alone, indicating that 5‐AZA was present in the particles. Next, we experimentally confirmed the effect of 5‐AZA in CiCMC‐NPs on the cell cycle (Figure 2 C–E). First, we confirmed cell cycle arrest using FACS. Compared with the NC group, cells treated with 5‐AZA‐containing CiCMC‐NPs (Figure 2 C, b, d) increased the proportion of cells in G2/M stage (black) and G1 stage (white) (Figure 2 C). When cell cycle arrest occurred, the levels of factors regulating the cell cycle also changed. Changes in Cdk2 and Cyclin D1, associated with G1/S phase, were observed at the protein level. Western blot analysis revealed that the intensities of the Cdk2 and Cyclin D1 bands (Figure 2 D, b, d) were reduced in the presence of 5‐AZA relative to the NC group (Figure 2 D, a). In a parallel experiment, protein level was confirmed using a confocal microscope: Cdk2 is shown in red, and Cyclin D1 in green; both fluorescence intensities were weaker than in NC (Figure 2 E, a). Quantitation of fluorescence intensity revealed that cyclin D1 levels were lowest in 5‐AZA‐treated cells, and that Cdk2 levels were lowest in cells treated with CiCMC‐NPs (Figure 2 E, b). Therefore, 5‐AZA was efficiently loaded into the nanogels, and when the particles were introduced into cells, they affected the cell cycle similarly to treatment with the drug alone. Thus, we confirmed that the cellular effects of 5‐AZA were preserved in CiCMC‐NPs. 2. 3 Cell Uptake of Human GATA4, TBX5, and MEF2C Expression Vectors Next, we tested CiCMC‐NPs complexed with both 5‐AZA and plasmids that induce cardiomyocyte differentiation ( Figure 3 ). GMT pDNAs were synthesized by cloning into key vectors, and then used to generate CiCMC‐NPs. Figure 3 a illustrates how GMT pDNA vectors were complexed with CiCMC‐NPs: specifically, GMT pDNA vectors were complexed with CMC‐NPs (CMC nanogels mediated with PEI). Due to the carboxyl groups in CMC, it was necessary to use materials with amine groups in order to achieve complexation with both nanogel and pDNA vectors. Because PEI has appropriate amine groups, the PEI/CMC nanogels complexed with anionic pDNA vectors, yielding complete CiCMC‐NPs (Figure 3 A, a). Gel retardation analysis revealed that the optimal vector concentration for CMC‐NP binding to pDNA vectors was 0. 1 µg. Hence, to confirm that CiCMC‐NPs were produced, GMT pDNA vectors were conjugated to CMC‐NPs at a concentration of ≥0. 1 µg (Figure 3 A, b). When we measured several types of samples, we could confirm that PEI/CMC nanogels complexed with pDNA had positive charges, whereas pDNAs and CMC had negative charges (Figure 3 B). Figure 3 Characterization of cardiogenic vectors (GATA4, MEF2C, and TBX5) in complex with CiCMC‐NPs by RT‐PCR, Western blot, FACS, and confocal microscopy. A) Zeta‐potential due to the formation of CiCMC‐NPs. B, a) Scheme of interaction between cationic CMC‐NPs and anionic GMT pDNA. C) Maps of expression vectors generated by recombinant PCR methods. D) FACS analysis of the efficiency of pEGFP‐encoding vectors. E) Transfection efficiency of NHDFs using CiCMC‐NPs, at the mRNA and protein levels, as determined by confocal microscopy. Vector: a) GATA4, b) MEF2C, and c) TBX5. Next, we cloned GMT pDNA vectors and fabricated complexes of these DNAs with CiCMC‐NPs. The expression vectors used in this study were generated by recombinant PCR methods. Briefly, human GATA4 and TBX5 cDNA clones were purchased from Dharmacon (Dharmacon, Lafayette, CO, USA), and a MEF2C clone was purchased from Korea Human Gene Bank ( https://genbank. kribb. re. kr, Korea Research Institute of Bioscience & Biotechnology, South Korea). The cDNA clones were amplified by PCR, and then ligated into the multiple cloning site (MCS) of pEGFP‐C1 followed by restriction digestion ( Bgl II/ Sal I for GATA4, Bsp EI/ Sal I for TBX5, and Bsp EI/ Bam H1 for MEF2C ). All constructs were verified by sequence analysis (Figure 3 C, a–c). Next, we tested the efficiency of introduction of GMT pDNA into NHDFs by CiCMC‐NPs (Figure 3 D, E). Because each vector also encoded EGFP, the efficiency of gene transfer could be confirmed by monitoring fluorescence by FACS or confocal microscopy. Based on these measurements, efficiency was 42% for GATA4, 27% for MEF2C, and 33% for TBX5 (Figure 3 D, a–c). The vectors were localized to the nucleus, and the CiCMC‐NPs (which had red fluorescence) were distributed around the nucleus (Figure S4, Supporting Information). In parallel, we confirmed expression of each gene at the mRNA and protein level, and found that vector efficiency did not differ significantly when CiCMC‐NPs were in complex with one vector (single, Sin) versus both vectors (multiple, Mul). Hence, the vectors were transfected into NHDFs using CiCMC‐NPs. The degree of expression from each vector was similar (Figure 3 E, a–c). When pDNA was introduced into cells using CiCMC‐NPs, the GMT vectors efficiently induced cardiomyocyte differentiation. Therefore, we conclude that CiCMC‐NPs are appropriate for use as nanoparticles to induce differentiation of fibroblasts into cardiomyocytes. 2. 4 Confirmation of Direct Conversion through Cardiogenic Markers In this study, we fabricated CiCMC‐NPs to generate a system capable of transferring genes and drugs that promote direct conversion of NHDF cells. In the experiments described above, we investigated the function and efficiency of these drugs and genes in nanoparticles. The nanoparticles were as efficient as either drugs or genes alone, indicating that they were suitable for induction of cardiomyocyte differentiation. Figure 4 A shows a simplified representation of differentiation markers. When CiCMC‐NPs were transferred to NHDF cells, GMT pDNA was expressed in the cells. Nkx2. 5 and MEF2C were expressed under the control of GATA4, as part of the core network. Expression of MEF2C, Nkx2. 5, and TBX5 induces differentiation into cardiomyocytes, in which various markers are expressed. Nkx2. 5 and MEF2C, early markers of heart development, were expressed at early times after induction. When NHDFs transfected with GMT were differentiated into induced cardiac cells (iCMs) 14 days after transfection, troponin I and α‐actinin were expressed. Over two different time periods, we confirmed that GMT‐transfected NHDF cells were differentiated into CiCMC‐NPs. Figure 4 Expression of cardiac‐specific genes after treatment with CiCMC‐NPs. A) Summary of the molecular mechanism of cardiomyocyte proliferation. B) RT‐PCR of cardiac‐specific genes, comparing early markers with late markers as a function of time. C) Immunofluorescence staining of early markers by time. D) Protein levels of MEF2C and Nkx2. 5 (%) about C: a) NC; b) CMC with 5‐AZA; c) CMC with GMT; and d) CiCMC. Figure 4 B shows mRNA expression of cardiac markers when iCM differentiation was induced by CiCMC‐NPs as well as PEI/CMC nanogels with 5‐AZA or PEI/CMC nanogels with GMT alone. Cardiac marker mRNAs were expressed in transfected NHDF cells when differentiation was induced with either CiCMC‐NPs or with 5‐AZA or GMT (Figure 4 B). CiCMC‐NPs induced the differentiation of NHDFs cells into iCMs more easily than CMCs loaded with 5‐AZA or coated with GMT pDNA. Cardiac markers were expressed at relatively high levels in both early and late stages of cardiomyocyte development. RT‐PCR analysis revealed that early markers of cardiomyocytes were strongly expressed on day 7, indicating that CiCMC‐NPs drove expression more strongly than cells induced with 5‐AZA or GMT alone. Because MEF2C was introduced via NPs, it was also slightly expressed on day 1. Expression of MEF2C was further amplified by expression of Nkx2. 5. Following induction, the late markers were strongly expressed on day 14, at a time point when expression of the early markers MEF2C and Nkx2. 5 had begun to fade away. In addition, CiCMC‐NPs drove higher levels of expression than 5‐AZA or GMT alone. This result indicates that CiCMC‐NPs are powerful tools for induction of differentiation. In addition, we monitored protein levels by immunofluorescence. The protein levels of early markers MEF2C and Nkx2. 5 mirrored the RT‐PCR results: Nkx2. 5 (green) and MEF2C (red) were highly expressed in cells transfected with CiCMC‐NPs (Figure 4 C). CMCs containing only one factor, such as 5‐AZA or GMT pDNA alone, yielded only weak expression, and ultimately did not reach the levels achieved by CiCMC‐NPs containing both factors. Thus, once again, the transfer of two factors was superior to the transfer of a single factor. Quantitative analysis of early expression markers as a function of time yielded the same results (Figure 4 D). Following transfection with CiCMC‐NPs, expression of MEF2C and Nkx2. 5 was significantly elevated on day 7. By day 14, however, MEF2C and Nkx2. 5 were almost undetectable, whereas the protein levels of late markers mirrored their mRNA levels (Figure S5, Supporting Information). Expression of early factors was higher with the dual delivery system than when only one factor was introduced, indicating that the late markers expressed when differentiated into myocardial cells were expressed compared to the positive control (PC). 2. 5 Expression of Late Cardiac Markers as a PC CiCMC‐NP‐induced iCMs expressed cardiogenic markers with the greatest efficiency among all sample groups examined. Early markers were strongly expressed on day 7, but declined by day 14, when late markers were expressed. Hence, we compared the experimental and benign models 14 days after induction of iCM differentiation ( Figure 5 ). The CiCMC‐NPs drove higher levels of marker genes than the NC, and the expression levels were similar to those in the PC. At the mRNA level, α‐actinin and cTnI were expressed more strongly in the PC than in the NC (Figure 5 A). These findings were mirrored at the protein level (Figure 5 B). Figure 5 Expression of late cardiac markers relative to positive controls. A) mRNA levels of α‐actinin and cTnI ( p < 0. 05). B) Protein levels of α‐actinin and cTnI ( p < 0. 05 and p < 0. 01). C) Immunofluorescence staining for α‐actinin and cTnI. D) Sarcomere staining to confirm myocyte morphology. a) NC; b) CiCMC; and c) PC. Immunofluorescence revealed that late markers of cardiomyocytes were expressed at similar levels in the experimental group and PC (Figure 5 C). As in previous experiments, the morphologies and marker expression patterns differed between the PC and NC. α‐Actinin (red) formed characteristic thread‐like fibers in the cytoplasm, and cTnI (green) was strongly localized to the nucleus. Cardiomyocyte cells are the muscle cells that make up the heart, and thus contain long chains of specialized organelles called sarcomeres. In this sense, cardiomyocyte cells have streak patterns similar to those of skeletal muscle cells and can be identified by immunofluorescence. To further confirm that iCMs were similar to the PC, we compared the sarcomere pattern of iCMs and PC by sarcomere dyeing. At a magnification of ×10, it was clear that the cells contained stripes, indicating that they were muscle cells containing sarcomeres (Figure 5 D). 2. 6 Evaluation of iCMs in Mouse Hearts Using Human‐Specific Staining To evaluate cardiac muscle function, we tested the effects of NHDFs transfected with genes and drug after isolation of mouse hearts and intramyocardial injection of NHDFs ( Figure 6 A). CiCMC‐NPs harboring GMT and 5‐AZA significantly increased cardiac beating relative to NHDFs treated with particles not containing either factor (Video S1, Supporting Information). The cardiac beating occurred as early as 1 h after injection, consistent with ex vivo monitoring, which revealed beating activity for 1 day. Figure 6 Ablation after injection of cardiomyocyte‐like cells engineered with CiCMC. A) Full view of the heart after CiCMC injection, Masson trichrome staining (MTS) of sectioned slides, and fluorescence staining. B) MTS and DAPI staining: a, c): NC and b, d) CiCMC. C) H&E staining: a): NC and b): CiCMC. In addition, we conducted histological analyses to verify the presence of implanted NHDFs at injection sites, and to monitor the presence of muscle in hearts (Figure 6 B). Consistent with the video observations, NHDFs transfected with genes and drug exhibited high levels of collagen deposition; this production of extracellular matrix (ECM) in the heart indicated that the transplanted cells had survived (Figure 6 B, b), in contrast to NHDFs treated without GMT genes and 5‐AZA (Figure 6 B, a). Also, the presence of transplanted cells in the heart significantly increased the deposition of collagens near the transplant sites. The transplanted cells, observed by 4′, 6‐diamidino‐2‐phenylindole (DAPI) staining (Figure 6 B, d), were clearly present at injection sites. To further verify the presence of exogenous cells, we stained the transplanted cells with H&E and observed them by confocal microscopy (Figure 6 C). The images confirmed that exogenous cells were present in isolated mouse hearts (Figure 6 C, b). They were compactly settled, whereas vacant spaces were observed in control hearts (Figure 6 C, a, b). 3 Conclusion In summary, we used nanogels complexed with drug (5‐AZA) and pDNA (GMT) to directly convert NHDFs into cardiomyocyte‐like cells. Our novel nanoparticle material could effectively deliver both pDNA and drugs into cells, resulting in successful cardiomyocyte differentiation. We anticipate that this approach will be useful in regenerative medicine. 4 Experimental Section Material CMC (MW = 90 000 and substitution degree = 0. 8–0. 9), 5‐AZA, and branched poly(ethyleneimine) (bPEI, MW = 25 000) were obtained from Sigma (Steinheim, Germany). pDNA‐encoding enhanced green fluorescent reporter protein (pEGFP) was obtained from Clontech. Dulbecco's modified Eagle's medium, high glucose (DMEM‐high), fetal bovine serum (10%), and Dulbecco's phosphate‐buffered saline (DPBS, pH 7. 4) were purchased from Hyclone. Antibiotics and trypsin–EDTA (0. 05%) were obtained from Invitrogen (Carlsbad, CA, USA). Primary antibodies against MEF2C and α‐actinin were obtained from Santa Cruz Biotechnology (Dallas, TX, USA), and primary antibodies against Nkx2. 5 and troponin I were purchased from Abcam. Secondary antibodies were obtained from Bio‐Rad Laboratories (Hercules, CA, USA). All other chemicals were of analytical grade and used without further purification. Synthesis of pDNA‐ and 5‐AZA‐Coupled CMC Nanogels CMC was dissolved in filtered water. bPEI (MW = 25 000; Sigma) was dissolved in filtered water (50 mg per 10 mL). 5‐AZA was dissolved in filtered water (100 mg per 1 mL) and passed through a 0. 2 µm filter. CMC (0. 5 µg, 0. 1% solution) was mixed with 5‐AZA (20 µM) and vortexed, and then the CMC/5‐AZA solution was mixed with bPEI (0. 5 µg) at a 1:1 ratio and vortexed again. Cardiogenic vectors‐encoding GMT were complexed with CMC nanogels coupled to 5‐AZA. Each GMT vector also encoded EGFP. An NLS‐EGFP expression plasmid (pEGFP) was generated by ligating the EGFP open reading frame derived from pEGFP‐N3 into pcDNA3. 1 (Invitrogen) followed by insertion of the SV40 NLS into the N‐terminus of green fluorescent protein (EGFP). All plasmid constructs were verified by DNA sequencing. Characterization of CiCMC‐NPs Sizes of CiCMC‐NPs were measured using a Zetasizer Nano ZS (Malvern, Southborough, MA, USA). Briefly, the nanogels were suspended in deionized water, and mean hydrodynamic diameter was determined by accumulation analysis. The zeta‐potential values were predicted based on the electrophoretic mobility of bPEI‐modified CMC nanogels in deionized water, which was evaluated using folded capillary cells in automatic mode. The viscosities of the various CiCMC nanogels were measured on a Brookfield Viscometer DV‐III Ultra (Brookfield Engineering, Middleboro, MA, USA) equipped with a programmable rheometer and circulating baths with a programmable controller (TC‐502P, Brookfield Engineering). The T‐F spindle was set to rotate at 0. 2 rpm over a temperature range of 4, 25, 37, and 48 °C. To analysis of hydrogen ion by proton that was measured the hydrogen ion concentration range of pH 4, 5. 5, 7. 4, and 8. Analysis of Cell Cycle Arrest by Flow Cytometry NHDFs were seeded in six‐well plates at 1. 5 × 10 6 cells per well and cultured at 37 °C and CO 2 (5%), after which they were rinsed twice and preincubated for 30 min with serum‐free DMEM‐high (2 mL) at 37 °C. CiCMC‐NPs were added to the cells and incubated for 4 h at 37 °C. After incubation, the NHDFs were washed twice with DPBS (1 mL) and treated with trypsin–EDTA (0. 05%) for 5 min. The cells were resuspended in bovine serum albumin (BSA, 1 mL, 0. 1%) and centrifuged at 13 000 rpm for 3 min. The cells were fixed for 1 h in ethanol (2 mL, 70%) on ice, after which they were centrifuged to removed supernatant. After three washes in DPBS, the cells were suspended in propidium iodide staining solution (1 mL, 10 µg mL –1, Molecular Probes) and RNase A (100 µg mL –1 ) in DPBS and incubated at 37 °C for 10 min. Samples were transferred to 96‐well plates, and fluorescence was measured by flow cytometry. Evaluation of Transfection Efficiency with Cardiogenic Transcription Factors NHDFs were seeded in six‐well plates (3 × 10 5 cells per well) and cultured at 37 °C and CO 2 (5%), after which they were rinsed twice and preincubated for 1 h with DMEM‐high medium (2 mL) at 37 °C. CiCMC‐NPs were added to the cells and incubated for 4 h at 37 °C. The NHDFs were then washed three times with PBS (1 mL) to remove any free gene complexes, suspended in PBS, and incubated for an additional 24 h. To determine transfection efficiency, the cells were harvested and analyzed on a flow cytometer (Guava Technologies, Hayward, CA, USA) equipped with a 488/642 nm laser. Data represent mean fluorescence signals from 5000 cells. For confocal microscopy, the cells were fixed with paraformaldehyde (4%). The cells were mounted in mounting medium (DakoCytomation, Hamburg, Germany) and visualized using a confocal laser scanning microscope (LSM 880 Meta; Zeiss, Germany). Fluorescence was monitored in the DAPI (excitation, 358 nm; emission, 461 nm), EGFP (excitation, 488 nm; emission, 530 nm), and CiCMC‐NP channels (excitation, 610 nm; emission, 655 nm). For Western blotting, the cells were rinsed twice in DPBS and resuspended in RIPA buffer (50 µL). Following electrophoresis on 10% (w/v) acrylamide SDS–PAGE gels, resolved proteins were transferred to membranes using a wet system. Membranes were incubated for 4 h at room temperature (RT) with primary antibodies diluted in blocking solution (cTnI, 1:500, Abcam; α‐actinin, 1:500, Santa Cruz Biotechnology), and then incubated for 1 h at RT with secondary antibodies. Binding was visualized using the Amersham ECL reagent (GE Healthcare, Pittsburgh, PA, USA), and signals were recorded on X‐ray film. Protein bands were quantified using the Image Lab software (version 4. 0, Bio‐Rad) and normalized against β‐actin, used as a loading control. Total RNA was extracted using TRIzol (Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription was performed using synthesized cDNAs. Marker expression was confirmed by RT‐PCR using primers for GMT. RT‐PCR products were run on agarose gels (1. 5%) and stained with ethidium bromide. Immunofluorescence Analysis iCMs were incubated for 4 h at RT with primary antibodies against cTnI (1:100, Abcam) and α‐actinin (1:500, Santa Cruz Biotechnology). Antibodies conjugated with Alexa Fluor 488 and Alexa Fluor 555 (Thermo Fisher Scientific) were used as secondary antibodies for cTnI or α‐actinin, respectively. Sections were mounted in mounting medium (DakoCytomation) and visualized on an LSM 880 confocal laser scanning microscope. Fluorescence was monitored in the Alexa Fluor 555 (excitation, 555 nm; emission, 565 nm), Alexa Fluor 488 (excitation, 495 nm; emission, 519 nm), and DAPI channels (excitation, 358 nm; emission, 461 nm). Injection in Mouse Heart The animal study was approved by the Institutional Animal Care and Use Committee (IACUC) of CHA (approval number: IACUC180104). After ICR mice (7‐week‐old, male; Orient) were prepared for surgery, the hair was removed, and the chest was cut to allow the heart to be pulled through the ribs. The prepared iCMs and nanogel were mixed 1:1 and injected through a syringe into the heart muscle, and the heart was collected 3 days later. Evaluation of iCMs in Mouse Heart Collected hearts were fixed for at least 1 day. The fixed organs were dehydrated and paraffin‐embedded. For observation of morphology, the samples were treated with Masson's trichrome (Sigma), Harris hematoxylin (Merck) and eosin, and nuclear stain. Statistical Analysis Data are representative results or the means of at least three independent experiments, ± SD. Statistical analyses were performed using two‐tailed Student's t ‐test. Differences were considered significant at p ≤ 0. 05, and p ‐values are shown in figures as needed. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file.
|
10. 1002/advs. 201901878
| 2,020
|
Advanced Science
|
A 3D‐Printed Hybrid Nasal Cartilage with Functional Electronic Olfaction
|
Abstract Advances in biomanufacturing techniques have opened the doors to recapitulate human sensory organs such as the nose and ear in vitro with adequate levels of functionality. Such advancements have enabled simultaneous targeting of two challenges in engineered sensory organs, especially the nose: i) mechanically robust reconstruction of the nasal cartilage with high precision and ii) replication of the nose functionality: odor perception. Hybrid nasal organs can be equipped with remarkable capabilities such as augmented olfactory perception. Herein, a proof‐of‐concept for an odor‐perceptive nose‐like hybrid, which is composed of a mechanically robust cartilage‐like construct and a biocompatible biosensing platform, is proposed. Specifically, 3D cartilage‐like tissue constructs are created by multi‐material 3D bioprinting using mechanically tunable chondrocyte‐laden bioinks. In addition, by optimizing the composition of stiff and soft bioinks in macro‐scale printed constructs, the competence of this system in providing improved viability and recapitulation of chondrocyte cell behavior in mechanically robust 3D constructs is demonstrated. Furthermore, the engineered cartilage‐like tissue construct is integrated with an electrochemical biosensing system to bring functional olfactory sensations toward multiple specific airway disease biomarkers, explosives, and toxins under biocompatible conditions. Proposed hybrid constructs can lay the groundwork for functional bionic interfaces and humanoid cyborgs.
|
1 Introduction The rapid growth of the biomedical engineering field in the past few decades has enabled the emergence of bionic organs, a highly serviceable asset for replicating specific organ functions and increasing the accuracy of in vitro test models, with the aim to eventually replace native organs. To create bionic organs, high‐precision and functional electronic devices need to be integrated into viable engineered biological systems. While the two stem from fully disparate classes of materials and fabrication techniques, recent research has been focused on closing this gap by introducing techniques and protocols optimized for co‐fabrication of biocompatible and implantable bionic organs. 1 The current state of development of bionic organs is still impeded by several challenges, particularly in mimicking the complexity of tissue or organ structure and functionality. The intricate heterogeneity in biological structures and the physical properties of human organs has intrigued a need for multi‐material and multi‐cellular hybrid designs. Various studies have introduced a well‐established protocol to fabricate macro‐scale tissues and organ‐like constructs with resemblance to the biological structure and physical properties of the target tissue or organ. 2 However, tissue‐specific functionalities such as sensory and auditory capabilities, olfaction, and vision are often chiefly missing from the replicas. Among the various sensory organs in the body, the nose has a great potential to be targeted for bionic organ engineering, as both the tissue and function can be mimicked through tissue engineering techniques and odor‐detecting electronic devices. The nasal cartilage tissue is majorly composed of hyaline cartilage, which consists of densely packed collagen and proteoglycan‐based extracellular matrix (ECM) embedded with chondrocytes. 3 Moreover, it has a relatively simple structure along with mechanically robust and elastic properties compared to other tissues in the body. 4 These attributes facilitate engineering 3D nasal cartilage tissues using a combination of biomaterials and microfabrication techniques. On the other hand, the sensing and olfactory system in this organ is endowed with intricate and unique physical and biological properties. With the ability to discriminate thousands of volatile compounds and chemical structures, the olfactory system plays an important role in assisting humans with perception of the outer environment. Through studying the chemical structure of the olfactory receptors (ORs) and their function, the specific binding of each of these proteins to odorant compounds with similar chemical structures can be achieved, paving the way for engineering an odor‐perceptible artificial nose. Two distinct research communities have dedicated their efforts to nose organ reconstruction. In the first group, where the focus is on emulating the function of the nose via electronic nose sensor devices, the biological features and physical functions of a native nose are neglected. 5 As such, interfacing the electronic sensing system with the tissue is also ignored. The electronic noses often consist of a sensing unit, which is immobilized with odor‐specific receptors which capture the target molecule and transduce the binding event to an electrical signal. Sensing mechanisms such as field effect transistors (FETs), quartz crystal microbalance, and electrochemical (EC) sensor units have proved successful. 6 Electronic noses have been developed for a range of applications to detect toxic gases, 7 food quality monitoring 8 as well as being used as a diagnostic tool for diseases such as pneumonia 9 and lung cancer. 10 An obstructive factor in integrating the commonly developed sensor devices with tissues is pertained to the harsh conditions in which the sensor devices operate. Moreover, selection of toxic chemicals (e. g. , specific electrolytes) and electrode materials to improve the sensing capability can create a noncytocompatible environment for most tissues. 11 The second community, on the other hand, has focused merely on modification or reconstruction of the nasal anatomy for cosmetic purposes, or for after the organ has been physically damaged (e. g. , those caused by traumatic accidents). Given the structural features and replication feasibilities of the nasal organ, engineering and development of cartilage‐mimicking tissues has proved practically successful in several studies. 12 Additionally, some groups in this category have targeted cartilage regeneration and tissue repair. [qv: 12c, 13] The elaborate structure of the nasal passageway plays an integral part in conditioning air for olfaction and respiration. This conditioning includes precise controlling of the gaseous fluidic mechanics of the inhaled air and adjustments to humidity and temperature, and filtering and air flow characteristics such as flow rate, intermittence, and regime (laminar, transitional, or turbulent). 14 Moreover, this structure allows the airflow velocity to be decreased near the olfactory epithelium, facilitating full contact between gas molecules and the lined surface of the olfactory epithelium thus increasing the sensitivity of odor detection. [qv: 14c, 15] To achieve enhanced artificial olfaction with close relevance to the native structure, such anatomical elaborations should be integrated into the bioelectronic noses. Here, we aim to bridge this gap between the two communities to create an effective biocompatible functional hybrid nose‐like tissue construct. Previously, a 3D printing–based approach was pursued to bring auditory functionality to auricular cartilage via developing a 3D‐printed bionic ear. 16 Following a similar strategy, we are targeting the retention of olfaction to develop a hybrid nasal tissue. To the best of our knowledge, there have not been any attempts toward the creation of a hybrid nasal tissue with the capability of odor sensing and olfactory system retention. Specifically, we have developed a chondrocyte‐laden 3D‐bioprinted cartilage‐like construct with an electronic olfaction‐mimetic biosensor. Employing 3D bioprinting technology is especially beneficial in this approach as it allows for the creation of 3D hybrid platforms with desired geometries and high precision. While 3D printing parameters can be readily engineered to fit the cellular environment requirements, the structure can be coupled with an electronic functional element to create a viable hybrid device. 16, 17 Moreover, photo‐crosslinkable hydrogel‐based bioinks composed of gelatin methacryloyl (GelMA) and polyethylene glycol dimethacrylate (PEGDMA) can form a mechanically robust and biocompatible 3D microenvironment to support cartilage cell growth and their differentiation by tuning the mechanical properties of the bioinks. As such, two distinct bioinks with stiff and soft mechanical properties composed of different hydrogel concentrations could be optimized to mimic the mechanical properties of the native ECM of nasal cartilage to allow for nasal cartilage tissue formation. To integrate olfaction (biosensing based on OR immobilization) and odor sensing (e. g. , biosensing using peptides and aptamers) into the 3D‐printed cartilage tissue constructs while preserving biological culture conditions, EC‐based biosensing presents an applicable choice with advantages such as high sensitivity, label‐free detection, a wide linear detection range, and excellent detection limit. Additionally, EC‐based biosensors possess flexibility in detection ability; for instance, the sensor can be conveniently functionalized with odorant receptor proteins to detect a wide range of human‐detectable odors and chemicals. Furthermore, scalability and ease of miniaturization of the EC‐based biosensors on various substrates (e. g. , glass, paper, flexible polymeric substrates, etc. ) facilitate their integration into live tissues and organs. In this paper, the biosensing mechanism is based on label‐free EC impedance spectroscopy (EIS) to further improve sensitivity and circumvent acute cytotoxicity issues induced through labeling odorant receptors with electrochemically active specific molecules. Nonetheless, common electrolytes, such as potassium ferricyanide (K 3 Fe(CN) 6 ), are often toxic to cellular environments and affect cell‐laden system viability. To operate the biosensor under biological culture conditions, we employ a culture media‐based electrolyte system with a weak and nearly negligible EC redox activity, lower conductivity compared to K 3 Fe(CN) 6, but adequate sensitivity to pico‐level concentrations of the analyte. 18 Developing such cytocompatible biosensors can open a pathway toward augmented sensing in hybrid organs; for instance, it is possible to functionalize the system with airborne pathogenic biomarkers, 19 volatile disease metabolites, 20 or odorless contaminants, 21 and thus achieve a cyborg olfactory organ. Such a bionic nose can prove useful in applications such as the detection of explosives, illegal drugs, or food poisoning. 2 Results and Discussion 2. 1 Mechanically Tunable Dual Bioink System The dimensions of common self‐standing cell‐laden 3D‐printed constructs often do not exceed few centimeters. Several studies that created constructs beyond a few centimeters tall faced problems, such as losing the precision in geometry as short as a few minutes after printing, due to the swelling or low stiffness of the bioink. 22 To overcome this challenge, one solution is to increase the viscosity or stiffness of the bioink to enhance the printability and reduce swelling. 23 However, in many cases, increasing the bioink stiffness results in reduced cell viability and proliferation imposed by the physical constraints of the now stiffer matrix. 24 Consequently, the trade‐off between geometry precision and cell behaviors remains an ongoing challenge. To overcome this challenge, two distinct bioinks with stiff and soft mechanical properties were used to form a mechanically robust and biocompatible 3D‐printed construct to support cartilage cell growth and differentiation within desired geometries. To enable two distinct mechanical properties within a single printed construct, we employed a multi‐material 3D bioprinting technique to create a self‐standing 3D hybrid platform with desired stiffness and geometries (>1. 5 cm height of printed constructs). Figure 1 shows the procedure for fabrication of the hybrid nasal cartilage using a dual nozzle printing system. In the current study, a combination of GelMA and PEGDMA was employed to create a viable scaffold supportive of cartilage regeneration while being capable of maintaining the geometry of the structure even weeks after culture. High hydrophilicity, biocompatibility, and considerably low immunogenicity are some of the features of the mentioned polymers. 25 Owing to the reversible thermal gelation of GelMA pre‐polymer solutions, the viscosity of the bioinks can be further reinforced prior to printing through physical gelation. Post‐printing chemical crosslinking can be achieved through photo‐crosslinking with UV light. Figure 1 Schematic diagram showing the procedure of printing the nose. The soft ink including the chondrocyte‐laden solution and the stiff ink are both loaded in the printer. Afterward, dual printing is performed on top of the microfabricated sensor functionalized with a TNT‐specific peptide. This hybrid system can be tuned to detect a range of targets such as natural odors, airborne pathogens, and odorless explosives. Two bioinks consisting of different compositions of PEGDMA and GelMA were selected and denoted as “soft” and “stiff, ” with the latter containing higher concentrations of each material. The former promotes cell adhesion and creates a suitable stiffness for chondrocyte growth while the latter creates a structural scaffold to retain the shape and higher stiffness for mechanical strength. Following pre‐polymer solution preparation, the soft ink was mixed with the cells to create a cell‐laden bioink, while the stiff ink was prepared and printed in the original acellular conditions. Addition of sacrificial biomaterials, such as gelatin, to the pre‐polymer solution of both inks allowed for improved printability and a geometrically defined structure post‐printing. Later the gelatin was dissolved from printed constructs during cell culture processes. To develop two different soft and stiff bioinks, we first tested the printability of bioinks with various combinations of gelatin, GelMA, and PEGDMA, reaching good printability. At the same time, we fabricated bulk hydrogels and then measured the mechanical properties (data not shown). We then selected two different compositions of soft and stiff bioinks: 5% GelMA/5% PEGDMA/6. 5% gelatin and 8% GelMA/10% PEGDMA/3% gelatin, respectively. We confirmed two distinct levels of mechanical stiffness for crosslinked soft and stiff bulk hydrogels as 20. 3 ± 1. 1 and 67. 3 ± 1. 1 kPa, respectively (Figure S1a, Supporting Information). The fracture point of stiff bulk gel was found to be 125 kPa at 50% strain, five times higher than the soft bulk gel. In addition, the microstructures of soft and stiff crosslinked bioinks were observed by scanning electron microscopy (SEM), and were found to be micro‐porous structures. However, while the soft gel tended to easily collapse during the freeze‐drying process due to its weak mechanical stiffness, the stiff gel was able to maintain its micro‐porosity (Figure S1b, c, Supporting Information). To understand and characterize the temperature‐sensitive and fluidic behavior of each ink, the physical properties of the pre‐polymer solution of soft and stiff bioinks were measured. Due to the temperature‐dependent sensitivity of the gelatin and GelMA pre‐polymer solutions, the viscosity of the inks was directly influenced by temperature, resulting in the creation of straight fibers and a more precise geometry under 19 °C and highly reduced printability above 27 °C ( Figure 2 a). Therefore, the incubation time for each ink required careful optimization to reach suitable printable viscosity between 17 and 22 °C. Frequency sweep measurements showed that both soft and stiff bioinks possessed appropriate rheological properties to be used in bioprinting. The higher storage modulus ( G ' > G ”) across the range of frequencies studied indicated gel‐like behavior, and therefore, the bioinks' suitability to hold the shape of the bioprinted structures, especially for the stiff ink, in which G ' was nearly sevenfold greater than that of the soft ink (Figure 2 b). Moreover, both inks presented a non‐Newtonian shear‐thinning behavior, with the power law index ( n ) being < 1. Such an index value is favorable for bioprinting due to the capacity of the material in behaving as a fluid under high shear stress, allowing a smoother nozzle extrusion, and behaving as a rigid body under low shear, after bioprinting (Figure 2 c and Table S1, Supporting Information). Figure 2 Rheological properties of the soft and stiff inks. a) Temperature‐dependent behavior of soft and stiff inks affecting the storage ( G ') and loss ( G ”) moduli, thus printability of each ink ( N = 3). b) Storage ( G ') and loss ( G ”) moduli of soft and stiff inks across a frequency sweep ( N = 3). c) Change of viscosity with increments of shear stress for both inks showing the shear thinning behavior of the inks ( N = 3). d) Fluorescence imaging of dyed extruded fibers using the soft (red) and stiff (green) bioinks across a variation of nozzle printing speeds. The soft ink creates thicker fibers due to lower viscosity. e) Characterization of fiber diameter by nozzle printing speed for the extruded fibers shown in (d). The extruded fiber diameters ranged from 200 to 480 µm for the stiff ink and 550 to 1200 µm for the soft ink with lower viscosity ( N = 4). f) Printability of the GelMA and gelatin optimized for printing. Due to the lower viscosity of the soft ink, the extruded fiber was thicker compared to the stiff ink at the same printing speed. By adjusting the moving speed of the nozzle, the diameter of the extruded lines was optimized to reach similar printability in both inks (Figure 2 d, e). Speeds higher than 24 mm s −1 resulted in breakage in lines and thus loss of printability. Therefore, we chose 18 mm s −1 for the soft ink nozzle and 6 mm s −1 to print the stiff layers. The air pump pressure was kept constant around 30 kPa for both inks. The diameter of the extruded fiber using stiff ink ranged from 200 to 480 µm while the soft ink with lower viscosity exhibited a higher extruded fiber diameter range (≈550 to ≈1200 µm). As such, the diameters of the extruded fibers in the mentioned optimized nozzle speeds were found to be around 800 µm for the soft layers and 400 µm for the stiff layers. Moreover, by adjusting the gelatin to GelMA ratio in ink compositions, the printability of the inks was optimized to create continuous fibers without needle blockage (Figure 2 f). Since the cell‐laden layers require delivery of oxygen and nutrients through the thick printed construct, the permeability of the constructs had to be adjusted to allow for the circulation of nutrients and oxygen. We modified the printing parameters such as 3D slicing and infill ratio of the geometry to create a macro‐porous structure without compromising the shape integrity after printing and during culture. Using a 15–20% infill ratio in slicing the 3D model into layers met both criteria. After optimizing the printing parameters, the PEGDMA and GelMA pre‐polymer chains in the printed constructs were crosslinked and solidified upon UV radiation. Figure 3 a depicts the crosslinking procedure of the bioinks. Moreover, the formation of covalent bonds between the GelMA and PEGDMA among the layers helped to avoid any lamination issues between the stiff and soft printed layers despite their dissimilar mechanical properties. As a result, a multi‐layered cube composed of soft (red) and stiff (green) printed gel layers could be obtained (Figure 3 b). Comparing the 3D computer‐aided design (CAD) model dimensions with those of the 3D‐printed product, we observed a 21% increase in the area accompanied with a 3% reduction in height before swelling the printed construct in the biological media. Assuming that both the stiff/soft hydrogels of the printed construct swelled in all three dimensions and the effects of swelling and collapsing could be superposed, we can conclude that the original collapse of the gel was around 18% of the original height. This reduction could be pertained to a greater collapse in the soft layers than that of the stiff layers as a result of low stiffness and less crosslinking density of the soft hydrogel. Also, gravitational forces of thick whole constructs might lead to the closure of internal pores in the soft hydrogel. 26 We then tested the printed constructs with various ratios of soft and stiff printed layers in terms of structural integrity and mechanical properties (Figure 3 c–f). In terms of the effect of the printing process on the mechanical properties, the existence of print infill relatively reduced the stiffness of fully stiff (0:1) and fully soft (1:0) printed constructs with ≈57 and ≈3 kPa Young's modulus, respectively (Figure 3 f) compared to the nonprinted bulk crosslinked gels (≈20 kPa, soft and ≈67 kPa, stiff; Figure 3 g and Figure S1a, Supporting Information). This reduction happened due to the creation of larger cavities within the microgrid structure as well as the propagation of cracks on the edge of the printed lines which caused breakage. Due to the presence of infill and patterned cavities in all printed constructs, the Young's moduli of the printed constructs were not directly comparable with those of bulk gels. Therefore, an “infill factor” was introduced to all obtained Young's moduli to normalize the printed constructs by the nominal print area, actual area, and weight of the printed samples. More specifically, fully stiff (0:1) and fully soft (1:0) printed constructs were normalized by their bulk gel counterparts through calculating the approximate density of each construct using the measured weight and dimensions. Next, two infill factors were obtained for the fully soft and fully stiff hydrogels (i. e. , soft infill and stiff infill). Composite infill factors for 1:1, 1:2, and 2:1 were introduced by composing different ratios of fully stiff and fully soft infill factors (e. g. , 1:1 infill was made with (1X soft infill+1X stiff infill)/2). Figure 3 a) Schematic of each ink composition after the UV crosslinking procedure and 37 °C incubation. The dots represent crosslinked sites upon UV exposure. The stiff inks include more crosslinked sites due to higher GelMA concentration. After incubation at 37 °C, the gelatin is dissolved, leaving the construct with a porous structure. b) A cube printed with soft (red) and stiff (green) inks (1:1) and crosslinked at 150 mW cm −2. c) Different ratios of soft (red) to stiff (green) were selected and tested. (2:0) and (0:2) represent the softest and stiffest constructs, respectively. d, e) Stress–strain curve of the different print ratio composites depicted in (c). f) Young's modulus of the printed composite constructs ( N = 4). g) Young's modulus for soft and stiff bulk gels (nonprinted). h) Percentage of weight loss in the hydrogel samples during the 17 day culture. The UV crosslinking intensity was optimized to reduce the degradation rate and 300 mW cm −2 represents the least degradation. i) One of the cubes crosslinked at 100 mW cm −2 before (top) and after (bottom) degradation within the 17 day culture. Scale bar is 5 mm. j) SEM imaging of printed soft (i, ii) and stiff (iii, iv) gels. Red line indicates the printed fiber edge. Scale bar is 200 µm. By increasing the soft:stiff ratio, or in other words, increasing the number of stiff layers, higher geometrical accuracy and mechanical properties can be obtained. For instance, fully stiff (0:1) hydrogels showed a significant difference in Young's modulus compared to the composite hydrogels (1:0, 1:1, and 1:2). This tunability can be especially useful for 3D‐bioprinted constructs with complex printing patterns and nozzle movements in all X‐, Y‐, and Z ‐directions, namely, those in a 3D nose model. However, increasing the overall stiffness of the constructs can inhibit the growth and ECM secretion of the cells. In this case, using higher soft:stiff ratios would provide a more appropriate substrate for chondrocyte growth. Mechanical compression tests on the samples with different soft:stiff ratios proved no significant difference between the Young's modulus of 1:0, 1:1, and 1:2 ratios. However, the Young's modulus increased significantly between 2:1 and 1:2 ratios. As a result, the (1:2) ratio was selected to create a composite construct with structural integrity while at the same time providing mechanical properties comparable to those of the ECM required for chondrocyte growth and cartilage development. Although the mechanical stiffness of the native mature nasal cartilage matrix was reported as 100–400 kPa, 27 studies have shown greater chondrocyte maturation and high cartilage matrix diffusion in engineered tissues with lower mechanical properties (10–30 kPa) compared to the native cartilage. 16, 28 For instance, a recent study by Levett et al. showed that GelMA–hyaluronic acid (HA)–chondroitin sulfate (CS) scaffolds with mechanical properties ranging from 20 to 40 kPa can induce chondrogenic differentiation and high production of cartilage ECM including collagen II (COLII) and glycosaminoglycan (GAG) content. Moreover, embedding chondrocytes in the composite GelMA‐HA‐CS scaffolds led to a significant increase in mechanical properties up to 150 kPa after 8 weeks of culture. It is hypothesized that optimizing initial conditions such as large pore sizes and proper initial mechanical properties can lead to a large matrix secretion and higher mechanical properties in the engineered cartilage‐like tissues over longer culture periods. As such, it is suggested that initially softer constructs can transform to the stiffest ECM over time more effectively than initially stiff hydrogels, since the latter can impede the formation and dispersion of the new ECM produced by the cells. 29 Nonetheless, the degradation test of the printed construct with 1:2 ratio confirmed that structural integrity was preserved even after 17 days of culture while the final stiffness was enough to provide a suitable growing environment for the cells (Figure 3 h). The percentage weight loss of the GelMA‐PEG‐gelatin acellular constructs was caused by both dissolving of gelatin and degradation of GelMA chains at 37 °C upon incubation. GelMA polypeptide chains were chemically crosslinked via conjugated methacrylate groups, however, most of the peptide bonds in the GelMA polypeptide chains could be broken by hydrolysis from salt ions existing in a water‐based solvent (ea. biological media) combined with an elevated temperature of around 40 °C. 30 In terms of cell‐laden structures, the protein structure of the GelMA hydrogels permits cells to enzymatically degrade and remodel the gel for cell spreading and expanding in the degraded spaces. 31 Moreover, by tuning the UV intensity, the printed construct could obtain a tunable range of degradation properties as well as maintain its shape integrity to minimize weight loss (Figure S2, Supporting Information). Increasing the crosslinking intensity resulted in a higher elastic modulus, however, exceeding 150 mW cm −2 led to the reduction of cell viability due to the release of free radicals as well as possible damages and mutations in the cell DNA (Figure S3, Supporting Information). The optimal UV intensity for simultaneous targeting of structural integrity and high cell viability was found to be 100 mW cm −2 with an exposure time of 80 s at an 8 cm distance from the light source. Figure 3 i shows the degraded hydrogel constructs crosslinked at 100 mW cm −2 on day 1 and day 17 of culture at 37 °C. By incubating the structure at 37 °C after UV crosslinking, the gelatin was removed from the structure and dissolved in the culture media. As a result, a highly microporous structure was achieved, as confirmed by SEM imaging (Figure 3 j). Interestingly, printed hydrogels maintained a relatively higher porous structure compared to bulk crosslinked gels (Figure S1b, Supporting Information). This event can be pertained to the local chemical and physical crosslinking of the extruded fibers as exposed to the printer halogen light and lower temperature of the bed, thus creating a structurally robust construct. As seen in Figure 3 j‐i, ii, the printed soft layers maintained a highly porous structure with larger pore sizes compared to the stiff layers (Figure 3 j‐iii, iv, thus promoting the perfusability of the constructs. 2. 2 A Supportive Microenvironment for Chondrocyte Culture and Growth After optimizing printing parameters, cell‐laden bioinks were characterized to study the feasibility of cell growth and functionality in the multi‐material printed structure. Chondrocytes, a representative cell type composing the nasal cartilage, were cultured and encapsulated in the soft bioink for cell‐laden bioprinting. A major challenge in expanding the primary chondrocytes in conventional 2D monolayer dishes has been the loss of chondrogenic phenotypes and the occurrence of “dedifferentiation” after a few passages. Dedifferentiation of chondrocytes is often followed by the gradual decay of chondrocyte‐specific molecular markers such as GAGs, aggrecan (ACAN), and COLII. 32 At this stage, chondrocytes would lose the original rounded morphology and tend to a more extended fibroblast‐like morphology. 33 Several studies have proven that “re‐differentiation” can be achieved by embedding and culturing the chondrocytes in high‐density 3D gel suspensions such as agarose, 34 alginate beads, 35 or 3D collagen‐based matrices. 36 Moreover, the chondrocyte markers such as COLII deposition are shown to be restored through 3D culture. In the present study, the printing and culture of the encapsulated chondrocytes in the GelMA‐based soft bioink led to the maintenance of a typical rounded morphology in the chondrocytes (Figure S4, Supporting Information). The presence of RGD (Arg‐Gly‐Asp) binding sites in the GelMA hydrogel was previously shown to enable high cellular adhesion. 37 The 3D‐printed constructs composed of cell‐laden soft and stiff printed layers can be a suitable scaffold for seeding and encapsulation of chondrocytes to improve chondrogenic behavior in 3D and to promote the deposition of cartilage ECM. We first encapsulated the chondrocytes in the soft bioink and printed the multi‐material construct using the cell‐laden soft ink and acellular stiff ink in a dual nozzle printing process ( Figure 4 a). Next, the cell‐laden constructs were allowed to stabilize and were cultured in vitro for 2 days. To enhance the cell viability and increase the cell density throughout the macroscale construct, additional amounts of chondrocytes were seeded on the printed constructs on day 2 of culture post printing. Seeding the chondrocytes on the printed constructs also allowed for higher cell signaling and interconnectivity across the construct. The printed constructs cultured with encapsulated cells sustained a high cell viability and adhesion to the biomaterial even after 30 days of culture in vitro (Figure 4 b–d). The viability of encapsulated chondrocytes in the soft bioink was found to be approximately 80% on day 1 due to the loss of a fraction of cells to shear stress and the lack of cell culture conditions (Figure 4 d). However, the cells recovered within 7 days of printing to an average viability of 95%. To perform the viability assay such as to be able to compare the values on day 1 and day 7 regardless of the effect of the seeding process, no cells were seeded in these constructs. Moreover, the encapsulated cells in the innermost layers of the constructs showed original rounded chondrocyte morphology (Figure S4a, b, d, Supporting Information) as opposed to the highly elongated morphology often observed in 2D cultures (Figure S4c, Supporting Information). Moving from the innermost to the outer layers of the constructs, the encapsulated cells gradually showed a relatively elongated morphology with a higher area of phalloidin (F‐actin) expression (Figure 4 b, c). In terms of seeded cells on the printed constructs, Figure 4 e, f shows the 3D reconstructed confocal images of cell morphologies cultured on the thick printed construct after 30 days. Accordingly, the cells covered the 3D construct along the printed fibers, creating a fully cellular macro‐scale viable construct. Moreover, the metabolic activity study of the cell‐laden printed and seeded constructs exhibited an increase over the course of 14 days (Figure 4 g). The constructs were re‐plated after seeding and during culture to avoid media consumption by the excess cells migrating to the outside environment of the constructs and those attaching to the bottom of the plate. Therefore, the results of the metabolic assay were recorded merely based on the cells inside or on the constructs. Supplementing the chondrocytes with L‐ascorbic acid can promote COLII and GAGs deposition and lead to the effective differentiation of chondrocytes. 38 Expression of COLII in the printed constructs supplemented with L‐ascorbic acid was observed after 14 days of culture (Figure 4 h, i). The amount of collagen deposition decreased by moving from the edges of the constructs to inner sections (Figure 4 j). This reduction can be pertained to the reduced delivery of L‐ascorbic acid and nutrients to the inner parts of the printed construct. In conclusion, the soft gel exhibited a supportive platform for 3D culture of chondrocytes. Figure 4 Characterization of cell‐laden ink during in vitro culture. a) Schematic diagram of cell integration in the bioprinted construct. b, c) Confocal images of 3D soft gels immunostained with F‐actin/DAPI on day 10 and day 30 of culture. The scale bars in the insets are 50 µm. d) Cell viability after printing shows a small decrease but the cells revive until day 7 ( N = 3). e, f) 3D reconstruction of Z‐stack images of the printed construct. The thickness of the scanned layers is 890 µm. Scale bar in the phase contrast image is 200 µm. g) PrestoBlue assay to analyze the metabolic activity of the cells in the cell‐laden and cell‐seeded printed constructs over the course of 10 days of culture in vitro ( N = 4, * p < 0. 05, ** p < 0. 01). h, i) COLII staining and confocal imaging of soft ink layers on the edge of (h) and inside (i) the gel after 20 days of culture. Scale bar in the inset image is 50 µm. j) Percentage of collagen production per cell decreased slightly from the edge of the printed construct into the inner layers of the hydrogel. The fluorescence intensity of the confocal layers was measured and the ratio of this intensity to total numbers of DAPI gave out the percentage of collagen deposition per cell. 2. 3 Integration of the Biosensor and Hydrogel The next step after optimization of the printing and chondrogenic maturation was creating the hybrid nose structure coupled with the sensing capability. As such, a CAD file of the nose was modified to include open nostril cavities for biosensor embedding ( Figure 5 a–c). Printing the convex geometry of nose models along with nostril cavities required a relatively convoluted printing pattern with both convexities and concavities. The ink ratio of soft:stiff was therefore optimized to 1:2 to create a free‐standing nose to minimize the final structural (total layers: 10) collapse to less than 3% and to achieve geometrical integrity while maintaining porosity for nutrient perfusion as well as maintaining a chondrocyte‐supportive ECM through the soft layers. Next, the dual ink cell‐laden nose was printed on top of two functionalized sensors while the sensor chambers were left exposed (Figure 5 a, d). To better illustrate the multi‐material printing of the nose using color contrasting, the nose in Figure 5 c was printed using a 1:1 ink ratio (stiff layers shown in green and soft layers in red). Figure 5 Integration of the biosensing system with the 3D‐printed construct. a) Dual ink nozzles printing a nose on top of the sensor electrodes. b, c) CAD 3D drawing of the nose with dual ink layers (top) and the printed construct using the same code (bottom). d) Optical image of the microfabricated Au biosensor with three electrode sensing system (top) and 3D‐printed dual‐ink nose integrated with the biosensors in each nostril (bottom). e, f) Live/dead assay images of chondrocytes seeded on the biosensor to test biocompatibility on Au (e) and Ag (f) electrodes. g) Quantified cell viability graph on Au and Ag electrodes on day 1 after integration with cells. h) F‐actin/DAPI staining of the sensor 7 days after integration. Interfacing tissues with electronics is still hampered by biocompatibility and cytotoxicity caused by the highly conductive but toxic elements used in the biosensing components, especially for the electrode materials and electrolytes. As such, it is essential to develop and optimize the biosensors in a cell‐friendly environment where the target biomimetic organ can remain viable while the functionality is maximized. Several groups have attempted to solve this issue by fabricating biocompatible electroconductive substrates such as graphene, carbon nanotubes, poly(3, 4‐ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), and gold (Au) to provide a minimally toxicogenic environment for the integrated cell‐laden system. 39 To achieve our aim, a metal‐based three‐microelectrode system with a working electrode (WE, Au), counter electrode (CE, Au), and reference electrode (RE, silver, Ag) was of specific interest for fabricating a robust EC biosensing system due to its decent stability, beneficial electron‐transfer kinetics, and capability to covalently bind with various chemical functional groups, namely, thiol‐based structures. 40 In addition, to fabricate biocompatible microelectrodes, an e‐beam deposition method was selected to avoid using any toxic organic chemicals or compounds compared with other microfabrication methods. To evaluate the cytotoxicity of the electrode materials (Au and Ag), chondrocytes were seeded on the bare microelectrodes and their viability was analyzed. Using Au as a highly biocompatible conductive substrate, we observed 97. 7 ± 1. 3% cell viability (Figure 5 e–g). The Ag RE showed decreased viability (79. 7 ± 1. 0%) compared with that of the Au electrode. However, the cytocompatibility of the Ag electrode was still good enough to allow for cell growth and proliferation after 7 days of culture as confirmed by F‐actin staining (Figure 5 h). 2. 4 Biosensing Explosive Molecules EIS is a label‐free detection method which has recently been employed as a fast and reliable way to detect a myriad of chemical odors, airborne pathogens, and biomarkers with high sensitivity. 41 Similar to the sensing mechanism in the olfactory epithelium, the EIS system can be functionalized to mimic the odor binding mechanism by creating a sensing platform similar to that performed through nasal mucus. We employed this detection method to develop a biosensing component for the hybrid nose ( Figure 6 a). The method of fabrication and functionalization was developed and discussed in detail in a previously published paper where an Au‐based microelectrode was used as a highly conductive and effective substrate for label‐free monitoring of cell secretomes from in vitro cultured tissues. [qv: 41c] Figure 6 Biosensing mechanism and procedure. a) Schematic diagram of the functionalization procedure of the biosensor using the TNT‐specific peptide. b, c) Nyquist plot of measurement using different concentrations of TNT, sensing using ferrocyanide. Upon capturing TNT by the peptide, impedance increases in the ferrocyanide‐based measurement system. d) Bode plots using 0. 1–1000 pg mL −1 TNT, sensing using ferrocyanide. Plots (c) and (d) share the same legend. e) Calibration curve for TNT sensing using ferrocyanide. f) Nyquist plots of measurement using 10 pg mL −1 of TNT, sensing using culture media. g) Bode plots using 0. 1–1000 pg mL −1 TNT, sensing using culture media. h) Calibration curve for TNT sensing using cell culture media. i) Diagram of degradation of TNT peptide over 8 days of incubation at 37 °C. Au‐based biosensing provides high flexibility in biosensor functionalization with a variety of natural or synthetic receptors to detect a wide range of chemical structures including explosives and human‐imperceptible biomolecules. Real‐time in situ detection of explosives is considered a major demand for security purposes and environmental concerns. Therefore, timely and sensitive detection of such chemicals can prevent a number of unfavorable and destructive events such as mine site blast injuries caused by explosive debris post explosion. 42 2, 4, 6‐Trinitrotoluene (TNT) is one of such prevalent explosives used for various industrial practices which can be captured and sensed using the EIS system. We tested the biosensing system with TNT using a TNT‐specific peptide chain (EPQLKM) developed in a previous paper. 43 The mechanism of detection in the EIS‐based sensing system is premised on changes brought on by the interfacial electron transits between the redox probe [Fe(CN) 6 ] 4−/3− and the substrate electrode upon creation of specific bonds between the antibody (here, TNT‐specific peptide) and the antigen (TNT). 44 At a specific range of TNT concentrations, the amount of TNT captured by the receptor antibodies functionalized on the sensor corresponded to those present in the solution, and could be characterized via recording the electrode impedance. 44 Therefore, we obtained the Nyquist curves in frequencies ranging from 10 −1 to 10 5 Hz (potential 0. 10 V, and modulation amplitude, 5 mV) using a 50 × 10 −3 m K 3 Fe(CN) 6 electrolyte solution. As shown in Figure 6 b, upon functionalization of a self‐assembled monolayer (SAM) and blocking media solution on the sensor, a significantly larger semicircle was seen in the high‐frequency region compared to the bare signal. Nonetheless, the signals recorded at low frequencies (i. e. , Warburg impedance) were shown to be eliminated compared to the bare sensor signal. Subsequent functionalization of the sensor with analytes led to the binding of new molecules to the sensor surface, thus increasing the insulating layer between the probe and the electron transit. As such, the impedance at each functionalization step showed an increment compared to the former step (Figure 6 b). Similarly, the diameter of the Nyquist semicircle (i. e. , electron transfer resistance, R ct ) at each step was also increased by immobilizing new layers. Next, the impedance was measured before and after the exposure of TNT molecules to the biosensor (Figure 6 c, d). Similar to the receptor immobilization steps, exposing higher concentrations of TNT to the sensor led to the capture of more TNT molecules by the sensor and thus an increase in the recorded impedance and R ct values. The results proved that the system can detect a range of 1–1000 pg mL −1 of TNT (Figure 6 e). In addition, the biosensor exhibited capability to detect TNT with a limit of detection of 0. 38 pg mL −1 with a sensitivity of 8. 6 (log(ng mL −1 )) −1. Recently, Gao et al. implanted a TNT‐receptor functionalized bioelectronic nose sensor in the green fluorescent protein‐labeled olfactory sensory neurons (OSNs) of transgenic mice, showing detection sensitivity of up to 10 × 10 −6 m TNT in vivo (11 ng mL −1 ). 45 Potassium ferrocyanide (K 3 Fe(CN) 6 ) is a common electrolyte mediator used for the EIS‐based measurements due to its strong redox activity. 46 However, exposure of such electrolytes to cells and biological systems can cause cytotoxicity and lead to erroneous cell behavior. 18 One solution to this problem is to seclude the measurement chamber from the incubation platforms to avoid direct cell contact with K 3 Fe(CN) 6. 47 This solution, however, would not fit the purpose of a hybrid bionic nose which aims to monitor and measure the analytes in real‐time continual monitoring after being integrated with the engineered tissue. As an alternative, the biocompatibility of the system can be increased through changing the mediator electrolyte to a biocompatible nontoxic platform such as chondrocyte culture media (Figure 6 f, g). 18 As an initial step, we tested and compared the impedance response of the bare electrode to Dulbecco's phosphate buffer saline (PBS) and cell culture media, which contain ≈150 × 10 −3 m salts and have an electrical conductivity of ≈1. 5–2 S m −1 (Figure S5a, b, Supporting Information). 48 However, in PBS and culture media, the semicircle curve disappeared due to the lack of or weakness of a redox reaction. The magnitude of the impedance recorded from the bare electrode in K 3 Fe(CN) 6 was found to be 1000 times lower than PBS and culture media due to the relatively high resistance of media and PBS compared with that of K 3 Fe(CN) 6. Next, the sensor was functionalized with an EPQLKM peptide and the results were measured in media and compared to those measured in the presence of K 3 Fe(CN) 6 (Figure 6 f–i). To compare the impedance ( Z ) values at any arbitrary frequency, such values can be extracted and drawn for different functionalization steps. As an example, we chose 2. 09 Hz to compare the Z values of the sensors. As seen in Figure 6 g, the impedance at 2. 09 Hz follows an increasing trend upon increasing the concentration of TNT from 0. 1 to 1000 pg mL −1. The calibration curve was calculated based on the different concentrations of TNT measured in media (Figure 6 h). As a result, the sensitivity of the sensor was found to be 0. 1 pg mL −1. Considering that the developed hybrid system needs to be functional after days of culture at 37 °C, it is important to ensure that the functionalization of the sensor is not disturbed by the culture duration and environmental factors. Therefore, we performed a degradation test to confirm that the peptide can stay viably active in the culture environment for up to 8 days after printing and before being degraded (Figure 6 i and Figure S5c, Supporting Information). The results proved that the peptide can stay active for up to 5 days after functionalization. Afterward, the gradual degradation of functionalized layers occurred, and the peptide coating layer was gradually destroyed from the substrate, leading to an impedance reduction to a value lower than the blocking stage. Therefore, for prolonged sensing, the sensor should be functionalized again or replaced by a new electrode after 5 days. Consequently, the hybrid construct exhibited a potential to detect secretions of bioactive targets and disease biomarkers in humid mucus conditions. In later works, the detection mechanism can be fitted to provide biosensing in an integrated engineered mucus. Although the proposed approach introduces a different method for the integration of electronics with biomaterials, several challenges need to be addressed to reach a fully functional reconstructed nasal organ. For instance, embedding engineered co‐culture systems of neural and endothelial networks into the hybrid devices could be a first step toward compensating for the absence of vasculature and nerves in the thick printed constructs as opposed to the native organ. Moreover, full retention of the human olfactory system requires the culture and integration of hundreds of ORs, combinations of which drive the smell recognition in the native olfactory epithelium. 49 Although the emergence of the bioelectronic nose has pushed science toward artificial odor perception, there are still several challenges in reaching a full functional and implantable artificial nose. First, although the developed biosensors are capable of distinguishing their target odorants with high sensitivity and selectivity, they still fail to identify complex mixtures and reach a comprehensive odor perception system similar to that of humans. 50 In fact, the natural olfactory‐taste system is stimulated by the combinatorial pattern recognition of various ORs, 49, 51 and thus, the activation of several ORs would lead to the perception of a specific odor or taste. To reach higher levels of complex odor perception, the bionic nose devices require the implementation of a complete set of natural ORs which would jointly contribute to differentiation of diverse smells. Further, to enable odor discriminatory function in 3D bionic noses, it is essential to convert the single‐target biosensing system into a multisensory array functionalized with a range of ORs and peptide receptors. Recently, a multiplexed biosensor was proposed by Son et al. 50 that immobilized human ORs onto a multichannel carbon nanotube‐based FET to analyze four types of taste‐ and odor‐causing compounds produced by food. Implementing such sensors in our hybrid system can generate a tunability to detect other targets such as a wide range of complex human‐perceptive odors using only a small OR matrix. In another study, an artificial multiplexed superbioelectronic nose was recently introduced by Kwon et al. to mimic the human odor discrimination in mixtures by functionalizing micro‐patterned graphene FETs with human ORs. [qv: 5h] Toward augmented detection of the bioelectronic nose through multiplex biosensing, Peveler et al. have developed a multichannel quantum dot array to detect five explosives including TNT up to 0. 2 µg mL −1. 52 Lastly, a crucial step toward biomimicking human olfaction is to standardize bionic nose devices by creating a universal code for gauging odors. This procedure can be done by choosing a set of primary odor molecules to build up odor mixtures and patterns, similar to the role of three primary light colors in building a wide spectrum of lights. 53 Creation of hybrid bionic noses may contribute to understanding the fundamental mechanisms employed by the natural olfactory system for reaching selectivity. Eventually, the developed hybrid nose could promote a minimally invasive diagnosis tool for airway diseases, such as imminent asthmatic attacks, through sensing the airborne biomarkers in the breath. Moreover, 3D‐bioprinted hybrid nasal constructs could be instrumental in development of “cyborg” organs which possess augmented functionalities over their native human counterparts. Lastly, the developed hybrid system may offer enhanced precision to nasal cartilage reconstruction strategies as well as reviving the olfactory sensation which is often permanently lost during traumatic injuries. Therefore, the hybrid nasal construct can provide a potential solution to the drawback of current rhinectomy methods. Being in its infancy, full nose organ retention still requires extensive studies and the addressing of several challenges. An artificial nose, which is to be implanted on the surface of the body, directly interacts with exogenous factors and the extrinsic environment, and thus cannot be fully maintained merely through the internal biological system for oxygen and nutrient supplementation, temperature regulation, and waste removal through blood circulation. A complementary external support is required to ensure the maintenance of environmental factors for tissue survival, or else the implant would be transformed into a massive necrotic block. To address this challenge, one solution is to add skin grafts on top of the artificial implant to isolate the organ from the outer environment and create uniform integration with the internal biosystem. This approach would require additional studies including the adhesion between the bionic nose implant and the skin graft. 54 To obtain effective and accelerated integration with the host body on the initial days of implantation, engineered nose constructs must be supplemented with an array of pre‐vascularized networks which can connect with the host's vascular system. Additionally, another approach to integrate skin grafts could be suggested by co‐culturing the bionic nose with a skin‐mimetic layer (i. e. , artificial skin) 55 or 3D bioprinting a layer of tissue engineered skin on top of the bioprinted nose organ. 56 In such a case, the bionic skin would potentially provide environmental necessities (e. g. , temperature and humidity control, oxygen supply) for self‐maintaining the implant in the initial few weeks as well as releasing specific biological cues to induce skin regeneration. Finally, full replication of the anatomy and function of the nose will require reconstruction of the native structural and cellular heterogeneity including but not limited to cartilage, bone, OSNs, epithelial layers, and vasculature. This step would require concurrent advancement and research in multiple areas of tissue engineering including neural, skin, bone, and cartilage engineering. 3 Conclusion We introduced a new hybrid device consisting of a dual bioink‐printed nose construct with an integrated biosensing system. The multi‐material printed construct consisted of several soft layers engineered for chondrogenic growth and adhesion, and multiple stiff layers with higher mechanical properties for promoting the mechanical robustness and macroscale geometrical integrity. The cell‐laden constructs supported chondrocyte growth and secretion of cartilage ECM after 14 days of culture in vitro. Moreover, the system showed capability in sensing TNT and could be further functionalized and tuned to detect a variety of natural odors, chemical structures, and disease biomarkers. The developed hybrid device lays the groundwork for a 3D‐printed viable cartilage‐like tissue with an integrated enhanced electronic olfactory system which can eventually become a viable humanoid cyborg nose organ. While these approaches can provide a more inclusive model of the native biological entities, they bring a myriad of new capabilities and novel applications in medicine, organ mimicry, and humanoid robotics by introducing controlled or even enhanced functionalities over the original biological functions. Lastly, future integration of a multiplex biosensing platform in the developed hybrid system and recapitulation of the multi‐layered heterogenous structure of the nasal cartilage can be a potential pathway toward achieving full nasal regeneration and nasal implants in the future. 4 Experimental Section Materials : Gelatin from porcine skin, PEGDMA (1000 k), L‐ascorbic acid, and photo‐initiator (PI) 2‐hydroxy‐4′‐(2‐hydroxyethoxy)‐2‐methylpropiophenone (Irgacure D‐2959) were purchased from Sigma‐Aldrich. Hanks' balanced salt solution (HBSS), fetal bovine serum (FBS), PBS, and Dulbecco's modified Eagle medium (DMEM) were purchased from Thermo Fisher Scientific. Medium degree GelMA was synthesized based on the previous protocols. [qv: 25b] Briefly, 10% w/v gelatin was dissolved in PBS at 50 °C. Afterward, 5% v/v methacrylic anhydride (Sigma‐Aldrich) was added to the solution dropwise at a constant rate of 0. 5 mL min −1. The reaction was mixed for 2 h. Next, PBS (40 °C) was added to the solution and the mixture was dialyzed (12–14 kDa dialysis membranes) at 40 °C for 7 days. The mixture was then placed at −80 °C for 2 days followed by lyophilization for 3 days. The resulting GelMA foam was stored at room temperature until use. SAM was prepared by dissolving 11‐mercaptoundecanoic acid (11‐MUA) (Sigma‐Aldrich) in >99% pure ethanol (Sigma‐Aldrich). N ‐(3‐Dimethylaminopropyl)‐ N '‐ethylcarbodiimide hydrochloride (EDC), N ‐hydroxy‐succinimide 98% (NHS), and potassium ferricyanide [K 3 Fe(CN) 6 ] were purchased from Sigma‐Aldrich. Ag epoxy glue (MG Chemicals) was used to connect the electrodes to the measurement system. A TNT‐specific peptide sequence (EPLQLKMGGGGWFVI) was purchased from Peptide 2. 0. Peptide was diluted in deionized (DI) water at a concentration of 1 mg mL −1. Preparation of the Bioink : To prepare the soft ink, 5% GelMA (medium degree of MA), 5% PEGDMA, 6. 5% gelatin, and 0. 5% PI were used. For the stiff ink, 8% GelMA (medium degree of MA), 10% PEGDMA, 3% gelatin, and 0. 5% PI were used. To prepare the soft ink, PEGMDA was dissolved in HBSS and incubated at 80 °C for 20 min. Next, the GelMA, gelatin, and PI were added to the solution and the vial was covered by aluminum foil and incubated at 80 °C for 30 min. The ink was then moved to the 37 °C incubator for 20 min before being added to the cells. The preparation of stiff ink began by dissolving the gelatin in DI water and incubating at 80 °C for 10 min. Afterward, GelMA and PI were added to the ink, covered by aluminum foil and incubated at 80 °C for 20 min. Next, the PEGDMA was mixed with the solution and incubated at 37 °C for 30 min. The inks were allowed to stabilize at room temperature. Soft ink rested for 15 min prior to printing and Stiff ink required 40 min at room temperature before printing. Both times were coordinated to reach printable conditions at the same time. Mechanical Characterization : Compression stress tests were performed using a parallel plate platform (ADMET, MTESTQuattro). Samples with around 0. 9 cm circular diameter and 0. 5 cm thickness were loaded and tested until rupture. All measurements were performed at room temperature. Prior to all measurements, the zero gap was determined. Four samples of each condition were tested. For the printed samples, an infill factor was introduced to normalize all the samples based on the theoretical (in the g‐code) and practical infill (after the print) by comparing the weights and strain of printed samples to bulk gels with the same dimensions. This infill factor was averaged and defined for the soft and stiff ink separately. For instance, to calculate the infill factor for 1:2, 1X soft and 2X stiff were used. All results of the composite ratios were normalized by their corresponding infill factors (0:1, 1:1, 1:2, 2:1, 1:0 and 0. 89, 0. 79, 0. 75, 0. 82, 0. 68, respectively). To calculate Young's modulus, the elastic part (10–20% strain) of the stress–strain curve was used. Rheology tests were performed on a TA Instruments DHR‐3 rheometer. A 20 mm diameter parallel plate geometry was used for all measurements. Samples were approximately 500 µm thick and the gap size was approximately 500 µm in all cases. All rheological measurements except the temperature sweep experiment were performed at 25 °C. To measure the viscosity by sweeping the temperature, a temperature control attachment (PolyScience) was used. Prior to measuring each sample, the measuring system inertia of the upper geometry as well as the motor friction was calibrated. All samples were allowed for relaxation and reached equilibrium for 30 min after loading and before initiating the test. An amplitude sweep was done on the samples to determine the linear viscoelastic window at three frequencies, ω = 0. 3, 1, and 10 rad s −1, in 0. 01–2% strain. 3D Bioprinting : 3D bioprinting was performed using a Cellink Inkredible bioprinter. A CAD file of a nose was sliced using Slic3r software to generate gcode readable by the bioprinter. The printing infill was chosen as 20% for both inks. A nozzle speed of 15 mm min −1 was used to print the final constructs. A 3 mL syringe (BD) and 27 gauge needle (Fisnar) were used for printing both inks. The nozzle was covered by foil during the process to avoid random crosslinking of the gel. An air pump was used to create the extrusion force for the nozzle. The pump rate was adjusted to initiate the printing and then decreased to 40 kPa for the soft ink and 90 kPa for the stiff ink. UV crosslinking was induced using a OmniCure S2000 machine and a 5 mm diameter light guide UV lamp. The stage was adjusted to 8 cm distance from the light guide and the UV intensity was calibrated to 100 mM cm −2. All constructs were crosslinked for 80 s, flipping the construct carefully half‐way through the process. Fabrication of the Sensors : The sensors were fabricated at the Harvard Center for Nanoscale Systems. Ag RE, Au CE, and Au WE were created. After cleaning the square glass substrates (25 mm × 25 mm) with oxygen plasma, a shadow mask process was used in order to manufacture the microelectrodes (Figure S1, Supporting Information). In this process, metal layers were selectively deposited over a shadow mask which has apertures in a metal film of 0. 25 mm thickness. In this process, the first shadow mask for WE and CE was attached to cleaned glass substrate. The 20 nm thick titanium (Ti), 20 nm thick palladium (Pd), and 500 nm thick Au were deposited on the glass using e‐beam evaporation. Also, the Au electrodes were not patterned with a passivation layer. Next, the second shadow mask for RE was attached to the Au deposited glass wafer. The alignment was achieved by using alignment keys. 20 nm thick Ti, 20 nm thick Pd, and 500 nm thick Ag were then deposited. After peeling off the shadow mask from the wafer, the required patterns were realized without the need for any wet processing. Sensor Functionalization to Detect TNT : The functionalization of the sensor followed the previously established protocol. [qv: 41c] Briefly, the surface of the sensor was first functionalized with SAM using 11‐MUA. Next, an EDC/NHS conjugation was coated on the sensor to create a covalent bond between the surface and SAM where the carboxylterminated alkyl surface was converted to an active NHS ester when reacting with 11‐MUA. Next, the peptide was coated on the sensor at a concentration of 1 µg mL −1. Surface passivation was done using cell culture media. Dilutions of TNT were prepared in fresh cell culture media ranging from 0. 1 pg mL −1 to 1. 0 ng mL −1. The measurements were taken using K 3 Fe(CN) 6 and cell culture media. In this case, the culture media was left at room temperature to reach pH equilibrium. Nyquist plots were obtained in the frequencies ranging from 10 −1 to 10 5 Hz under a potential value of 0. 1 V and modulation amplitude of 5. 0 mV in 50 × 10 −3 m K 3 Fe(CN) 6. Toxicity Test of the Biosensor : All sensors were washed with 70% ethanol followed by 30 min incubation at 37 °C with 1% Anti‐Anti in PBS solution. In Vitro Culture : Primary chondrocytes from 1 month old calves (Astarte bio) were cultured and passaged for 1 week at 37 °C and 5% CO 2. DMEM, 10% FBS, and 1% p/s were used as the culture media. Cells were encapsulated in the bioink at a density of 20 million cells mL −1. Furthermore, additional cells with 0. 5 million cells cm −2 were seeded on the constructs on day 2 of printing. After printing, the culture media was supplemented with 50 µg mL −1 l ‐ascorbic acid to induce collagen II production. To increase the cell viability during the printing process, the concentration of HBSS in the soft ink was adjusted to include 20% FBS added directly to the cell pellet before cell encapsulation (1:9 ratio of HBSS to FBS). The constructs were supplemented with media after printing and the media were changed every 2 days. Cell‐Laden Characterization : Cell viability and proliferation of the printed constructs were assessed using live/dead assay (Thermo Fisher scientific) and PrestoBlue kit (Thermo Fisher scientific). The colorimetric assays were measured using a plate reader (Infinite 200 pro, Tecan Austria GmbH) by measuring the absorbance at 570 nm with reference to 600 nm. The results were normalized by day 1 of culture. Fixation of cell‐laden constructs was done on days 14 and 30. Briefly, the constructs were treated with 4% paraformaldehyde solution (Thermo Fisher scientific) for 30 min and then permeabilized using 0. 1% Triton X (Sigma‐Aldrich). F‐actin (1:40, Invitrogen) and COLII antibody (1:100, Invitrogen) were added to the constructs for staining. The constructs were incubated with the primary antibody overnight on a mild shaker at 4 °C. Secondary antibody goat anti‐mouse Alexa Fluor 594 (Invitrogen) was used to image the COLII staining and was added to the constructs, followed by incubation at 4 °C for 6 h. DAPI (4′, 6‐diamidino‐2‐phenylindole, 1:1000, Sigma‐Aldrich) was added 30 min prior to imaging. Confocal imaging was performed using a Zeiss LSM 880 airyscan microscope. Images were processed and analyzed in Fiji software. Statistical Analysis : All tests were performed in triplicates and the average and standard deviation were calculated in Graphpad Prism. The results of the PrestoBlue assay were analyzed using one‐way analysis of variance method. Error bars represented mean ± standard deviation of measurements in each group. To compare different treatment groups and study the existence of significant differences among groups, Tukey's multiple comparison method with p < 0. 05 was employed. Statistical Significance in all graphs was indicated as not significant (ns) ( p > 0. 0. 5), *( p < 0. 04), and ****( p < 0. 0001). Conflict of Interest The authors declare no conflict of interest. Supporting information Supplementary information Click here for additional data file. Supplemental Movie 1 Click here for additional data file.
|
10. 1002/advs. 201902295
| 2,019
|
Advanced Science
|
Developmentally Engineered Callus Organoid Bioassemblies Exhibit Predictive In Vivo Long Bone Healing
|
Abstract Clinical translation of cell‐based products is hampered by their limited predictive in vivo performance. To overcome this hurdle, engineering strategies advocate to fabricate tissue products through processes that mimic development and regeneration, a strategy applicable for the healing of large bone defects, an unmet medical need. Natural fracture healing occurs through the formation of a cartilage intermediate, termed “soft callus, ” which is transformed into bone following a process that recapitulates developmental events. The main contributors to the soft callus are cells derived from the periosteum, containing potent skeletal stem cells. Herein, cells derived from human periosteum are used for the scalable production of microspheroids that are differentiated into callus organoids. The organoids attain autonomy and exhibit the capacity to form ectopic bone microorgans in vivo. This potency is linked to specific gene signatures mimicking those found in developing and healing long bones. Furthermore, callus organoids spontaneously bioassemble in vitro into large engineered tissues able to heal murine critical‐sized long bone defects. The regenerated bone exhibits similar morphological properties to those of native tibia. These callus organoids can be viewed as a living “bio‐ink” allowing bottom‐up manufacturing of multimodular tissues with complex geometric features and inbuilt quality attributes.
|
1 Introduction Tissue‐engineered advanced therapy medicinal products (TE‐ATMPs) are poised to revolutionize health care by replacing or restoring the function of damaged organs. Although major advances in the field of cell therapy manufacturing have been witnessed, only a small fraction of TE‐ATMPs exhibit quality attributes that could guarantee predictive performance in vivo and hence support clinical translation. 1, 2, 3 To tackle these hurdles, a conceptual and technical merging of developmental biology and engineering principles is taking place within regenerative medicine. These “developmental engineering” strategies strive to mimic developmental events while guaranteeing robustness and predictive outcomes in a clinical setting. 4, 5, 6, 7 According to this strategy, cellular self‐assemblies and condensations of the appropriate length scale are key initiators for the formation of transient tissue structures capable of executing developmental programs with a high level of independence leading to organogenesis processes. 8, 9 These processes are regulated through the activation of tissue‐specific genes and pathways characterized by a high degree of autonomy resulting in tissues that are able to undergo a similar cascade of processes even ex vivo. 10 This type of recapitulation of developmental events has previously been demonstrated with human adult stem cells, for example, for the formation of epithelial 1 and liver 11 organoids. In the context of bone tissue engineering, fracture healing of long bones includes the formation of a cartilaginous “soft callus” that subsequently is transformed into bone, 12 a process that resembles the well‐described and tightly synchronized process of endochondral ossification in the growth plate during development. 13, 14, 15 The autonomy of the growth plate cartilage in embryonic cartilage anlagen was previously reported, and even when the cartilage anlage was decomposed into single cells and re‐implanted subcutaneously, re‐organization occurred and a growth plate‐like structure was formed. 4, 10 Furthermore, investigations inspired by “developmental engineering” demonstrated recapitulation of endochondral ossification in ectopic environments using embryonic stem cells 16 or bone marrow mesenchymal stromal cells (BM‐MSCs) 17, 18 and orthotopically using rat 19 or human 20 BM‐MSCs. However, only partially successful results have been demonstrated due to scalability challenges and uncontrolled complexity in 3D cell culture formats currently used for inducing chondrogenic differentiation. 18, 21 The use of scaffold‐free microspheroid cultures could provide a more homogeneous 3D culture format to precisely engineer soft callus‐like microenvironments. 22, 23, 24, 25 The ability to produce populations of small functional modules will constitute a major step toward the incorporation of design principles in skeletal living implant manufacturing. 2 The formation of high‐throughput cell microspheroid populations of defined size and their use as building modules for bottom‐up tissue formation strategies is gaining momentum for various TE applications. 22, 26, 27, 28 However, the construction of complex engineered tissues possessing multicomponent tissue architecture is still elusive. Although bottom‐up approaches have been suggested in recent years in order to build larger tissue structures from micromodules, the majority of these studies used cell microspheroids with minimal cell‐secreted extracellular matrix (ECM). 26 Regarding long bone defect regeneration, there is scarce literature on the potential of modular bioengineering strategies to generate larger implants while there is no understanding yet of how this architecture dictates whole tissue function after implantation. Ideally, modules for regeneration of long bone defects should possess an autonomy that would guarantee that the repeated functional units can synergistically contribute to the regenerative process, resulting in a predictive clinical outcome. In this work, we present a developmental bioengineering strategy based on self‐assembly of human‐periosteum‐derived cells (hPDCs). hPDCs have great promise for regeneration of long bone defects, since the majority of cells forming the “soft callus” during fracture healing are derived from the periosteum. 14, 15, 29 In addition, recently published studies demonstrated the presence of skeletal stem cells within the periosteum with improved capacity to regenerate bone as compared to BM‐MSCs. 15, 30 Herein, self‐assembly of hPDCs allowed scalable production of semiautonomous callus organoids that formed bone microorgans upon implantation. The in vitro maturation toward callus organoids was linked to gene expression patterns encountered in the embryonic growth plate and during fracture healing. Furthermore, an assembly of multiple callus organoids resulted in multimodular constructs that formed large bone organs ectopically and healed critical‐sized long bone defects in mice. In both cases, bone organs were formed in the absence of contaminating fibrotic tissue and exhibited a well‐developed bone marrow compartment, thus demonstrating the potential of this modular approach ( Figure 1 a) for future clinical applications. Figure 1 Long‐term culture of periosteal microspheroids. a) Schematic overview of the bioengineering process starting with cellular aggregation, condensation, and differentiation followed by callus organoid assembly and implantation in ectopic and orthotopic environment. b) Projection area of microspheroids over time (87–400 microspheroids, 10–90 percentiles). c) Representative bright‐field images of microspheroids over time. d) Representative 3D renderings of confocal images of stained with DAPI (nucleus) and Phalloidin (F‐actin) over time. e) DNA quantification of microspheroids over time, normalized to day 0 (5 h) ( n = 6, 10–90 percentiles). f) Representative confocal z ‐projection images of LIVE (green)/DEAD (red) staining over time. g) Semiquantification of cell proliferation in microspheroids over time. EdU fluorescent area was normalized to DAPI fluorescent area (10–15 microspheroids per condition, 10–90 percentiles). h) Representative fluorescent images of proliferating cells (EdU, red) in microspheroids over time, blue represents the nucleus. ** p < 0. 01; *** p < 0. 001; one‐way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. Scale bars: c, d, f, h) 50 µm. 2 Results 2. 1 Long‐Term Culture of Microspheroids Follows Early Pattern of Endochondral Ossification Endochondral ossification is initiated with cell aggregation and condensation, followed by chondrocyte specification, differentiation, and formation of a cartilage tissue intermediate that subsequently is replaced by bone. 31 Here, cell aggregation, condensation, and differentiation of hPDC microspheroids were studied over a period of 4 weeks (Figure 1 b, c). The self‐aggregation process comprised two steps. Initially, over a course of 5 h (day 0), hPDCs self‐assembled to form a stack of cells until a spheroid shape was attained (Figure 1 c, d; Movie S1, Supporting Information). Filamentous‐actin (F‐actin) staining demonstrated changes in the actin cytoskeleton by formation of stress fibers during the first week as well as compaction of microspheroids with a more confined cortical actin network over time and its thinning after 3 weeks (Figure 1 d). 3D visualization of cell nuclei showed the presence of nuclear condensation and fragmentation indicating occurrence of apoptosis in some cells starting from day 14 32 (Figure S1a, white arrows, Supporting Information). Furthermore, DNA quantification suggested a stable number of cells during 2 weeks followed by a 44% decrease after 3 weeks (Figure 1 e). The majority of cells in the microspheroids were viable; however, an increase in dead cells was observed during the last week of the culture period (Figure 1 f). Messenger ribonucleic acid (mRNA) transcripts of the marker of proliferation Ki‐67 ( MKI67 ) declined after 21 days (Figure S1d, Supporting Information) and 5‐ethynyl‐2′‐deoxyuridine (EdU) staining confirmed this trend by revealing a high number of proliferating cells (46%) during the first weeks, which subsequently decreased and was almost absent after 4 weeks in culture (Figure 1 g, h). This decrease in proliferation is also seen during endochondral ossification 31 indicating chondrocyte differentiation and maturation of the microspheroid cells. To further define the differentiation stages of the microspheroids, gene expression of relevant markers was analyzed ( Figure 2 a). The early chondrogenic transcription factor sex‐determining region Y box ( SOX)9 was upregulated (5‐fold) the first 14 days in culture followed by a downregulation while the cartilage matrix marker collagen type II alpha 1 ( COL2A1 ) was highly upregulated (6100‐fold) after 21 days in culture. The early osteogenic and pre‐hypertrophic marker runt‐related transcription factor 2 ( RUNX2 ) was upregulated after 7 (10‐fold) and 14 days (16‐fold) where after a downregulation was seen. The transcription factor osterix ( OSX or SP7 ), which is directly regulated by RUNX2 and expressed in pre‐hypertrophic chondrocytes and osteoblasts, followed a similar expression trend. 33, 34 Distinct upregulation of the hypertrophic markers collagen type X alpha 1 chain ( COL10A1 ) (1340‐fold) and Indian hedgehog signaling molecule ( IHH ) (33‐fold) was detected at day 21. In addition, alkaline phosphatase ( ALP ) gene expression was upregulated (19‐fold) at day 14 and integrin binding sialoprotein ( BSP or IBSP ), linked to matrix mineralization and osteoblast differentiation, 35, 36, 37 was upregulated 8400‐fold, after 21 days in culture. No significant upregulation of the analyzed genes ( SOX9, COL2A1, RUNX2, OSX, COL10A1, IHH, ALP, and BSP ) was detected between day 21 and day 28 (Figure S2a, Supporting Information). Therefore, the following analyses were performed until day 21. In summary, the above results demonstrated a proliferation phase that was interchanged with cellular differentiation and maturation defined by genes associated with both hypertrophic chondrocyte and osteogenic differentiation. Figure 2 Microspheroids follow endochondral ossification patterns toward pre‐hypertrophic callus organoids able to form bone in vivo. a) Quantification of mRNA transcript of chondrogenic and pre‐hypertrophic/hypertrophic gene markers was performed and normalized to D0 ( n = 6 mean value ± SEM). * p < 0. 05; ** p < 0. 01; *** p < 0. 001; one‐way ANOVA followed by Tukey's multiple comparison test. b–e) Representative sections of: b) Alcian Blue, c) Safranin O, d) IHH immunostaining, and e) confocal z ‐projection image of OSX immunostaining over time. f) Schematic view of individual callus organoid implantation. g) 3D rendering of nano‐CT images after 4 weeks in vivo implantation. h) Safranin O, i) Masson's Trichrome, and j) TRAP staining after 4 weeks in vivo. k) Bright‐field image of invading blood vessels (black arrow, #: microwell) and l) CD31 immunostaining 4 weeks after implantation (* represents the agarose mold). Scale bars: b–e) 50 µm; g) 100 µm; h–j) the upper row represents 100 µm and the lower row 50 µm; k) 400 µm; and l) 50 µm. 2. 2 Microspheroids Mature toward Pre‐Hypertrophic Callus Organoids That Form Bone Microorgans In Vivo The gene expression analysis indicated chondrogenic differentiation toward hypertrophy in combination with osteogenic differentiation at day 21 (Figure 2 a). Furthermore, Alcian Blue staining at low pH, specific for glycosaminoglycan (GAG), confirmed an increased presence of cartilage‐like ECM within the microspheroids, and pre‐hypertrophic like cells were visible after 3 weeks in culture (Figure 2 b, black arrows). Safranin O staining demonstrated slight presence of cartilage‐specific sulfated GAGs after 21 days in culture (Figure 2 c) and immunostaining confirmed the presence of IHH, OSX, and COL2 protein after 14 days in culture (Figure 2 d, e; Figure S1b, d, e, Supporting Information). The gene expression and histological analysis demonstrated that the microspheroids, containing ≈250 aggregated cells (day 0, Figure 2 b, c), matured into microtissues with differentiated cells and ECM (day 14, Figure 2 b, c). Based on the upregulation of hypertrophic gene markers (Figure 2 a) and the presence of pre‐hypertrophic cells (Figure 2 b), day 21 microtissues were chosen to be implanted subcutaneously to evaluate their capacity to mature into bone in vivo. Implantation of whole agarose microwell platforms with a diameter of 5 mm was carried out in immunodeficient mice to ensure that microtissues would remain entrapped in their microwells (Figure 2 f). After 4 weeks of ectopic implantation, nano‐computed tomography (nano‐CT) scans demonstrated the formation of distinct mineralized spheres (Figure 2 g) with a volume of (5. 4 ± 3. 54) × 10 5 µm 3 and an average diameter of 209 ± 41 µm (24 spheres quantified from three explants). Histological sections further demonstrated the presence of bone matrix (Figure 2 h, i) surrounding a marrow compartment with osteoclast activity (Figure 2 j) and blood vessels (Figure 2 k, l). These data demonstrated the development of microtissues with proteoglycan‐rich ECM positive for IHH and COL2. Furthermore, these microtissues were able to form bone microorgans in vivo (Figure 2 g), confirming that these implants behaved as single semiautonomous bone‐forming modules in vivo acting as callus organoids. This defines a maturation process from microspheroids (day 0) to microtissues (day 14) and finally callus organoids (day 21) (Figure 2 e). 2. 3 Callus Organoids Fuse into Larger Constructs In Vitro In order to demonstrate that the above‐mentioned microtissues and callus organoids can be used as building modules to form larger constructs, we initially studied the fusion process of two callus organoids. Despite long‐term culture as microtissues, creating a substantial amount of secreted ECM, the callus organoids spontaneously fused over 24 h (Movie S2, Supporting Information). Subsequently, ≈3000 modules were flushed out of their microwells and assembled in an agarose well (2 mm diameter and depth) for fusion into a multimodule construct ( Figure 3 a; Figure S2b, Supporting Information). 14 (microtissues) and 21 day (callus organoids) modules were chosen for further analysis based on chondrogenic ( SOX9 and COL2A1 ) and hypertrophic ( COL10A1, IHH, and ALP ) gene markers (Figure 2 a, d, e), as well as cell morphology (Figure 2 b–d). Both 14 day and 21 day modules fused into larger constructs that could be handled and transported (Figure 3 a); however, single callus organoid structures were still visually discernible in 21 day constructs. As a control to these structures, a macropellet formed with the same number of cells and cultured for 3 weeks in the same media formulation was introduced (Figure 3 a, Macropellet). Figure 3 Assembly of cartilage intermediate microtissues into larger bone forming constructs. a) Schematic drawing demonstrating module assembly into an agarose macrowell (left) and representative photographs of the day 14, day 21 constructs, and Macropellet (right). b) Alcian Blue staining of fused constructs and Macropellet. c) 3D rendering of nano‐CT images 4 and 8 weeks after implantation. d) Quantification of mineralized tissue 4 and 8 weeks after implantation (mean value ± SEM, n = 3–6). e) Representative images of CD31 immunostaining (black arrows demonstrate blood vessels), and f) quantification 8 weeks after implantation (mean value ± SEM, n = 3–6). ANOVA followed by Tukey's multiple comparison test. Scale bars: a) 500 µm, b) 500 and 100 µm, c) 500 µm, and d) 100 µm. Alcian Blue staining demonstrated increased module fusion within the day 14 constructs as compared to day 21 constructs (Figure 3 b), albeit both module constructs contained positive staining thoughout their structures. In contrast, the macropellet, not assembled with modules, did only show Alcian Blue staining at the periphery (Figure 3 b). Safranin O staining corresponded to the Alcian Blue staining seen in macropellets. In contrast, Safranin O positive areas were found throughout the day 21 constructs (Figure S3a, Supporting Information). None of the constructs demonstrated positive staining for Alizarin red or von Kossa (Figure S3b, c, Supporting Information), indicating that mineralization was not present in the constructs. In conclusion, these results demonstrated the formation of larger constructs through assembly of micromodules resulting in more homogenously distributed GAG‐rich ECM as compared to macropellets (Figure 3 b). 2. 4 Assembled Callus Organoids Form Single Large Bone Organs In Vivo Next, day 7, 14, and 21 constructs, as well as the macropellets were implanted ectopically in immunodeficient mice to investigate their capacity to form bone in vivo (4 and 8 weeks). None of the day 7 constructs were retrieved ( n = 4). However, mineralization was detected with nano‐CT in the other three conditions after 4 week implantation (Figure 3 c). No significant difference in mineralization percentage was seen between the conditions after 4 or 8 weeks. However, a nonmineralized core was detected in the macropellet at both timepoints (Figure 3 c, white arrows). Furthermore, after 8 weeks' ectopic implantation, the day 21 constructs and macropellets contained a mineralized cortex, while the mineralized tissue in the day 14 constructs appeared porous hence less mature. The number of blood vessels was quantified with CD31 immunostaining, and no significant difference between the constructs was detected although a larger number of day 21 constructs (5/6) contained a high amount (>50 blood vessels mm −2 ) of blood vessels as compared to day 14 constructs (2/4) and macropellets (0/3) (Figure 3 e, f; Figure S3d, Supporting Information). Safranin O and Massons's Trichrome staining on histology sections after 4 weeks' implantation revealed that day 21 constructs contained bone, bone marrow, as well as remodeling cartilage indicating the occurrence of endochondral ossification ( Figure 4 a; Figure S3f, Supporting Information). Although no significant difference was detected, limited bone marrow compartments were seen in the day 14 constructs and macropellets in contrast to the day 21 constructs (Figure 4 a, b, e). Strikingly, significant areas of fibrotic tissue were detected in both day 14 constructs and macropellets as compared to day 21 constructs (Figure 4 f). Furthermore, tartrate‐resistant acid phosphatase (TRAP) staining (Figure 4 c) demonstrated osteoclast activity in all constructs although more prominent in day 21 constructs. Human osteocalcin (hOCN) staining demonstrated that implanted cells contributed to the bone formation in all constructs (day 14 construct: 74 ± 10%, day 21 construct: 58 ± 18%, and macropellet: 72 ± 2%) (Figure 4 d, g; Figure S3e, Supporting Information). Figure 4 Histological assessment after ectopic implantation. a) Safranin O, b) Masson's Trichrome, c) TRAP, and d) hOCN staining of day14, day 21 constructs, and Macropellet 4 weeks after in vivo implantation. e–g) Quantification (mean value ± SEM) of: e) bone marrow compartment ( n = 5–6), f) fibrous tissue area ( n = 5–6), and g) hOCN positive cells ( n = 3–4) 4 weeks after implantation. h) Hematoxylin and eosin (H&E), i) Safranin O, j) Masson's Trichrome (M's T), and k) hOCN staining after 8 weeks in vivo implantation. * p < 0. 05; ** p < 0. 01; *** p < 0. 001; one‐way ANOVA followed by Tukey's multiple comparison test. Scale bars: a–c, h–j) 500 µm (left) and 100 µm (right) and d) 100 µm. Furthermore, hematoxylin–eosin (H&E), safranin O, and Masson's Trichrome staining revealed mature bone in all conditions after 8 weeks' implantation, but the day 14 constructs and macropellets still contained domains of fibrotic tissue which were absent in the day 21 constructs (Figure 4 h–j). In addition, OCN positive cells of human origin were present also after 8 weeks' implantation (Figure 4 k). Although mineralized, some of the day 21 constructs maintained a hypertrophic chondrocyte phenotype after both four (three of six implants) and eight (one of six implants) weeks in vivo implantation (Figure S3f, Supporting Information). Taken together, these results supported that callus organoids fused into larger day 21 constructs in vitro and further developed into bone organs in vivo. 2. 5 Temporal Gene Expression Patterns during Callus Organoid Formation Follow the Endochondral Ossification Process toward a Niche for Matrix Remodeling and Bone Organ Formation To better explain the differentiation pathway of the callus organoids, an RNA sequencing analysis of D0 (5 h), D7, D14, and D21 modules was performed demonstrating a similar trend as the limited gene expression analysis (Figure 2 a; Figure S4a, b Supporting Information). Furthermore, the number of significant ( p < 0. 05 and log 2 ‐fold > 1) differentially expressed genes decreases over time from 3949 (D0–D7) and 847 (D7–D14) to 55 (D14–D21) and 84 (D21–pellet) ( Figure 5 a) indicating that the most dramatic changes occurred at the early stages of differentiation. Figure 5 RNA sequencing analyses of spheroids and Macropellet ( n = 3–4). a) Volcano plots of differentially expressed genes from RNA‐seq data between the different spheroids and Macropellet. X ‐ and Y ‐axes represent log 2 fold change and log 10 p‐ value, respectively, and green dots represent genes with log2FoldChange > 1 and padj < 0. 05. b) Gene Ontology (GO) Biological Processes (2017b) of genes within the five different clusters of the 400 most variable genes (log2FoldChange > 1 and padj < 0. 05). c) Heat map of spheroid mRNA transcript levels of genes regulating endochondral ossification (relative expression value for each gene). d) Venn diagram of the number of significant differentially expressed genes (green dots in (a)) for the different spheroid maturations. Green text represents genes associated with endochondral ossification and gray text genes associated with angiogenesis. Pathway analysis using Enrichr 38, 39 with WikiPathway (2016) grouped upregulated genes (D0–D21) into endochondral ossification (WP474, adj. p ‐value 1. 0e‐13) and embryonic skeletal system development (Gene Ontology (GO): 00 48706, adj. p ‐value 1. 379e‐8) from GO Biological Process (BP) (Data File S1, Supporting Information). 38, 39, 40 Next, unsupervised clustering was performed on the 400 most variable genes in order to gain a holistic overview of signaling action during the callus organoid maturation process and the GO enrichment for each cluster was defined (Data Files S2 and S3, Supporting Information). The first cluster with a continuously upgoing trend included genes enriched to skeletal and cartilage development and regulation of mitogen‐activated protein kinases (MAPK) and ERK1/2 signaling (where ERK = extracellular signal‐regulated kinase) involving the wingless‐INT (WNT), bone morphogenetic protein (BMP), and fibroblast growth factor (FGF) signaling (Figure 5 b). These are crucial signaling pathways driving endochondral ossification working in a converging manner toward chondrocyte hypertrophy. 41 Interestingly, genes related to WNT signaling were also present in the transient downregulated cluster 3. This cluster included genes related to transforming growth factor beta (TGF‐β)/BMP related SMAD and WNT signaling motivating a converging crosstalk during callus organoid maturation. In addition, Notch, which is important for stem cell maintenance, suppression of chondrocyte differentiation, and proliferation, 42 was represented in the downregulated cluster 3. These two clusters (clusters 1 and 3) indicate cell signaling regulation analogous to the molecular cascade of events present during endochondral ossification. 41 Cluster 2, with constant downregulation, included genes associated with DNA transcriptional activity correlating with the decrease in cell proliferation occurring during the transition from proliferative to hypertrophic chondrocytes 43 which was indicative also in our data (Figure 1 g, h). Genes associated with ECM disassembly ( MMP13 ) and produced by hypertrophic chondrocytes ( COL10A, COL9, and SPP1 ) were grouped in the constantly upregulated cluster 4 supporting maturation toward hypertrophic callus organoids that exhibited a high turnover and capacity to remodel the surrounding matrix. In addition, genes linked to calcium‐ion regulation were highly represented in the upregulated cluster 5, suggesting a gradual transition to a pre‐hypertrophic niche favoring mineralization, although no in vitro mineralization was detected (Figure S3b, c, Supporting Information). 44 The GO enrichment of the unsupervised clusters demonstrated that the callus organoid maturation followed signaling pathways regulating endochondral ossification. This was further supported by analysis of well‐known regulators. During the first phase (D7), important regulators of chondrocyte proliferation, differentiation, and organization were upregulated, including IGF1 and its receptor IGF1R, 45 GLI3, 46, 47, 48 PTHrP, 49 and the SOX trio ( SOX 5 / 6 / 9 ) 50 (Figure 5 c). From day 14 onward, the PTHrP positive state converted into a IHH positive state, followed by increased expression of chondrocyte hypertrophy activators, such as GLI1, FOXA2, MEF2C, OSX, RUNX2, and RUNX3 48, 51 (Figure 5 c). This pattern was mirrored in the gene expression of matrix proteins and regulators involved in endochondral ossification with a distinct upregulation of collagens ( COL2A1, COL9A1, COL10A1, and COL11A1 ) and signaling factors correlated to pre‐hypertrophic/hypertrophic chondrocytes and osteoblasts ( SPP1, IBSP, DMP1, and ALPL ). 48, 51 These data demonstrate a regulatory “switch” between D7 and D14 plausibly crucial for bone formation in vivo. Genes significantly upregulated between both D7–D14 and D14–D21 included FOXA2, DMP1, and SCIN which are crucial for chondrocyte hypertrophy, 52 cartilage–bone transition, 53 and bone resorption, 54 respectively, indicating their significant role in bone formation and also in the current manufacturing approach (Figure 5 d). Subsequently, comparison from D14 to D21 indicated further maturation linking to the in vivo formation of a bone organ with matrix remodeling and the presence of bone marrow (Figure 4 ). Since only a limited number of genes were significantly changed (55 up/downregulated genes, Figure 5 a) during this period, individual analysis was performed for these genes (Figure S4c, Supporting Information). Around 15 of the 43 differentially upregulated genes (Figure S4c, Supporting Information) have been associated with endochondral ossification and pre‐hypertrophic/hypertrophic chondrocytes (Table S1, Supporting Information) whereof IBSP, 36, 37 chondromodulin ( CNMD or LECT1 ), 55, 56 and WNT4 57 were exclusively significant for D14–D21 (Figure 5 d). Interestingly, 15 of the 55 significantly changed genes D14–D21 (Figure S4c, Supporting Information) have been associated with regulation of angiogenesis (Table S2, Supporting Information), a pivotal event during the transition from cartilage to bone in endochondral ossification. Of these angiogenic genes, six were exclusively differentially expressed between D14 and D21 (Figure 5 d). Conclusively, the RNA‐seq analysis demonstrated endochondral maturation from initial microspheroids (aggregated cells) to callus organoids (cells and ECM) exhibiting pre‐hypertrophic characteristics and active remodeling of the secreted ECM resulting in bone organ formation in vivo. 2. 6 Assembled Callus Organoids Heal Critical‐Sized Long Bone Defects Based on the ectopic implantation and RNA sequencing results, day 21 modules were defined as “callus organoids” and selected as modules for the formation of larger constructs and orthotopic implantation in a murine, critical‐sized long bone defect. 58 An agarose mold based on the dimensions of the critical‐sized defect and with the decrease in size during fusion of callus organoids accounted for was fabricated ( Figure 6 a). Next, ≈6000 callus organoids were seeded into the agarose mold (Figure 6 b) and fused during 24 h resulting in a construct (≈4. 5 mm length and 2 mm wide) (Figure 6 c) that was fitted into the tibia defects of immunodeficient mice (Figure 6 d). Figure 6 Healing of murine critical‐sized long bone defect. a) Schematic visualization of implant formation. b, c) Bright‐field image 1 h (b) and 24 h (c) after callus organoid assembly. d) Photograph of a 4 mm tibia defect after healing. e) X‐ray images of tibia defect with a day 21 construct. f) Negative control: X‐ray of empty defect after 8 weeks. g) Quantification of mineralized volume in defects with day 21 construct and empty defects ( n = 4 animals for each condition, two‐way ANOVA followed by Tukey's multiple comparison test). h) Nano‐CT 3D rendering images over time of defect with day 21 construct. i) Cross section of 3D rendering of native tibia and defect 8 weeks after day 21 construct implantation. j–l) In vivo CT quantification of structure: j) thickness, and k) linear density over time visualized with l) in vivo CT 2D images. m–p) Comparison between native tibia and healed defect 8 weeks after construct implantation was demonstrated by ex vivo nano‐CT quantification of: m) mineralized tissue (%), n) medullary cavity (%), o) structure thickness, and p) structure linear density ( n = 4, unpaired t ‐test). q) H&E, r) Masson's Trichrome, and s) hOCN immunohistological staining of defect 8 weeks after day 21 construct implantation. * p < 0. 05; ** p < 0. 01; **** p < 0. 0001. Scale bars: b, c, e, h, i) 1 mm, l) 500 µm, q–s) overview 1 mm and zoom‐in 100 µm. X‐ray images and 3D renderings of in vivo CT scans demonstrated occurrence of mineralization after 2 weeks, and bridging of defects was detected after 4 weeks followed by increased corticalization until week 8 (Figure 6 e, h; Figure S5a, Supporting Information). No bridging was detected in the empty defects after 8 weeks (Figure 6 f) while quantification of in vivo CT images confirmed the increase in mineralized tissue over time in experimental conditions (Figure 6 g). Cross section of the nano‐CT 3D rendering from week 8 demonstrated the presence of cortical bone in the defect with a nonmineralized compartment in the center, suggesting a defined bone marrow cavity (Figure 6 i). Structure thickness increased significantly from week 2 to 4 correlating to the time of defect bridging (Figure 6 j; Figure S5a, Supporting Information). Furthermore, the number of structures decreased from week 4 to 8 indicating remodeling from a trabecular to a more cortical structure, which was also visible on in vivo CT images (Figure 6 k, l). Next, the healed defects at week 8 were compared to native tibia at the same location as the defect, in mice of the same age and gender. No significant differences were found regarding mineralized percentage, volume (Figure 6 m; Figure S5b, Supporting Information), structure linear density (Figure 6 p), or medullary cavity occupancy (Figure 6 n), while the medullary volume in healed defects was significantly larger than in native bone (Figure S5c, Supporting Information). In addition, structure thickness was lower in the healed defects as compared to native bone indicating that longer healing time may be necessary for full regeneration (Figure 6 o). H&E and Masson's Trichrome staining after 8 weeks confirmed full bridging (3/4) with the presence of mature bone and bone marrow (Figure 6 q, r; Figure S5d, e, Supporting Information), and hOCN staining (Figure 6 s) revealed the contribution of donor cells to the bone formation process. In conclusion, the assembly of multiple callus organoids into an easy‐to‐handle scaffold‐free implant resulted in full bridging of a critical‐sized long bone defect by the formation of cortical‐like bone tissue with a medullary cavity containing bone marrow with the absence of fibrous tissue. In addition, structural characterization of the regenerated defect showed high similarities to native tibia. 3 Discussion In this work, we developed a bottom‐up modular strategy for scalable biofabrication of cartilage intermediate tissues that were able to form ossicles without contaminating tissue compartments while exhibiting a unique capacity to heal critical‐sized long bone defects. During native fracture healing, cells from the periosteum are the main contributors of the callus. 14, 29 These cells have recently been shown to possess a higher regenerative capacity than bone marrow mesenchymal cells and contain a skeletal stem cell population with distinct functions during endogenous bone repair. 15, 30 Moreover, it was recently reported that periosteum contains not only renewable skeletal stem cells forming membranous, cortical bone, but also endochondral bone upon damage. 59 Hence, this understudied progenitor cell source possesses critical advantages in terms of clinical application for the design of engineered ATMPs aiming to heal large long bone defects. To date, the formation of cartilage intermediates in vitro was obtained through the use of pellets containing large amount of cells (>2 × 10 5 cells). 17, 18, 60, 61 However, the use of such methods has resulted in diffusion‐related challenges such as the formation of undifferentiated tissue compartments in vitro which hinder the concerted progression of tissue maturation to their final phenotype upon implantation. 18 This was also detected in our study by the large fibrotic compartments encountered within macropellet explants (Figure 3 ). In addition, when chondrogenic 61 and hypertrophic 62 pellets were fused into larger structures, limited remodeling in vivo was shown. Hence, we designed cell microspheroids (comprised of 250 cells) that would not exceed 150 µm in diameter to match with the length scale that diffusible signals can be transported and to mimic the initial developmental event of growth plate formation (condensation) whereby only a few hundred cells are needed. 63 During differentiation, we observed that cells underwent a cascade of molecular and cellular events that reflect endochondral ossification allowing them to transform from cellular spheroids to semiautonomous microtissue structures, callus organoids, capable of undergoing organogenesis (Figures 1, 2, 3 ). In addition, the assembly of callus organoids into larger tissue structures resulted in implants containing active regenerative components throughout their structure. Ultimately, the populations of callus organoids described in our study could be viewed as a living “bio‐ink” that also allows the formation of scaffold‐free tissue structures with intricate geometric features (Figure S6, Supporting Information). 64 In order to obtain quality attributes that could be linked to the functionality of the callus organoids, high‐depth transcriptomic profiling was carried out. Sets of genes were determined to provide signatures to identify whether engineered microtissue niches have attained the degree of autonomy required for bone organ formation. These were compared to recent studies focused on the identification of transcription factor panels that control differentiation transitions from one zone to the other in the growth plate. 48, 51 With this comparison, we were able to discern similar temporal gene regulation kinetics and link the phenotypic state and semiautonomous function of our microtissues to that of an “early pre‐hypertrophic” stage for day 14 modules (microtissues) and that of “late pre‐hypertrophic” stage for day 21 modules (callus organoids). In addition, GO term analysis of the 400 most variable genes revealed additional etiologies (Figure 5 b) for the striking bone organ formation observed in our study. Upregulated clusters indicated gradual transition to pre‐hypertrophy, favoring mineralization as well as ECM disassembly and organization (Figure 5 b). Apart from the relevance of ECM disassembly and organization in the transition from hypertrophic cartilage to bone, 65 this property could also be key in regulating the orchestrated transition of the multimodular constructs into a single ossicle by facilitating ECM reorganization and vascular invasion across the implant. This could also explain the rapid vascularization and bone marrow formation of day 21 constructs observed as early as 4 week postimplantation (Figure 4 ) as well as host integration in the long bone defect (Figure 6 ). Chan et al. 66 have previously demonstrated the importance of endochondral ossification for the formation of hematopoietic stem‐cell (HSC) niches, and here we provide a set of metrics that would allow fine tuning and robust bone organ formation. Although scaffold‐free constructs are beneficial for mimicking native tissue morphology, a combination of the callus organoids with suitable biomaterials could further enable upscaling into centimeter‐sized implants and even enhance their performance. 67, 68, 69 Functionalized biomaterials possessing molecular signatures relevant to the timescales of the differentiation cascades and the proper length scale could interact and support endochondral ossification events. Petersen et al. recently demonstrated that the architecture of collagen scaffolds can direct endochondral fracture healing in vivo. 70 They showed that scaffold pores oriented along the defect resulted in ECM alignment and controlled invasion of progenitor cells and blood vessels leading to the onset of endochondral ossification. In the present work, we observed rapid vascularization and bone formation which was attributed to active ECM remodeling, a dynamic property that could further be supported by properly designed scaffolds through the delivery of relevant enzymes. 71 Localized delivery of growth factors through tailored biomaterials could further direct tissue maturation in vivo while avoiding release of supraphysiological levels, which for BMP‐2 has been proven to cause severe side effects including swelling and heterotopic bone formation. 72 Herberg et al. demonstrated that a combination of BMP‐2 and TGF‐β1 releasing microparticles in cell‐based constructs resulted in mineralized bridging in tibia defects which was further enhanced by mechanical stimulation of the defect. 73 Furthermore, nanoscale fibronectin coatings on polycaprolactone scaffolds were shown to allow incorporation of ultralow dose BMP‐2 (100 ± 8 ng cm −2 ) resulting in bone formation in vivo. 74 However, it is of note that the use of biomaterials could also have adverse effects for tissue regeneration when their properties are not coupled to the precise regenerative context. For example, collagen I scaffolds used in both clinical and research applications for bone regeneration were recently shown to impede osteogenic differentiation and fracture healing. 75 This highlights the importance of thorough understanding of the interaction between the scaffold material and the biology for a specific application. There are still a number of technical challenges that need to be addressed for future biomanufacturing of callus organoids for mass production. The transition of the static process developed in this study to bioreactor systems where thousands of organoids could be generated could aid in its full automation and enhance its capability. In addition, the transfer of this process to stirred bioreactor systems could potentially allow increased flexibility in terms of achievable scale. 76 At the same time, already available technologies for isolating single microtissue modules for at‐line quality controls could provide an ideal method for real‐time evaluation of their degree of autonomy 77 allowing the implementation of real‐time potency monitoring as envisaged in the quality by design paradigm for cell therapy. Bioprinting technologies with the capacity to manipulate single spheroids have been developed through laser‐induced forward transfer, a high‐resolution method using laser pulses. 78 Finally, robotic devices have been shown to possess the capacity to manipulate single spheroids and positioning them in preordered grids allowing them to fuse 79, 80 or depositing them in printed scaffolds. 78 Another technical bottleneck that will need to be addressed is the vascularization of multicentimeter‐sized implants. Although chondrocytes possess resistance to stress conditions found at the implantation site such as hypoxia and low nutrient availability, it is expected that vascularization will be a prerequisite for cell survival in large implants. 81 Recently, vascularized structures based on the concomitant use of mesenchymal condensations, of similar dimensions to the ones presented here and endothelial cells, exhibited improved in vivo functionality. 8 Moreover, sacrificial writing into functional tissue (SWIFT) bioprinting with direct fabrication of vasculature in organoid suspensions could also be employed for introducing vasculature patterns when upscaling to larger callus‐organoid‐based implants. 82 In addition, using purified stem cell populations recently described by Chan et al. 83 could substantially enhance the potential and efficiency of the strategy described in this work. 4 Conclusion In conclusion, the described callus organoids provide an engineering approach for predictive design of large‐scale living implants. The callus organoids exhibited a deterministic behavior by reaching autonomy thresholds attributed to synchronized activation of molecular pathways providing robustness and potentially facilitating regulatory approval and safety. Furthermore, this process is scalable both in terms of production of single callus organoids and in terms of tissue implant size and at the same time allowing the design of intricate geometric features. Importantly, the in vivo functional assessment of orthotopic bone formation with bridging of the long bone defect took place within the timelines of natural fracture healing and resulted in a bone structure highly resembling native long bone. 12 With these advancements, we believe that future biofabrication of skeletal implants using callus organoids will follow design principles resulting in achieving “bone by design”. This will eventually pave the way for the biomanufacturing of clinically relevant implants possessing robust functionality and causal connection with the clinical outcome. This can revolutionize the mitigation of currently unmet clinical challenges such as healing of critical‐size long bone defects. 5 Experimental Section Cell Expansion : hPDCs were isolated from periosteal biopsies of nine different donors, and two different cell pools were created (ages of 29 ± 12 and 14 ± 3 years) as previously described. 84 The hPDC pools were expanded (5700 cells cm −2 ) until passage 7 (in vivo, RNA‐seq) and 10 (in vitro) at 37 °C, 5% CO 2, and 95% humidity in Dulbecco's modified Eagle medium (DMEM, Life Technologies, UK) with 10% fetal bovine serum (HyClone FBS, Thermo Scientific, USA), 1% antibiotic–antimycotic (100 units mL −1 penicillin, 100 mg mL −1 streptomycin, and 0. 25 mg mL −1 amphotericin B), and 1 × 10 −3 m sodium pyruvate (Life Technologies, UK). Medium was changed every 2–3 days, and cells were harvested with TrypLE Express (Life Technologies, UK) at a confluence of 80–90%. TrypLE Express was used for all passaging and harvesting steps during cell handling. The ethical committee for Human Medical Research (Katholieke Universiteit Leuven) approved all procedures, and patients' informed consent forms were obtained (ML7861). Formation of Microspheroids : Agarose microwell inserts for formation of a high number of microspheroids with homogeneous size distribution were created as previously described by Leijten et al. 85 Briefly, 3 % (w/v) Agarose (Invitrogen, Belgium) was poured onto a polydimethylsiloxaan (PDMS, Dow Corning Sylgard 184 elastomer, MAVOM Chemical Solutions) master mould containing pillars with a diameter of 200 µm. The agarose was let to solidify where after microwell inserts with an area of ≈1. 8 cm 2 were punched out, placed in 24‐well plates, 1 mL of phosphate‐buffered saline (PBS; Lonza, Verviers, Belgium) was added and the wells were sterilized under UV for 30 min. Each well insert contained ≈2000 microwells. hPDCs were harvested and seeded with a concentration of 500 000 cells per well to obtain ≈250 cells per spheroid after self‐aggregation. Microspheroids were differentiated into microtissues in a serum‐free chemically defined chondrogenic medium (CM) containing LG‐DMEM (Gibco) supplemented with 1% antibiotic–antimycotic (100 units mL −1 penicillin, 100 mg mL −1 streptomycin, and 0. 25 mg mL −1 amphotericin B), 1 × 10 −3 m ascorbate‐2 phosphate, 100 × 10 −9 m dexamethasone, 40 µg mL −1 proline, 20 × 10 −6 m of Rho‐kinase inhibitor Y27632 (Axon Medchem), ITS+ Premix Universal Culture Supplement (Corning) (including 6. 25 µg mL −1 insulin, 6. 25 µg mL −1 transferrin, 6. 25 µg mL −1 selenious acid, 1. 25 µg mL −1 bovine serum albumin (BSA), and 5. 35 µg mL −1 linoleic acid), 100 ng mL −1 BMP‐2 (INDUCTOS), 100 ng mL −1 growth/differentiation factor 5 (GDF5) (PeproTech), 10 ng mL −1 TGF‐β1 (PeproTech), 1 ng mL −1 BMP‐6 (PeproTech), and 0. 2 ng mL −1 basic FGF‐2 (R&D systems). 86 Half of the media volume was changed every 3–4 days. Viability Assay : Cell viability in microspheroids was assessed qualitatively with LIVE/DEAD Viability/Cytotoxicity Kit (Invitrogen, USA) for mammalian cells by following the manufacturer's protocol. Briefly, microspheroids were rinsed with PBS, where after they were incubated in 2 × 10 −6 m Calcein AM and 4 × 10 −6 m ethidium homodimer‐1 for 30 min at 37 °C, 5% CO 2, and 95% humidity. Stained microspheroids were visualized with a confocal microscope ZEISS LSM 510 META (Cell imaging core facility of KU Leuven) with 4 µm thick slices. Cell Proliferation Assay : Cell proliferation during microspheroid differentiation was visualized with Click‐iT EdU Imaging Kit (Life Technologies, USA) according to the manufacturer's protocol. Briefly, 10 × 10 −6 m EdU was added to the microspheroids during 4 days for each time point. Next, samples were fixed in 4% paraformaldehyde (PFA), EdU was detected with Alexa Fluor azide, stained with Hoechst 33 342 (5 µg mL −1 ) followed by visualization with a Leica M165 FC microscope (Microsystems, Belgium). The percentage of EdU/Hoechst (proliferating per all cells) stained area was quantified using ImageJ software 87 for 10–15 microspheroids per time point. Cytoskeleton and Nuclei Visualization : Cell nucleus and F‐actin distribution within microspheroids was visualized by staining with 2. 5 µg mL −1 4′, 6‐diamidino‐2‐phenylindole (DAPI) (Invitrogen) and 0. 8 U mL −1 Alexa Fluor 488 phalloidin (Invitrogen) during 1 h at room temperature. Stained spheroids were imaged with an inverted laser scanning fluorescence confocal microscope ZEISS LSM 510 META (Cell imaging core facility of KU Leuven) with 1 µm thick slices using an argon ion 488 nm and MaiTai laser. DNA Quantification, Total RNA Extraction, and Quantitative Reverse Transcription–Polymerase Chain Reaction Analysis : Quantitative real‐time polymerase chain reaction (qRT‐PCR) was used to quantify mRNA of markers relevant for endochondral ossification. Pooled microspheroids (≈2000 microspheroids represent n = 1) were washed in PBS followed by cell lysis in 350 µL RLT lysis buffer (Qiagen, Germany) and 3. 5 µL β‐mercaptoethanol (Sigma Aldrich, Germany), vortexed and stored at −80 °C. DNA assay kit QuantiT dsDNA HS kit (Invitrogen) was used to quantify the DNA content for each condition. Cell lysate was spun down and the DNA assay was performed according to the manufacturer's protocol. RNeasy Mini Kit (Qiagen) was used to isolate the total amount of RNA from lysed cells. After RNA extraction, the RNA concentration was quantified with NanoDrop 2000 (Thermo Scientific), and sample purity was evaluated at A260/A280 (protein purity; ≈2. 0+) and A260/A230 (salt purity; 2. 0–2. 2). RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Scientific, USA) was used for reverse transcription; 500 ng of RNA was mixed with 1 µg of oligo (dT18) for each reaction (5 min at 65 °C). The reaction mixture (4 µL 5× reaction buffer, 1 µL ribolock ribonuclease inhibitor, 2 µL dNTPmix (10 × 10 −3 m ), and 1 µL RevertAid H Minus M‐MuL VRT) was added to the samples and run in Applied Biosystems Veriti 96‐Well Fast Thermal Cycler (60 min at 42 °C followed by 10 min at 70 °C). qRT‐PCR was further performed with the cDNA, SYBR Green (Life Technologies) and primers designed for the specific human markers in cycling: 95 °C, 3 s; 60 °C, 20 s. Glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) was used as house‐keeping gene and relative differences in expression were calculated using the 2 −ΔΔ Ct method. 88 RNA Sequencing : RNA isolation from samples ( n = 3–4) was performed as described above. The Genomics Core Leuven performed the sequencing and the RNA‐seq expression analysis as follows. Library preparation was performed with the Illumina TruSeq Stranded mRNA Sample Preparation Kit, according to the manufacturer's protocol. Denaturation of RNA was performed at 65 °C in a thermocycler and cooled down to 4 °C. Samples were indexed to allow for multiplexing. Sequencing libraries were quantified using the Qubit fluorometer (Thermo Fisher Scientific, MA, USA). Library quality and size range were assessed using the Bioanalyzer (Agilent Technologies) with the DNA 1000 kit (Agilent Technologies, CA, USA) according to the manufacturer's recommendations. Each library was diluted to a final concentration of 2 × 10 −9 m and sequenced on Illumina HiSeq4000 according to the manufacturer's recommendations generating 50 bp single‐end reads. A minimum of 14M reads per sample were produced. Quality control of raw reads was performed with FastQC v0. 11. 5. Adapters were filtered with ea‐utils v1. 2. 2. 18. Splice‐aware alignment was performed with TopHat v2. 0. 13 against the human hg19. The number of allowed mismatches was 2. Reads that mapped to more than one site to the reference genome were discarded. The minimal score of alignment quality to be included in count analysis was 10. Resulting sequence alignment map (SAM) and binary alignment map (BAM) alignment files were handled with Samtools v0. 1. 19. 24. Quantification of reads per gene was performed with HT‐Seq count v0. 5. 3p3. Count‐based differential expression analysis was done with R‐based (The R Foundation for Statistical Computing, Vienna, Austria) Bioconductor package DESeq. Reported p ‐values were adjusted for multiple testing with the Benjamini–Hochberg procedure, which controls false discovery rate (FDR). A list of differentially expressed genes was selected at an FDR of 0. 05. Formation of Microtissue Constructs : Macrowells with a diameter and a depth of 2 mm (ectopic implantation) and a length of 5 mm, a width of 3 mm, and a depth of 2 mm (orthotopic implantation) were created with 3% w/v agarose (Invitrogen, Belgium) and sterilized under UV. Microtissues were recuperated from their microwells by gently pipetting up and down several times. The microtissue suspension was concentrated with centrifugation to a volume corresponding to the macrowells. Next, the microtissues were added into the macrowells (≈3000 for ectopic and ≈6000 for large bone defect implantation) and incubated for 1 h to sediment, where after CM was added and constructs were incubated for additional 23 h to fuse into constructs. In Vivo Implantation of Microtissue Constructs : Subcutaneous implantation was used to validate the construct's autonomy to form cartilage and bone tissue. Bone and cartilage do not naturally form in this location and chondro‐ and osteo‐inductive signals must therefore arise from the construct itself. After 24 h fusion, the microtissue constructs were implanted subcutaneously in immune compromised mice ( Rj :NMRI nu/nu ). Explants were taken out 4 and 8 weeks after in vivo implantation and fixed in 4% PFA for subsequent nano‐CT and histological analysis. A large bone defect mouse model, described elsewhere, 58 was used to assess the impact of the environment and mechanical loading on the bone forming potential of the day 21 microtissue constructs. Briefly, a custom‐made Ilizarov fixator was fixed to the tibia using 27 G steel needles. The tibia was exposed, and a 4 mm mid‐diaphyseal segment was removed with a diamond saw. Custom‐made constructs (≈6000 callus organoids per construct, n = 4) were placed into the defect, and the skin was sutured to close the wound. An empty defect was used as control ( n = 4). Defects were monitored with in vivo micro‐CT (SkyScan 1076, Bruker micro‐CT, BE) 1, 2, 4, 6, and 8 weeks after surgery (voxel size of 9 µm). Animals were sacrificed after 8 weeks; the tibia was fixed in 4% PFA and analyzed with ex vivo nano‐CT and processed for histology. All procedures on animal experiments were approved by the local ethical committee for Animal Research, KU Leuven. The animals were housed according to the regulations of the Animalium Leuven (KU Leuven). Quantification of Mineralized Tissue from In Vivo Micro‐CT and Ex Vivo Nano‐CT : Ex vivo nano‐CT (Pheonix Nanotom M, GE Measurement, and Control Solutions) was used for 3D quantification of mineralized tissue in each explant. Explants were scanned with a diamond target, mode 0, 500 ms exposure time, 1 frame average, 0 image skip, 2400 images, and a 0. 2 mm aluminum filter. Subcutaneous explants were scanned at a voltage of 60 kV and a current of 140 µA resulting in a voxel size of 2 µm. Large bone defect explants and native tibia were scanned at a voltage of 60 kV and a current of 390 µA resulting in a voxel size of 5. 6 µm. CTAn (Bruker micro‐CT, BE) was used for all image processing and quantification of mineralized tissue based on automatic Otsu segmentation, 3D space closing, and despeckle algorithm. Percentage of mineralized tissue was calculated with respect to the total explant volume. CTvox (Bruker micro‐CT, BE) was used to create 3D visualization. Histochemistry and Immuno‐Histochemistry : Retrieved subcutaneous explants were fixed in 4% PFA overnight and decalcified in ethylenediaminetetraacetic acid (EDTA)/PBS (pH 7. 5) for 10 days at 4 °C followed by paraffin embedding. Tibias were fixed in 2% PFA overnight and decalcified in EDTA/PBS (pH 7. 5) for 3 weeks then dehydrated and embedded in paraffin. Ectopic samples were sectioned at 5 µm and tibias at 6 µm. Histology was performed according to previously reported methods of H&E, Alcian Blue, Masson's Trichrome, and Safranin O staining. 10 Immuno‐histochemistry was performed on PFA‐fixed microtissues (Osterix), paraffin‐embedded PFA‐fixed microtissues (Indian Hedgehog), and paraffin‐embedded EDTA‐decalcified explants (human osteocalcin, CD31). Epitope retrieval was performed with Uni‐Trieve (INNOVEX Bioscience, USA) for 30 min at 70 °C. Quenching of endogenous peroxidase activity was performed with 3% H 2 O 2 for 10 min. Next, sections were blocked in serum for 30 min and incubated overnight at 4 °C with the primary antibodies human osterix (R&D Systems, MAB7547: dilution 1:300), human osteocalcin 29 (a gift from E. Van Herck, Legendo, KU Leuven, BE; dilution 1:5000), rabbit polyclonal anti‐Ihh antibody–N‐terminal (Abcam, ab80191; dilution 1:50), rabbit anticollagen type II (Merck Millipore, AB761; dilution 1:50), or purified rat antimouse CD31 (BD Biosciences, USA, 550 274; dilution 1:50). Next, slides were blocked and incubated with the secondary antibodies Alexa 488 antimouse (Thermo Fisher Scientific, A11001; dilution 1:500), horseradish peroxidase (HRP) conjugated goat anti‐guineaPig or—rabbit (Jackson ImmunoResearch, UK; dilution 1:500) for 30 min and peroxidase activity was determined using 3, 3′‐diaminobenzidine (DAB) (K3468, Dako, USA). For detection of CD31, the secondary antibody Biotin conjugated Goat‐anti‐Rat Ig (BD Biosceinces, USA, 559 286) and a tyramide signal amplification (TSA) Biotin detection system (PerkinElmer, USA) were used. Stained histology sections were visualized with a Leica M165 FC microscope (Microsystems, Belgium) or an inverted laser scanning fluorescence confocal microscope ZEISS LSM 510 META (Cell imaging core facility of KU Leuven). Histomorphometry was performed in ImageJ software using ROI manager 87 on three to four nonconsecutive sections per sample, and mean values from these sections were used as data point for one sample. Transcriptomics Analysis : An unsupervised analysis of the RNA‐seq data and subsequently gene visualization was performed. For this, a [gene × experimental condition] matrix was obtained from the bulk RNA‐seq data. First genes were ranked based on variance, and then the gene expression profile of 400 most variable genes across four time points was selected for downstream analysis. Gene expression values were mean and log2‐normalized. Then, k ‐means clustering was used to computationally cluster these genes based on their expression profiles. In order to select the number of clusters, the elbow method was applied and determined that k = 5 was the optimal parameter for achieving the most robust partition. Clustering results were visualized in order to provide insight into the patterns of correlation between samples and expression levels. A profile plot, also known as parallel coordinate plot was plotted using ggplot2—a package for data visualization within the R ‐statistical computing environment ( http://www. r-project. org/ ) in order to visualize the expression levels of a total of 400 gene transcripts across all four time points including k ‐means cluster information. Subsequently, Gene Ontology enrichment of Biological Processes (2017b) for each cluster was performed with Enrichr. 38, 39 Statistical Analysis : All experiments were performed with at least three replicates per condition. Data were represented as mean ± standard error of the mean (SEM) or box‐plot with 10–90 percentiles, if otherwise not stated. Data were compared with one‐way or two‐way ANOVA and Tukey's Multiple Comparison test or Student's t test. Results were considered statistically different for p‐ values lower than 0. 05 (* p < 0. 05, ** p < 0. 01, *** p < 0. 001). Statistical analysis was performed with GraphPad Prism 8 (GraphPad Software, Inc. , USA) unless otherwise stated. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file. Supplemental Movie 1 Click here for additional data file. Supplemental Movie 2 Click here for additional data file. Supporting Information Click here for additional data file. Supporting Information Click here for additional data file. Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902307
| 2,020
|
Advanced Science
|
Multi‐Material 3D and 4D Printing: A Survey
|
Abstract Recent advances in multi‐material 3D and 4D printing (time as the fourth dimension) show that the technology has the potential to extend the design space beyond complex geometries. The potential of these additive manufacturing (AM) technologies allows for functional inclusion in a low‐cost single‐step manufacturing process. Different composite materials and various AM technologies can be used and combined to create customized multi‐functional objects to suit many needs. In this work, several types of 3D and 4D printing technologies are compared and the advantages and disadvantages of each technology are discussed. The various features and applications of 3D and 4D printing technologies used in the fabrication of multi‐material objects are reviewed. Finally, new avenues for the development of multi‐material 3D and 4D printed objects are proposed, which reflect the current deficiencies and future opportunities for inclusion by AM.
|
1 Introduction The world's major industrial countries are promoting 3D printing or additive manufacturing (AM) as a technology foundation of future manufacturing. Due to special characteristics of AM such as facile and customizable manufacturing, this method is being broadly used in many areas such as electronics, aerospace, robotics, and textile. [ 1 ] With the emerging of smart materials, attempts to combine them with AM led to 3D parts that are activated by external stimuli and/or environment over time (i. e. , 4D (4‐dimensional) printing). [ 2 ] Current initiatives in the development of AM tools involve development of multi‐material 3D and 4D printing. Using multi‐material 3D and 4D printing, it is feasible to ameliorate the quality of parts by altering composition or type of materials within the layers; that is not easy to obtain by conventional manufacturing methods. A wide range of materials such as polymers, metals, ceramics, and biomaterials has been used in various AM methods to obtain multi‐material products. Therefore, a thorough understanding of multi‐material 3D and 4D printing is required. The number of research pertaining the additive manufacturing of multi‐material parts has steadily increased since 2010. Some surveys of multi‐material AM already exist in the literature. For example, review articles [ 3, 4, 5 ] covered some of the research done on the multi‐material printing prior to 2017. In response to the growing interest in this area, the present article aims to provide a broader and updated review on 3D and 4D printing of multi‐material parts, by providing a comprehensive list of multi‐material additive manufacturing methods published in the literature. Here, we review the technologies and applications of multi‐material 3D and 4D printing. We first consider the main technologies for printing multi‐material objects. Next, we describe the multi‐material 3D and 4D printing for different types of materials: polymers, metals, ceramics, and biomaterials. Finally, we discuss the limitations of current technologies and the challenges for future research. To limit the scope of our survey, the emphasis has been on the additive manufacturing of parts made of discrete multiple materials. Publications directly related to other aspects of multi‐material additive manufacturing, such as the raw materials are premixed or composited before the 3D printing, or porous materials suitable for secondary material infiltration have been excluded from this survey. In the current survey, some of the publications may have multiple citations. For instance, a paper in which both polymer and biomaterial study are presented will be cited in the “Multi‐material additive manufacturing of polymers” as well as the “Multi‐material additive manufacturing of biomaterials” sections. 2 Multi‐Material Additive Manufacturing Technologies Multi‐material additive manufacturing systems may be classified based on the technology, feed stock, source of energy, build volume, etc. Based on the ISO/ASTM 529000:2015 standard, AM methods can be classified into seven different categories and examples of AM processes are depicted in Figure 1. Figure 1 Classification of additive manufacturing technologies; the seven categories: material extrusion, vat photopolymerization, binder jetting, material jetting, sheet lamination, directed energy deposition, and powder bed fusion. Multi‐material additive manufacturing technology can reduce production time with no extra cost for manufacturing parts with complex morphology. As shown in Figure 1, seven AM technologies are currently available commercially, with each having its own advantages and limitations. An overview of some of these technologies and a summary of their main advantages and disadvantages are provided in Table 1. Although there is a variety of commercially available 3D printers, only a limited number enables the production of 3D parts with multiple materials. Tables 2 and 3 list some of the commercially available multi‐material 3D printers and their specifications for polymers and biomaterials, respectively. Table 1 Additive manufacturing technologies Technology Method Process description Advantages/disadvantages Application areas Vat photopolymerization Stereolithography (SL) Synonyms: SLA SL makes use of a photopolymer liquid as the source material in a vat. This liquid plastic is transformed into a 3D object layer‐by‐layer by lowering the build platform into the vat and curing using a UV laser. ⊕ Can build large parts with very good accuracy and surface finish ⊖ Works with photopolymers which are not stable over time and do not have well defined mechanical properties. Prototypes, casting patterns, jewelry, dental, and medical applications Digital light processing (DLP) DLP technology is very similar to SL but uses a different light source and makes use of a liquid crystal display panel. ⊕ Higher print speed compared with SLA ⊕ Excellent accuracy of laying ⊕ Low cost printers ⊖ Insecurity of the consumable material ⊖ High cost of materials Prototypes, casting patterns, jewelry, dental, and medical applications Continuous direct light processing (CDLP) CDLP works similar to DLP except it relies on the continuous motion of the printing bed in the z ‐direction (upward). Faster build times are possible as the printer does not have to stop and separate the object from the printing bed after each layer is printed. ⊕ High print speed ⊕ Excellent accuracy of laying ⊕ Low cost printers ⊖ Insecurity of the consumable material ⊖ High cost of materials Prototypes, casting patterns, jewelry, dental, and medical applications Material extrusion Fused deposition modeling (FDM) Synonyms: Fused filament fabrication, FFF Fused layer modeling/manufacturing, FLM A plastic filament is melted and extruded through a nozzle. Objects are built layer‐by‐layer. ⊕ Can build fully functional parts in standard plastics ⊖ Printed parts have an anisotropy in the z ‐direction (vertical direction) and a step‐structure on the surface Prototypes, support parts (jigs, fixtures), small series parts Direct ink writing (DIW) Synonyms: Robocasting (RC), direct‐write assembly (DWA), or microrobotic deposition (μRD), bioplotting, low‐temperature deposition manufacturing (LDM), freeform 3D printing, extrusion freeform fabrication (EFF) Material in a semi‐liquid or paste form can be extruded through a nozzle and used to print the cross sections of a sliced 3D model. ⊕ Highest resolution for an extrusion system ⊕ Ideal for research environments and medical (bone) applications ⊖ Limited part geometry ⊖ High cost of system ⊖ Small build volume Solid monolithic parts, scaffolds, biologically compatible tissue implants, tailored composite materials, ceramics Binder jetting (BJ) 3D printing, BJ Inkjet printing heads jet a liquid‐like bonding agent onto surface of powder. By bonding the particles together, the object is built up layer‐by‐layer. ⊕ A rather fast and cheap technology ⊕ Wide range of material types ⊕ Parts in full color are possible ⊖ Parts coming directly from the machine have limited mechanical properties Prototypes, casting patterns, molds and cores Material jetting (MJ) Multijet modeling, drop on demand, DOD, thermojet, inkjet printing Inkjet printing head jets molten wax onto a printing bed. Once the material is cooled and solidified, it allows to fabricate layers on top of each other. ⊕ Can achieve very good accuracy and surface finishes ⊖ Only works with wax‐like materials Prototypes, casting patterns Polyjet modeling, multijet modeling, polyjetting, multijetting, jetted photopolymer Similar to multijet except printing head jets liquid photopolymers onto a printing bed. The material is immediately cured by UV light and solidified which allows to build layers on top of each other. ⊕ Different materials can be jetted together to achieve multi‐material and multi‐color objects ⊖ Works with UV‐active photopolymers which are not durable over time Prototypes, casting patterns, tools for injection molding Powder bed fusion (PBF) Laser sintering (LS) Synonyms: Selective laser sintering, SLS SLS has some similarities with SL. A thin layer of plastic powder is selectively melted by a laser. The parts are built up layer‐by‐layer in the powder bed. ⊕ Can manufacture parts in standard plastics with good mechanical properties ⊕ A constantly growing set of materials available ⊖ Parts do not have exactly the same properties as their injection molded counterparts Prototypes, support parts, small series parts Selective laser melting, SLM; direct metal laser sintering, DMLS; laser cusing The LS process is very similar to the LM process. A thin layer of metal powder is selectively melted by a laser. The parts are built up layer by layer in the powder bed. ⊕ Can manufacture parts in standard metals with high density, which can be further processed as any welding part ⊖ Is rather slow and expensive ⊖ Surface finishes are limited Prototypes, support parts (jigs, fixtures, etc. ), small series parts, tools Electron beam melting (EBM) A thin layer of metal powder is selectively melted by an electron beam. The parts are built up layer by layer the in the powder bed. ⊕ Parts can be manufactured in some standard metals with high density by electron beam melting ⊖ The availability of materials is limited ⊖ The process is rather slow and expensive Prototypes, small series parts, support parts Multijet fusion (MJF) MJF is basically a combination of the SLS and MJ technologies. A carriage with inkjet nozzles deposits fusing agent on a thin layer of plastic powder in which it selectively melted with a high‐power IR energy source. ⊕ High production speed ⊖ The availability of materials is very limited Prototypes, production parts, housings Directed energy deposition (DED) Laser engineered net shaping (LENS) Uses a high power laser to melt metal powder that is deposited onto the table. Metal is sprayed onto the focal point on the laser where the metal becomes fused together. An inert gas is used to shield the metal from atmospheric gases. It uses a layered approach to manufacture the components. ⊕ Can be used to repair parts as well as fabricate new ones ⊕ Has a very good granular structure ⊕ Powder forming methods have only few material limitations ⊕ The properties of the material are similar or better than the properties of the natural materials ⊖ Some post‐processing involved ⊖ The part must be cut from the build substrate ⊖ Has a rough surface finish, ⊖ May require machining or polishing ⊖ Low dimensional accuracy Fabrication and repair of injection molding tools, fabrication of large titanium and other exotic metal parts for aerospace applications Electron beam additive manufacture (EBAM) Uses an electron beam as the heat source to weld and create metal parts using wire or metal powder. The method is similar to LENS, however, electron beams are more efficient than lasers. ⊕ A wider selection and greater availability of wire products versus powder ⊕ Wire feedstock is cheaper than powder ones ⊕ Less safety and procurement issues compared with LENS ⊕ Significantly less energy consumption compared with powder‐feed method ⊖ Limited to single material printing Fabrication and repair of injection molding tools, fabrication of large titanium and other exotic metal parts for aerospace applications Sheet lamination Laminated object manufacturing (LOM) Layers of paper, plastic, or metal laminates are coated with adhesive and welded together using heat and pressure and then cut to shape with a computer controlled laser or knife. ⊕ Ability to produce larger‐scaled models ⊕ Uses very inexpensive paper ⊕ Fast and accurate ⊕ Good handling strength ⊖ Need for decubing, which requires a lot of labor, can be a fire hazard, and finish, accuracy and stability of paper objects ⊖ Not as good as materials used with other rapid prototyping methods Prototypes, large parts John Wiley & Sons, Ltd. Table 2 Multi‐material polymer and polymer composite 3D printers Technology 3D printer commercial name/Manufacturer (Country) Build volume [mm 3 ] Nozzle type Layer resolution [mm] Stock materials Open source Material extrusion (FDM) Duplicator 5/Geeetech (China) 230 × 150 × 150 Dual 0. 1–0. 3 Filament: ABS/PLA/flexible PLA/wood /nylon No Creater Pro/FlashForge (China) 227 × 148 × 150 Dual 0. 1 ≈ 0. 5 ABS/PLA No CraftBot3/CraftBot (Hungary) 270 × 250 × 250 Dual (separate) 0. 1 ≈ 0. 3 N/A No BCN3D SIGMA R19/BCN3D Technologies (Spain) 210 × 297 × 210 Dual (separate) 0. 05–0. 5 PLA/ABS/nylon/PET‐G/TPU/PVA/composites/others Yes Zortrax Inventure/Zortrax (Poland) 135 × 135 × 130 Dual 0. 09–0. 29 Model materials (Z‐PETG, Z‐PLA, Z‐SEMIFLEX, Z‐ULTRAT Plus) and support materials (Z‐SUPPORT, Z‐SUPPORT Plus) No Makergear M3‐ID/Makergear Head 1: 203 × 232 × 203, Head 2: 180 × 232 × 203 Dual (separate) 0. 02–0. 35 ABS, ASA, HIPS, Nylon, PET‐G, PET‐T, PLA, polycarbonate, polypropylene, PVA, TPE, TPU, metal composites, wood composites, carbon fiber composites No Ultimaker 3/Ultimaker 197 × 215 × 200 Dual 0. 02–0. 6 PLA, tough PLA, ABS, nylon, CPE, CPE+, PC, PP, TPU 95A, PVA Yes 3DWOX 2X/Sindoh 228 × 200 × 300 Dual (separate) 0. 05–0. 4 PLA, ABS, flexible, PVA No Raise3D Pro2/Raise3D 280 × 305 × 300 Dual N/A PLA/ABS/HIPS/PC/TPU/TPE/NYLON/PETG/ASA PP/glass fiber enforced/carbon fiber enforced Metal particles filled/wood fille No LulzBot TAZ Workhorse/LULZBOT (USA) 280 × 280 × 285 Dual 0. 05–0. 4 PLA, ABS, nylon, polycarbonate, carbon fiber reinforced blends, TPU 85A and 95A (flexible), PETG, PETT, copolyester, PVB (polycast), PVA, HIPS, and many more 3rd party filaments Yes ZMorph VX/ZMorph (Poland) 250 × 235 × 165 Dual 0. 05 ≈ 0. 4 ABS, PLA, PVA, PET, ASA, nylon, HIPS, thermochrome, TPU, flex materials No CEL RoboxPRO/CEL (UK) 210 × 300 × 400 Dual 0. 05 ≈ 0. 5 ABS, PETG, PC, nylon, PVOH No Ultimaker S5/Ultimaker 330 × 240 × 300 Dual 0. 02–0. 6 PLA, tough PLA, ABS, nylon, CPE, CPE+, PC, PP, TPU 95A, PVA Yes Material jetting ProJet® MJP 5600/3D systems (USA) 518 × 381 × 300 N/A. 013–016 Flexible and rigid photopolymers within the VisiJet family of materials No Objet260 Connex3/Stratasys (USA) 255 × 252 × 200 N/A 0. 016 Variety of materials such as Vero family No J735/Stratasys (USA) 350 × 350 × 200 N/A 0. 014 Variety of materials such as Vero family No J750/ Stratasys (USA) 490 × 390 × 200 N/A 0. 014 Variety of materials such as Vero family No OBJET1000 PLUS/Stratasys (USA) 1000 × 800 × 500 N/A 0. 016 Variety of materials such as Vero family No Objet Connex350/ Stratasys (USA) 342 × 342 × 200 N/A 0. 016 Variety of materials such as Vero family No Objet Connex500/Stratasys (USA) 490 × 390 × 200 N/A 0. 016 Variety of materials such as Vero family No F900/Stratasys (USA) 914. 4 × 609. 6 × 914. 4 N/A 0. 127–0. 508 Variety of materials such as Vero family No Multi‐Fab/Computational Fabrication Group, Massachusetts Institute of Technology (USA) N/A N/A N/A Variety of materials Yes FDM and MJ (curing by UV) 3Dn DDM/nScrypt (USA) 300 × 300 × 150 Up to 5 0. 0005 Variety of third party materials for both UV assisted and FDM processes Yes FDM and continuous filament fabrication (CFF) Onyx Pro (Desktop)/Markforged (USA) 320 × 132 × 154 1 0. 1 Onyx fiber materials: continuous fiberglass No Mark Two (Desktop)/Markforged (USA) 320 × 132 × 154 1 0. 1 Onyx fiber materials: carbon fiber, fiberglass Kevlar, HSHT fiberglass (high‐strength high‐temperature fiber‐glass) No MARKFORGED X5 (Desktop) / Markforged (USA) 330 × 270 × 200 1 0. 1 Onyx fiber materials: continuous fiberglass No MARKFORGED X7 (Desktop)/Markforged (USA) 320 × 132 × 154 1 0. 1 Onyx fiber materials: carbon fiber, fiberglass Kevlar, HSHT fiberglass (high‐strength high‐temperature fiber‐glass) No John Wiley & Sons, Ltd. Table 3 Multi‐material biomaterial 3D printers Technology 3D printer Build volume [mm 3 ] Printing head Layer resolution [mm] Stock materials Open source Material extrusion (DIW) 3D‐Bioplotter Starter series/ EnvisionTEC (Germany) 150 × 150 × 80 2 0. 1 Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes 3D‐Bioplotter Developer series/EnvisionTEC (Germany) 150 × 150 × 140 Up to 3 0. 1 Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes 3D‐Bioplotter Manufacturer series/EnvisionTEC (Germany) 150 × 150 × 140 Up to 5 0. 1 Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes BioFactory/RegenHU (Switzerland) 60 × 55 × 55 Up to 8 N/A Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes 3Ddiscovery (Bench‐top)/RegenHU (Switzerland) 130 × 90 × 60 Up to 7 N/A Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes BioScaffolder 3. 2 and 4. 2/GESIM (Germany) N/A 3 N/A Any liquid, melt, paste, or gel can be used to be dispensed through a needle tip Yes John Wiley & Sons, Ltd. 3 Multi‐Material Additive Manufacturing of Polymers There have been significant efforts in the scientific community to fabricate multi‐material polymer composites. In this section, we review the related works on polymers and their composites. 3. 1 Vat Photopolymerization A vat of liquid photopolymer (resin) is used by vat photopolymerization, and the model is printed layer by layer using some types of light sources. Stereolithography (SL), digital light processing (DLP), and digital light synthesis (DLS) are the three main vat polymerization techniques. The vat photopolymerization process is not generally a candidate for multi‐material 3D printing. It constructs parts from a vat of photopolymers, and thus using multiple materials in vat photopolymerization provides difficulties with controlling contamination between each vat. However, due to its advantages such as surface finish, accuracy of dimensions, and the options for a variety of materials, vat photopolymerization has been adapted to support multi‐material printing. [ 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 ] This is achieved by using multiple vat systems with different UV‐curable polymers. These systems can provide high printing resolution, but changing materials during printing significantly slows down the printing process. [ 17 ] 3D printing of short and continuous fiber‐reinforced polymer composites using SL was studied by Sano et al. [ 14 ]. Glass powder and fiberglass fabric were used as the short and continuous fiber reinforcement of light‐cured resin materials. The tensile strength and Young's modulus were 7. 2 and 11. 5 times higher than those of the pure resin specimens, respectively. Digital light projection micro‐stereolithography (PµSL) is an additive micro manufacturing method capable of manufacturing arbitrary 3D micro‐scale structures. Using PµSL, Chen and Zheng [ 16 ] fabricated multi‐material metamaterials with big and tailorable negative Poisson's ratios ( Figure 2 ). Their multi‐modulus metamaterials were comprised of encoded elasticity ranging from soft to rigid. The authors found that, in contrast to ordinary architected materials whose negative Poisson's ratio is governed by their geometry, these metamaterials are capable of exhibiting Poisson's ratios from large negative to zero, independent of their 3D micromechanical structure. Figure 2 3D printing of multi‐material microscale lattice with dissimilar materials using digital light projection micro‐stereolithography approach (PµSL): a) 3D multi‐material microscale lattice, b) PµSL setup, c) bimaterial microlattice, d–g) isotropic microscale lattice comprised of different polymers. Reproduced with permission. [ 16 ] Copyright 2018, Springer Nature. 3. 2 Material Extrusion The core principle of material extrusion‐based technologies is that any material that is in a paste or semi‐liquid form can be extruded through a nozzle and used to build layer‐by‐layer a sliced 3D model. Depending on the temperature required or suitable for the extrusion, it can be classified into two main sub‐groups: fused filament fabrication (FFF) or fused deposition modeling (FDM) for extrusion of melted thermoplastic polymers and direct ink writing (DIW) for extrusion without melting. Different terminologies associated with these two categories are provided in Table 1. The material extrusion technology can be easily extended to multi‐material 3D printing through the use of multiple nozzles. 3. 2. 1 Fused Filament Fabrication The FFF uses filaments made of thermoplastic polymers which are melted and extruded through a nozzle on the desired substrate in a layer‐by‐layer manner. Dual or multi‐extruder printing heads are often used in material extrusion systems to print multi‐material parts at once. For example, it is a common method to use one extruder to print dissolvable supports that can be easily removed from the main printed structure, or use them to print in two colors, or two materials that will be present in the end print (see Figure 3 a ). Many multi‐material FFF printers are listed in Table 2. However, dual and multi‐extruder printers typically come with a few limitations: the presence of the additional extruder (second one or more) will reduce the printing area that would be available for printing with a single extruder; the chances of oozing and stringing become higher; and finally layer‐shifting defects may be observed if one of the extruders causes material deposited by the other to warp. Figure 3 Extrusion based multi‐material additive manufacturing: a) traditional FDM; b) in situ fusion of fibers with molten thermoplastic in the nozzle; c) extrusion of pre‐impregnated fibers. Recently, some research groups have improved the mechanical performance of 3D printed polymers by reinforcing them with continuous fibers. Printing using continuous fibers have also been tried using customized FDM printers using polylactic acid (PLA), [ 18, 19, 20, 21 ] Acrylo Butadien Styrene (ABS), [ 20, 21, 22, 23 ] nylon, [ 24, 25, 26, 27, 28, 29, 30 ] and epoxy resin [ 31 ] as the matrix, while carbon, [ 18, 19, 20, 21, 23, 24, 26, 27, 28, 29, 30, 31, 32 ] glass, [ 27, 33 ] and Kevlar [ 25, 27 ] fibers have been used as the reinforcements. Two main configurations for printing continuous fiber‐reinforced composites have been developed as i) in‐situ fusion of fibers with thermoplastic in the nozzle [ 18, 20, 22, 23, 24, 26, 27, 28, 30, 32 ] and ii) extrusion of pre‐impregnated fibers. [ 21, 29 ] The former approach (See Figure 3b ) can be performed by modifying the printing head to receive the continuous fiber and thermoplastic filament, simultaneously. One of the main challenges of the first approach is to have a proper bonding between the reinforcement and the matrix. This is mainly because the printing head cannot generate enough pressure to push the melting resin onto the reinforcing fiber; besides, the short time of impregnation of continuous fibers is another reason. The second approach (see Figure 3c ) is much more complicated compared with the first one, and good impregnation of long fibers is not certain. The method of in situ fusion of fibers with molten thermoplastics was often used by researchers for additive manufacturing of continuous fiber composites. This approach was used by the Markforged company (USA) as the manufacturer of the most common commercially available multi‐material continuous fiber composite printers. [ 24, 25, 26, 27, 28 ] The Markforged line of 3D printers is limited to their own 3D printing Eiger software, each printer only uses one type of specialized and expensive filament, the carbon fiber inlay method and a few other parts of the printer are locked‐down by some patents. Several authors used Markforged 3D printers (Mark One, [ 24, 25, 26, 27 ] Mark Two, [ 28 ] and Mark X [ 30 ] ) for the fabrication of continuous fiber‐reinforced multi‐material composite. For instance, Peng et al. [ 30 ] studied the effect of synergistic reinforcement on the mechanical properties of additively manufactured polyamide‐based composites filled with continuous and short carbon fibers. Morphological, thermal, and mechanical testing for the printing tows were first characterized to obtain the properties of the printing materials. The mechanical properties of laminated composites showed to be higher with increasing continuous carbon fiber content. Aside from Markforged printers, there have been other customized 3D printers that used the approach of in situ fusion of fibers with molten thermoplastics for additive manufacturing of continuous fiber composites. [ 18, 20, 22, 23, 32 ] For instance, Nakagawa et al. [ 32 ] improved the strength of printed thermoplastic parts by sandwiching continuous carbon fibers between upper and lower ABS layers. An FDM‐based 3D printer for manufacturing of continuous fiber‐reinforced thermoplastic composites was developed by Yang et al. [ 23 ]. The authors also developed an extrusion head for the continuous fiber hot‐dipping purpose. Their work showed that the bending and tensile strength of these 10 wt% continuous carbon fiber/ABS specimens were improved to 127 and 147 MPa, respectively. These values were far greater than the ABS parts and close to the continuous carbon fiber/ABS composites manufactured by injection molding process with the same fiber content. The extrusion of pre‐impregnated fibers was also used by some researchers due its advantages such as achieving better bonding between the matrix and the continuous fibers. Fabrication and 3D printing of continuous carbon fiber prepreg filaments were performed by Hu et al. [ 21 ]. The flexural properties of parts printed with the filament were studied. It was found that layer thickness has a significant influence on the final strength and modulus, while the printing temperature and speed had minor influences. By using FDM approach, combined with a continuous toolpath (G‐code), Dickson et al. [ 29 ] produced woven continuous carbon fiber composites. Studies on open hole tensile coupons were conducted in which 6 mm holes were routed into the fiber‐reinforced composite structure and the resulting mechanical performance of the parts were compared with specimens which had been die‐punched as well as an un‐notched control group. The latter showed a strength equivalent to 49% that of unnotched specimen. FDM‐based multi‐material additive manufacturing through a single extruder was also reported in the literature. [ 34 ] In this direct feed FDM technology ( Figure 4 ), multiple materials in any available form can be co‐fed into a single‐screw extruder and subsequently deposited onto the print bed. This technology potentially has the capability to print a structure with controllable and variable compositions. Figure 4 Schematic of a single screw extrusion‐based AM system. Reproduced with permission. [ 34 ] Copyright 2016, AIP Publishing. Khondoker et al. [ 35 ] proposed a customized bi‐extruder for FDM multi‐material additive manufacturing of functionally graded materials (FGMs) made of immiscible thermoplastics (see Figure 5 ). The proposed bi‐extruder can print two thermoplastic polymers through a single nozzle with a static inter‐mixer to enhance the adhesion between feeding materials. The dual‐extruder was characterized by printing parts using PLA, ABS, and high impact polystyrene (PS). It was also observed that the mechanically interlocked extrudates substantially reduce adhesion failures within and between filaments. Figure 5 Custom bi‐extruder for FDM multi‐material additive manufacturing of FGM objects: a) an exploded view of the designed bi‐extruder, b) assembly of the manufactured bi‐extruder, c‐i) the 3D model of the passive inter‐mixer, c‐ii) image of the inter‐mixer fabricated by DMLS, c‐iii) an image showing the inter‐mixer inserted into the bi‐extruder channel. Adapted with permission. [ 35 ] Copyright 2018, Emerald Publishing Limited. The capability of printing complicated patterns of multiple materials by combining multiple printing heads has been useful for antenna engineers. This approach provided them with a powerful tool to rapidly prototype new antenna concepts and the ability to explore ideas not realizable using standard fabrication processes. [ 36, 37, 38 ] An FDM‐based multi‐material additive manufacturing printer (3Dn‐300) sold by nScrypt Inc (see Table 2 ) was used by some authors to fabricate antennas. [ 36, 37, 38 ] For instance, Pa et al. [ 36 ] fabricated a low‐profile antenna that includes an integrated artificial magnetic conducting (AMC) ground plane. This system integrates dual deposition heads in which one print head dispenses Polycarbonate (PC) using a filament extrusion process to print all the dielectric components and the second head prints silver conductive elements using a micro‐dispensing technology. 3. 2. 2 Direct Ink Writing In multi‐material DIW, the paste‐like materials are extruded which do not necessarily need to be polymers or even to be heated. The material filament is deposited using dispensers (usually pneumatic dispensers) that are mounted onto a motion‐controlled positioning stage or a dispensing robot ( Figure 6 ). The printing materials, such as epoxy resins, however, requires certain viscoelastic and rheological characteristics to be smoothly extruded from the printing head. The majority of ink solutions made using such materials show a shear‐thinning rheological behavior characterized by decreasing viscosity with increasing shear rate. Figure 6 DIW printers use pressurized air, piston, or screw for extrusion of materials. [ 115 ] Oxman et al. [ 39 ] proposed a DIW‐based AM platform for the fabrication of FGMs. In their study, a printing head consisting of a nozzle with a mixing unit was installed to the z ‐direction of a gantry robotic machine. The authors printed silicone in two different colors to demonstrate the performance of their 3D printer. Rocha et al. [ 40 ] fabricated graphene‐based electrodes for electrochemical energy storage using inks with thermoresponsive properties. Reduced chemically modified graphene (rCMG) was incorporated in their polymer composites (e. g. , Pluronic F127; BASF) for enhancement of thermo‐electrical properties. The electrochemical performance of their rCMG‐based electrode demonstrated the potential of multi‐material printing in energy applications. Kikkinis et al. [ 41 ] used a DIW‐based 3D printer (3D discovery from RegenHU, see Table 3 ) for multi‐material additive manufacturing of heterogeneous composites under an external magnetic field. Bastola et al. [ 42, 43 ] fabricated multi‐material hybrid magnetorheological elastomers using a BioFactory 3D printer made by RegenHU (Switzerland). In their work, a controlled volume of a magnetorheological (MR) fluid was encapsulated layer by layer into an elastomer (silicone) matrix as shown in Figure 7. Figure 7 a) 1) Schematic of the MR fluid printing system including a piston‐cylinder unit and a printing nozzle, 2) printing cartridges with MR fluid (black) and elastomer matrix (clear); b) schematic for printing of hybrid MR elastomer via DIW; c) the steps involved in printing of hybrid MR elastomer: 1) elastomer matrix deposition to form a bottom layer, 2) bottom layer curing with UV light, 3) printing of MR fluid, 4) elastomer matrix deposition to cover MR fluid patterns, 5) curing with UV light; and d) 3D printed hybrid MR elastomers: 1) dot pattern, 2) line pattern, 3) line pattern with mesh, 4) asterisk shaped pattern, 5) circular pattern; Adapted with permission. [ 43 ] Copyright 2018, Elsevier. 3D printing of fiber‐reinforced thermosetting composites were also reported in the literature. A custom‐made 3D printing platform was used in the work of Hao et al. [ 31 ] to print continuous fiber‐reinforced epoxy composites. The mechanical properties of the composite lamina were characterized in their study. Their results indicated that the mechanical properties of the fabricated epoxy composite were better than that of similar PLA and short carbon fiber reinforced composite ones. Li et al. [ 44 ] combined DIW and microfluidics to manufacture a multi‐material 3D printing system for printing textured composites with liquid inclusions of programmable compositions and distributions. The printing system used was based on commercial LulzBot (Aleph Objects) printer with its original printhead replaced by Objet350 Connex3 printer. Microfluidic chips and the nozzle were integrated to the printhead. The proposed multi‐material microfluidic 3D printing framework could be used to fabricate soft robotic devices. Nassar et al. [ 45 ] used a fully open‐source DIW 3D printer to fabricate flexible smart sensors as shown in Figure 8. The authors modified a RepRap Pro Ormerod 2 desktop 3D printer to include a second printing head for the extrusion of pastes and inks. A silver palladium paste mixed with ethanol was used as the conductive material and Glassbend Flexi was used as the flexible substrate material. With a single‐step procedure for simultaneous printing of structural and functional materials, the authors demonstrated the feasibility of fabricating complex packages with embedded sensing and electronic components. Figure 8 Smart flexible sensing circuit using a modified 3D printer: a) 1) extrusion system for conductive paste, 2) syringe and custom housing for the extrusion mechanism; b) CAD design of the printed structure: 1) the bottom layer with empty cavities, 2) placement of the colored LEDs, 3) silver‐palladium paste printed, 4) structure with the top plastic layer printed embedding the sensor and electronics, 5) y ‐axis bending, 6) x ‐axis bending; c) 1) fabricated multi‐material 3D printed smart sensing structures with the fully embedded blue LED, 2) testing of the fully embedded red LED; and d) Bending test set up to evaluate the embedded printed strain sensor. Adapted with permission. [ 45 ] Copyright 2018, IEEE. Coextrusion of inks has led to 3D printing of wearable textile and sensors. Zhang et al. [ 46 ] developed a single‐step printing of fiber‐reinforced smart patterns for electronic textile (E‐textile) using a Anycubic I3 MEGA 3D printer equipped with a coaxial spinneret as shown in Figure 9. The authors used silk fibroin and CNT ink as the shell and core layer, respectively. In another work, Bodkhe et al. [ 47 ] used DIW to 3D print piezoelectric sensors with their coextruded silver electrodes in a single step (see Figure 10 ). In their work, an I&J 2200–4 (I&J Fisnar) robotic 3D printer was used to coextrude PVDF/BaTiO 3 nanocomposites with a commercially available silver ink to fabricate piezoelectric sensors (Figure 10b ). Their printed piezoelectric sensors successfully worked and the produced voltage was linearly proportional to the applied strain. Figure 9 DIW of core‐shell patterns on fabrics: a) a schematic depicting the coaxial 3D printing; b) picture of the 3D printing process; c) some printed patterns; d) a picture showing the flexibility of the printed textile. Reproduced with permission. [ 46 ] Copyright 2019, Elsevier. Figure 10 a) Schematic of the coaxial printing process, inset: cross section of the coextruded filament illustrating the piezoelectric in the core and the conductive inks as the shell; b) SEM image of cross section of the coextruded filament (scale bar = 1 mm); c) picture of the coextruded piezoelectric thread (scale bar = 500 µm); d) freestanding whiskers printed on a FDM printed cat (scale bar = 10 mm); e) conformal sensors printed on a hemisphere (scale bar = 5 mm); and f) spanning filaments (scale bar = 10 mm). Reproduced with permission. [ 47 ] Copyright 2018, John Wiley and Sons. Multi‐Material FGMs in Material Extrusion FGMs are characterized by composition variation across the part. [ 48, 49, 50 ] The design of heterogeneous compositional gradients is illustrated in Figure 11 a and it can be categorized according to 1D, 2D, and 3D as shown in Figure 11b. Distribution of the materials can also be uniform or through special patterns. Figure 11 a) Schematic of combination of density and compositional gradation within a heterogeneous material, and b) types of gradients classification. Adapted with permission. [ 116 ] Copyright 2018, Elsevier. Both FDM and DIW methods have been used for extruding FGM materials. Different materials can be mixed in a static mixer to form a uniform paste. The directions of depositing each layer and gap sizes between filaments are the important printing parameters that affect the mechanical properties. [ 51 ] Two identically shaped FDM models, but with different deposition densities and orientation of printing were fabricated by Li et al. [ 51 ] to demonstrate the differences in stiffness along the horizontal axis. Oxman et al. [ 39 ] fabricated an FGM made of soft blue silicone (Shore 00–10) mixed with a harder red silicone (Shore 00–50) to fabricate gradients in both color and durometer. Other combinations of materials were tested in their work for potential use on the platform including UV‐curable silicones and polyurethanes. Ren et al. [ 52 ] fabricated polyurethane objects with various gradient patterns (see Figure 12 ). Multiple nonlinear 1D/2D/3D color/Al 2 O 3 concentration gradient objects were successfully manufactured. In their study, the results of the cantilever bending test and simulation showed that the material gradient can effectively relieve the stress concentration. Figure 12 3D printing of objects with spatially non‐linearly varying properties. Reproduced with permission. [ 52 ] Copyright 2018, Elsevier. Figure 13 shows a freeze‐form extrusion fabrication process aimed at printing FGM 3D parts by Leu et al. [ 53 ] The main concept is to mix multiple pastes according to object material composition requirements and to extrude the mixed paste to manufacture a 3D part layer by layer in an environment below the water freezing temperature. On this basis, a triple‐extruder system including the mechanical machine, electronics, and computer software have been developed by the authors. The capability of the developed system was verified by observing the transitions between green and pink colored CaCO 3 pastes and relating them to the measured velocities of the corresponding plungers. Figure 13 3D printing of FGMs using a triple extruder a) schematic; b) the triple‐extruder system in a temperature‐controlled enclosure: three servo motors control linear cylinders for paste extrusion and a three‐axis gantry system controls nozzle movement; c) extrusion of pink and green colored CaCO 3 pastes. The color of the fabricated part starts at pink (c‐A) and shifts to brown (c‐B), then green (c‐C), then brown (c‐D), then pink (c‐E), and finally green (c‐F); d) a fabricated test bar that was graded from 100% Al 2 O 3 to 50% Al 2 O 3 + 50% ZrO 2. Adapted with permission. [ 53 ] Copyright 2012, Elsevier. 3. 3 Powder Bed Fusion Powder bed fusion (PBF) is an AM technology whereby a heat source (e. g. , laser, heated printing head) is used to consolidate a material powder to form 3D parts. The heat source is applied to powder particles which gradually indexes down as each layer is finished and new powder is spread over the build area. The PBF process for polymers includes the following common printing techniques: electron beam melting (EBM), selective heat sintering (SHS), selective laser melting (SLM), and selective laser sintering (SLS). One of the benefits of multi‐material 3D printing when compared to the standard single‐material printing is less possibility of powder cohesion which usually leads to inaccurate part dimensions and poor surface finish. Therefore, a process in which a “build” powder (e. g. , a polymer) is co‐deposited with a non‐fusible “support” powder (e. g. , a different polymer or ceramic) would completely avoid this issue. With multi‐material powder deposition, expensive polymer powders could be placed only where needed, and cheap, fully reusable ceramic powder would form the surroundings to provide mechanical support during the build process. It is clear that such a process could significantly reduce powder waste. Recently, Aerosint SA company (Belgium) has developed a low‐waste, multi‐material 3D printing process based on powder bed fusion technology compatible with both polymer and metal powders. Their prototype is a retrofitted industrial SLS printer in which they have integrated their patterning drums (see Figure 14 ). The process is based on the selective deposition of voxels of powder in a layer‐by‐layer way, with sintering happening uniformly for polymers or via laser for metals that require higher temperatures. Powder bed fusion with multiple polymer powders was used in the work of some researchers. [ 54, 55 ] Laumer et al. [ 54 ] used a simultaneous laser beam melting (SLBM) technique to additively manufacture parts consisting of different polymer powders within one building process. By applying a simultaneous illumination with changeable intensity distribution over a large area, different polymeric powders deposited next to each other within a layer can be transferred simultaneously from a solid into a molten form. Nevertheless, the accurate preparation of arbitrary multi‐material powder layers to still be achieved perhaps by using advanced coating/deposition methods. Another technology to generate multi‐material powder layers can be the electrophotography also known as xerography. [ 55 ] Stichel et al. [ 55 ] demonstrated the application of electrophotographic polymer powder transfer for the SLS‐based preparation of multi‐material layers. An experimental setup with two chambers was designed that enabled the investigation of the electrophotographic powder transfer at typical process conditions of SLS. Their results confirmed the beneficial application of electrophotography for multi‐material powder deposition. Figure 14 A selective powder recoating technology by Aerosint SA. Reproduced with permission. [ 69 ] Copyright 2019, Aerosint SA. 3. 4 Material Jetting Material jetting (MJ) is another AM process that can print multiple materials in the same printing job. MJ creates objects in a similar method to a 2D ink jet printer. Material is jetted onto a build platform using either a continuous or drop on demand (DOD) approach. A list of commercially available multi‐material jetting printers are listed in Table 2. PolyJet (Stratasys Ltd. , USA) is probably the most common commercially available multi‐material jetting process. In this system, the nozzles are able to switch between different materials, including support material. Schematic of material jetting is shown in Figure 15 with the print tray and the respective print head movement (Figure 15a ) and two examples of printed structures (Figure 15b, c ). The hardware and software architectures for these multi‐material printers are locked‐down. Figure 15 a) Schematic of material jetting process. Reproduced with permission. [ 117 ] Copyright 2014, Aerosint SA, and b, c) 3D printed bicycle helmet and shoe. Adapted with permission. [ 119 ] Copyright 2020, Stratasys. Several authors used PolyJet 3D printers for fabrication of multi‐material systems. [ 56, 57, 58, 59, 60, 61, 62 ] Connex3 Objet260, [ 58, 60, 61 ] Connex3 Objet350, [ 59 ] Connex3 Objet500, [ 59 ] and Objet1000 Plus [ 57 ] were used in these studies. Some details of these printers are listed in Table 2. For instance, Keating et al. [ 56 ] used a Stratasys Objet500 Connex multi‐material 3D printer to fabricate a 3D printed multi‐material microfluidic proportional valve. The developed microfluidic valves enabled the development of programmable, automated devices for controlling fluids in a precise manner. Compared to previous single‐material 3D printed valves that are stiff, the multi‐material valves developed by the authors constrain fluidic deformation spatially. This has been done through combinations of stiff and flexible materials, to enable intricate geometries in an actuated, functionally graded device. Cazón‐Martín et al. [ 60 ] analyzed a novel approach that combines lattice structures and a multi‐material additive manufacturing for the design and manufacturing of soccer shin pads. The shin pads were consisting of a sandwich structure: two rigid layers (inner and outer) and a middle layer having a lattice structure that works as a shock‐absorbing geometry. A Connex3 Objet260 printer was used in their study to fabricate the specimens. The developed shin pads were dynamically tested along with two commercially available shin pads using drop weight impact tests. The results showed that two of the specimens have acceleration reductions between 42% and 68% with respect to the commercial ones, while the penetration was reduced by 13–32%. Another commercial multi‐material jetting process is ProJet (3D Systems, USA). The main difference between PolyJet and ProJet is that PolyJet uses a water‐soluble photopolymer as the support material while ProJet uses a wax. As discussed earlier, researchers have studied the PolyJet 3D printers extensively, however, less works can be found on ProJet 3D printer in the open literature. Yang et al. [ 63 ] evaluated the building performance of the ProJet 5500X multi‐material machine. The authors measured the dimensional error and surface roughness of the printed parts and analyzed them using a microscope, a 3D coordinate measuring machine, and a surface profilometer. They found that by using wax as the support material, fine features and lateral features with dimensions as small as 250 µm could all be built properly. Features with high depth and diameter ratios were also possible to be built. The authors also found that the printing accuracy of a material jetting system mainly affected by the accuracy of the printer machinery (such as droplet size and print heat positioning), material property such as shrinkage, and the size and structure of product. Electrohydrodynamic jet (E‐jet) has also been adapted for multi‐material printing using a multi‐nozzle head. E‐jet is a high resolution material jetting printing technology where the printed liquids are driven by an electric field. E‐jet printed droplet ranges from nano‐ to micro‐scales. During the past decade, there has been various applications for E‐jet printing, primarily for biosensing and printed electronics applications. Pan et al. [ 64 ] proposed a multi‐level voltage approach to perform the addressable E‐jet printing utilizing multiple nozzles in parallel with high consistency. The multi‐level voltage approach controls the electric field on each of the nozzles. A good dimensional and position consistency was observed in the printed objects. The authors showed that multi‐level voltage approach is an efficient way to perform the addressable E‐jet 3D printing with several parallel nozzles with high consistency. Custom‐made material jetting printers can also be found in the literature. [ 65, 66 ] MultiFab (see Figure 16 ) is a machine vision assisted platform for multi‐material 3D printer developed by Sitthi‐Amorn et al. [ 17 ]. MultiFab can print simultaneously up to ten different materials. The platform achieves a resolution of at least 40 µm by utilizing piezoelectric inkjet printheads adapted for 3D printing. Unlike previously discussed commercial printers, the hardware and software architectures of MultiFab are extensible and reconfigurable. Moreover, none of the commercial 3D printers uses machine vision system for calibration, 3D scanning, closed feedback loop, and alignment with auxiliary objects. Figure 16 (left) Multifab multi‐material 3D printer and, (right) a set of fabricated materials and objects. Reproduced with permission. [ 17 ] Copyright 2015, ASSOCIATION FOR COMPUTING MACHINERY. 3. 5 Sheet Lamination Laminated object manufacturing technique (LOM) includes layers of adhesive‐coated paper, plastic, or metal laminates that are successively glued together and cut to shape with a knife or laser cutter. Sheet lamination process categories based on the mechanism employed to achieve bonding between layers are gluing or adhesive bonding, thermal bonding, clamping, and ultrasonic welding. Sheet lamination approaches exhibit the speed benefits of a layer‐wise process while still utilizing a point‐wise energy source. In the multi‐material LOM, the material feed comes from dissimilar materials. [ 3 ] Limited studies have focused on the application of multi‐material LOM. For instance, Mohammadzadeh et al. [ 67 ] combined xurography [ 68 ] with LOM to create multi‐material microfluidic devices. In their process, 2D layers were placed upon each other to fabricate a 3D object. 4 Multi‐Material Additive Manufacturing of Metals and Ceramics 4. 1 Powder Bed Fusion As discussed in Section 3. 3, powder bed fusion machines use thermal energy such as laser for melting the powder into the designed shape. Commonly used printing techniques in powder bed fusion are SLS, SLM, EBM, and direct metal laser sintering (DMLS). One of the major disadvantages of current powder bed fusion methods is that they are inherently mono‐material. The current focus of AM is to simplify and streamline manufacturing by enabling the production of geometrically complex, functional parts that can effectively replace the entire assemblies made from many simple components. Such assemblies are often made of a variety of materials. Hence, a future direction for AM of metals should be to produce parts made of multiple materials. Currently, the patented spatially selective, multiple‐powder deposition system of Aerosint SA (Belgium) seems to be the only available multi‐material 3D printing system based on powder bed fusion technology adaptable to metal, ceramic, and polymer powders (see Figure 14 ). [ 69 ] FGMs in Powder Bed Fusion of Metals and Ceramics The powder bed fusion methods such as SLS can be also used to produce multi‐material FGM parts. Based on SLM technology, Mumtaz et al. [ 70 ] fabricated an FGM component blending Waspaloy and Zirconia materials using a high powered laser. The graded specimens initially consisted of 100% Waspaloy with subsequent layers containing increased volume compositions of Zirconia (0–10%). It was found that specimens contained an average porosity of 0. 34% and a gradual change between layers without any major interface defects. 4. 2 Directed Energy Deposition Directed energy deposition (DED) is an AM process in which, the focused thermal energy is used to bond materials by melting as they are being deposited. Powder feed and wire feed systems are two major subcategories of DED. Other popular terms for DED include laser engineered net shaping (LENS), directed light fabrication (DLF), direct metal deposition (DML), laser metal deposition (LMD), laser deposition welding (LDW) and 3D laser cladding, Wire+Arc additive manufacturing (WAAM). The build volumes of these systems are generally larger than powder bed fusion. Various metallic alloys are available and it is possible to gradually and continuously change from a material to another one while manufacturing. This particularity makes possible the manufacturing of multi‐material parts. The multi‐material DED can encompass several different technologies that are identified by the way the material is being fused, each suited for different and specific purposes. The techniques based on powder bed fusion technology can only create discrete material gradient. The DED, on the other hand, is capable of fabricating multi‐material with continuous gradient within and across the layers. However, parts made by DED require multiple steps of post‐processing to acquire desired shape and dimensional accuracy. Li et al. [ 71 ] employed the SLM technique for AM of 12 wt% nano TiN‐modified CoCrFeNiMn. The TiN nanoparticles led to a uniform distribution in the FCC (face‐centered cubic) matrix. 4. 2. 1 Powder Feed Systems The powder feed technologies (such as LENS) use thermal energy (e. g. , laser) to print parts layer by layer from metals, alloys, ceramics, or composites in powder form. The LENS technology has been used to fabricate FGM objects such as the composite of stainless steel 316L and Stellite Grade 12 Co‐Cr alloy. [ 72 ] Both continuous and sharp/discrete compositional gradient parts could be fabricated in periodic multilayered structures, and the transition zone thickness was controllable by process variables. Muller et al. [ 73 ] modeled a powder flow rate by a first order transfer function with the capability of material composition in each layer to be adjusted by varying the powder flow rate of different primary materials. AM of an Inconel 718‐Copper alloy bimetallic structure was studied by Onuike et al. [ 74 ] using LENS. The bimetallic structure was fabricated with the goal of improving the thermal and mechanical properties compared with the Inconel 718 alloy. The average thermal diffusivity of the bimetallic structure was measured at 11. 33 mm 2 s −1 for the temperature range of 50–300 °C; a 250% increase in diffusivity was observed when compared to the pure Inconel 718 alloy at 3. 20 mm 2 s −1. Conductivity of the bimetallic structures increased by almost 300% compared to Inconel 718 as well. Brueckner et al. [ 75 ] used a similar technique to fabricate linearly graded material combination SS AISI 316L and INC718 ( Figure 17 ). Their studies showed that the linearly graded transitions for combining SS AISI 316L and INC718 were beneficial using LMD. Figure 17 Powder fed multi‐material LMD processing: several different powder materials (e. g. , Material A and B) can be mixed in situ by an integrated powder‐mixing chamber in the nozzle tip. Reproduced with permission. [ 75 ] Copyright 2019, AIP Publishing. 4. 2. 2 High‐Entropy Alloys High‐entropy alloys (HEAs) are formed by mixing equal or relatively large proportions of five or more elements. [ 76, 77, 78 ] The HEAs are popular for their superior properties such as better strength‐to‐weight ratios, with a higher degree of fracture resistance, tensile strength, as well as corrosion and oxidation resistance than conventional alloys. [ 79 ] Hence, HEAs are expected to be high‐performance novel structural materials, substituting for conventional alloys such as Ni‐based superalloys and stainless steels. [ 79 ] Some research works have been performed on AM of parts made of HEAs. [ 71, 76, 79, 80, 81 ] For instance, Gao and Lu [ 76 ] used a coaxial powder feeding laser 3D printing system (see Figure 18 ) to print CoCrFeMnNi alloys. The authors investigated the microstructure (Figure 18b ) and mechanical properties of fabricated HEA. An equiaxed‐to‐columnar transition structure was observed in the melt pool of the printed sample. The printed HEA exhibited an outstanding combination of high strength and excellent ductility. The ultimate tensile stress of the printed CoCrFeMnNi HEA was stronger than that of the as‐cast alloy while its ultimate tensile elongation was comparable. Figure 18 a) A schematic of coextrusion of powders for printing of HEA; b) microstructure of the printed HEA; c) as‐printed CoCrFeMnNi HEA sample. Reproduced with permission. [ 76 ] Copyright 2019, Elsevier. 4. 2. 3 Wire Feed Systems In multi‐material wire feed direct deposition, wires of desired materials are fed and then melted using an energy source (laser or an electron beam). The energy source solidifies the wires on the bed along a preferred path. The part is then built in a layer‐by‐layer fashion until a complete component is made. Syed et al. [ 82 ] investigated the process characteristics of simultaneous wire‐ and powder‐feed direct metal depositions for possible higher build rate and higher material usage efficiency while maintaining the geometry accuracy (see Figure 19 ). Their study compared the process characteristics, advantages and disadvantages of wire‐ and powder‐feed DED and showed that by adding powder and wire, the deposition rate can be increased. Figure 19 Simultaneous wire‐ and powder‐feed direct metal deposition. Reproduced with permission. [ 82 ] Copyright 2006, AIP Publishing. FGMs in DED of Metals and Ceramics Fabrication of metal and ceramic FGMs in DED is usually performed by using multiple chambers with different powder materials to be deposited on different layers in order to make the desired FGM component. Caroll et al. [ 83 ] fabricated an FGM part by powder‐feed DED with an FGM structure from SS304L to the nickel‐based alloy IN625 ( Figure 20 ). Microparticles of a secondary phase responsible for development of cracks in fabrication and microhardness were observed near the SS304L end of the gradient zone (≈82 wt% SS304L). Figure 20 a) Schematic of gradient alloy specimen, b) Photograph of specimen after fabrication by laser‐based powder feed DED. Reproduced with permission. [ 83 ] Copyright 2016, Elsevier. 4. 3 Sheet Lamination Ultrasonic AM (UAM) is a solid‐state metal seam welding method that utilizes sound to bond layers of metal. The technique creates strong bonds with high density and works with different metals. UAM was used by several authors to produce metal FGMs. [ 84, 85, 86 ] For instance, Kumar [ 84 ] studied the joining of different metallic foils using stainless steel, Al, and Cu foils. The foils were joined by ultrasonic welding using a machine that mechanically vibrates the welding head at 20 kHz. In the work of Bisadi et al. , [ 87 ] by UAM, the authors lap joined dissimilar sheets of 5083 Al alloy and commercially pure copper method. It was shown that joint defects appear at very low or high welding temperatures. 5 Multi‐Material Additive Manufacturing of Biomaterials In a broader prospective, use of AM for printing tissues and organs made of biomaterials can be classified as a) biomaterials without cell (acellular biomaterials) such as scaffolds made of natural or synthetic polymers and b) biomaterials with cell (cellular bio‐inks). Once the scaffold is printed, the cells are deposited using a 3D printing technology. Bio‐ink containing live cells controls positioning and the amount of cells. Natural polymers (e. g. , alginate‐gelatin, collagen, chitosan, cellulose) are beneficial for manufacturing of scaffolds, but synthetic polymers (e. g. , polycaprolactone [PCL], ABS, PLA, PA, polydimethylsiloxane [PDMS], polyether ether ketone [PEEK]) are sometimes preferred for their high mechanical strength, controlled degradation rate, and processability. Ceramic polymers (e. g. , hydroxyapatite) can be used for fabricating scaffolds for bone regeneration due to their desired mechanical properties and biocompatibility. Bioceramic scaffolds have bioactive component to support the growth of bones. The main technology used for deposition and patterning of multiple biomaterials is extrusion, also known as bioplotting. [ 88, 89, 90, 91, 92, 93, 94, 95 ] Bioplotting is based on extruding a material with specific viscosity from a syringe to fabricate 3D shape of biomaterials, as shown in Figures 6 and 21. Bioplotting allows for the production of a wide variety of practical biomedical tissues with different shapes and material compositions. [ 96 ] The extrusion system based on syringe achieves relatively low resolution. However, the key advantage of this technique is material flexibility. Biomaterials in the form of pastes, solutions, and hydrogels can all be fed into 3D bioplotters. A temporary, sacrificial material may be needed to support the printed structure since viscous raw materials have low stiffness that may result in the collapse of complex structures. A list of commercially available bioplotters is provided in Table 3. Figure 21 Schematic of multi‐material bioplotting of biomaterials: a) a biomaterial ink palette for fabrication of tissues and organs, b) printable inks with different compositions, c) computer model of an organ, d) biomaterial inks, and e) an example of a functional construct. Reproduced with permission. [ 97 ] Copyright 2016, IOP Publishing, Ltd. Multi‐material 3D printed scaffolds, tissues, and organs require different bio‐inks, all of which must demonstrate cell‐compatibility and printability. For instance, the printing and post‐processing of the acellular ink should be cell‐compatible for multi‐material printing of cellular and acellular inks. Therefore, the use of organic solvents or extreme temperatures is not recommended as it would compromise cell viability within the printed structure. [ 97 ] Bakarich et al. [ 92 ] developed a material extrusion‐based gradient printing system, and its function was demonstrated by 3D printing a range of tough hydrogel composites. A spectrum of soft and wet to hard and dry particulate‐reinforced composites were prepared by changing the ratios of a soft alginate/polyacrylamide‐based hydrogel to a hard UV‐curable ink in the materials. The printed materials were mechanically characterized in tension and were modeled by composite laminate theory. Sears et al. [ 95 ] reported the development of a biodegradable, fumarate‐based emulsion ink for bioprinting robust bone grafts with designed, hierarchical porosity. A combinatory approach that utilized thermoplastic polyester printing to reinforce the emulsion ink prints was then developed by the authors to enhance the compressive properties and illustrate the potential of this technique to improve scaffold biomimicry. In the authors’ work, the addition of either a PCL or PLA shell resulted in a significant increase in compressive modulus and yield strength with the PLA shell resulting in constructs with compressive properties in the range of trabecular bone (see Figure 22 ). A multichannel open source hardware and software 3D bioplotter was designed by Lee et al. [ 93 ]. Hybrid scaffolds with synthetic polymeric materials and cell laden hydrogels were printed and the authors verified the performance of the 3D bioplotter. Figure 22 a) Combinatorial printing process with layer by layer deposition of the thermoplastic polyester outer shells and high internal phase emulsions (HIPE) emulsion ink inner material; b) integration between the emulsion ink and thermoplastic (PCL) shell. Adapted with permission. [ 95 ] Copyright 2017, Elsevier. Multi‐material 3D printing of inkjet‐based systems were also used for 3D printing of biopolymers. Poellmann et al. [ 98 ] 3D printed micropatterned, multi‐material hydrogrels using E‐jet direct ink writing. The authors fabricated polyacrylamide features in microscale integrated in another hydrogel of a different composition. Once photopolymerization was done, the droplets were backfilled with a second polyacrylamide mixture, the second mixture was polymerized and the sample was peeled off the substrate. Fluorescent and confocal microscopies verified multi‐material patterning, while scanning probe microscopy revealed a patterned topography with printed spots forming shallow wells. 6 Multi‐Material 4D Printing Multi‐material 4D printing uses AM technologies to fabricate stimulus‐responsive parts that can actively change their properties when subject to appropriate stimuli. The use of 4D printing is expected to become more popular with applications across biomedical, aerospace, and electronic industries. Geometrical transformation after 3D printing is the main feature of multi‐material 4D printing. Shape changes (or shape memory effect) in 3D printed parts can be induced by different external stimuli, to cause shrinkage, expansion, or folding of the printed parts as the fourth dimension. Shape memory polymers (SMPs) and shape memory alloys (SMAs) are two different kind of materials that are utilized for 4D printing. SMPs may be preferred over SMAs due to their wide range of glass transition temperatures, T g, from −70 to 100 °C, permitting their elastic properties to be tailored. [ 99 ] Some common problems with SMAs are their complex manufacturing, higher costs, toxicity, and limited recovery. For instance, SMPs can obtain a shape recovery property up to four times of plastic strain, whereas SMAs are around 7–8%. The challenge for multi‐material 4D printing is to utilize materials that are strong and malleable in the presence of stimuli. Ideally, when material is induced by different stimuli should also exhibit different behavior. [ 100 ] From stimulus point of view, multi‐material 4D printing can be classified by their stimuli like temperature, humidity, or solvents, as well as pH or light. Wu et al. [ 101 ] used different thermal response SMP fibers to print composite materials having different shape changes with rising temperature. The authors used two fibers with different ( T g of ≈57 and ≈38 °C). After a simple single‐step thermomechanical programming process, the fiber families could be sequentially activated to bend when the temperature was increased. Figure 23 shows the active motion of printed objects consisting of SMP fibers that have different T g. Mao et al. [ 102 ] utilized multiple thermal response SMPs with different T g to fabricate more complex motion of printed parts using Objet Connex 260. A composite strand was printed with hinges comprised of nine SMPs for shape recovery process sequentially. Figure 24 indicates the folding of the 1D strand with the hinge section after immersing in hot water (≈90 °C). Bodaghi et al. [ 103 ] designed and fabricated a self‐expanding/shrinking mechanism by fabricating two types of SMPs with low and high T g in fiber forms into a flexible matrix. SMP fibers were eccentrically located in the beam‐type actuator unit and their arrangement was changed along the beam length. Self‐expansion and shrinkage characteristics were verified both experimentally and numerically. The authors fabricated planar and tubular shapes composed of periodically arranged actuating units. The actuating units were made from TangoBlackPlus and VeroWhitePlus available in the Objet 500 printer material library. In another study, Bodaghi et al. [ 104 ] explored 4D printing of triple SMPs with self‐bending feature. The concept was on the basis of arranging hot–cold programming with FDM printing technology to engineer triple SMPs. Their experiments revealed that the printed SMPs have elasto‐plastic response at low temperatures while they behave hyper‐elastically at high temperatures in the large deformation regime. Figure 23 Multi‐shape memory effects of a printed active composite strip: a) the design and dimensions of the sample. The enlarged drawing is the cross section of the structure. b) The original printed sample. The length scale in the bottom is in mm. c–f) Shape change of the sample at different temperatures. Reproduced with permission. [ 101 ] Copyright 2016, Springer Nature. Figure 24 a–c) The schematic of sequential self‐folding strand. Series of photographs showing the shape recovery process of the helical SMP component. Reproduced with permission. [ 102 ] Copyright 2015, Springer Nature. Using a high resolution PµSL and incorporating a family of photo‐curable copolymer networks, Ge et al. [ 105 ] printed high‐resolution multi‐material thermal response SMPs. They fabricated a multi‐material gripper and the microscale resolution achieved using the PµSL method. Owing to photo‐curable thermoset with different crosslinkers, the authors could print devices with a variety of T g. The same research group in Singapore, has extended the concept of multi‐material 4D printing to active hinges [ 106 ] and SMA wires. [ 107 ] For instance, Akbari et al. [ 107 ] fabricated soft actuators by embedding SMA wires into various soft matrices manufactured by multi‐material 3D printing (see Figure 25 ). In order to achieve a wide range of deformations, ten different printing materials were characterized and used in their actuators design. In addition, the authors developed a finite element model to simulate complex deformations of the printed actuators and facilitate the design process. Figure 25 Snapshots of grabbing an object using a multi‐material printed bending soft actuator by embedding thin SMA wires eccentrically into a polymeric soft matrix (FLX9940). Reproduced with permission. [ 107 ] Copyright 2019, Elsevier. Another challenge in the field of 4D printing is the controllable morphing. In a study conducted by Wang et al. [ 108 ] 4D printing of continuous carbon fiber‐reinforced composites was introduced. The composite fabricated by this method could realize programmable deformation with a high deformation accuracy (see Figure 26 ). The deflection of the printed composite was initiated by the difference of coefficients of thermal expansion (CTEs) between continuous carbon fibers and flexible Polyamide66 matrix. Figure 26 4D printing method by embedding continuous fibers in matrix, realizing deformation of complex surfaces. Reproduced with permission. [ 108 ] Copyright 2018, Elsevier. Moreover, 4D printing technique can provide the opportunity of printing 3D electronic circuits. Deng et al. [ 109 ] designed a mechanism ( Figure 27 a ) to obtain self‐folding 3D circuits (Figure 27b ) using a polystyrene film sensitive to thermal stimuli based on DIW method. As depicted in Figure 27a, resin was used for one side of the film as a constraint layer, and the other side of the film was left empty. As a result, by rising the temperature, the empty side was folded on the hinge. Figure 27 S elf‐foldable 4D printing using DIW a) self‐foldable design and b) folding of structure under heat, c) LED cube case produced by self‐folding mechanism after printing. Reproduced with permission. [ 109 ] Copyright 2016, IEEE. Simulation of multi‐material 4D printed objects can also be found in the literature. [ 104, 110, 111 ] For instance, to enhance design capabilities of dual/triple SMP by 4D printing, a phenomenological constitutive model was developed by Bodaghi et al. [ 104 ]. The authors incorporated crucial elements including SMP phase transformation, hyper‐elasticity, elasto‐plasticity, and hot‐cold programming in the framework of large deformation regime in their models. Their developed model was then coupled with the finite element formulation to simulate 4D printed dual/triple SMP structures through an elastic‐predictor plastic‐corrector return map algorithm. The experimental and numerical results in their study demonstrated the potential applications of dual/triple SMPs in mechanical and bio‐medical engineering devices like self‐bending gripers/stents and self‐shrinking/tightening staples. Sossou et al. [ 110 ] proposed a modeling framework for simulating smart and conventional materials behaviors on a voxel basis. This allowed for arranging materials in any distribution and rapidly evaluating the distribution behavior. Homogeneous and heterogeneous objects made of conventional and smart materials were modeled and simulated in their work. A printed smart valve and a theoretical actuator were used as test cases in the authors’ work. 7 Discussion, Evaluation, and Future Directions Many different 3D and 4D printing processes for fabrication of multi‐material polymer, metal, ceramic, and biomaterials have been reviewed in this manuscript. It was noticed that the vat photopolymerization process is not generally a candidate for multi‐material 3D printing. It prints parts from a vat of UV curable resins, and the use of multiple materials in vat photopolymerization provides challenges with contamination management between material systems. However, due to its advantages such as high‐quality surface finish, dimensional accuracy, and a variety of material options that includes transparent materials, vat photopolymerization has been adapted to support multi‐material printing. The material extrusion technique can be easily extended to multi‐material 3D printing through the use of multiple nozzles. PBF methods also suffer from inherent mono‐material processability. A future direction for AM of metals should be to produce parts made of multiple materials. Currently, the patented spatially selective, multiple‐powder deposition system of Aerosint SA (Belgium) seems to be the only available multi‐material 3D printing system based on powder bed fusion technology adaptable to metal, ceramic, and polymer powders. Material jetting is a common method for multi‐material printing. PolyJet (Stratasys Ltd. , USA) is probably the most common commercially available multi‐material jetting process. The hardware and software architectures for these multi‐material 3D printers are often proprietary and inextensible. Due to the complicated technology, less custom‐made 3D printer based on material jetting is available (e. g. , Multifab [ 65 ] ). To the best of authors’ knowledge, no work on multi‐material 3D using binder jetting was found in the literature. [ 112 ] DED methods seem to be good candidate for multi‐material printing. The build volumes of these systems are generally larger than powder bed fusion. Various metallic alloys are available and it is possible to gradually and continuously change from a material to another one while manufacturing. Powder feed and wire feed systems are two major subcategories of DED, however, no work on multi‐material 3D and 4D printing using wire feed system was found in the literature. In multi‐material LOM, the material supply either comes from two different materials or comes from blended multiple materials. However, the studies focused on the application of multi‐material LOM are very limited as discussed in the text. Advanced 3D printed composite materials and 4D printed responsive materials were described with their specificities and their ontologies. Moreover, various functions of these products in different industries including medical devices, electronics, biomedical implementations, sporting goods, and robotics have been summarized. It was seen that part performance could be enhanced by the utilization of multiple material systems and complex geometries. The use of composites can also target functional regions within a part; applying the most appropriate materials in the most appropriate areas. Topological optimization is employed to propose new complex geometries which are easier to fabricate by AM with weight gain while keeping a relatively high mechanical behavior. The predictive computational models to simulate the geometry of 3D‐4D printing filaments is currently being studied to improve the properties of multiple material systems. Although considerable advancements were achieved via 3D and 4D multi‐material printing, the potential has not been fully explored yet. 7. 1 Mechanical Properties The mechanical performance of multi‐material additively manufactured parts is usually better in comparison with those printed by single‐material printing. Formation of voids between subsequent layers of printed parts can affect their mechanical performance due to a decrease in interfacial bonding between printed layers. Different mechanical behavior under vertical tension or compression compared to that of the horizontal direction is another common challenge of multi‐material AM. Robust 3D printing processes such as micro‐additive layering are important to provide stability between layers and improve surface finish to the resolutions that meet their specific applications. 7. 2 Production Efficiency Efficiency in production of parts may be another potential research direction in multi‐material AM systems. A balance between the production efficiency (e. g. , production rate) and part quality is needed all the time. One may use a higher energy power or faster scanning speed to raise the production rate, but the part quality may be influenced. To overcome this problem, one solution is to optimize the printing parameters. Another issue that affect the production efficiency is the complex post processing methods. Effective methods for post‐processing, including support material removal and processes related to heat treatment, need to be improved. 7. 3 Micro/Nano Multi‐Material AM While the majority of commercial multi‐material 3D printers create parts in macroscale, variety of sizes may be needed. Recently, micro/nano multi‐material AM processes have attracted a significant attention because of their influence on many applications such as MEMS/NEMS and nanomanufacturing. [ 113 ] As discussed earlier, the electrohydrodynamic printing technique seems to be a promising printing method for 3D printing of parts from micro to nanoscale of different materials. [ 114 ] 7. 4 Multi‐Material Metamaterials and Lattice Structures Some research groups are concentrated on AM processes with multi‐material manufacturing capabilities and new internal structures based on high performance computers and optimization tools. Metamaterials and advanced lattice structures have applications in flexible materials that seem to have potentials in variety of disciplines such aerospace, civil, textile, and tissue engineering applications. The utilization of micro/nano multi‐material AM with metamaterials and lattice structures can lead to innovative functional parts. 7. 5 AM of Continuous Fiber‐Reinforced Composites As indicated by several researchers, the mechanical performance of continuous fiber‐reinforced composites is expected to be more significant than short fibers. Existing challenges are mainly related to processing of the materials, the bonding between fibers and the matrix, and interlaminar properties of the 3D printed continuous fiber‐reinforced composites. Choice of the appropriate 3D printing technology and finding proper binder are essential to achieve a better mechanical performance. 7. 6 4D Multi‐Material Printing This approach is a relatively novel and fascinating research area and has great capability for extension. As 4D multi‐material printing rooted in the 3D one, some similar challenges such as limited choice of materials, printing resolution, slow, mechanical performance, and potential to obtain dimensional accuracy. New developments could result in smart structures for actuation or motion following a predetermined program. FGMs can also tailor the microstructure properties of a 4D printed part to create more complex geometrical transformations by strategically controlling the density and directionality of stimuli‐responsive materials. It can also improve the inter‐material bonding of heterogeneous smart compositions, and even disregard the material properties of being active or non‐active. The multi‐material 4D printing development is the decisive point to accelerate the growth of the smart material area. In summary, there are many significant achievements of multi‐material AM technology, while, some challenges still remain, including the production efficiency, mechanical properties, and applications that are mentioned above. Multidisciplinary research and development will be crucial to overcome those challenges and fully realize the potential of multi‐material AM in different applications. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201902326
| 2,019
|
Advanced Science
|
Cell Adhesion, Morphology, and Metabolism Variation via Acoustic Exposure within Microfluidic Cell Handling Systems
|
Abstract Acoustic fields are capable of manipulating biological samples contained within the enclosed and highly controlled environment of a microfluidic chip in a versatile manner. The use of acoustic streaming to alter fluid flows and radiation forces to control cell locations has important clinical and life science applications. While there have been significant advances in the fundamental implementation of these acoustic mechanisms, there is a considerable lack of understanding of the associated biological effects on cells. Typically a single, simple viability assay is used to demonstrate a high proportion of living cells. However, the findings of this study demonstrate that acoustic exposure can inhibit cell attachment, decrease cell spreading, and most intriguingly increase cellular metabolic activity, all without any impact upon viability rates. This has important implications by showing that mortality studies alone are inadequate for the assessment of biocompatibility, but further demonstrates that physical manipulation of cells can also be used to influence their biological activity.
|
1 Introduction The manipulation of biological matter has become widespread in microfluidics. 1, 2, 3 The benefits of operating at a reduced size scale have enabled highly efficient techniques for selective patterning and manipulating cells. 4, 5 Such approaches have been used for a wide range of tasks, including single cell analysis, 1, 3, 6, 7 tissue engineering, 8, 9, 10, 11, 12 studying cell–cell interaction and signaling, 2, 13, 14 sorting, 5, 15, 16, 17 and drug screening. 18 As the applications of this technology are focused within the clinical and life sciences, a thorough understanding of the associated biological impact imposed by these manipulation techniques is necessary. The desire to manipulate suspended matter within microfluidic systems has inspired a range of techniques both passive 5, 19, 20 and active. 21, 22 Passive approaches rely heavily on the geometry of the microfluidic channels and inertial forces. Flow profiles are altered by the introduction of sudden expansions and contractions, weirs and pillars to impede and divert the flow into desired streamlines. This reliance on the resultant flow profile restricts the flexibility of the chip, being single task specific. In contrast, active methods are significantly more robust, capable of on‐demand actuation and offer the ability to change functionality ad‐hoc, leading to an increased selectivity. To this end, a range of active methods have been developed using magnetic, 23, 24 optical, 25, 26 electrical 16, 27 and acoustic 28, 29, 30 excitation. Acoustofluidics is the use of acoustic forces to manipulate suspended matter within microfluidics, 31, 32 and has the advantage of uniquely combining ease of on‐chip integration and simple, yet dextrous establishment of force fields in a noncontact manner. 29, 33, 34 As a direct result, it has been extensively used to capture, 33, 35 pattern, 36, 37, 38 and sort particles, 15, 34, 39 cell sonoporation, 40 synthesize nanomaterials, 41 as well as to mix fluids. 42, 43 Although acoustofluidics has been widely accepted as a biocompatible technique, substantiated with cell viability studies; 44, 45, 46, 47, 48 there have been no extensive viability studies at elevated frequencies (30–600 MHz). Indeed, typically studies are corroborated with a single viability method, most commonly live/dead staining, 6, 49, 50, 51 or trypan blue exclusion 52, 53 and in some instances proliferation studies (MTT). 36 In contrast to these singular approaches, in passive microfluidic systems in which cells are predominantly subjected to shear forces arising from the flow field, a wide range of multifaceted cell viability studies 54 have been conducted showing, for example, shear‐dependent regulation of the von Willebrand factor of human umbilical vein endothelial cells 55 and the potential for circulating tumor cell apoptosis at high shear levels. 56 This lack of biological knowledge may result in potential unrecognized adverse effects (i. e. , “false positives”), or underexploited beneficial effects, based on the current biocompatible analysis methods, hampering direct translation of these technologies within clinical and life science applications. To address this inadequacy and to understand the associated biological effects of high frequency ultrasound, we probe a range of cell lines, using a suite of techniques chosen to determine postacoustic exposure effects on proliferation, membrane permeability, metabolic rate, cell attachment, and morphology. Four distinct cell lines were used, with known variation in cell stiffness and from two species (human and mouse). They were HaCaT (human keratinocytes), L929 (mouse fibroblast), MSCs (mesenchymal stromal cells, human bone marrow‐derived primary cells), and MG63 (mouse osteosarcoma). The selected cells represent not only species and source tissue variability but also vast differences in predicted cell stiffness values, ranging from 0. 8 kPa (MSCs 57, 58 ) to 10 kPa (HaCaT 59 ) to better understand the impact of acoustic exposure on cell heterogeneity. Experiments we have performed show no significant differences in nuclear morphology and proliferation rates across all cell types and conditions studied. However, we did observe significant variations in cell attachment, spreading and metabolic activity, over a range of shear stresses induced by flow and acoustic exposures, differences that would otherwise have remained undetected due to the consistently high viability percentage observed in standard live/dead assay data. Investigation of the effect of acoustic excitation powers, flow rates, and channel dimensions on cell behavior is found to have a consistent increase in metabolic activity across cell lines in response to acoustic exposure. 2 System Principles A major reason to use high frequency (30–600 MHz) ultrasound in a microfluidic system is that the wavelength is in the same order as that of a typical cell (i. e. , 2–50 µm). This is a prerequisite for patterning of single cells, 6 but has also been shown to provide the possibility of high sensitivity sorting, 15, 33, 60 patterning using either standing waves 61, 62 or traveling waves 32, 63 and fluid mixing protocols. 43 Excitation can be achieved using surface acoustic waves, these waves are substrate bound until they encounter a fluid body whereupon energy is efficiently coupled. An alternating current is applied, at the operational frequency (48. 5 MHz used here; 80 µm surface acoustic waves (SAW) wavelength, λ SAW ), to a set of opposing interdigitated transducers (IDTs) patterned on a piezoelectric substrate (128° Y‐cut X propagating Lithium Niobate used; LiNbO 3 ), resulting in a standing wave that couples directly as a pressure field within the fluid medium containing cells. 6 Typically, two main forcing mechanisms are present, namely acoustic radiation forces and acoustic streaming induced drag forces. The former acts on the suspended matter as a result of its interaction with the incident and scattered acoustic waves, whereas, the latter drags the particle in steady state fluid flows driven by Reynolds stresses. 31 The relative significance of these two forces can be altered based on the excitation parameters such as the frequency, as well as the channel geometry, particle and fluid properties. 64 Here, the streaming effects are minimal in comparison to the radiation force based on the frequency and channel dimensions used. The microchannel was designed as a serpentine, as shown in Figure 1 (the experimental device has 11 turns as opposed to the 3 turns shown). The curves of the serpentine were sufficiently large that the shear stress profile is largely unchanged from the straight part of the channel (see Figure S2, Supporting Information). It was necessary to have a continuously flowing system to ensure that sufficient numbers of cells could be passed through the sound field, while the serpentine design facilitated prolonged exposure times. Both the shear stress and exposure times were varied by using combinations of two different channel heights (25 and 50 µm) and two flow rates (5 and 10 µL min −1 ). The microfluidic chip was temperature controlled using a Belektronig Benchtop Temperature controller to avoid heating of the cells during acoustic excitation. Figure 1 Design concept depicting the serpentine channel and IDT design. Distinct cell types are 1) fed (indicated by the blue arrow) into the system at a constant flow rate, 2) exposed to the ultrasound, and 3) retrieved (indicated by the green arrow) prior to seeding into tissue culture plastic (TCP) for various assessment techniques (see Figure S1 in the Supporting Information for the process diagram). 3 Results 3. 1 Cell Morphology and Substrate Attachment As an initial, simple readout of cell behavior in response to acoustic exposure, the adhesion and morphology of cells retrieved from the device was determined. Cells subjected to acoustic exposure were compared to i) cells that were simply plated into standard tissue culture wells (tissue culture plastic (TCP) control) and ii) cells that were passed through the device in the absence of acoustic stimulation (flow control). These controls provided a benchmark to account for any effects of detaching and manipulating the cells or exposing them to fluid flow within the device. For the L929 cell line ( Figure 2 a), no differences in cell adhesion were observed in response the flow control or at the lower level of 400 mV acoustic excitation. However, cell adhesion and spreading were affected when an excitation amplitude of 800 mV (indicated by the arrows in Figure 2 a) was used. Under these conditions, the cells remained circular with no spreading—this is a particularly striking alteration as no signs of recovery were observed even after 72 h as demonstrated by quantification of cell area and cell aspect ratio as shown in Figure 2 b, c, respectively. Interestingly, no differences were observed in the nuclear circularity (Figure 2 d). It is well established that nuclear abnormalities are associated with a diseased state (most prominent in cancer; nuclear blebbing and enlarged nuclei), thus, assessment of nuclear circularity was conducted to rule out this adverse scenario. Figure 2 Acoustic exposure resulted in decreased cell attachment speeds for L929 cells, while presenting minimal impact on nuclear morphology. a) Phase contrast images of L929 cells 24, 48, and 72 h postexposure across each experimental condition. Arrows depict cells displaying inhibited substrate attachment, while asterisks depict normally spread cells. Quantification of observed b) cell area, c) cell aspect ratio, and d) nuclear circularity for L929 cells across 24 and 72 h postexposure. Data are presented as mean ± SD from triplicate samples, n = 9 (> 600 cells per time point) for each condition tested, with data analyzed using one‐way ANOVA supplemented with Tukey post hoc testing. Scale bar, 50 µm. Statistically different samples are denoted by * p < 0. 05. The data for the other three cell types, MSCs, MG63, and HaCaT, are shown in Figure 3 (images shown in Figure S3, Supporting Information). For the HaCaT cells (Figure 3 a), no significant variation in morphology was observed across cell area, aspect ratio and nuclear circularity, for all conditions and time points. The higher resistance to morphological change as a result of an external stressor may be attributed to the abundant expression of keratin within these cells, 65 which makes for a relatively stable structure. For the MG63 cells (Figure 3 b) and MSCs (Figure 3 c), no significant change was observed for the flow control samples over 24 h, while the highest acoustic power resulted in a complete inability for cells to attach to the substrate. For the MSCs, which are known to be extremely mechanosensitive and thus relatively more susceptible to external stressors, attachment did not occur even at the lower acoustic power level (see Figure 3 c). Figure 3 Acoustic exposure resulted in limited changes to cell phenotypes. Quantification of i) cell area, ii) cell aspect ratio, and iii) nuclear circularity for a) HaCaT, b) MG63, and c) MSC cells across 24 and 72 h postexposure. Labels of no cell attachment denote scenarios in which cells could not adhere to the growth substrate postexposure and thus could not be assessed. Data are presented as mean ± SD from triplicate samples (>600 cells per time point), with data analyzed using one‐way ANOVA with Tukey post hoc testing. Statistically different samples are denoted by * p < 0. 05, ** p < 0. 005. 3. 2 Cell Viability and Metabolic Activity Cell viability is commonly measured either using live/dead staining as a simple way to discriminate viable cells or assays that use cellular metabolism as a surrogate marker, such as MTS (a novel tetrazolium compound [3‐(4, 5‐dimethylthiazol‐2‐yl)‐5‐(3‐carboxymethoxyphenyl)‐2‐(4‐sulfophenyl)‐2H‐tetrazolium, inner salt; (MTS (a) ]). We therefore performed both of these assay types to examine the impact of acoustic stimulation upon the different cell populations. Importantly, although the live/dead data showed very little variation across treatment and cell type ( Figure 4 a‐i–d‐i and Table S1, Supporting Information), the metabolic data revealed several significant effects. First, we saw that simply by passing the cells through the microfluidic chip, there was a dip in metabolic activity (Figure 4 a‐ii–d‐ii) which lasted for up to 72 h, compared to the TCP control. However, this was mitigated when acoustic actuation was applied and the metabolic readings were comparable to the TCP control at the highest power level. Two possible hypotheses are that 1) the acoustic field decreases the effects of shear induced by the fluid flow—this could occur due to acoustophoretic particle migration toward the center line of the channel, 15 and hence away from the high shear regions at the periphery of the channel, or 2) the acoustic fields are stimulating an increase in metabolic activity irrespective of shear. This could occur either directly or by indirectly acting upon currently undefined cellular mechanotransduction signaling pathways. Although the observed metabolic activity trend was similar across all the data obtained, the data set is not full, as the reduced adhesion of MSCs and MG63s under acoustic stimulation mean that data could not be collected for these conditions. Figure 4 Variability is evident between viability assays designed to target membrane permeability and metabolic activity. Live/dead (green/red) fluorescence staining i) 24 h postexposure (see Table S1 in the Supporting Information for tabulated live cell percentage) and ii) formazan absorbance (MTS assay; metabolic activity) are presented for a) MSCs, b) MG63, c) HaCaT, and d) L929 cells. MTS assay data are presented as mean ± SD from triplicate samples ( n = 9) for 24, 48, and 72 h time points postexposure, with data analyzed using one‐way ANOVA with Tukey post hoc testing. Labels of no cell attachment denote scenarios in which cells could not adhere to the growth substrate postexposure and thus could not be assessed. Scale bar, 50 µm. Statistically different samples are denoted by * p < 0. 05, ** p < 0. 005. 3. 3 Cell Proliferation Although MTS (and similar) assays can be used over time as an indicator of cell proliferation, they are not strictly markers of cell proliferation and rely on the assumption of equal metabolic activity in all cells and under all conditions. For this reason, we also assessed a more direct marker of cell proliferative capacity—Ki67 staining. Ki67 is expressed during all active phases of the cell cycle (G1, S, G2, and mitosis), but is absent in resting (quiescent) cells (G0). Overall expression of Ki67 increases during cell progression through S phase of the cell cycle. Therefore we used staining for Ki67 to determine the number of actively dividing cells at a particular snapshot in time (i. e. , at 24 h ( Figure 5 a) and at 72 h (Figure 5 b)); Ki67 marker shown in red). Figure 5 Postexposure assessment revealed no significant changes in cell proliferation rates for each cell type tested. Fluorescence observations of each cell type a) 24 and b) 72 h postexposure. Staining depicts f‐actin (green), nuclei (blue), and Ki67 (red). Labels of no cell attachment denote scenarios in which cells could not adhere to the growth substrate postexposure and thus could not be assessed. Scale bar, 20 µm. Quantification of Ki67 positive cells are presented for c) MSCs, d) MG63, e) L929, and f) HaCaT cells across 24 and 72 h postexposure. Data are presented as mean ± SD from triplicate samples, n = 9 (>600 cells per time point) for each condition, with data analyzed using one‐way ANOVA supplemented with Tukey post hoc testing. By comparing the proportion of Ki67‐positive (Ki67 + ) nuclei, it was clear that the proliferation rate of different cell types naturally varies. This can be expected as these cells are inherently different in terms of functionality and tissues are well known to have different turnover rates. 66 For example, 24 h postexposure, only 10% of MSCs were shown to express detectable levels of Ki67 in comparison to MG63 cells in which 55% of cells were Ki67 +. However, within a single cell type no significant difference in Ki67 activity, and hence proliferation rate, was detected (Figure 5 c–f) while maintaining a consistent cell number across all time points considered (see Figure S4, Supporting Information). This confirms that cell proliferation was not affected by flow through the device or acoustic stimulation. Furthermore, these data collectively indicate that the variations in MTS assay data, as reported in Figure 4 a‐ii–d‐ii, are due to specific variations in metabolic activity rather than proliferation rate. 3. 4 Probing Shear Effects via the Modification of Channel Design and Flow Rates Two hypotheses could explain the data presented in Figure 4 : 1) acoustophoretic migration away from the channel periphery mitigates the effects of shear induced by flow and 2) the acoustic excitation directly causes an increase in cell metabolic activity. To probe these further, a set of experiments were conducted using a larger channel with a height of 50 µm for the larger cell types (MSCs and MG63s), while maintaining the same flow rate of 10 µL min −1. This effectively reduces the wall effects on the cell but also reduces the average velocity of the cells meaning that the duration of acoustic exposure is doubled. Based on this, we expected to see an increase of the acoustic‐based effects, particularly if hypothesis 2 of a direct influence of acoustic excitation upon metabolic rate is correct (the migration of hypothesis 1 can be expected to occur quickly so exposure time would not play a role). The smaller L929 and HaCaT cells were again passed through the initial 25 µm height channel, but with the flow rate halved to 5 µL min −1 to lower the shear stresses and increase the exposure time. If hypothesis one regarding a shear‐induced effect is correct, we should see little change, as the acoustics act to reverse the shear effects. However the hypothesis of a direct effect of acoustic stimulation on cellular metabolism is correct, we would expect to see a further increase in the metabolic rate for cells exposed to ultrasound for longer. The associated velocities and shear stresses within these channels are numerically solved and reported in the Supplementary Information (see Figure S2, Supporting Information). 3. 5 Untangling the Effects of Shear from Acoustic Stimulation on MSCs and MG63 Using the modified conditions, experiments using the MTS assay and live/dead staining were conducted, again comparing cells plated directly onto TCP to cells exposed to flow‐only and those under acoustic stimulation. An increase of the channel height to 50 µm, effectively reducing wall effects on the cells, resulted in a full set of data, mitigating issues related to cell attachment as observed in Figure 4. A consistently high percentage of viable cells was observed using a live/dead staining, for both MSCs and MG63s, with the exception of MSCs stimulated with 800 mV in which a large proportion (93 ± 6%) of cell death was observed ( Figure 6 a‐i and Table S2, Supporting Information). This can be attributed to excessive acoustic excitation, in terms of the input power and exposure time, effectively identifying the upper limit of exposure. MSCs are widely known to be extremely mechanosensitive, 67, 68 thus unsurprisingly show the lowest tolerance to a persistent external stress from the cell lines investigated here. As expected, due to the low proportion of live cells, the metabolic activity for the 800 mV acoustic exposure, across all three time points is significantly reduced. It is important to note that although exposure below 800 mV does not result in cell death, cell health is not necessarily unaffected. Strikingly, and consistent with the initial experiments (Figure 4 a‐ii, b‐ii), a trend of decreased metabolic activity was observed for the flow control condition and an increased metabolic activity in cells exposed to acoustic excitation was reported (Figure 6 a‐ii, b‐ii). All stimulated cell populations had higher metabolic activity than those exposed to flow alone (Figure 6 a‐iii, b‐iii), with the exception of the 800 mV—treated MSCs in which the majority of the cell population was dead. Figure 6 Increasing channel dimensions (i. e. , 50 µm channel height) presents enhanced acoustic effects while enabling a) MSC and b) MG63 cell extraction. Decreasing the flow rate (i. e. , 5 µL min −1 ) modulates metabolic activity of c) L929 and d) HaCaT while retaining a high degree of cell viability. a–d) Live/dead (green/red) fluorescence staining i) 24 h postexposure (see Tables S2 and S3 in the Supporting Information for tabulated live cell percentage) and ii) formazan absorbance (MTS assay; metabolic activity) are presented. MTS assay data analyzed using one‐way ANOVA with Tukey post hoc testing. iii) Growth rate extracted from MTS data is presented for all cell lines. All data presented as mean ± SD from triplicate samples ( n = 9). Scale bar, 50 µm. Statistically different samples are denoted by * p < 0. 05, ** p < 0. 005. 3. 6 Untangling the Effects of Shear from Acoustic Stimulation on L929 and HaCaT To understand the effects of reduced shear and extended acoustic exposure, the L929 and HaCaT cell lines were passed through the system at a reduced flow rate of 5 µL min −1. The expected exposure time was thus increased to 19 s, similar to that of the modified MSC and MG63 channels reported in Figure 6 a, b. Consequentially, the flow control experimental results indicate a consistently low cell mortality rate (Figure 6 c‐i, d‐i) and depict a reduced variation in metabolic activity, relative to the TCP control (Figure 6 c‐i–iii, d‐i–iii), when compared to the data presented in Figure 4 c‐ii, d‐ii, suggesting that the effects of shear induced by the flow are relatively insignificant and that any variations observed are likely caused by the acoustic excitation. L929s were not able to adhere after exposure to 800 mV stimulation—this suggests that the limit of tolerance to acoustic exposure was reached under these conditions. While a living HaCaT population was retrieved, as shown by live/dead staining (Figure 6 d‐i and Table S3, Supporting Information), the significant drop in metabolic activity (Figure 6 d‐ii–iii) can be attributed to an excessive stress inflicted by the prolonged exposure of higher acoustic intensities, rendering the cessation of normal biological function while preserving the cell membrane and structure. This ability to preserve its structural integrity can be attributed to the high levels of keratin in HaCaT cells 65, 66, 69 that act to safeguard the cell under their normal function in the skin which can be exposed to harsh environments. Acting as double‐edged sword in this instance, the higher density of keratin renders the cell relatively stiff. We propose that this stiffness results in an increased exertion of acoustic radiation, due to the inherently higher acoustic contrast factor, making it more susceptible to an acoustic stimuli. 44 This observation is potentially the clearest indication of the shortfall presented by simplistic live/dead assay as evidence for cell viability. Here, the reported live/dead percentage would serve as a “false positive” when used to substantiate cell viability in an acoustic microfluidic platform as it is contrasting in nature when compared to the cell metabolic activity. The other notable pattern in this data are that both cell types again showed increased metabolic activity at 400 mV acoustic stimulation (Figure 6 c‐iii, d‐iii). Especially the HaCaT cell type (Figure 6 d‐ii–iii), which reported a higher metabolic activity than that of the TCP control. Overall, this leads to a pattern where flow appears to moderately decrease the metabolic activity but acoustic stimulation either mitigates this effect or overrides it by independently increasing the metabolic activity. The latter supports hypothesis 2, linking a mechanotransductive mechanism to ultrasonic excitation. 3. 7 Flow Control and Acoustic Comparison As a further investigation of the effects of flow versus acoustic stimulation, the changes relative to the TCP control were calculated ( Figure 7 ). We observed a smaller reduction in the normalized metabolic activity (percentage difference to TCP control) at the lower flow rate for both cell L929 (Figure 7 a‐i) and HaCaT cells (Figure 7 a‐ii) at 24 h postexposure. Any differences in metabolic activity were diminished after 72 h (i. e. , 0%; indicated by the horizontal dash line), indicating that the effects are caused by an acute stress that is recoverable. This reaffirms the hypothesis that shear stress experienced by the cell decreases the metabolic activity. Figure 7 A percentage comparison between the high flow rate (i. e. , 10 µL min −1 ; in black) and the low flow rate (i. e. , 5 µL min −1 ; in red) settings normalized to the TCP control. a) Flow control comparisons for i) L929 and ii) HaCaT at different flow rates. b) Varying levels of acoustic exposure for i) L929 and ii) HaCaT at different flow rates. We next compared the impact of acoustic exposure, independently to flow rate. The 400 mV acoustic exposure settings consistently showed a higher metabolic rate than the TCP, for both flow rates (Figure 7 b). When the detrimental effects of shear are lessened by operating at a lower flow rate and the length of exposure to acoustic effects is increased, the increase in metabolic rate is such that the values recorded are higher than those of the TCP. This offers firm support of hypothesis 2 whereby acoustic exposure has an impact on the cell metabolic activity. 4 Discussion Here we fabricated a system that facilitates assessment of cell behavior following acoustic exposure and used a wide range of methodologies to examine the impact upon cell behavior. The results demonstrate sensitivity to acoustic exposure in a manner that is highly context‐dependent. Metabolic rate was seen to increase (benchmarked against a TCP control), cell adhesion prevented and morphology changed—all depending upon the conditions used and cell lines studied. Importantly, many of these occurred when no difference was seen in live/dead assays, highlighting the urgent need for a more nuanced approach when evaluating the biological impact of acoustic exposure. Based on the cell morphology data (Figures 2 and 3 ), we show adverse effects under some acoustic conditions in which cells were unable to attach or spread less effectively, so altering critical adherent cell functionality. This occurred to varying degrees, ranging from cells that could adhere to the substrate, but not spread (for the case of L929 cells; Figure 2 ) to cells that showed a complete inability to adhere to the substrate (e. g. , MSCs as in Figure S3a and MG63 in Figure S3b, Supporting Information). As shown in Figure 8 a, we postulate a link with cell stiffness. As stiffness increases, we observe a better resistance to morphological changes resulting from the exposure conditions tested. The highly mechanosensitive cell considered, MSCs are more susceptible to external stressors, the MG63 cells slightly less so, both of which succumbed to cell attachment issues arising from acoustic exposures (see Figure 3, Figure S3a, b, Supporting Information) when passed through the 25 µm high channel at 10 µL min −1. This trend continues with the stiffer, L929 cells which succumbed to alteration in cell aspect ratio postexposure (i. e. , cell spreading was affected) but successfully attached (Figure 3 c). Finally, the stiffest cell considered, HaCaT was unaffected by all exposure conditions tested. Figure 8 a) Description of distinct cell fate in terms of morphological changes as a function of cell line and exposure condition as the cells were passed through the 25 µm high channel device at 10 µL min −1 (see Figures 2 and 3 and Figure S3, Supporting Information). b) Percentage difference in cell metabolic activity normalized to TCP as a function of cell line with increasing stiffness based on results reported in Figure 6 at a 400 mV acoustic exposure at 72 h postexposure. A similar trend is suggested for the reported variations in metabolic activity as a function of cell stiffness. We observed variations in the cell metabolic activity as a result of shear induced by flow as well as acoustic exposure, and showed clearly that this was not due to a change in absolute cell number from altered proliferation rates. First, the cellular metabolic activity was suppressed as a result of shear stress induced by the flow. Second and more importantly, we report an increase in cellular metabolic activity as a direct result of acoustic exposure (see Figure 6 d‐ii–iii) for the first time. However, a further increase in acoustic exposure beyond a threshold (cell line‐dependent) results in a detrimental effect. This either 1) directly compromises the cells membrane, effectively killing the cells as in the case of MSCs (Figure 6 a‐i), or 2) stresses the cells excessively such that the cell's metabolic activity is significantly suppressed, potentially due to the cessation of normal biological function (Figure 6 d‐ii–iii), in the absence of changes to structural integrity. The latter occurs while measurements also yield a low mortality percentage, showing the inadequacy of using live/dead assays as a single readout. Stamp et al. 70 observed enhanced cell migration within a tissue under the influence of ultrasonic excitation. In this context they hypothesized possible mechanisms including mechanical actuation, better delivery of nutrients, and thermal and electrical effects. While this study does not deal with metabolic activity of individual cells, and the conditions within our microfluidic channel are more easily controlled eliminating some of the possible mechanisms, others could potentially apply. Furthermore, we suggest a link with cell stiffness. A trend of increasing metabolic activity as the cell stiffness increases was observed, (Figure 8 b). Here, the percentage difference in metabolic activity as compared to TCP is reported. This is consistent with an increase in the resultant acoustic radiation force as the stiffness of the suspended matter increases. 44, 64 Alternatively, we hypothesize that the cells are actively attempting to improve resistance to an external stress by increasing their structural stiffness. The MTS assay used as a surrogate marker of metabolism functions through a formazan reduction accomplished by nicotinamide adenine dinucleotide phosphate (NADPH) or nicotinamide adenine dinucleotide (NADH). 71 Due to NADPH's link to the regulation of cholesterol synthesis, 72 we could assume that our findings may be a result of cell stiffness increase in response to acoustic exposure. Studies suggesting roles for cholesterol in cell membrane stiffness and tension have been conducted; 73 however, further investigation would be required to validate these theories. Interestingly, the data variability of each cell consistently decreases as cell stiffness increases, showing the smallest variance in HaCaT (10 kPa 59 ) and L929 (4 kPa 74 ) and larger variance for the softer MG63 and MSCs (0. 3–0. 8 kPa 57, 58 ) respectively (see Table S4 in the Supporting Information for tabulated values). This is indicative of the inherent heterogeneity which is known to be more prevalent in softer cell types (i. e. , MSCs) than stiffer ones (i. e. , HaCaT). 5 Conclusion In conclusion, our data clearly show differences in cell behavior in response to acoustic stimulation, aspects that are not evident when using the standard live/dead stain as a single readout of biocompatibility. This has critical implications for the methodologies used to evaluate the biocompatibility of acoustofluidic devices and platforms and makes a strong case for the inclusion of a broad panel of cellular readouts. The variations and chronic thresholds in response are cell type‐specific and so safe operation ranges should be considered while developing acoustic based microfluidic platforms with reference to the cell type used. Our findings also reveal a tantalizing hint toward a correlation between acoustic exposure, cell stiffness and cellular metabolism, which if understood, could be harnessed for therapeutic applications in the future. 6 Experimental Section Cell Culture : Mesenchymal Stromal Cells (human, bone marrow derived, Lonza) and MG63 (human osteosarcoma, ATCC) cells were maintained in Dulbecco's modified Eagle medium (DMEM) containing [1 g L −1 ] d ‐glucose and [110 mg L −1 ] sodium pyruvate (Life Technologies), supplemented with [100 U mL −1 ] penicillin‐streptomycin (Life technologies) and 10% (v/v) fetal bovine serum (FBS) (Scientifix). HaCaT (human keratinocyte, ATCC) and L929 (mouse fibroblast, ATCC) cells were maintained in DMEM containing [4. 5 g L −1 ] d ‐glucose and [110 mg L −1 ] sodium pyruvate (Life Technologies), supplemented with [100 U mL −1 ] penicillin‐streptomycin (Life technologies) and 10% (v/v) FBS (Scientifix). Maintenance was undertaken in humidified conditions at 37 °C with 5% CO 2. All cells were routinely screened for and confirmed free of mycoplasma every 3 months. 24 h prior to experimentation, cells were serum‐starved overnight in standard culture media supplemented with only 0. 25% FBS. Acoustic Exposure : The IDTs were designed such that the SAW wavelength, λ SAW at the desired frequency was twice the pitch of the electrodes. Here a wavelength of 80 µm was used, resulting in an operational frequency of 48. 5 MHz. The λ SAW was selected such that the half acoustic wavelength, λ ac, in fluid was approximately the size of the cells considered (i. e. , λ ac /2 = 15 µm). Energy coupled into the fluid was lost from the substrate wave; hence, a pair of IDTs was used to offer a more uniform field than could be expected from a single IDT. Following serum starvation, cells were lifted using TrypLE Express (LifeTech), collected at the desired density and transferred to microfluidic devices for exposure. Cells were maintained at a constant flow rate of 5 or 10 µL min −1 as per the experimental data set, while being excited at the designed frequency and power of either 400 mV (500 mW; accommodating for amplification via power amplifier and s11 values) or 800 mV (1375 mW; accommodating for amplification via power amplifier and corresponding s11 values) typical of similar SAW‐based manipulation platforms at these frequencies, accommodating for the flow rates. 6, 28, 60, 75, 76 To circumvent issues related to clogging, cell aggregation was avoided via frequent agitation of the syringe and a 180° rotation every 3 min to avoid sedimentation. To minimize any external influences, the chip was actively cooled to maintain its temperature via the aid of a peltier cooler, accompanied by a heat sink and fan. The cell numbers across each experimental run were maintained at either 2500 cells cm −2 (MSCs and MG63 cells) or 8000 cells cm −2 (HaCaT and L929 cells) throughout. MTS Metabolic Assay : Metabolic activity was screened using a CellTiter 96 AQueous One Solution assay kit (Promega, USA) with a staining solution made up according to the manufacturer instructions. MTS assays were conducted 24, 48, and 72 h following exposure in which cells were rinsed with phosphate‐buffered saline (PBS) and left in MTS staining solution for 3 h. Solutions were then removed and absorbance quantified at 490 nm using a Multiskan spectrum plate reader (Thermo). Live/Dead Staining Assay : Cell viability was assessed using a live/dead assay kit (Life Technologies, USA) with a staining solution made up according to the manufacturer instructions. Live/dead staining was conducted at time points of 24 h postexposure. Samples were imaged using a Nikon eclipse Ts2 microscope followed by further quantitative analysis using ImageJ. Viability was calculated using the following formula: (live cell count / (live cell count + dead cell count)) × 100. Immunofluorescence Staining : To validate cell proliferation alongside cell/nuclear morphology changes, immunofluorescence staining was performed. Cell monolayers were washed with PBS and fixed in 4% Paraformaldehyde (PFA) (Sigma, USA) diluted in PBS for 30 min at room temperature (RT). Cells were permeabilized with 0. 5% Triton X‐100 (Sigma, USA) diluted in PBS for 15 min, then blocked in 3% (w/v) bovine serum albumen (BSA) diluted in PBS for 30 min. Cells were then incubated in primary antibody (Ki67, AbCam [1:1000]), diluted in blocking solution for 1 h at RT. Cells were subsequently washed 3 × 5 min in PBS and further incubated in a mixture of secondary antibody (antimouse alexa fluor 555, Sigma, USA), Actin‐green (Life Technologies, USA) and Hoechst 33 342 (Life Technologies, USA) at a dilution of 1:300, 2 drops mL −1 and 1:2000 respectively for 1 h at RT. Images were taken as described for live/dead staining assays. Quantification of Cell and Nuclear Phenotypes : For the assessment of both cell and nuclear phenotypes, immunofluorescence images counterstained with Phalloidin (cell phenotypes) and Hoechst (nuclear phenotypes) were processed in ImageJ. Images were thresholded, counted, and analyzed using the shape descriptors plugin, with analysis parameters maintained as to enable direct comparisons between conditions. Cell area and aspect ratios provide detailed information on cell shape and orientation in 2D, whereas nuclear circularity index is a measure of how closely the nuclear shape resembles that of a mathematically perfect circle (scale from 0 to 1, with the perfect circle denoted as 1). Nuclear circularity index of < 0. 5 are potential indicators of a diseased state, e. g. , nuclear blebbing in cancer or progeria development. Statistical Analysis : A Kolmogorov–Smirnov test was used to test data for normal distribution and Levene's test was used to determine homogeneity of variance. Data with a normal distribution were analyzed by one‐way ANOVA and Tukey (equal variance) or Games‐Howell (unequal variance) post hoc tests. Nonparametric data were analyzed by Kruskal–Wallis test unless otherwise stated. All statistical analysis was performed using GraphPad Prism v7. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902359
| 2,019
|
Advanced Science
|
Recent Advances in Chemical Functionalization of 2D Black Phosphorous Nanosheets
|
Abstract Owing to their tunable direct bandgap, high charge carrier mobility, and unique in‐plane anisotropic structure, black phosphorus nanosheets (BPNSs) have emerged as one of the most important candidates among the 2D materials beyond graphene. However, the poor ambient stability of black phosphorus limits its practical application, due to the chemical degradation of phosphorus atoms to phosphorus oxides in the presence of oxygen and/or water. Chemical functionalization is demonstrated as an efficient approach to enhance the ambient stability of BPNSs. Herein, various covalent strategies including radical addition, nitrene addition, nucleophilic substitution, and metal coordination are summarized. In addition, efficient noncovalent functionalization methods such as van der Waals interactions, electrostatic interactions, and cation–π interactions are described in detail. Furthermore, the preparations, characterization, and diverse applications of functionalized BPNSs in various fields are recapped. The challenges faced and future directions for the chemical functionalization of BPNSs are also highlighted.
|
1 Introduction With the isolation of graphene in 2004, 2D materials have become a prime focus in material science research because of their extraordinary physicochemical properties such as large specific surface area, good mechanical strength, and high optical transparency. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 Inspired from the wide applications of graphene, a series of 2D layered materials such as transition metal dichalcogenides (TMDs), 20, 21, 22, 23 hexagonal boron nitride ( h ‐BN), 24, 25 graphitic carbon nitride ( g ‐C 3 N 4 ), 26, 27 layered metal oxides, 28 layered double hydroxides 29, 30 2D polymers, 31 transition metal carbides or carbonitrides (MXenes), 32, 33 and elemental analogues of graphene 34 have been established within the past few years. However, there are few inadequacies with these materials which limit their ideal performance in practical applications. For example, for field effect transistors (FETs), graphene is a zero‐bandgap semiconductor 35 whereas TMDs have a tunable bandgap with a lower charge carrier mobility (10–100 cm 2 V −1 s −1 ). 36 Black phosphorous (BP), an allotrope of phosphorous, can be exfoliated into few‐layer nanosheets using the scotch tape based microcleavage. 37 Monolayer BP shows a puckered structure along the armchair direction and a bilayer configuration along the zigzag direction ( Figure 1 a). The phosphorus atoms are chemically connected to each other through sp 3 ‐hybridized covalent bonds within the layer and different layers of BP are stacked together by weak van der Waals interaction. 38 BP shows thickness dependent direct bandgap (Figure 1 b) from 0. 3 eV (bulk BP) to 2. 0 eV (monolayer), which is significantly larger than monolayer graphene and similar to TMDs (1. 2–1. 8 eV). Monolayer BP possesses a charge carrier mobility of 1000 cm 2 V −1 s −1 at room temperature which is much larger than layered TMDs. 36 As a result, BPNSs show an excellent on/off ratio of 10 3 –10 4 in FETs and are considered as a promising material which can potentially bridge the gap between graphene and 2D TMDs. 39, 40, 41 Importantly, compared to other 2D materials, BPNSs possess high in‐plane anisotropic properties due to their exceptional puckered structure. 42, 43, 44, 45, 46, 47, 48, 49, 50 Because of these exciting properties, BPNSs are being utilized for various applications including photocatalysis, 51, 52 biomedicines, 53, 54 lithium and sodium ion batteries, 55, 56, 57 lithium–sulfur batteries, 58 supercapacitors, 59, 60 FETs, 61 optoelectronics devices, 62, 63, 64, 65 sensing, 66, 67, 68 etc. Figure 1 a) Schematic structure of three‐layer BPNSs. Reproduced with permission. 98 Copyright 2015, John Wiley and Sons. b) Band structures obtained from DFT calculations for monolayer (1L), bilayer (2L), trilayer (3L), and bulk BP. Reproduced with permission. 99 Copyright 2014, American Physical Society. Although BP is unique in many ways, chemical degradation of phosphorous into phosphorus oxides in the presence of ambient oxygen and water results in the rapid loss of semiconducting properties, due to the high reactivity of the lone pair electrons in BP. 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79 According to theoretical calculations, O 2 molecule prefers a perpendicular configuration to approach the surface of BPNSs and tends to dissociate with exothermic energy of −4. 07 eV per O 2 molecule. 71 The calculated lower dissociation barrier of 0. 54 eV facilitates the oxidation readily at room temperature. After dissociation, atomic oxygen prefers a dangling configuration on the BP surface. 80 Due to the high polar nature, H 2 O molecules prefer to bind the surface through hydrogen bonding to the lone pair of electrons in BPNSs. Based on calculations, H 2 O molecules will not dissociate on the surface directly as this is associated with a higher endothermic energy. However, for the oxidized BPNSs, the endothermic energy is significantly decreased, which enables dissociation of H 2 O on BPNSs surface easily. 71 However, there are controversial reports on degradation, in which BPNSs can react with water even in the absence of oxygen. 67 In addition to this, light irradiation can further speed up the degradation through photooxidation. 70, 81, 82, 83 Based on ab initio electronic structure calculations and molecular dynamics simulations, the light induced ambient degradation of BPNSs involves three steps. 81 In the initial stage, superoxide (O 2− ) is generated through a charge transfer reaction between BPNSs and oxygen under ambient light. In the next stage, the superoxide dissociates at the surface of BPNSs and forms two P—O bonds. Finally, water molecules interact with the oxygen atom in the P—O bond through hydrogen bonding, remove the O atom and its bonded P atom from the surface, and break the P—P bond connected to the bottom layer. In addition, an enhanced degradation was observed experimentally with the decrease in the thickness of BPNSs. 69, 84 The bandgap of monolayer BPNSs matches with the redox potential of O 2 /O 2− and thereby, charge transfer rate from BPNSs to O 2 is enhanced and thus leads to a higher oxidation rate. To this end, continuous efforts have been devoted from researchers to overcome the poor ambient stability of BPNSs and various techniques such as Al 2 O 3 protective layer coating, 85, 86, 87, 88, 89 h ‐BN encapsulation, 90, 91, 92, 93 hybrid Al 2 O 3 /BN encapsulation, 94, 95 ionophore coating, 96 and chemical modification, 97 have been demonstrated. For example, Al 2 O 3 protection can preserve the properties of BPNSs over 7 d of ambient exposure, however, a long‐term stability is still challenging. Furthermore, in term of using BPNSs toward solution processing, chemical functionalization is more preferred and required. Chemical functionalization of BPNSs using various covalent and noncovalent approaches is an effective and controllable approach to passivate and modify the properties of BPNSs. For instance, chemical functionalization utilizes the lone pair of electrons present on the phosphorous atom to form direct chemical bonds, can thus, protect BPNSs away from oxygen, to achieve a long‐term stability for BPNSs at ambient conditions. As the research on BPNSs continues to grow rapidly, critical reviews on the recent advances in various aspects of BPNSs are essential for further developments. Recently, Hui and co‐workers reported a review on the progress of preparation and electrochemical energy storage applications of 2D BPNSs. 100 Furthermore, the advancements in different applications of BPNSs including photocatalysis, sensors, optoelectronics, photovoltaics and, biomedicines were also reviewed. 36, 39, 52, 57 Considering the poor ambient stability of BPNSs, Abate and co‐workers reviewed different passivation techniques, particularly on inorganic coating and encapsulation methods. 77 In addition, the review was also focused on anisotropic photophysical surface properties of BPNSs. Hirsch and co‐workers reviewed MoS 2 and BP functionalization as post graphene chemistry with a general discussion on the reactivity of BP toward chemical functionalization. 101 In this regard, there is an urgent need for a timely, comprehensive, in‐depth and critical review focusing on various latest development of chemical functionalization strategies of BPNSs, which hold a great potential not only for passivation but also for tuning the properties of BPNSs. Herein, an overview on the chemical functionalization of BPNSs based on various covalent and noncovalent approaches is presented ( Scheme 1 ). For the covalent functionalization methods, different reaction mechanisms involving radical addition, nitrene addition, nucleophilic substitution, and metal coordination are described. In addition, various noncovalent strategies such as van der Waals, electrostatic, and cation–π interactions are also included. Preparation, characterization, and applications of functionalized BPNSs are also discussed. This timely review on the various chemical functionalization approaches will contribute to the future development of research in BPNSs. Scheme 1 Chemical functionalization of BPNSs using covalent and noncovalent methods. 2 Preparation of BPNSs Mono‐ and few‐layer BPNSs can be prepared through both top‐down and bottom‐up approaches. In the top‐down method, bulk BP crystals are exfoliated into BPNSs by breaking the weak van der Waals interactions between stacked layers using external forces. Depending on the external force, the exfoliation can be mechanical exfoliation, liquid phase ultrasonic exfoliation, and electrochemical exfoliation. On the other hand, for the bottom‐up approaches, chemical vapor deposition (CVD) is the most common used method. 2. 1 Top‐Down Methods 2. 1. 1 Mechanical Exfoliation BPNSs can be mechanically peeled off from bulk BP crystals using scotch‐tape. 37 Although this method can produce high quality 2D BPNSs in a simple and cost‐effective manner, the obtained BPNSs show inhomogeneous size, poor repeatability, and extremely low production yield. To obtain BPNSs in a high yield, several alternative methods were employed including metal assisted mechanical exfoliation 102 and exfoliation using polydimethylsiloxane (PDMS) stamp, 70 viscoelastic stamp using blue Nitto tape, 103 and poly(methyl methacrylate)/poly(vinyl alcohol) stack. 104 In another approach, Zhu and co‐workers reported a large‐scale preparation of relatively stable BPNSs through high energy solid‐state mechanical ball milling of bulk BP in the presence of an additive LiOH. 105 The high‐energy mechanical milling facilitates the generation of free radicals at the edges by breaking the P—P bonds and thus leads to the formation of stable hydroxyl functionalized BPNSs. In addition, this method has been widely utilizing as an efficient method to synthesize functionalized BPNSs for various applications. 106, 107 2. 1. 2 Liquid Phase Ultrasonic Exfoliation In this method, bulk BP is added into a solvent or solvent mixture, which is subjected to ultrasonication‐assisted exfoliation followed by centrifugation/standing by to yield mono‐ and few‐layer BPNSs. In this approach, the choice of solvent plays a key role in the yield and stability of the as‐exfoliated BPNSs. It was found that solvents with surface tension of 35–40 mJ m −2 facilitate the exfoliation of bulk BP. 108, 109, 110 With respect to the solvent system, liquid phase exfoliation can be categorized into organic solvent exfoliation, water phase exfoliation, surfactant assisted exfoliation, polymer or ionic liquid (IL) assisted exfoliation. BPNSs dispersions can be prepared by exfoliation of bulk BP in organic solvents such as N ‐methyl‐2‐pyrrolidone (NMP), 111, 112 N ‐cyclohexyl‐2‐pyrrolidone (CHP), 67 γ‐butyrolactone, 113 dimethylformamide (DMF), 114 dimethyl sulfoxide, 114 and isopropanol. 113, 115, 116 Hersam and co‐workers have demonstrated that the size of BPNSs can be controlled by processing time and sonication power with NMP as the optimal solvent. 111 In order to further improve the efficiency of exfoliation, several modified methods have been developed. For instance, the addition of NaOH in NMP results in a high yield exfoliation of BPNSs with a good stability in water ( Figure 2 a). 98 Furthermore, the concentration of BPNSs in CHP by probe sonication instead of bath sonication can reach as high as 1 mg mL −1. 67 Small organic molecules assisted liquid phase exfoliation can also produce high quality BPNSs for photocatalysts. 117 Figure 2 a) Schematic illustration of the fabrication process of NaOH‐NMP‐exfoliated BPNSs. Reproduced with permission. 98 Copyright 2015, John Wiley and Sons. b) Electrochemical exfoliation procedure of BP in an acidic aqueous solution. The starting BP crystals (left) and the exfoliated BPNSs in DMF (right) are also shown. c) No potential applied. d) After 20 min of applying a voltage of +3 V. e) After 2 h of applied voltage. Reproduced with permission. 126 Copyright 2017, John Wiley and Sons. f) Strategy for the synthesis of thin BP films using CVD and schematic apparatus for the deposition of red phosphorous (RP) film. Reproduced with permission. 135 Copyright 2015, Institute of Physics (IOP) Publishing Ltd. In addition to organic solvents, water has been employed as a medium for exfoliation of BPNSs. 118, 119 The surface tension of water can be tuned by adding surfactants into water. For example, Hersam and co‐workers utilized deoxygenated water with amphiphilic sodium dodecyl sulfate as the surfactant and achieved a higher concentration of BPNSs in water with flake thickness thinner than the ones produced using anhydrous organic solvent exfoliation. 120 In another case, Peng et al. has reported polymer assisted exfoliation of bulk BP in polyvinylpyrrolidone solution. 121 When bulk BP was exfoliated in ILs or polymer ionic liquids (PILs), mono‐ or multilayer BPNSs show high oxidation resistance at ambient conditions due to their interaction with ILs or PILs. 122, 123 Although an improved stability is achieved, the high boiling point of ILs/PILs may hinder the application in surface deposition since it is not so easy to remove the ILs/PILs completely. 2. 1. 3 Electrochemical Exfoliation In electrochemical exfoliation, parameters such as electrolyte, operating voltage, and precursors, are very important for a successful exfoliation. 124 Erande and co‐workers used bulk BP as working electrode and Na 2 SO 4 as the electrolyte to yield BPNSs of 3–15 stacked layers. 125 In the mechanism, oxygen and hydroxyl radicals (produced from the oxidation of water) were generated by the positive bias on the working electrode and inserted between adjacent layers of bulk BP, thus reducing the interlayer interactions. Subsequently, BPNSs were separated out from the bulk BP when the free radicals were oxidized to generate oxygen. The obtained yield of BPNSs is high (>80%), indicating that this approach can be potentially applied to large scale production. Modified electrochemical exfoliation approaches have been demonstrated by Pumera and co‐workers using bulk BP as working electrode, Pt foil as the counter electrode, and H 2 SO 4 as the electrolyte (Figure 2 b–e). 126 Huang et al. have reported an electrochemical cation (tetrabutylammonium) insertion method to obtain few‐layer BPNSs. 127 The number of layers can be controlled by adjusting the applied potential. Feng and co‐workers have reported an electrochemical delamination strategy to produce BPNSs: the one using tetra ‐ n ‐butylammonium cations and bisulfate anions can have an exfoliation yield up to 78% and flake sizes on a micrometer scale. 128 Despite progress has been made, 129, 130, 131, 132, 133, 134 it is still challenging to prepare BPNSs mainly consisting of monolayer flakes through electrochemical exfoliation approach, which requires a better understanding of the mechanism of electrochemical exfoliation. 2. 2 Bottom‐Up Method CVD is a well‐established bottom up technique for fabricating large area layered materials. 136, 137, 138, 139, 140, 141, 142 Ji and co‐workers have demonstrated an in situ CVD approach to fabricate BPNSs with a thickness of four layers and average areas of >3 µm 2. In their method, amorphous red phosphorous (RP) thin film was first grown on a silicon substrate by heating RP powder or bulk BP at 600 °C. Then both high temperature and high pressure were applied to convert the RP thin film to BPNSs. 143 Similar method was reported by Xia and co‐workers, RP film was thermally evaporated onto a flexible polyethylene terephthalate substrate and then converted into BP nanofilm under high pressure at room temperature (Figure 2 f). 135 Although CVD method has been successfully used to fabricate 2D nanomaterials including graphene, TMDs, and h ‐BN, 140, 144, 145 the progress for the synthesis of BPNSs using CVD approach is still in the early stage and more efforts are required to grow much thinner BPNSs. Another widely used bottom up method is the direct chemical synthesis of 2D materials from small precursors by hydrothermal, solvothermal, or templated synthesis methods. 146, 147, 148, 149, 150 This method provides an alternative way to synthesize/prepare 2D materials on a large scale and in a relatively low‐cost manner. Recently, partially oxidized BPNSs have been prepared through a one‐step solvothermal approach by reacting white phosphorus in ethylenediamine at 100 °C for 12 h. 151 However, this is an largely unexplored area in the preparation of BPNSs, more suitable precursors and synthetic methods are highly demanded for the efficient and cost‐effective synthesis of BPNSs. Although various top‐down and bottom‐up approaches have been developed, large scale production of high quality BPNSs with precise control over the number of layers is still lacking. Therefore, development of more powerful synthetic methods for producing high quality BPNSs is required for further research progress in this field. 3 Characterization of BPNSs As with other 2D materials, the quality of BPNSs can be characterized using various spectroscopic and microscopic techniques. Raman spectroscopy is a very useful and powerful technique for the characterization of BPNSs. 152, 153, 154 The Raman spectrum of BPNSs exhibits three peaks located at 361, 437, and 465 cm −1, which are assigned to out‐of‐plane phonon mode (A 1 g ), in‐plane modes along the zigzag direction (B 2g ), and in‐plane modes along the armchair direction (A 2 g ), respectively ( Figure 3 a). 105, 155 The Raman peaks of BPNSs are thickness dependent and show redshifts with increasing layer number, especially for the in‐plane mode A 2 g. 98 In addition, the intensities of the three vibrational modes increase with layer number (Figure 3 b). The intensity ratio of A 2 g to A 1 g was found to be higher for monolayer than multilayer BPNSs. 156 The ratio of A 1 g /A 2 g is highly indicative for oxidative status of BPNSs and shows an exponential decay with air exposure time, 69 which is also recapped afterward in the chemical functionalization part. Figure 3 a) Raman spectra of bulk BP and BPNSs with different numbers of layers. b) Layer‐dependent Raman peak enhancement. Reproduced with permission. 98 Copyright 2015, John Wiley and Sons. c) XPS P2p peaks of BPNSs. Reproduced under the terms of the Creative Commons Attribution 4. 0 International License. 159 Copyright 2017, The Authors, Published by Springer Nature. Morphology of few‐layer BPNSs: d) TEM image, e) AFM image, and f) corresponding height image. Reproduced with permission. 119 Copyright 2015, American Chemical Society. g, h) HRTEM images with different crystal lattices and i) SAED patterns. Reproduced with permission. 98 Copyright 2015, John Wiley and Sons. X‐ray photoelectron spectroscopy (XPS) is often used to determine the chemical composition of BPNSs, for example, the presence of elements and functional units. In the high resolution P2p spectra, BPNSs show two intense peaks centered at 129. 6 and 130. 4 eV correspond to P2p 3/2 and P2p 1/2 signals of P—P bonds and a weak broad peak centered at 134. 3 eV from PO x species due to the oxidation of BPNSs (Figure 3 c). BPNSs typically show a 15–20% oxide shoulder with respect to the P2p peak corresponding to the P—P bonds. 67 Fourier‐transform infrared spectroscopy (FTIR) measurements can provide information about functional groups present on BPNSs, which show two weak vibrational peaks at 1183 and 1005 cm −1 due to the presence of P=O and P—O groups from surface oxidation. 157 When BPNSs are further functionalized, the foreign functional groups attached/bounded onto BPNSs can show distinct signals in the FTIR spectrum to help with the structure analysis. 31 P NMR spectroscopy is another possible characterization tool for BPNSs. The characteristic peak for BPNSs is at ≈ 20. 15 ppm ascribed to the P—P bond. 158 The peak position can change when BPNSs are chemically modified, 107 which will be discussed later in the chemical functionalization part. The morphology and crystallinity of BPNSs can be examined by transmission electron microscope (TEM) and selected area electron diffraction (SAED) measurements (Figure 3 d, g–i). The characteristic lattice parameters for BPNSs are about 3. 23 Å and 2. 24 Å, corresponding to the (012) and (014) plane of BP crystal, respectively. 160 The SAED pattern indicates the highly crystalline nature of the BPNSs. Besides TEM, scanning electron microscope can be also used to study the morphology of BPNSs, possibly with the combination of elemental mapping using energy dispersive X‐ray (EDX). 161 Thickness and size of BPNSs can be investigated by atomic force microscopy (AFM). The thickness of a monolayer phosphorene is considered to be 0. 53 nm. 160 However, in practice the thickness may vary depending on the substrate used and whether it has been oxidized (Figure 3 e, f). 162 AFM can provide further information about surface functionalization on BPNSs (an increase in thickness) and oxidative degradation by the loss of smooth surface morphology. 97, 162, 163, 164 Due to the thickness dependent bandgap energies, BPNSs have a wide range of UV–vis absorption spectra from visible to infrared region. BPNSs show highly anisotropic absorption properties originate from the symmetry forbidden selection rule. 166, 167 Because of the anisotropic effect, BPNSs also exhibit dichroism which means the extent of absorption depends on the polarization state of incident light. 168 It was found that BPNSs absorb polarized visible light more easily in the armchair direction with a larger absorption coefficient than in the zigzag direction ( Figure 4 a, b). 45, 169 Because of the tunable bandgap, BPNSs also exhibit highly layer dependent photoluminescence from mid‐infrared to near‐infrared (NIR) wavelengths (Figure 4 c). 165, 170 Since the photoluminescence is originated from anisotropic excitons generated in BPNSs, the light emission is largely polarized along armchair direction (x direction in Figure 4 d). In addition, the photoluminescence intensity also depends on excitation light polarization. Photoluminescence quenching effect was also observed for BPNSs when heterostacked with MoS 2 or WS 2. 171 Figure 4 Typical absorbance spectra of a) a thin (9 nm) and b) a thick (225 nm) BP flake with incident light polarization along the armchair and zigzag directions. Reproduced with permission. 45 Copyright 2015, American Chemical Society. c) Photoluminescence spectra of 2L, 3L, 4L, and 5L BPNSs. Reproduced with permission. 165 Copyright 2014, American Chemical Society. d) Photoluminescence peak intensity as a function of polarization detection angle for excitation laser polarized along x (gray), 45° (magenta) and y (blue) directions. Reproduced with permission. 166 Copyright 2015, Springer Nature. Besides the above techniques, thermogravimetric analysis (TGA) can be used to study the thermal stability of BPNSs and also determine the content of functional units in chemically modified BPNSs. 105, 172 X‐ray diffraction (XRD) can reveal the crystalline structure and interlayer distance of (functionalized) BPNSs. 118, 173 4 Covalent Functionalization on BPNSs Various functional molecules/polymers or materials can be chemically attached onto BPNSs through direct P—C and/or P—O—C bond formation. So far, reactions involving different reactive intermediates, such as free radicals, nitrenes, and carbocations, have been reported for covalent functionalization of BPNSs. Formation of P—C and/or P—O—C bonds can not only passivate BPNSs but also potentially introduce new properties. On the other hand, covalent functionalization may largely alter the electronic properties of BPNSs due to direct breaking of P—P bonds. 4. 1 Radical Addition In the past, covalent functionalization of nanomaterials such as graphene, transition metal dichalcogenides, and carbon nanotubes using free radicals derived from diazonium salts has been well established. 174, 175, 176, 177, 178 As in the case of BPNSs, radical intermediates can be generated using different methods including diazonium reaction, ball milling methods, and P—P bond breaking reactions. For example, diazonium salts can accept one electron from BPNSs and generate a highly reactive phenyl radical after release of nitrogen molecules. The radical subsequently added onto the surface of BPNSs to form a P—C covalent bond ( Scheme 2 ). 97, 179 Scheme 2 Mechanism of P—C bond formation on BPNSs using diazonium chemistry. 4. 1. 1 Reactions Using Diazonium/Diiodonium Salts Hersam and co‐workers have used diazonium chemistry for the covalent functionalization of BPNSs ( Figure 5 a). 97 Few‐layer BPNSs were first prepared by mechanically exfoliating bulk BP on Si/SiO 2 substrates and were then chemically modified with 4‐nitrobenzenediazonium (4‐NBD) and 4‐methoxybenzenediazonium (4‐MBD) tetrafluoroborate salts. From density functional theory (DFT) calculations, the formation of P—C bonds was found to be thermodynamically favorable with two aryl moieties per supercell of 16 phosphorous atoms. AFM measurements showed an increase of ≈1. 5 nm in the height of the BP flake and an increased surface roughness upon functionalization (Figure 5 b, c). 4‐NBD and 4‐MBD showed different conversion rates for the functionalization of BPNSs: the former is higher than the latter. In the P2p measurements, a broad peak at ≈133 eV corresponding to the P—C bonds was observed for 4‐NBD after 30 min of functionalization, whereas it was only evident after 180 min for 4‐MBD. Similarly, the doublet at ≈130 eV in the P2p spectra shows a more pronounced decrease in the intensity for 4‐NBD than for 4‐MBD after 180 min. From confocal Raman spectroscopy, the A 1 g mode for BPNSs modified with 4‐NBD diminishes more rapidly than that for BPNSs modified with 4‐MBD. In the proposed reaction mechanism, a single electron transfers from BPNSs to the diazonium salts, leads to an elimination of N 2 and the generation of reactive aryl radicals. The electron deficient radicals subsequently attack the basal BPNSs plane yielding the formation of P—C bonds. The electron transfer from the BPNSs to aryl diazonium ion is the rate‐limiting step for aryl diazonium reactions. The lower reduction potential of 4‐MBD relative to 4‐NBD results in a less‐favorable electron transfer, which is manifested in a slower P—C bond formation. After 3 weeks of ambient exposure, the aryl diazonium functionalized BPNSs exhibited an increased stability against oxidative degradation compared with the unmodified BPNSs. The chemical modification of BPNSs resulted in a controllable p‐type doping effect with an enhanced hole mobility and on/off ratio up to 10 6 in FETs. Figure 5 a) Reaction of benzenediazonium tetrafluoroborate derivatives and mechanically exfoliated few‐layer BPNSs (light blue) on a Si (gray)/SiO 2 (purple) substrate. b) AFM image (top) of BPNSs prior to functionalization, along with the height profile extracted along the blue line (bottom). c) AFM image (top) of the same flake after 30 min of exposure to 10 × 10 −3 m 4‐NBD. Reproduced with permission. 97 Copyright 2016, Springer Nature. In order to overcome the poor air stability and improve dispersibility of BPNSs, Chen and co‐workers synthesized a conjugated polymer derivative modified BPNSs, poly[(1, 4‐diethynylbenzene)‐alt‐9, 9‐bis(4‐diphenylaminophenyl)fluorene] (PDDF)‐covalently grafted BP (PDDF‐ g ‐BP), by using 4‐bromobenzene‐diazonium (4‐BBD) functionalized BP (4‐BBD–BP) ( Figure 6 ). 180 Initially, BPNSs with an average thickness of 10. 4 nm were prepared by liquid phase exfoliation in NMP and then the 4‐BBD moieties were covalently attached to the BPNSs surface using diazonium chemistry. Then the PDDF‐ g ‐BP was synthesized by Sonogashira coupling reaction of 4‐BBD–BP and 1, 4‐diethynyl benzene along with 9, 9‐bis(4‐diphenylaminophenyl)‐2, 7‐dibromofluorene. The formation of P—C bond was confirmed by the peak at 284 eV of C1s XPS spectra and IR frequency at 828. 53 cm −1. The absence of Br3d signal at 71. 5 eV in the wide scan XPS spectra of PDDF‐ g ‐BP suggests that PDDF was successfully end‐capped with 4‐BBD–BP to produce PDDF‐ g ‐BP. An electron transfer from PDDF to BPNSs was observed for PDDF‐ g ‐BP and was confirmed by steady state fluorescence studies. The photoinduced charge transfer was further explored using light induced electron paramagnetic resonance technique. A significant decrease in the intensity of the EPR signal was observed for the PDDF‐ g ‐BP after illumination, attributed to the formation of PDDF ∙+ —BP ∙– radical ion‐pair. Because of the improved solution processability and homogenous film formation, the Au/PDDF‐ g ‐BP/indium tin oxide (ITO) device showed a good nonvolatile memory performance with an on/off current ratio of 10 4 compared to that of PDDF/BP blends. Interestingly, the on and off current was kept the same even after 200 cycles. Figure 6 Synthesis of PDDF and PDDF‐ g ‐BP. a) Pd[(PPh 3 ) 4 ], CuI, Et 3 N, dry acetonitrile, 80 °C, 72 h. b) N ‐methyl‐2‐pyrrolidone (NMP), ultrasonic radiation for 6 h (200 W). c) 4‐Bromobenzenediazonium tetrafluoroborate, tetrabutylammonium hexafluorophosphate, acetonitrile, room temperature (RT), 3 h. d) 1, 4‐Diethynyl benzene, 9, 9‐bis(4‐diphenylaminophenyl)‐2, 7‐dibromofluorene, Pd[(PPh 3 ) 4 ], CuI, Et 3 N, dry acetonitrile, 80 °C, 72 h. Reproduced with permission. 180 Copyright 2018, John Wiley and Sons. In another case, Yu and co‐workers described a modification strategy utilizing a fluorescent dye Nile Blue 690 via diazonium chemistry. 181 The BPNSs with an average lateral size of ≈35. 0 nm were prepared by liquid phase exfoliation in NMP. The diazonium tetrafluoroborate salt of the dye (NB‐D) was reacted with the exfoliated BPNSs to form stable P—C bonds on the BPNSs surface by aryl diazonium coupling ( Figure 7 ). The NB@BPs showed a broad peak at 133. 3 eV in the P2p XPS spectra and a peak at 284. 2 eV in the C1s XPS spectra, corresponding to P—C bonds in the NB@BPs. The NB@BPs exhibited better photothermal performance and greater fluorescence intensity than the bare BPs, suggesting successful synthesis of fluorescent BPs by NB‐D covalent modification. This study paves the way to develop novel modified BPNSs with exciting properties by attaching functional molecules such as photo‐, electro‐, thermo‐, or piezo responsive materials and other functional dyes. 182 Figure 7 Schematic illustration of a) the synthesis of NB‐D and b) the preparation of NB@BPs. Reproduced with permission. 181 Copyright 2017, American Chemical Society. Collines and co‐workers reported a covalent functionalization of liquid exfoliated few‐layer BPNSs using aryl iodonium salts and demonstrated superior ambient stability compared with arylation using diazonium salts. 183 The highly electron deficient diaryliodonium salts have the ability to arylate both O and P nucleophiles by leaving aryl iodides, which enables covalent functionalization of BPNSs at room temperature with excellent ambient stability ( Figure 8 ). The authors observed that the arylation using iodonium salts showed a greater degree of functionalization than the one using diazonium salts. The mechanism of the aryl iodide reaction with BPNSs is still not clear yet. In general, the P or O nucleophile can either react through a ligand exchange and then by a ligand coupling with the removal of aryl iodide leaving group, or through the radical mechanism as in the case of diazonium salts. In terms of stability, after 1 week of ambient exposure, the arylated BPNSs using aryl iodonium salt showed a 9% increase in the intensity of oxide peak while diazonium functionalized BPNSs showed a 30% increase in the intensity. Compared with diazonium salts approach, the enhanced stability of functionalized BPNSs using iodonium salts are attributed to the higher degree of functionalization of BPNSs. Iodonium salts can be applied to both aryl and alkyl groups providing an efficient route to covalent modification of BPNSs with increased ambient stability. Figure 8 Covalent functionalization of BPNSs using aryl iodonium salts. Reproduced with permission. 183 Copyright 2018, American Chemical Society. 4. 1. 2 Reactions Using Ball Milling Method Solid state mechanochemical ball milling was established as a facile method for the exfoliation of graphite into few‐layer graphene nanosheets. 184, 185, 186, 187, 188 Motivated from this, Yang and co‐workers have prepared stable few‐layer BPNSs (BP‐ball milling (BM)) by ball milling of bulk BP using anhydrous lithium hydroxide (LiOH) as an additive. 105 In the absence of any noble metal cocatalyst, the BP‐BM exhibited significantly enhanced visible light photocatalytic H 2 evolution rate (512 µmol h −1 g −1 ), much better than that of the bulk BP (18 times) and even higher than that of g ‐C 3 N 4. It was found that, the presence of LiOH additive is essential to obtain stable BPNSs and good H 2 evolution activity. In contrast, ball milling without any additive or with an additive of NaCl, resulted in undesirable products such as BP oxides ( Figure 9 ). The stabilization of BP‐BM compared to the bulk BP is due to the formation of edge selective hydroxyl functionalized BPNSs. The formation of hydroxyl groups involves the generation of reactive species such as free radicals via cleavage of P—P bonds during ball‐milling. The functionalization was confirmed by the high‐resolution O1s XPS spectra: the presence of an additional peak at 533. 0 eV for BP‐BM is due to the P—OH bond, while the two intense peaks at 530. 6 and 532. 0 eV correspond to the P—O and P—O—P bonds, respectively. It should be pointed out that, the BP–BM photocatalyst was found to be relatively less stable due to the photooxidation of exfoliated BPNSs, which requires further functionalization strategy to increase the stability. Figure 9 Scheme for the ball milling of BPNSs with additives (LiOH or NaCl) and without any additives. Reproduced with permission. 105 Copyright 2017, John Wiley and Sons. Chemical functionalization of BPNSs on the edges can largely preserve the intrinsic properties of BPNSs, while at the same time can potentially introduce new properties to BPNSs. As an example, edge selective functionalization of BPNSs by covalently bonding of C 60 molecules was reported by Yang and co‐workers for the effective passivation of BPNSs without losing its surface integrity. 107 The BPNSs–C 60 hybrid was prepared using a facile one‐step solid‐state mechanochemical route by ball‐milling of bulk BP and C 60 powders without any additives ( Figure 10 a). The average thickness of BP–C 60 hybrid is ≈2. 5 nm which corresponds to ≈4‐layer nanosheets. An average C 60 molar content of 19 per 1000 P atoms was estimated from TGA. From XRD analysis, the crystal structure of BPNSs was found to be preserved after ball‐milling. In contrast to the ball milled BPNSs (Figure 10 b), the TEM image of the BP–C 60 hybrid shows lattice fringes of (020) plane at the edges of the BPNSs without any amorphous coating (Figure 10 c), indicating that the BP–C 60 hybrid possesses an improved structure stability at ambient condition. In addition, along the edges of the BPNSs, hollow nanospheres with diameter of ≈1. 0 nm were observed in the high‐resolution TEM (HRTEM) image of the BP–C 60 hybrid, which matches well with the morphology of C 60 molecules. This indicates that C 60 molecules are primarily attached at the edges of BPNSs in the BP–C 60 hybrid. In scanning transmission electron microscopy energy dispersive X‐ray (STEM‐EDX) spectroscopic measurements of the BP–C 60 hybrid, a much lower content of the C elements was found than the P elements. The presence of C 60 moiety in the BP–C 60 hybrid was further confirmed by FTIR spectra, showing four characteristic vibrational peaks of C 60 at 526, 576, 1182, and 1428 cm −1. Besides, three additional vibrational peaks (707, 770, and 795 cm −1 ) attributed to the P—C bonds were only observed in the BP–C 60 hybrid but not in the BP/C 60 physical mixture. In the solid‐state 13 C‐NMR spectra of the BP–C 60 hybrid, an additional weak peak at 75. 75 ppm was found due to the formation of sp 3 carbon on C 60, along with intense signals in the 130–160 ppm corresponding to the sp 2 carbon of C 60. The weak peak was absent in the spectra of both BP/C 60 mixture and pure C 60, suggesting the formation of sp 3 ‐carbon on the C 60 cage. The formation of the P—C bonds was further confirmed by the P2p XPS spectra with a new peak centered at ≈133. 5 eV. The possible mechanism for the formation of the hybrid involves the generation of reactive species such as radicals at the edges via a mechanochemical cleavage of P—P bonds in the bulk BP during its exfoliation into few‐layer BPNSs using high‐energy ball milling. The activated C 60 molecules were attached onto the activated edges of BPNSs via covalent P—C bonds. Notably, the degradation rate of BP–C 60 in water was reduced by a factor of 4. 6 compared with that of BPNSs. Since C 60 molecules possess high stability toward light, oxygen, and water, the bonding of C 60 molecules onto the edges of BPNSs can act as a sacrificial shield which effectively prevents BPNSs from the attack of light, oxygen, and water. The chemical attachment of C 60 molecules in the BP–C 60 hybrid results in a photoinduced electron transfer process from BPNSs to C 60, which can suppress the recombination of charge carriers and subsequently enhance the photoelectric conversion property (ten times higher than that of the BP ball milling sample) and also the photocatalytic activity (much higher than the BP/C 60 mixture). Figure 10 a) Schematic representation for the preparation of BP–C 60 hybrid. HRTEM (scale bar: 5 nm) and TEM (inset, scale bar: 100 nm) images of the b) ball‐milled BPNSs and c) BP–C 60 hybrid. Reproduced under the terms of the Creative Commons Attribution 4. 0 International License. 107 Copyright 2018, The Authors, Published by Springer Nature. 4. 1. 3 Reactions Involving P—P Bond Breaking In another strategy, BP nanoflakes (BPNFs) were covalently functionalized with carbon free radicals from azodiisobutyronitrile (AIBN) molecules. 158 According to theoretical predictions, AIBN carbon free radicals are covalently attached to BPNFs through the breaking of P—P bonds (2. 77 Å) and the formation of P—C bonds (2. 01 Å) onto a supercell of 36 phosphorus atoms ( Figure 11 ). The BPNFs‐AIBN with a size of 200–400 nm was prepared by reacting AIBN with BPNFs dispersed in a mixture of NMP and toluene (v/v 1:3) at 75 °C for 4 h under argon atmosphere. Compared with bulk BP, both the A 1 g and B 2g modes of BPNFs and BPNFs–AIBN do not display any changes, while the A 2 g mode showed a blueshift. In the FTIR spectra, the vibration frequencies of the cyano group in AIBN at ≈2242 cm −1 was redshifted to ≈2235 cm −1 after the functionalization reaction with BPNFs. The formation of P—C bonds was proved by the corresponding peak at −9. 7 ppm in the 31 P solid‐state NMR spectra. However, the peak intensity of the P—C bonds for BPNFs–AIBN was only 1. 29%, suggesting that only few carbon free radicals were covalently attached onto BPNFs. In the XPS spectra, the appearance of a new peak at 133. 1 eV attributed to the P—C bonds confirms the covalent functionalization of BPNFs. AFM and HRTEM measurements also indicated that the original morphology and crystalline state of BPNFs were retained after modification by carbon free radicals. Importantly, after functionalization, the stability of BPNFs in air and aqueous solution was significantly improved. However, besides the theoretical predictions, there is no direct experimental evidence that the P—P bonds were breaking during the reaction. Figure 11 Scheme for the preparation of BPNFs–AIBN. Reproduced with permission. 158 Copyright 2018, Royal Society of Chemistry. 4. 2 Nucleophilic Substitution Because of the lone pair electrons on the phosphorous atom, BPNSs can act as nucleophiles in the presence of electrophiles. Under certain conditions, oxygen atoms on hydroxyl groups introduced on BPNSs also make nucleophilic substitution possible. Toward this direction, Pumera and co‐workers have demonstrated covalent modifications of BPNSs to form P—C and/or P—O—C bonds based on nucleophilic substitution. 157 In their approach, BPNSs were prepared by combination of ultrasonication and shear force milling in organic solvents under argon atmosphere. BPNSs can undergo direct reaction with alkyl halides ( Figure 12 ). Alternatively, BPNSs can be first treated with thionyl chloride to form chlorinated BPNSs, which can be further subjected to reaction with aliphatic alcohols. The success of the nucleophilic substitution reactions was proved by XPS, Raman, and FTIR. For Example, the XPS spectra show evidence for the presence of long perfluorinated alkyl chains of nucleophilic reagents and formation of P—O—C bonds. In contrast, the application of highly reactive (S)‐bromomethyl ethanethioate led to the direct formation of the P—C bonds. Furthermore, the appearance of the new modes/bands in the Raman/FTIR spectra also suggests the formation of new bonds. Compared with nucleophilic substitution, other approaches like reaction with organometallic reagents and radical reaction with diazonium tetrafluoridoborates were found to give a lower degree of covalent functionalization. Figure 12 Proposed mechanism of the reaction on BPNSs through nucleophilic substitution with alkyl halides. Reproduced with permission. 157 Copyright 2017, John Wiley and Sons. Taking advantages from the reductive graphene chemistry, Hirsch and co‐workers prepared BP intercalation compounds (BPICs) using alkali metals such as K and Na, which was then reacted with electrophilic alkyl halides to form P—C covalent bonds ( Figure 13 ). 189 The activated BPIC was dispersed in tetrahydrofuran (THF) by ultrasonication and then quenched with alkyl halides to yield the functionalized BPNSs as a dark gray powder. Statistical Raman spectroscopy showed a new band at about 145 cm −1 and additional new modes in the range of 250–300 cm −1, suggesting the formation of the P—C bonds. The same results were observed in the in situ experiments, in which hexyl iodide was slowly evaporated onto the synthesized BPICs under ultrahigh‐vacuum conditions. DFT calculations reproduce the above Raman bands and reveal the formation of radical species and P—P bond breaking in the BP lattice. Thermogravimetric analysis coupled to mass spectrometry (TG‐MS) measurements were performed: the two mass losses between 100 and 300 °C were assigned to the defunctionalization of the hexyl chain from the BP lattice and a sharp mass loss above 400 °C was due to the decomposition of BPNSs. The functionalized BPNSs showed a higher decomposition temperature compared to the pristine BPNSs due to the increased thermal stability. Importantly, the control experiments by functionalization of BPNSs with hexyl iodide without alkali metal intercalation under the same conditions did not show relevant features in the Raman and TG‐MS. Regarding to the stability, the functionalized BPNSs can be stable up to 15 d when exposed to oxygen and moisture. Quantitative magic‐angle spinning 31 P solid‐state NMR measurements indicate that around 7% of negatively charged P atoms react quantitatively in the substitution reaction. Figure 13 Reaction of BP intercalation compounds with alkyl halides through nucleophilic substitution. Reproduced with permission. 189 Copyright 2019, John Wiley and Sons. 4. 3 Nitrene Addition In the diazonium chemistry, the formation of P—C single bonds retains one unpaired electron in the phosphorus atom, which may recede the passivation effect. To solve this problem, Yang and co‐workers reported the covalent azide functionalization of BPNSs, leading to the formation of P=N double bonds with phosphorus atoms and thus largely passivating the highly reactive BPNSs. 155 In their method, BPNSs were prepared by liquid exfoliation of bulk BP in DMF and then mixed with 4‐azidobenzoic acid. The reaction mixture was then stirred in an inert condition at 140 °C for 48 h to yield the functionalized BPNSs ( Figure 14 ). The reaction mechanism involves the in situ generation of nitrene as a reactive intermediate, which attacks the lone pair electrons on the phosphorus atom and results in the formation of the P=N bonds. The as‐prepared functionalized BPNSs were fully characterized by FTIR, solid‐state 31 P NMR, XPS, with additional support from DFT calculations, indicating the existence of P=N double bonds. For example, the shoulder peak at −6. 35 ppm in the 31 P NMR spectra and the intense peak at 401. 5 eV in the N1s XPS spectra are attributed to the P=N bonds. The morphology of both the BPNSs and functionalized BPNSs were studied by TEM/HRTEM and STEM‐EDX, suggesting that the crystallinity of BPNSs was remained after chemical functionalization. Importantly, through azide functionalization, the stability of BPNSs was significantly improved at ambient conditions. Figure 14 Reaction between liquid phase exfoliated BPNSs and 4‐azidobenzoic acid. Reproduced with permission. 155 Copyright 2019, John Wiley and Sons. 4. 4 Metal Coordination The free electron pairs responsible for the oxidative degradation of BPNSs can be effectively protected by sharing with the vacant orbitals of electropositive metals or with electron deficient molecules, through coordination bonds. Chu and co‐workers have designed a titanium sulfonate ligand (TiL 4, L referring to the sulfonic ester group) to react with BPNSs to form TiL 4 ‐coordinated BP (TiL 4 @BP) to enhance the stability of BPNSs in water and humid air ( Figure 15 a). 190 The TiL 4 ligand was synthesized by reacting titanium tetraisopropoxide [Ti(O i Pr) 4 ] with p ‐toluenesulfonic acid in ethanol at 50 °C for 3 h. The ultrasmall BPNSs were prepared in NMP using a liquid exfoliation technique and the TiL 4 @BP with an average size of 3. 3 nm was generated by reacting BPNSs with TiL 4 in NMP at room temperature for 15 h. In the HRTEM image of TiL 4 @BP, the lattice fringes of 2. 1 Å corresponding to the (014) plane of the BP crystal were observed. The 1 H NMR spectra of TiL 4 @BP were found to be similar to those of TiL 4 indicating successful coordination of TiL 4 onto the BPNSs. The binding energy of P2p corresponding to the Ti—P coordination bond was found to be at 132. 4 eV for TiL 4 @BP. For the TiL 4 @BP, the Ti2p 1/2 (463. 5 eV) and Ti2p 3/2 (458. 0 eV) peaks were detected, whereas no Ti2p peak was observed for the bare BP sample, confirming the effective functionalization of BPNSs. To evaluate the role of Ti coordination in the stability of BPNSs, the optical absorbance of TiL 4 @BP and bare BP were monitored. After 72 h, the absorbance of the bare BP at 450 nm was decreased by 55% compared to the original value, whereas the absorbance of TiL 4 @BP was maintained as high as 95%. In contrast to the severe degradation for the bare BP sample (Figure 15 b), the optical images of TiL 4 @BP demonstrated that TiL 4 coordination protects the BPNSs from oxidation in air with a relative humidity as high as 95% (Figure 15 c). After TiL 4 coordination, BPNSs maintain the photothermal performance even after 72 h. The TiL 4 @BP in a physiological environment exhibited excellent biocompatibility with different cells and good photothermal effect in killing the cancer cells when illuminated by an 808 nm laser. The present method provides a simple and efficient way to improve the stability of BPNSs against degradation at ambient conditions. Later on, the authors have employed TiL 4 @BPNSs and TiL 4 @BPQDs to regulate the aggregation of Amyloid‐β (Aβ), providing some insight into the development of functionalized BPNSs to prevent amyloidosis. 191 Figure 15 a) Synthesis of TiL 4 and surface coordination of TiL 4 onto BPNSs. Optical images of b) bare BP and c) TiL 4 @BP after exposure to the humid air at room temperature for different time period (12 and 24 h). Reproduced with permission. 190 Copyright 2016, John Wiley and Sons. Similarly, later on Yu and co‐workers used a lanthanide trifluoromethanesulfonates (LnL 3 ) for the surface coordination of BP nanostructures. 192 The electrophilic ligands strengthen the coordinating ability of Ln 3+ and modification of BP nanostructures with LnL 3 through coordination of the lone‐pair electrons of phosphorus with the empty d‐orbital of the lanthanide ions ( Figure 16 ). The BP quantum dots (BPQDs) were prepared by ultrasonic liquid exfoliation method in NMP and then, the GdL 3 @BPQDs hybrid with a diameter of about 3. 7 nm were prepared by stirring the BPQDs with excessive amount of GdL 3 for 20 h under argon atmosphere. The HRTEM image reveals lattice fringes of 2. 3 Å corresponding to the (014) plane of BP and energy‐dispersive X‐ray spectroscopy (EDS) analysis exhibited characteristic peaks of P, Gd, F, S, C and O. After GdL 3 modification, the Raman scattering peaks was blueshifted and the zeta potential of BPQDs changes from −32. 5 ± 1. 4 to +28. 6 ± 1. 6 mV, confirming the efficient functionalization of BPQDs. Compared with GdL 3, the binding energies of Gd4d 5/2 and Gd4d 3/2 of GdL 3 @BPQDs shift negatively by 1. 5 and 1. 1 eV, respectively and the binding energy of the P2p signal at 132. 9 eV along with the normal BP peaks suggests the coordination of metal with phosphorus without influencing the BP structure. The effect of GdL 3 coordination on the stability of BPQDs was examined by monitoring the absorption intensity where in contrast to BPQDs, the GdL 3 @BPQDs complex showed a minimal decrease in the absorption intensity after exposing to air for 8 d. The passivation is further confirmed by the P2p XPS spectra of GdL 3 @BPQDs, which is nearly unchanged under ambient conditions over 8 d. The surface coordination strategy was also successfully extended to BPNSs and BP microflakes and it is found that the present strategy can effectively passivate different types of BP from oxidation and degradation. The lanthanide functionalization of BP not only prevents the oxidation in water and humid air but also enables them to be fluorescent and possess high R 1 relativities in magnetic resonance imaging (MRI). In a recent example, modified cisplatin—Pt—NO 3 [Pt(NH 3 ) 2 (NO 3 ) 2 ] was used for the surface modification of BPNSs to generate Pt@BPNSs, which was then interacted with DNA both in vitro and in vivo. 193 Compared with the unmodified cisplatin, the Pt@BPNSs showed a good cellular uptake rate and dramatically improved the drug sensitivity of cisplatin‐resistant cancer cell lines, holding great potential for applications in photothermal/chemotherapy. Figure 16 a) Schematic diagram for the synthesis of lanthanide functionalized BP nanosheets/QDs. b) The chemical formula of LnL 3 and the atomistic model of the lanthanide ligand coordinated BP surface. Reproduced with permission. 192 Copyright 2018, John Wiley and Sons. 5 Noncovalent Functionalization of BPNSs Covalent functionalization may largely change the intrinsic properties of BPNSs, such as the P—P bonds breaking. In order to preserve the pristine properties of BPNSs to a great extent, noncovalent functionalization of BPNSs has been developed by taking advantage of electrostatic interactions, van der Waals interactions, and cation–π interactions. 5. 1 Electrostatic Interactions Liquid phase exfoliated BPNSs can possess a negative zeta potential, 195 which renders it possible to noncovalently functionalize BPNSs with positively charged species (e. g. , small molecules, polymers) through electrostatic interactions. 195, 196, 201, 202, 203, 204 For instance, Sabherwal and co‐workers employed nanostructured BP as a platform in an electrochemical sensor for the detection of Myoglobin (Mb), a cardiac disease biomarker. Stable few‐layer BPNSs were prepared by surfactant‐assisted liquid‐phase exfoliation of bulk BP in aqueous medium. The BPNSs were noncovalently functionalized with cationic polymer poly‐ l ‐lysine (PLL) to give PLL‐BPNSs. Negatively charged DNA aptamers for Mb were then immobilized onto the PLL‐BPNSs via Coulomb interactions between PLL and DNA. The sensor has a very low limit of detection of ≈ 0. 524 pg mL −1 with a sensitivity of 36 µA pg −1 mL cm −2 for Mb spiked in serum samples. The sensor exhibited high specificity and sensitivity due to the high affinity screened aptamers and the enhanced electrochemical properties of the nanostructures, paving the way to achieve better cardiac biomarker detection for point‐of‐care diagnosis. Jain et al. have utilized cetrimonium bromide (CTAB), an ionic surfactant with a long hydrophobic chain and nonbulky charged headgroups to prepare few‐layer BPNSs by liquid phase exfoliation in deoxygenated water. 195 The ionic group present in the surfactants can interact with the lone pair electrons of phosphorous and the surfactant orient laterally at low concentration and vertically at high concentration over the BPNSs surface. This assembly resulted in a stable dispersion of BPNSs in water at highly concentrated surfactant solutions. The BPNSs exfoliated in the CTAB (P/CTAB) showed a smaller thickness of 3–10 nm in contrast to the BP nanostructures exfoliated in surfactant free deoxygenated water (P) and those in the presence of a structurally variant surfactant tetrabutylammonium hydroxide (P/TBAOH), indicating the favorable structure of CTAB for the efficient exfoliation of BPNSs in water. The higher redshifts (≈3. 0−3. 5 cm −1 ) of the A 2 g peak in the Raman spectra than that of P/TBAOH (0. 5 cm −1 ) also confirmed the better intercalation and exfoliation capabilities of CTAB. In the XRD measurement, P/CTAB with vertically oriented CTAB molecules showed an increase in the interlayer distance between BPNSs compared with the BP nanostructures exfoliated in surfactant free deoxygenated water (P). Furthermore, DFT calculations suggested a strong and stable vertically oriented adsorption of CTAB on the BPNSs surface than the surfactant TBAOH possessing bulky alkane groups. Partial density of states of P/CTAB showed a decreased bandgap of 0. 67 eV while for the pristine BPNSs the value is 0. 82. Bader charge analysis revealed that noncovalent functionalization of BPNSs by CTAB led to a p‐type doping effect (BPNSs can donate 0. 46 electrons to CTAB). The interaction between the surfactants and the BPNSs was investigated by NMR measurements. The retarded diffusion rate probed by 2D diffusion ordered spectroscopy, suggesting the presence of noncovalently bonded CTAB over the BPNSs surface. 2D nuclear overhauser effect spectroscopy (NOESY) further revealed the interdigitated arrangement of the CTAB surfactant. Importantly, the P/CTAB suppressed ambient degradation rate of BPNSs by 70−80% as indicated by XPS measurement. In another case, Zhang and co‐workers have synthesized highly stable functionalized BPNSs by exfoliating bulk BP in the presence of dilute solutions of PILs. 123 The highly charged molecular chains of the PILs were wrapped onto the BPNSs through electrostatic interactions and acted as a shield to prevent BPNSs from air/moisture exposure. In contrast to exfoliation approach using pure ILs, 122, 205 BPNSs can be produced in high yield and cost efficiency using diluted PILs solutions. Functionalization was carried out by ultrasonication of bulk BP in the PIL solutions namely P([VPIm]Br) in water, P([VPIm]PF 6 ) or P([VPIm]TFSI) in DMF ( Figure 17 a TFSI refers to bis(trifluoromethane)sulfonimide). The extent of exfoliation was compared by exfoliating bulk BP in solvents alone (water or DMF), in pure IL monomers [EmIm]BF 4 and [BmIm]BF 4, and in aqueous solutions of [EmIm]BF 4 and [BmIm]BF 4. Dilution of pure monomer ILs with water resulted in significantly lower exfoliation yield compared with the one using pure monomer ILs, attributed to the reduced interactions between BPNSs and ILs. Compared with the above control samples, the exfoliation in the diluted solutions of PILs was significantly improved: using P([VPIm]TFSI can obtain the highest exfoliation concentration up to 0. 19 mg mL −1 (Figure 17 b, c). The ambient stability of the BPNSs was investigated by monitoring time‐dependent optical absorbance changes. After 7 d of ambient exposure, the decrease in absorbance for the unmodified BPNSs dispersed in water and DMF was 83. 42% and 42. 86%, respectively, whereas that of BPNSs/P([VPIm]Br) dispersed in water was 21. 68%, and for BP/P([VPIm] PF 6 ) and BP/P([VPIm]TFSI) dispersed in DMF the value was only 12. 63% and 23. 07%, respectively. In addition, after 21 d, the PILs modified BPNSs showed a A 1 g /A 2 g ratio of 0. 6 indicating a low oxidation degree of BP planes. 69 XPS analysis further confirms the successful noncovalent functionalization and passivation of BPNSs. After 8 d of exposure in air, the intensity of P—O bond in the PILs functionalized BPNSs was found to be much weaker than that of the BPNSs, suggesting the good function of PILs in protecting BPNSs against ambient degradation. TEM images of the PILs modified BPNSs showed smooth surface and sharp edges over a period of 100 d suggesting highly crystalline structure with reduced oxidation. The retention of the high crystalline quality of the PILs modified BPNSs was further confirmed by SAED patterns. In contrast, a series of bubbles were observed on the surface of BPNSs/water and BPNSs/DMF samples after exposed to ambient conditions for a week, indicating the large extent of oxidation of BPNSs. Flexible photodetector devices based on the P([VPIm]TFSI) functionalized BPNSs showed good performances with a responsibility of 4. 6 µA W −1 under a bias potential of 3 V and excellent device stability compared to the unmodified BPNSs based devices, demonstrating the great potential in the practical application of flexible optoelectronics. However, future efforts are required to further improve the on/off ratios (e. g. , increasing the illumination intensity and fabricating photodetector devices with a narrower channel length). Figure 17 a) Chemical structures of the PILs and ILs. b) Photographs of BPNSs exfoliated with different solvents and c) comparison of their absorption intensity (wavelength at 600 nm). Reproduced with permission. 123 Copyright 2018, John Wiley and Sons. 5. 2 van der Waals Interactions Besides electrostatic interactions, recently there are also some reports on noncovalent modification of BPNSs by taking advantage of van der Waals interactions between BPNSs and foreign molecules. 206 For example, redox active anthraquinone (AQ) has been used to noncovalently functionalize BPNSs, reported by Pumera and co‐workers. 199 Few‐layer BPNSs with a thickness of 19. 1 ± 9. 7 nm were prepared by the shear exfoliation of bulk BP in aqueous surfactant sodium cholate, which was then modified by AQ to obtain the BPNSs–AQ complex. In the high‐resolution XPS spectra of the P2p, the signal corresponding to P oxides centered at about 134. 1 eV is more intense for BPNSs than BPNSs–AQ suggesting that the BPNSs–AQ complex is more protected against oxidation. A relatively high abundance of C=O groups (≈19%) for the BPNSs–AQ in the XPS spectra of the C1s indicates the successful modification of BPNSs by AQ. After being exposed to ambient environment for 1 month, BPNSs exhibited much more extensive degradation than the BPNSs–AQ complex, indicating that noncovalent modification of BPNSs using the hydrophobic AQ can be a possible solution to reduce the degradation of BPNSs. Electrochemical measurements of the BPNSs–AQ complex showed that the AQ was stably immobilized onto the BPNSs and the redox peaks were stable over 100 cycles. The surface coverage of the AQ on the BPNSs was calculated to be ≈1. 25 nmol AQ per mg of BPNSs and the electron transfer rate constant between AQ and BPNSs was 33 s −1. Interestingly, the BPNSs–AQ complex possessed a much higher gravimetric capacitance than the starting bulk BP material, holding potential applications in energy storage devices. In another case, as shown in Figure 18, BPNSs have been noncovalently functionalized with 7, 7, 8, 8‐tetracyano‐ p ‐quinodimethane (TCNQ) or perylene diimide (PDI). 200 The presence of TCNQ or PDI on the BPNSs was confirmed by several spectroscopic and microscopic characterization techniques, including attenuated total reflectance spectroscopy, Raman, AFM, scanning transmission electron microscope (STEM)–electron energy loss spectroscopy, UV–vis absorption spectroscopy, fluorescence emission spectroscopy, etc. For instance, upon excitation at 455 nm, the fluorescence emission of the PDI showed a dramatic quenching (≈66%) in the presence of BPNSs, suggesting there are interactions between PDI and BPNSs at the excited state. The mutual interactions between BPNSs and TCNQ/PDI was further studied using DFT calculations. Both the experimental and computational studies revealed that there were noncovalent interactions between the molecules and BPNSs, thus facilitating the exfoliation of bulk BP into few‐layer BPNSs. Electrons can transfer from the BPNSs to the TCNQ. The anchor of PDI on the BPNSs largely improves the stability of BPNSs at ambient conditions. Figure 18 a) Functionalization of BPNSs using TCNQ and PDI. b) Schematic representation of TCNQ (top) and PDI molecules (bottom) adsorbed on BPNSs. Reproduced with permission. 200 Copyright 2016, John Wiley and Sons. BPNSs can be effectively passivated to maintain their unique properties by formation of van der Waals heterostructures with other 2D materials such as graphene, TMDs, or h ‐BN. 93, 207, 208, 209, 210 For example, Doganov et al. prepared few‐layer BPNSs based heterostructures using graphene or h ‐BN. 207 BPNSs with a thickness of 3–9 nm were transferred onto SiO 2 /Si wafers using micromechanical exfoliation in an inert atmosphere. Then the heterostructures were fabricated by transferring graphene or h ‐BN using a dry transfer method. The ambient stability was studied by AFM and Raman analysis on the covered and exposed areas of BPNSs based heterostructures. In contrast to the graphene or h ‐BN protected surface, the exposed region of BPNSs showed a significant increase in the roughness over time. The optical images showed that, after 48 h of ambient exposure, the unprotected region of the BPNSs were completely degraded and no corresponding Raman bands of BPNSs were observed. On the other hand, the protected surface by graphene or h ‐BN showed the three characteristics Raman peaks of BPNSs, along with Raman signals from the graphene or h ‐BN. Compared with unprotected ones, the passivated BPNSs heterostructure based FETs showed 10 to 100‐fold improvement in electron mobility at room temperature. Remarkably, van der Waals interaction results in an in‐plane assembly of functional materials on BPNSs. Moreover, functionalization of BPNSs with other types of 2D materials including TMDs, layered metal hydroxides, metal organic frameworks (MOFs), 2D polymers, and functional organic molecules, is essential to further explore this hot area. 5. 3 Cation–π Interactions From the structure of BPNSs, the lone pair electrons on the phosphorus atoms are evenly distributed on the two sides of each BP layer and they can further interplay with each other to form conjugated π bonds, which can interact with metal ions through cation–π interaction. 194 Yu and co‐workers employed metal ions such as Ag + to interact with the exposed lone pair of electrons of BP forming Ag + ‐modified BPNSs to enhance the stability against oxidation and degradation. 194 The BPNSs were produced by a mechanical exfoliation method using scotch tape from bulk BP and were transferred to a Si wafer with a 300 nm thick SiO 2 layer through a PDMS thin film. The samples were immersed in the NMP solution containing silver nitrate (1 × 10 −6 M ) for 2 h, washed with NMP, and dried with argon gas to produce the Ag + ‐modified BPNSs on the wafer. The metal ions can interact with the conjugated π bonds via cation–π interaction ( Figure 19 a, b). The combined energy between Ag + ions and BPNSs was calculated to be −41. 8 cal using DFT calculation, suggesting that free Ag + can stably adsorb on the BPNSs surface to obtain the Ag + ‐modified BPNSs. Then the Ag + ‐modified BPNSs and bare BPNSs were exposed to air for 3 d to investigate their stability against oxidation. In the XPS spectra of P2p, BPNSs showed a strong peak at 134. 0 eV corresponding to the PO x species, whereas this peak is not obvious for the Ag + ‐modified BPNSs, indicating a better stability for the latter at ambient conditions. In addition, a new peak at 133. 0 eV for the Ag + ‐modified BPNSs was assigned to the interaction between Ag + and BPNSs. In contrast to the bare BPNSs, the presence of a Ag3d 5/2 peak at 367. 8 eV for the Ag + ‐modified BPNSs confirms the successful modification of Ag + on the BPNSs surface. The ambient stability was further studied by AFM measurements (Figure 19 c–i). The BPNSs and the Ag + ‐modified BPNSs on the Si/SiO 2 wafer were kept in air at a relative humidity of 95% and room temperature for 5 d. After 2 d, bubbles appeared in the bare BPNSs surface whereas the surface morphology of the Ag + ‐modified BPNSs was preserved: after 5 d, no obvious bubbles, corrosions, or degradations were observed. The Ag + ‐modified BPNSs FETs showed a hole mobility of 1666 cm 2 V −1 s −1 and an on/off ratio of 2. 6 × 10 6, which is more than two times and 44 times higher than that of the bare BPNSs FETs, respectively. Based on DFT calculations, Ag + modification can enhance the initial current of the hole‐transport, suppress the electron transport, lower the off‐state current, and increase the on/off ratio. The current modification strategy can be extended to other metal ions including Fe 3+, Mg 2+, and Hg 2+, demonstrating an efficient approach to improve both the stability and transistor performance of BPNSs. Figure 19 a) Schematic representation of adsorption of Ag + on BPNSs. b) Three different views of the Ag + ‐modified BPNSs. AFM images of c–e) the bare BPNSs and f–h) the Ag + ‐modified BPNSs exposed to air for 1–5 d. i) Changes in the surface roughness before and after functionalization with exposure time. Reproduced with permission. 194 Copyright 2017, John Wiley and Sons. Table 1 summarizes and compares the preparation methods of BPNSs, various types of reactions/interactions used, reagents employed for functionalization and the stability of functionalized BPNSs. Obviously, they have shown different passivation effect in enhancing the stability of BPNSs upon exposure to ambient environment. The selection of chemical functionalization will reply very much on the applications: for example, if there is no requirement for the electrical performance of the BPNSs, covalent approaches involving P—P bond breaking can be considered. Table 1 Summary of preparation, chemical functionalization, and stability of BPNSs Refs. Synthetic method of BPNSs Functionalization Type of reaction/interaction Reagents for functionalization Stability of functionalized BPNSs 97 Mechanical exfoliation on Si/SiO 2 substrates Covalent Radical addition using aryldiazonium chemistry 4‐Nitrobenzenediazonium and 4‐methoxybenzenediazonium tetrafluoroborate salt Stable up to 25 d of ambient exposure in solution 180 Liquid exfoliation in NMP Covalent 1) Radical addition using aryldiazonium chemistry 2) Pd catalyzed polymerization 1) 4‐bromobenzenediazonium tetrafluoroborate 2) 1, 4‐diethynyl benzene and 9, 9‐bis(4‐diphenylaminophenyl)‐2, 7‐dibromofluorene Stable device performance even after 3 months ambient exposure 181 Liquid exfoliation in NMP Covalent Radical addition using aryldiazonium chemistry Diazonium tetrafluoroborate salt of fluorescent dye Nile Blue 690 Stable after 3 d of ambient exposure in water solution 183 Liquid exfoliation in NMP Covalent Radical addition using aryldiodonium salt Bis(4‐fluorophenyl)iodoniumtriflate, (perfluoro‐ n ‐propyl)‐phenyliodonium triflate and bis(4‐methylphenyl)iodonium hexfluorophosphate Superior stability compared to aryldiazonium method after 1‐week ambient exposure 105 Mechanochemical exfoliation using ball milling Covalent Free radical/ions reaction via mechanochemical P—P bond cleavage in BP LiOH NA 107 Mechanochemical exfoliation using ball milling Covalent Free radical/ions reaction via mechanochemical P—P bond cleavage in BP C 60 The degradation rate of functionalized BPNSs was inhibited by a factor of 4. 6 106 Mechanochemical exfoliation using ball milling Covalent Free radical/ions reaction via mechanochemical P—P bond cleavage in BP Urea Stable for 2 months in absolute ethanol; voltammogram in cyclic voltammetry remains unchanged for 1000 consecutive cycles 158 Liquid exfoliation in Isopropanol Covalent Free radical addition AIBN Oxidation degree was reduced by 17. 82% 155 Liquid exfoliation in DMF Covalent Nitrene addition generated from organic azides 4‐Azidobenzoic acid Stable after standing for 21 d in ambient conditions; stability is 4. 7 times higher than that of diazonium‐functionalized BPNSs 157 Shear force milling in DMF Covalent Nucleophilic substitution reactions Aliphatic alcohols or organometallic reagents NA 189 Liquid exfoliation of BPICs in THF Covalent Nucleophilic substitution reactions Alkyl iodides NA 190 Liquid exfoliation in DMF Covalent Coordinating lone pair electrons of phosphorus atoms in the empty orbital of metal atom Titanium sulfonate ligand Stable after 1 week of ambient exposure in water 192 Liquid exfoliation in DMF Covalent Coordinating lone pair electrons of phosphorus atoms in the empty orbital of metal atom Lanthanide sulfonate complexes Stable after 8 d of ambient exposure in water 194 Mechanical exfoliation with scotch tape Noncovalent Cation–π interaction Silver nitrate Stable up to 5 d in ambient conditions and at least for 3 weeks in drying oven 195 Surfactant assisted liquid exfoliation in water Noncovalent Electrostatic interaction Cetrimonium bromide (CTAB) Degradation rate decreased significantly by 70−80% 196 Surfactant assisted liquid exfoliation in water Noncovalent Electrostatic interaction PLL Excellent storage stability of the developed aptasensor after 21 d 197 Liquid exfoliation in saturated NaOH solution in NMP Noncovalent Electrostatic interaction Doxorubicin (DOX) High photostability upon 808 nm laser irradiation for 5 min 123 PIL assisted liquid exfoliation in water or DMF Noncovalent Electrostatic interaction P([VPIm]Br), P([VPIm]PF6), or P([VPIm]TFSI) Stable up to 100 d 198 Liquid exfoliation in NMP Noncovalent Electrostatic interaction Tripeptide Fmoc‐Lys‐Lys‐Phe Stable for 2 d of ambient exposure 199 Shear exfoliation in aqueous surfactant sodium cholate Noncovalent van der Waals interactions Anthraquinone Stable up to 30 d 200 Chemical thinning in THF Noncovalent van der Waals interactions Perylene diimide Stable after 2 d of ambient exposure and stable up to 6 months in glove box storage 207 Mechanical exfoliation onto SiO 2 /Si wafers Noncovalent van der Waals interactions Graphene or h ‐BN Stable up to 48 h at ambient conditions John Wiley & Sons, Ltd. 6 Applications of Functionalized BPNSs Due to the improved stability at ambient conditions, functionalized BPNSs have already shown potential applications in energy conversion and storage, electronic devices, biological field, nanocomposites, etc. Based on their high specific surface area and good electrical conductivity, BPNSs have been already employed as electrode materials in energy storage devices. 60, 100, 115, 118, 131, 132, 213, 214, 215, 216, 217, 218 The high theoretical capacity (2596 mAh g −1 ) of BPNSs makes it as a promising electrode material in lithium‐ion and sodium‐ion batteries (LIBs and SIBs). 61 BPNSs based anode materials for LIBs exhibit improved first charge and discharge capacities compared to existing graphitic anode materials. 219, 220 On the other hand, upon periodic lithiation and delithiation process, BPNSs show capacity decay due to the severe breakage of anode materials because of the large volume expansion. This problem can be tackled by chemical functionalization of BPNSs with carbon materials, the formation of P—C bonds makes BPNSs more stable during lithium insertion/extraction. 211, 221 For instance, Cui and co‐workers synthesized a BPNSs–graphite composite using a high energy mechanical milling process ( Figure 20 a). The P—C bonds formed between BPNSs and graphite maintained excellent electrical contact and stability during lithium insertion/extraction process. 211 The composite, as an anode material, exhibited an initial discharge capacity of 2786 mAh g −1 at 0. 2 C and cycle life of 100 cycles with a capacity retention of 80%. Compared to lithium ion, due to its larger size, sodium ion can result in a higher volume expansion of BPNSs during sodiation process. To reduce this effect, BPNSs have been functionalized with different carbon materials including graphene, which can function as a cushion to maintain the structural stability of BPNSs and also as a conducting nanofiller to improve the electrical conductivity of BPNSs. 222, 223 It was also demonstrated that covalently functionalized BPNSs with rGO could also improve cycle performance of BPNSs in SIB anode. 224 In another report, a sulfur‐doped BPNSs‐TiO 2 nanocomposite was prepared using a ball milling approach. 225 The nanocomposite was then employed as anode material for SIBs and it showed good electrochemical performance: the discharge capacity can be up to 490 mA h g −1 over 100 cycles at 50 mA g −1, and it was maintained to be 290 mA h g −1 over 600 cycles at 500 mA g −1. Flexible all‐solid‐state supercapacitors using BPNSs‐carbon nanotube nanocomposite as electrode materials have been fabricated, showing a high power density of 821. 62 W cm −3 and a stable electrochemical performance over 10 000 bending cycles. 226 Figure 20 a) Schematic representation of lithiation and delithiation process in BP‐graphite composite. Reproduced with permission. 211 Copyright 2014, American Chemical Society. b) Hole mobility and on/off ratio of BPNSs based FET device as a function of Ag + modification time. Reproduced with permission. 194 Copyright 2017, John Wiley and Sons. c) Schematic illustration of flame‐retardant mechanism for the epoxy resin (EP)/polyphosphazene‐functionalized BPNSs (BP‐PZN) nanocomposites during combustion. Reproduced with permission. 212 Copyright 2019, American Chemical Society. Since BPNSs showing thickness dependent bandgap, high carrier mobility and anisotropic transport, they have been employed as a channel material in FETs. 227, 228, 229, 230, 231, 232 However, the poor ambient stability of BPNSs reduces their practical utilization in FETs. Effective passivation of BPNSs through chemical modification needs to be developed to achieve high performance functionalized BPNSs‐based FETs. 61 As discussed in the previous sections, compared with unmodified BPNSs, both aryl diazonium functionalized BPNSs (covalent interaction) and Ag + modified BPNSs (noncovalent interaction) show an improved ambient stability and an enhanced device performance in FETs (Figure 20 b), 97, 194 which is crucial for the realization of advanced electronic and optoelectronic devices. For efficient usage of BPNSs with polymers for high‐performance nanocomposites, two important challenges have to be first addressed during processing: homogeneous dispersion of BPNSs in polymer matrix and strong interfacial interactions between BPNSs and polymer matrix. 233, 234, 235, 236 One of the promising solutions to tackle the challenges is to chemically functionalize BPNSs, which can not only improve the dispersibility of BPNSs but can also reinforce the interfacial interactions, leading to a better chemical compatibility with different media and interfaces. It is worth pointing out that the functional units and polymers can potentially protect BPNSs away from degradation at ambient conditions. For example, recently Hu and co‐workers have prepared cobaltous phytate‐functionalized BPNSs and mixed them with polyurethane acrylate (PUA). 237 The nanocomposite shows an increasement in the mechanical properties of PUA: an enhancement of tensile strength by 59. 8% and tensile fracture strain by 88. 1%. For the flame retardancy of PUA, there is an obvious decrease in the heat release rate (by 44. 5%) and total heat release (by 34. 5%), and generation of less pyrolysis products including highly toxic carbon monoxide. Importantly, Raman measurement and XRD analysis indicate that the nanocomposite was stable at ambient conditions for 4 months due to the isolation and protection effect on BPNSs. In another report, they have also fabricated a crosslinked polyphosphazene (PZN)‐functionalized BPNSs through a polycondensation reaction between 4, 4'‐diaminodiphenyl ether and hexachlorocyclotriphosphazene on the surface of BPNSs. 212 Then the polyphosphazene‐functionalized BPNSs were incorporated into epoxy resin to investigate the flame‐retardant property and smoke suppression performance (Figure 20 c). The epoxy resin nanocomposite with a loading of 2 wt% polyphosphazene‐functionalized BPNSs showed an obvious improvement in the flame‐retardant property, showing a decrease of 59. 4% and 63. 6% for the peak heat release rate and the total heat release, respectively, with a reduction in the diffusion of the pyrolysis products. Importantly, as proved by XRD and Raman measurement, the nanocomposite showed a good stability at ambient conditions over 4 months, due to the protection from the polymer matrix. As discussed in the introduction part, BP is a semiconductor with a layer‐dependent direct bandgap, facilitating a broad absorption in visible and NIR region, which renders BPNSs a good candidate as metal free photocatalyst. 238 Computational calculations show H 2 evolution process is possible as the conduction band (CB) of BPNSs is more negative than redox potential of H + /H 2 (0 V vs normal hydrogen electrode). 239 For example, as described in the above section, BP‐BM with negatively shifted CB and positively shifted valance band (VB) exhibited excellent H 2 production efficiency (0. 47%), which is higher than that of graphitic carbon nitride g ‐C 3 N 4 (0. 1%) ( Figure 21 a). In contrast to bulk BP, the VB level of BP‐BM has been positively shifted which avoids undesirable recombination of the photogenerated electron–hole pairs (Figure 21 b). In addition to this, so far, many other chemically modified BPNSs based photocatalysts have been successfully employed for H 2 evolution reaction, revealing that BPNSs will be a good photocatalyst for energy production to solve the future energy crisis. 240, 241, 242, 243, 244, 245, 246, 247, 248 Besides H 2 evolution reaction, Co 3 O 4 modified BPNSs have been used as an electrocatalyst for oxygen evolution reaction. 249 The Co 3 O 4 modified BPNSs showed a much better electrocatalytic performance than Co 3 O 4 and BPNSs, attributing to the electron transfer process between the two species and improved stability of BPNSs. This study provides some useful insights to design highly efficient water oxidation electrocatalysts. Figure 21 a) Photocatalytic activity of BP‐BM in H 2 evolution reaction from water and b) the mechanism. Reproduced with permission. 105 Copyright 2017, John Wiley and Sons. c) In vitro cell viability experiments of BP@Hydrogel. d) Fluorescence images of mice after the in vivo photothermal assay. Reproduced with permission. 250 Copyright 2017, National Academy of Sciences. Due to their large surface‐to‐volume ratio, high thermal and electrical conductivity, and good optical properties, BPNSs can be potentially attractive toward biological applications. 251 However, BPNSs suffer from degradation and gradually lose their properties when it is exposed to physiological environment, which is a great obstacle for their bioapplications. To passivate BPNSs and maintain its functions, effective functionalization strategies must be performed. The outstanding biocompatibility and biodegradability of functionalized BPNSs in physiological environment makes them attractive and promising for wide bioapplications in photodynamic therapy 197 and photothermal therapy, 181, 201, 250, 252, 253 tissue engineering, 254, 255 biosensors, and bioimaging. 196, 202, 256, 257 Qiu et al. have developed a new concept of light activation of BPNSs based hydrogel for cancer therapy. 250 In their approach, BPNSs was prepared by liquid phase exfoliation in isopropanol. Then positively charged polyethylene glycol–amine was used to functionalize BPNSs through electrostatic interaction to form PEGylated BPNSs, in order to enhance their biocompatibility and physiological stability. Finally, BPNSs based hydrogel was prepared by mixing PEGylated BPNSs with a low‐melting‐point agarose. By tuning the light intensity and exposure duration, the release rates of doxorubicin (DOX) can be accurately controlled. In vitro and in vivo tests indicated that the BPNSs based hydrogel possessed an extremely good killing ability for cancer cell combined with tumor ablation effect (Figure 21 c, d). The approach holds the potential to be transferred to practical research in curing cancer. In another report, Yang and co‐workers have reported tannic acid‐Mn 2+ chelate networks on BPNSs, 258 and the formed complex shows very good MRI contrast enhancement capability, excellent photoacoustic imaging performance, and high photothermal conversion efficiency, demonstrating great potential in imaging‐guided photothermal therapy. Importantly, balancing the stability and biodegradation of functionalized BPNSs is very important to achieve desirable biological functions with minimal residual during in vivo test. Although functionalized BPNSs have shown interesting and promising results in the abovementioned areas, there is still a long way to go toward their large‐scale practical applications. For example, the toxic effect of functionalized BPNSs in biological systems should be always taken into considerations. For energy storage applications, improvement in electrical conductivity and cycle stability needs to be further investigated. On the other hand, functionalized BPNSs have been less employed in sensing, which can be one of the future directions to be explored. 7 Conclusions and Outlook The unique properties (tunable bandgap, high charge carrier mobility, and in‐plane anisotropy) of BPNSs make them an ideal bridge between graphene and TMDs. However, poor ambient stability of BPNSs hinders its possible wide applications. Chemical functionalization of BPNSs through covalent or noncovalent approach using metal oxide, functional organic molecules, other 2D materials or polymers have been proved to be a useful strategy to protect BPNSs from degradation at ambient conditions: it can not only passivate BPNSs but also introduce new properties to BPNSs. The reaction can be carried out in solid‐state or through liquid phase process. Depending on different methods used, they have shown different effectiveness on the passivation of BPNSs. For covalent functionalization, free radical reaction using diazonium compounds or aryl iodonium salts or ball milling, nucleophilic substitution to form P—C and/or P—O—C covalent bonds, has been successfully developed. In contrast to the above covalent approaches, nitrene addition results in a better stabilization of BPNSs owing to the utilization of both unpaired electrons present in the phosphorous atom. Coordination of the lone‐pair electrons on phosphorus atoms with metal complex is another effective way to stabilize BPNSs. On the other hand, noncovalent approaches such as electrostatic interactions, cation–π interactions, have been employed to prepare BPNSs based complex. Through van der Waals interactions, BPNSs can be combined with other type of 2D materials or functional molecules to fabricate novel heterostructures possessing extraordinary properties and performance. 208, 209, 259, 260, 261, 262, 263, 264, 265, 266 Building BPNSs‐based heterostructures through solution processing is a promising route to prepare them in a large scale ( Figure 22 ), however, achieving a highly uniform and ordered nanostructures is still challenging. 267, 268, 269, 270, 271, 272 For example, a 2D BPNSs/platinum heterostructure has been developed and used as a highly efficient photocatalyst for solar‐driven chemical reactions. 268 Compared with unmodified BPNSs, functionalized BPNSs exhibit improved performance in different applications such as LIBs/SIBs, photocatalysis, (opto)electronic devices, photothermal/photodynamic therapy, and sensing. However, there is still a long way to go before functionalized BPNSs can find their practical applications in various fields at ambient conditions. Figure 22 Schematic representation of different potential functionalized BPNSs including edge functionalized BPNSs, BPNSs based heterostructures and 3D networks. Although rapid progress has been made on the chemical functionalization of BPNSs, the research area is still in its sprouting stage. There are several challenges to be resolved. The first prerequisite is to produce high‐quality BPNSs (with bigger sizes) in a large scale, which is yet to be resolved, largely relying on the further development of preparation methods. For example, other bottom‐up approaches such as hydrothermal/solvothermal method from molecule precursors needs to be developed. 151 One of the biggest challenges in BPNSs research is still to find effective ways to minimize the degradation of BPNSs at ambient conditions both in the solid form and in solvent media. It should also be mentioned that the mechanism of degradation of BPNSs is still in debate and requires further efforts from both computational and experimental research to deeply understand the exact degradation mechanism and thus design novel strategies for effective passivation. Development of stable BPNSs is particularly demanded for their practical applications in various fields. Nevertheless, achieving a long‐term stability of functionalized BPNSs over months or years is currently still challenging. Efficient passivation could be achieved by developing new chemical functionalization strategies, for example, reacting/interacting effectively with the paired electrons on BPNSs using different kinds of functional units. 273 This will largely broaden/enrich the chemistry of BPNSs. Special attention should be paid to that, for certain applications such as (opto)electronics devices, the intrinsic properties of BPNSs should be reserved at the most extent after chemical functionalization. For covalent functionalization, improving the selectivity (e. g. , edges or defects or certain sites/domains selectively functionalized, Figure 22 ), 107 will hold the potential to precisely control the properties of BPNSs. On the other hand, building functionalized BPNSs based 3D hierarchical nanostructures can be a useful way to prevent the aggregation of BPNSs and thus the properties of BPNSs can be fully utilized (Figure 22 ). 274 Further issues regarding to how the charge transfer happens and how the charges are spatially distributed between BPNSs and functional units, still need to be fully understood at the fundamental level. 275 Another major challenge is how to accurately determine the exact chemical structure of functionalized BPNSs through covalent modification, which is currently lacking direct evidence. This will rely very much on the development of advanced microscopic characterization techniques such as HRTEM, AFM, or scanning tunneling microscope, in order to be able to monitor the reactions in real time (in situ) and to “see” the true chemical structures of the functional units attached onto BPNSs. Additional techniques such as solid NMR or 2D NMR or MS can be also employed to determine the chemical structures or to probe the interactions between functional units and BPNSs. 195, 276 Further support may come from theoretical calculations such as DFT, which will help to predict the chemistry, process, and mechanism. 157, 189 It is reasonable to expect that, with the continuous effort in the research of chemical functionalization of BPNSs, this area will continue to grow in the coming years and more breakthroughs will emerge to realize their practical applications in various fields. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201902398
| 2,020
|
Advanced Science
|
Human Platelet Lysates‐Based Hydrogels: A Novel Personalized 3D Platform for Spheroid Invasion Assessment
|
Abstract Fundamental physiologic and pathologic phenomena such as wound healing and cancer metastasis are typically associated with the migration of cells through adjacent extracellular matrix. In recent years, advances in biomimetic materials have supported the progress in 3D cell culture and provided biomedical tools for the development of models to study spheroid invasiveness. Despite this, the exceptional biochemical and biomechanical properties of human‐derived materials are poorly explored. Human methacryloyl platelet lysates (PLMA)‐based hydrogels are herein proposed as reliable 3D platforms to sustain in vivo‐like cell invasion mechanisms. A systematic analysis of spheroid viability, size, and invasiveness is performed in three biomimetic materials: PLMA hydrogels at three different concentrations, poly(ethylene glycol) diacrylate, and Matrigel. Results demonstrate that PLMA hydrogels perfectly support the recapitulation of the tumor invasion behavior of cancer cell lines (MG‐63, SaOS‐2, and A549) and human bone‐marrow mesenchymal stem cell spheroids. The distinct invasiveness ability of each cell type is reflected in the PLMA hydrogels and, furthermore, different mechanical properties produce an altered invasive behavior. The herein presented human PLMA‐based hydrogels could represent an opportunity to develop accurate cell invasiveness models and open up new possibilities for humanized and personalized high‐throughput screening and validation of anticancer drugs.
|
1 Introduction The complexity and dynamism of the native tissue microenvironment are actively involved into distinct physiological and pathological events, where cell–cell and cell–extracellular matrix (ECM) communication plays a crucial role. 1, 2 In cancer, tumor cell signaling triggers the deregulation of the spatiotemporal ECM remodeling that maintain the tissues integrity, favoring tumor cell migration and invasion into neighborhood regions and supporting metastasis development. 3, 4 Effectively, a large number of cancer‐related deaths are derived from these cellular processes and therefore understanding tumor invasion mechanisms is a focus in cancer research. Conventional 2D in vitro cultures have been the most widely used platforms for anticancer drug screening and validation. Despite their simplicity and quickness, they have recognized limitations in recapitulating in vivo behaviors which is reflected in misleading preclinical results. 5, 6 3D cell culture systems have been explored in cancer research field as a more complex and predictable in vitro model able to mimic the human in vivo cellular behavior in a more realistic way. 6, 7, 8, 9 The biochemical, physical, and mechanical cues offered by in vitro 3D models maintain cell morphology and polarity while preserving cell surface receptors expression (e. g. , integrins) involved in signaling pathways such as tumor metastatic cascade and therapy resistance. 10, 11, 12, 13 The potential to provide fundamental insights on biomedical research has stimulated the interest of applying 3D cultures in preclinical assays, thus encouraging technological advances to achieve the implementation of high‐throughput and high‐content screening techniques. 14, 15, 16, 17, 18 Multicellular spheroids are the most relevant 3D platform, largely contributing to the current knowledge of the tissues pathophysiology. Concerning tumor spheroids, the in vivo‐like cellular arrangement stimulates a set of biological conditions similar to those found in a native tumor mass. Oxygen, nutrients, and pH gradients are some of the spheroid‐reproduced chemical aspects that resemble the in vivo phenotypic heterogeneity of an avascular tumor. 19, 20 These heterogeneous tumor cell populations are characterized by cellular proliferation and metabolic activity gradients modulated by distinct protein and gene expression profiles related with the invasiveness ability of tumor cells. 19, 21 Contrary to 2D cultures, the native cell polarity is mimicked in these classic 3D models and preserved during the cell invasion process. 21, 22 Boyden chamber is the most widely used in vitro standard assay to assess tumor invasion. 23, 24, 25, 26 However, spheroids embedded into an ECM hydrogel have been proposed as a highly reproducible and automated system to develop accurate in vitro invasion models. 18, 27 In vivo cell invasiveness ability is dependent on cell type and its environment and, with regard to the 3D in vitro invasion models, the method of spheroid generation and the physicochemical cues of the surrounding matrix have a huge effect on cell invasive behavior. 28, 29 Numerous biomimetic materials, from natural origin to synthetic polymers, have been used to recreate the ECM in 3D in vitro models. 30, 31 Typically, natural‐based scaffolds are biochemically similar to the native tissue to be reproduced/replaced. 32 Synthetic ones have easily controllable properties; however, chemical functionalization is critical to improve their biochemical cues. Therefore, the search for new biomimetic materials that accurately mimic the natural ECM and, at the same time, provide reproducible cell invasive behavior is essential to an in‐depth comprehension of invasive mechanisms and future application as platforms for drug screening. Reconstituted basement membrane (rBM) of murine Engelbreth‐Holm‐Swarm sarcoma and type I collagen are the most used natural biomaterials so far. However, they exhibit noteworthy drawbacks regarding its reproducibility, difficult handling, and the animal origin that can induce an immune response and limit their clinical application. 30, 32 Our research group recently proposed a novel 3D in vitro platform made of photo‐crosslinkable human platelet‐rich plasma (PRP)‐derived hydrogels for the development of complex and reliable tissue models. 33 This humanized biomaterial follows the animal‐free tendency, being rich in allogenic fibrous and adhesive proteins, growth factors, and other bioactive factors that provide specific cues of human tissues microenvironment. 34 Besides their remarkable biochemical properties for tissue engineering applications, methacryloyl platelet lysates (PLMA)‐based hydrogels demonstrated to be a simple and cost‐effective platform with highly tunable mechanical properties, supporting human‐derived cell growth and invasion. 33 The combination of human‐specific biochemical and mechanical cues provides a physiologically relevant tissue‐like environment to develop tissue models mimicking cell‐intrinsic behavior, which can improve the predictivity of pharmacokinetic responses. 35, 36 PLMA‐based hydrogels have the potential to increase the predictive value of drug screening and validation models over the existing animal‐derived platforms, which can include biochemical signals that can drive nonspecific cell behavior. Furthermore, the easy availability of autologous PL and their combination with patient‐derived cells can open up the possibility of applying PLMA‐based hydrogels as a 3D platform for personalized drug screening. In this study, we introduced the aforementioned new human‐based biomimetic material, PLMA hydrogels, as 3D platforms for cell spheroid invasiveness studies. In order to evaluate the applicability of PLMA hydrogels for this purpose, rBM of murine Engelbreth‐Holm‐Swarm sarcoma (Matrigel) was used as a positive control since, besides to be considered the “gold standard” for 3D in vitro disease models development, it is also constituted by a complex mixture of proteins. Spheroids of three tumor cell lines (osteosarcoma, MG‐63 and SaOS‐2, and lung cancer, A549) as well as human bone‐marrow mesenchymal stem cells (hBM‐MSC) were embedded into PLMA hydrogels at three different concentrations, Matrigel and poly(ethylene glycol) diacrylate (PEGDA), the latter used as negative control. The easy control over the mechanical properties of PLMA hydrogels offer the opportunity of evaluating the matrix conditions that more closely reproduce the micromechanical environment depending on the tissue under study. In this sense, the invasiveness of each cell type was systematically analyzed with the aim of determining the invasion kinetics and uniformity, and evaluate the PLMA hydrogels mechanical conditions that potentiate the best approximation to in vivo cell invasiveness behavior. 2 Results and Discussion Advances in 3D in vitro systems have been contributed to the recreation of the cellular heterogeneity and microenvironment dynamism of several tissues, providing robust platforms with potential to be applied in preclinical drug screening studies. Although the cell–ECM interaction is considered as crucial in physiological and pathological events, most of the 3D in vitro models are still based on multicellular tissue spheroids (MCTS) cultured in a scaffold‐free setting. 14, 28 However, regarding some pathologies as cancer, the understanding of how a growing tumor interacts with the ECM, invades through it and responds to chemotherapeutic treatments into this dynamic environment can expedite therapy development. 24, 37 To date, only a few studies have combined scaffold‐free spheroids with ECM‐resembling scaffolds. Nevertheless, the importance of a surrounding matrix in the recapitulation of cell–cell and cell–ECM interaction during early metastatic processes has been demonstrated, evidencing spheroid embedding into a matrix as a promising strategy to assess MCTS growth, invasiveness ability, and antimetastatic drug sensitivity. 38, 39 In an attempt to provide a reliable platform for 3D spheroid invasion, overcoming the recognized limitations of the currently applied biomaterials, we herein explore for the first time the potential of a newer photoresponsive humanized material, the PLMA hydrogels, to support MCTS growth and invasion ( Figure 1 ). Spheroids of stem cells (hBM‐MSC) and cancer cell lines (MG‐63, SaOS‐2, and A549) were embedded into PLMA hydrogels. The human origin and tunable mechanical cues of these hydrogels are expected to produce a more faithfully recapitulation of human cellular behavior, supporting the individual invasiveness ability of each cell line. 33 Osteosarcoma and lung cancer cell lines were chosen in order to demonstrate that PLMA hydrogels are able to sustain the growth of malignant tumors of mesenchymal and epithelial origin, respectively. 40 Although tumor models are a research hotspot, the role of mesenchymal stem cells (MSC) on fundamental physiological and pathological events has also been a topic of increasing interest on biomedical field. The aggregation of MSC into spheroids has been suggested as an improvement of stem cell‐based therapies once the enhancement of anti‐inflammatory and angiogenic properties, stemness and survival of MSC have been reported in this 3D configuration. 41, 42, 43 In this sense, the ability of PLMA hydrogels to support hBM‐MSC cell culture in a spheroid setting was also evaluated. Figure 1 Schematics of 3D spheroids encapsulation to study cell spheroid invasiveness. Spheroids are formed into round‐bottom ultralow attachment plates during 3 d. The PLMA precursor solution is obtained solubilizing lyophilized PLMA in a photoinitiator solution for UV photopolymerization. Generated spheroids are encapsulated into PLMA hydrogels at three different concentrations (10, 15, and 20% (w/v)), poly(ethylene glycol) diacrylate (PEGDA) at 10% (w/v), and Matrigel. Over the time of culture, the encapsulated cells start to invade the surrounding ECM‐mimicking matrix, developing the named invasive branches. Throughout the 14 d of culture, the spheroid area and invasion length are measured to evaluate the invasiveness ability of each cell type into the different biomaterials and stiffnesses. 2. 1 Uniformity of Spheroid Formation Matrix‐dependent cell spheroid invasion should be evaluated assuming the same conditions in terms of initial spheroid size for each cell type in order to assure accurate and reproducible results. Hanging drop and forced‐floating are robust scaffold‐free techniques to spheroid generation which are expected to yield tunable and reproducible spheroids with normally distribution of initial sizes. 6, 44 In this study, 96‐well round‐bottom ultralow attachment plates were used to generate spheroids of four cell types (hBM‐MSC, MG‐63, SaOS‐2, and A549). An optimal seeding density of 12 000 cells/spheroid was established to produce spheroids with a mean size of 300–600 µm in diameter at 72 h, promoting the establishment of a cell heterogeneity suitable for physiopathological studies. hBM‐MSCs and MG‐63 are cell types characterized by the production of a dense ECM and expression of cell‐mediated adhesion proteins, which increase the spontaneous aggregation ability that results into highly compact and circular spheroids (Figure S1A, B, Supporting Information). However, not all cell types demonstrate this ability, with nonspheroid forming cells producing loose aggregates as a result of weak intercellular interactions (Figure S1C, D, Supporting Information). 45 The lack of spheroid‐forming ability is related to the poor expression levels of cadherin and catenin protein families, which are involved in cell–cell adhesion via adherens junctions. 46, 47 To overcome this barrier in spheroid formation, biological and synthetic additives acting as cross‐linking agents, 48, 49 adhesion stimulators, 48, 50 or rheological modifiers and crowding agents 51, 52 have been applied to strengthen cellular aggregation. Herein, methylcellulose (MethoCel) was used as crowding agent to improve the spheroid formation from SaOS‐2 and A549 cell lines, leveraging its semisolid gel‐like properties that increase the culture medium viscosity. The supplementation with MethoCel strengthened cell–cell interactions, allowing the formation of stable but not highly compacted SaOS‐2 and A549 aggregates after 3 d of formation (Figure S1C, D, Supporting Information). Concerning to the area of the spheroids generated from each cell type, hBM‐MSC and MG‐63 spheroids are the smallest ((1. 12 ± 0. 211) × 10 5 and (1. 20 ± 0. 104) × 10 5 µm 2, respectively) due to the strongest intercellular interaction that produced compacted spheroids—see Table 1 and Figure S1A, B (Supporting Information). In SaOS‐2 and A549 spheroids, the cells were not highly compacted, resulting in a higher average size ((3. 19 ± 0. 289) × 10 5 µm 2 for SaOS‐2, and (3. 34 ± 0. 273) × 10 5 µm 2 for A549)—see Table 1 and Figure S1C, D (Supporting Information). Analysis of size distribution suggests that spheroids were uniformly generated producing a Gaussian distribution of their initial sizes, as confirmed by D'Agostino–Pearson normality test (Figure S2, Supporting Information). Table 1 Area and diameter of the four cell spheroids types. Spheroids were obtained from 12 000 cells cultured during 3 d on round‐bottom ultralow attachment plates. Spheroid diameter was obtained through spheroid area data considering the spheroids as perfect circumferences Cell type Spheroid area [× 10 5 µm 2 ] Spheroid diameter [µm] hBM‐MSC 1. 12 ± 0. 21 376. 30 ± 35. 57 MG‐63 1. 20 ± 0. 10 390. 85 ± 16. 61 SaOS‐2 3. 19 ± 0. 29 637. 05 ± 28. 61 A549 3. 34 ± 0. 27 651. 78 ± 26. 32 John Wiley & Sons, Ltd. 2. 2 3D Embedded Spheroid Culture and Morphological Characterization Collective cell movement into the surrounding tissues plays a crucial role in morphogenesis, physiological processes (e. g. , wound healing), and also in cancer spreading and metastasis development. 22 Several evidences have demonstrated that, in addition to cell‐intrinsic mechanisms, biochemical and mechanical cues of the ECM are fundamental factors in cell invasion. 29, 53 From this standpoint, recapitulation of the pathological processes triggered by the microenvironment components could elucidate about the involved molecular mechanisms and drive the development of new therapeutic targets. rBM and type I collagen‐based invasion models have been widely used in invasion‐related studies. 54, 55 PLMA‐based hydrogels were recently proposed as a new biomimetic material with tunable mechanical properties for 3D cell culture. 33 Herein, we aimed to demonstrate that PLMA hydrogels are also suitable to sustain spheroid heterogeneity and support cell spheroid invasion. To validate the cell spheroid behavior into PLMA hydrogels, Matrigel and PEGDA were chosen as positive and negative controls, respectively. The choice of Matrigel instead of type I collagen is related with the fact that Matrigel is characterized by a complex mixture of proteins, thus more closely resembling the composition of PLMA. Regarding the negative control, PEGDA is a photopolymerizable synthetic polymer via UV light exposure, such as PLMA, but its lack of adhesive motifs makes it unable to support cell invasion. In order to explore the effect of the mechanical properties on spheroid behavior and cell invasiveness, compressive mechanical tests were performed in PLMA hydrogels at 10, 15, and 20% (w/v), and the PEGDA hydrogel concentration was chosen to have similar mechanical properties (Figure S3, Supporting Information). The PLMA hydrogels herein explored have a Young's modulus in the range of 13–16 kPa and the PEGDA hydrogel at 10% (w/v) has a Young's modulus of 12 kPa. Regarding the mechanical properties of Matrigel, Davidson et al. 56 reported a Young's modulus of 401 Pa obtained through a compression test. For this study, a 14 d invasion assay was carried out on the encapsulated spheroids generated from the four types of cells already mentioned in the different types and concentrations of biomaterials (Figures S4–S7, Supporting Information). Cells in spheroids with a diameter range from 200 to 500 µm start to develop chemical gradients (e. g. , nutrients, oxygen, pH) that define a typical zone of proliferative cells at the surface and necrotic/hypoxic cells in the core, sometimes coexisting with quiescent cells in the middle. 19 In order to analyze this heterogeneity, Live/Dead assays were performed using Calcein‐AM and PI staining for all conditions (cell spheroid type vs biomaterial type/concentration). After 24 h of spheroid encapsulation, fluorescence microscopy evidenced that hBM‐MSC and MG‐63 spheroids presented a well‐defined necrotic core surrounded by a zone of metabolically active cells ( Figure 2 ). Effectively, the compacted structure of these two types of spheroids as well as their sizes is favorable for the development of this in vivo‐like phenotypic heterogeneity. At the end of the 14 d of culture, hBM‐MSC and MG‐63 showed high cell viability and invasion for all concentrations of PLMA hydrogels—see Figures 3 and 4. In the case of hBM‐MSC spheroids, the necrotic core is significantly smaller in comparison with tumor‐derived spheroids, correlating with the study developed by Bartosh et al. , 41 where the authors showed that only a reduced fraction of cells are apoptotic or necrotic in a hBM‐MSC spheroid culture. Spheroids embedded into PEGDA hydrogels maintained an outer ring of viable cells surrounding a well‐defined necrotic core (Figure 3 ). When embedded into Matrigel, MG‐63 spheroids with an evident necrotic core invaded the hydrogel, whereas cell–ECM interaction seemed to promote mesenchymal cells invasion in hBM‐MSC spheroids. However, the invading hBM‐MSCs undergone in a quiescent state, which is demonstrated by the overlap of red and green stains (Figure 3 ). The entry of MSCs into a quiescent state is known to be related with loose cell adhesion due to anchorage deprivation. 57, 58, 59 So, the cell behavior herein showed by Matrigel‐embedded hBM‐MSC spheroids was triggered by scaffold softness, an important physicomechanical cue recently explored by Rumman et al. 59 using polyacrylamide gels of varying stiffness to induce cells quiescence. Although this cellular state is interesting to cell differentiation, cell invasion studies must be conducted using an ECM‐resembling matrix with improved mechanical properties able to support the viability of invasive cells. Regarding this issue, PLMA hydrogels clearly demonstrated to support hBM‐MSC viability, offering the proper mechanical cues to allow the formation of robust actin stress fibers and focal adhesions, contrary to Matrigel (Figure 4 ). Figure 2 Fluorescence microscopy images of Live/Dead staining. hBM‐MSC, MG‐63, SaOS‐2, and A549 spheroids encapsulated into PLMA hydrogel (10, 15, and 20% (w/v)), PEGDA hydrogel (10% (w/v)), and Matrigel at 24 h post encapsulation. All spheroids were generated with an initial number of 12 000 cells per spheroid. The green and red channels represent the Calcein‐AM and PI staining of live and dead cells, respectively. Scale bar: 200 µm. Figure 3 Fluorescence microscopy images of Live/Dead staining. hBM‐MSC, MG‐63, SaOS‐2, and A549 spheroids encapsulated into PLMA hydrogel (10, 15, and 20% (w/v)), PEGDA hydrogel (10% (w/v)), and Matrigel at 14 d of culture. All spheroids were generated with an initial number of 12 000 cells per spheroid. The green and red channels represent the Calcein‐AM and PI staining of live and dead cells, respectively. Scale bar: 200 µm. Figure 4 Fluorescence microscopy images of DAPI/phalloidin staining. hBM‐MSC, MG‐63, SaOS‐2, and A549 spheroids encapsulated into PLMA hydrogel (10, 15, and 20% (w/v)), PEGDA hydrogel (10% (w/v)), and Matrigel at 14 d of culture. All spheroids were generated with an initial number of 12 000 cells per spheroid. The blue channel represents the nuclear staining by DAPI and the red channel demonstrates the actin filaments staining by Phalloidin‐Red probe. Scale bar: 200 µm. With regard to SaOS‐2 spheroids, the above‐discussed issues concerning the formation of noncompacted cellular aggregates resulted in a homogeneous distribution of live and dead cells in those spheroids at 24 h of encapsulation (Figure 2 ). Over the 14 d of culture, the cells freely organized inside the PLMA hydrogel and Matrigel, forming a necrotic core, what was not verified into the negative control, PEGDA hydrogels. The homogeneous distribution of live and dead cells at 24 h of encapsulation was also verified in A549 spheroids encapsulated into the three biomaterials (Figure 2 ). However, in A549 spheroid embedded into PLMA hydrogels, an outer ring of dead cells was observed. It can be explained with the occurrence of solid stress imposed by the surrounding ECM. 60, 61, 62 This mechanical compressive stress can induce genotypic and phenotypic changes associated with a higher tumor malignancy that can be reflected into an increased tumor invasiveness ability. 62 On the other hand, this physical constraint can affect cell viability and promote apoptosis, an event that is reported to be reversible. 61, 63 Effectively, the elastic modulus of PLMA hydrogels (13–16 kPa) is much higher than Matrigel (401 Pa), what means that the compressive stress applied by PLMA hydrogels is quite higher (Figure S3, Supporting Information). 33, 56 Cheng et al. 61 demonstrated that when an external stress (2 kPa) is applied on agarose‐encapsulated spheroids, the caspase‐3 activity increases. From spheroid area analysis, it is observable that A549 spheroid significantly compacted into PLMA hydrogels during the first 24 h post encapsulation, which demonstrate the compressive stress exerted by this PL‐derived matrix that drive cell death ( Figure 5 ). In this mechanical context, PLMA‐encapsulated A549 spheroids are able to overcome this inhibitory feedback exerted by the surrounding matrix once PLMA is a protein‐based hydrogel and, for that reason, can be degraded and softened by released matrix metalloproteinases. After 14 d of culture, the presence of dead cells in the periphery was still verified, but to a lesser extent (Figure 3 ). These results demonstrate that the cells were able to overcome the described negative effect of the ECM compressive stress. Contrary to that described for SaOS‐2, A549 spheroids did not compact into Matrigel and were able to develop a highly defined necrotic core. In PEGDA, the majority of the cells were still viable and dead cells were mostly located in the center of the spheroid (Figure 3 ). Figure 5 Progression of spheroids area during the 14 d of culture. A–D) Quantification of the area of hBM‐MSC, MG‐63, SaOS‐2, and A549 spheroids into the different biomaterials: PLMA hydrogel (10, 15, and 20% (w/v)), PEGDA hydrogel (10% (w/v)), and Matrigel. Data are presented as mean ± SD ( n ≥ 3). 4′, 6‐Diamidino‐2‐phenylindole (DAPI)/phalloidin staining was performed at 14 d of culture in order to analyze the morphology of actin filaments, which are involved in the regulation of cell shape and polarity, playing a crucial role in cell motility. Fluorescence images obtained from z‐stacks show that invading cells from hBM‐MSC and MG‐63 spheroids acquired an in vivo‐like cell polarity in PLMA hydrogels and Matrigel—see Figure 4. Nevertheless, the pattern of invasion is clearly distinct, with a more evident collective invasion pattern into PLMA‐based matrix contrasting with a higher number of cells individually migrating into Matrigel. Moreover, the actin fibers stress is strongest in PLMA hydrogel, as can be seen by the most pronounced cell elongation comparing with cells invading Matrigel. hBM‐MSC spheroids presented an open spheroid structure due to the interconnectivity between collective invading cells. MG‐63 spheroids showed to maintain their compact structure (denser zones), surrounded by an interconnective network of tumor invading cells. These structural features were not verified in Matrigel‐embedded spheroids, where the cells that constitute the spheroid dispersed inside the matrix without defined invasive branches (Figure 4 ). In SaOS‐2 and A549, the intercellular interactions between aggregated cells embedded into PLMA hydrogels are visible as well as some small invasions (Figure 4 ). A549 cell spheroids demonstrated to self‐organize and form round to oval‐shaped noncellular regions in PLMA hydrogels and Matrigel, also visible in Live/Dead fluorescence images (Figure 3 ) that, according to in vivo histology analysis in literature, corresponds to malignant glands of acinar adenocarcinoma with invasive phenotype. 64 It clearly evidences that PLMA hydrogel stimulates an in vivo‐like cell organization of lung cancer cells and promotes an invasive behavior. To the best of our knowledge, this is the first time that this invasive phenotype was recapitulated in a spheroid‐based 3D in vitro culture, what can be useful to perform pathological studies in a more realistic way. This different tumor cell behavior into Matrigel is probably related with the matrix stiffness since lung cells are originated from a soft environment, then the environment provided by Matrigel is mechanically closest to the native one. With regard to cell proliferation, ATP quantification indicates a decrease of this proliferative marker over 14 d in all cell types, except in MG‐63 (Figure S8, Supporting Information). ATP production is dependent on each cell energy demand and is related to the metabolic pathways adopted by cells to trigger specific pathophysiological processes. Some studies reporting the use of ECM‐mimicking matrices to embed cell spheroids have been demonstrating that the matrix can exert a compressive solid stress responsible for changes in gene expression related with cell proliferation, apoptosis, invasion, and migration. 60, 63, 65 That mechanical stress stimulates the expression of proliferation inhibitors, inducing cell‐cycle arrest, and was also demonstrated to drive cell phenotype switching between proliferative and invasive behavior. 62, 63 Effectively, some reports have evidenced the hypothesis of phenotype switching as a key event characterized by cell‐cycle arrest involved not only in cancer metastasis, but also in physiological processes as embryogenesis. 66, 67 These two central phenotypes in malignant behavior were already studied in vitro using Matrigel invasion assays 68, 69 and confirmed in vivo by immunohistochemistry of invasive and proliferative cells of patient tumor samples. 70, 71 The decrease of ATP in hBM‐MSCs, SaOS‐2, and A549 spheroids can be explained by the switching from proliferative to invasive phenotype. As expected, in PEGDA hydrogels, due to the inexistence of proteolytic degradation, no cell type was able to invade, however the compressive stress may also have an effect on proliferation inhibition. Concerning the increased ATP production of MG‐63 from 24 h to 14 d of culture (Figure S8, Supporting Information), we hypothesize that this cell line is more resistant to mechanical stress than other cells. Moreover, the most pronounced ATP increase in protein‐derived matrices can be related with a high expression of metalloproteinases (MMPs) comparing with other cells, as SaOS‐2. 72 Concerning the ability of PLMA hydrogels to sustain the formation of a necrotic core surrounded by an outer zone of viable cells, the results indicate that 3D tumor spheroids of MG‐63 and SaOS‐2 embedded into this matrix recapitulate the in vivo‐like phenotypic heterogeneity of solid tumors (Figure 3 ). In the case of hBM‐MSC, this heterogeneity was also verified, demonstrating the feasibility of PLMA hydrogels to support MSC spheroid culture and explore this approach for MSC‐based therapeutics research. 73 Although A549 spheroids did not develop a necrotic core, optical contrast and fluorescence microscopy at 14 d of culture demonstrated that the cells were able to self‐organize and develop acinar structures particularly related with an invasive phenotype, validating the applicability of PLMA hydrogels for spheroid‐based lung cancer‐related studies. As expected, the four cell spheroid types encapsulated into PEGDA hydrogels were not able to invade the matrix, and significant structural changes were not visualized. 2. 3 Matrix‐Dependent 3D Spheroid Area and Invasion Speed Besides the evaluation of the spheroid phenotypic heterogeneity and invasive cells morphology, the consistency of invasion is a prerequisite to the development of accurate and reproducible in vitro invasion models. 44 Therefore, spheroid core area throughout the invasion assay and uniformity of invasion of each cell spheroid type into the different matrices were assessed through optical microscopy images analysis. For spheroid area analysis, it is important to note that only the area corresponding to regions where the spheroid cells were compact was considered in order to investigate the compact spheroid area maintenance—see Figure 1. Spheroid area changes during the 14 d of culture were visually distinct between each cell type and highly dependent on the biomaterial where they were embedded (Figure 5 ). In hBM‐MSC spheroids, a diminishing in their area was verified into PLMA hydrogels and Matrigel as a result of the mesenchymal‐related invasiveness ability (Figure 5A ). Regarding their behavior into PLMA hydrogels of different stiffness, the area suffered a further decrease on the stiffest hydrogels (15% and 20% (w/v) of PLMA). Nonetheless, a more pronounced decrease in the area of the spheroid core was seen in Matrigel, what is related with the aforediscussed dispersity of the cells that constitute the compact area of the spheroid. In PEGDA, despite some oscillations during the culture time, the difference between the initial and final area was not significant. For MG‐63 spheroids, no significant changes in area values were verified when encapsulated into PLMA or PEGDA hydrogels (Figure 5B ). In Matrigel, an increase in spheroid area was observed, as clearly evidenced by 24 h and 14 d Live/Dead images (Figure 3 ). SaOS‐2 spheroids embedded into PLMA hydrogels remained confined over the 14 d with a minimal decrease in the area value, which is probably related with the maintenance of cell adhesion proteins expression (Figure 5C ). It demonstrates that the herein proposed new biomaterial is not stimulating alterations on the gene expression levels of this cell type. Into PEGDA hydrogels, the cell aggregates seem to have dispersed during the first 7 d; after that, the weak cell–cell interactions were enough to promote cell reaggregation, being initial and final spheroid sizes similar. These spheroids demonstrated to have a different behavior when encapsulated into Matrigel, where the cells quickly compacted as seen in Figure 4. Contrary to what was observed in PLMA‐based hydrogels, the very weak cell–ECM interactions along with an improved cell–cell interaction via E‐cadherin expression can explain this particular behavior. Effectively, it is well‐reported that the stiffening of the tumor‐surrounding ECM is highly related with an invasive and metastatic phenotype, where cell–ECM interaction is improved. 29, 74, 75, 76 It means that the softness of Matrigel can probably be inducing increased E‐cadherin and decreased N‐cadherin expression, epithelial and mesenchymal‐associated markers, respectively. 76, 77 Relatively to A549, the other cell line classified as nonspheroid forming cells, cell aggregates compaction in PLMA hydrogels was observed during the first 2–3 d of culture, what can be related with the compressive force exerted by the surrounding matrix that led to cell death on the periphery of the spheroids—see Figure 5D. Then, in accordance with the enhanced cell viability until 14 d, spheroid area significantly increased. When embedded into Matrigel, A549 spheroid area increased over the 14 d. This limited growth into PLMA hydrogels in comparison with Matrigel is related with the mechanical stress imposed by the higher stiffness of PLMA matrix (Figure S3, Supporting Information). 60, 63 In PEGDA hydrogels, A549 cells aggregates firstly dispersed inside the gel. Perhaps due to the lack of adhesion motifs in PEGDA network, the cells compacted in a spheroid structure where cell–cell adhesion prevailed. Altogether, the spheroid area throughout the invasion assay showed to be highly dependent on the biomaterial and cell type. Nevertheless, the spheroid area tendency demonstrated that PLMA hydrogels guarantee a spheroid behavior similar to verified into Matrigel and PEGDA hydrogels, evidencing the potential of this new humanized biomaterial for further fundamental and therapeutic studies. Furthermore, the matrix stiffness also seemed to contribute to this parameter but in a lesser extent, demonstrating that matrix biochemical features are more important with regard to changes in spheroids area. In this respect, PLMA hydrogels are evidently the currently available biomaterial that offer the matrix biochemical cues to better recapitulate cell behavior. In terms of invasion kinetics of each cell spheroid type into a matrix, it is important to develop a platform that could sustain a uniform invasion, ensuring reproducibility and correctness of the results. To explore this issue in the PLMA matrix‐based models, the invasion length of the spheroids was measured during the first 72 h and at 7, 10, and 14 d of culture, and compared with the results obtained on the negative and positive controls. The invasion length was analyzed, and the invasion speed was determined through a linear regression of the curves of invasion length versus time—see Figure S9 (Supporting Information). The invasion speed for each condition was expressed as mean ± standard deviation (SD) and the uniformity of invasion was analyzed recording the goodness of fit ( R 2 ) through the above‐referred linear regression—see Figure 6 and Table S1 (Supporting Information). We assumed that R 2 ≥ 0. 80 represents the range where the invasion is considered sufficiently uniform to validate the correspondent condition as an accurate and robust invasion model. Figure 6 resumes the invasion speed of the different cells cultured in the three hydrogels, considering the distinct stiffness conditions of PLMA hydrogels. Figure 6 Invasion speed of each spheroid type into the different biomaterials and stiffness conditions. Data results from a linear regression analysis of the invasion length up to 14 d of culture and are presented as mean ± SD ( n ≥ 3). Statistical analysis was performed between the different biomaterials used in each cell spheroid type. Two‐way ANOVA analysis of variance combined with Tukey's multiple comparisons test revealed significant differences between analyzed groups: ** p < 0. 01 and **** p < 0. 0001. # means significant differences with the groups on the right, inside the same cell spheroid type group. For the remaining groups, the statistical analysis revealed no significant differences. ♦ PEGDA‐embedded spheroid did not invade the hydrogels. As expected, into the hydrogel considered as the negative control, PEGDA, the lack of cellular adhesion motifs and domains sensitive to cell‐mediated proteolysis hindered cell invasion of all cell types studied here—see Figure 6. The two cell types with higher spheroid‐forming ability, hBM‐MSC and MG‐63, also demonstrated the greatest invasiveness capacity in PLMA‐ and Matrigel‐based hydrogels. hBM‐MSC showed to have a higher invasion kinetics in the three PLMA formulations than in Matrigel, displaying variability of invasion speed between distinct stiffnesses of PLMA. Regarding MG‐63 cells, different invasion capacity was also verified but, in this case, only the PLMA at 10% (w/v) showed to enhance the invasion kinetics in comparison to Matrigel. Although MG‐63 cells exhibit mesenchymal features as hBM‐MSC, the data from PLMA biomaterial show that this cell type prefer softer matrices than mesenchymal stem cells. Some reports in which invasion and migration of SaOS‐2 and MG‐63 were evaluated into Transwell chambers, both cell lines exhibited similar ability regarding these parameters. 78, 79, 80 However, the herein presented data suggest that these two osteosarcoma cell lines have a significantly different behavior of invasion. As showed in Figure 6 and unlike to what is described in literature, SaOS‐2 demonstrated to have less invasive ability in all PLMA concentrations comparing with MG‐63. It is probably related with the fact that cell invasion has been evaluated throughout an embedded cell aggregate, spheroid, and not as cell suspension seeded on a matrix. Besides that, it is reported that spheroid compaction is associated with an aggressive invasive phenotype, 81 which correlates with the presented data once SaOS‐2 formed noncompacted cell aggregates contrary to MG‐63 cells. Nonetheless, once spheroid is considered the most relevant 3D platform, cell invasion studies could be more accurate when evaluated through a cell spheroid due to the tridimensionality that more faithfully mimics an in vivo tumor invasion. In Matrigel ®, the value of SaOS‐2 invasion kinetics was negative, which is related with the huge spheroid compaction observed during the first hours (Figure 5B ). This means that, although small invasions have been visualized and were similar to the ones observed in PLMA hydrogels (Figure S6, Supporting Information), the radius of invasion was smaller than the radius of the spheroid measured at day 0. A549 is an epithelial cell line able to form primary tumors and pulmonary metastasis into immunocompromised mice. 82 However, 3D invasion models to recapitulate and study the metastatic process are not well‐explored. So, we intended to investigate the ability of A549 cells to invade throughout an PLMA‐embedded spheroid. Interestingly, A549 were able to create small protrusions in PLMA, contrary to what was verified in Matrigel, where the cells were not able to invade in the matrix—see Figure 6. Although the low invasion speed, the data suggest that PLMA‐based hydrogels can sustain A549 invasion, offering a 3D platform to develop a lung invasion model to more faithfully study pulmonary metastasis. To the best of our knowledge, this is the first time that spheroid‐forming A549 cells demonstrated the ability to invade an involving matrix, even in a small extension, in the absence of stromal cells. Some studies have reported the importance of stromal cells (e. g. , endothelial cells) into this metastatic process of A549 and established the key role of growth factors, such as epidermal growth factor (EGF) and transforming growth factor beta 1 (TGF‐β1), in epithelial–mesenchymal transition (EMT). 83, 84 Taking into account the abundance of growth factors present in the raw material of the hydrogels here proposed, PLMA hydrogels are probably triggering the EMT of A549, stimulating its invasion through the matrix. 33 The goodness of fit ( R 2 ) obtained through the linear regression to determine the invasion kinetics is a useful parameter to analyze how uniformly the spheroids invade into the distinct matrices. This parameter, along with the value of invasion kinetics, is essential to choose the matrix whose biochemical and mechanical properties offer the best condition to support an in vivo‐like invasion of a cell type and, by this way, develop a faithful invasion model. In general, the results show that PLMA hydrogels support a uniform invasion of the four cell types. Only the invasion of A549 spheroids into PLMA at 20% (w/v) cannot be considered uniform. Regarding the spheroid invasiveness into Matrigel, hBM‐MSC and MG‐63 spheroids were able to invade uniformly, although it was found that hBM‐MSC entered in a quiescent state. The same does not hold true for SaOS‐2 spheroids, for which R 2 < 0. 80. Overall, PLMA hydrogels demonstrate to support the invasiveness ability and uniformity of the different spheroids in comparison to positive control, Matrigel. The data also show that the invasion kinetics can be controlled through hydrogel stiffness, demonstrating the ease of adapting PLMA hydrogels to the biomechanical cues of the native tissue under study. 2. 4 Growth Factors Release from PLMA Hydrogels Platelet lysates (PL) are protein concentrates rich in fibrous and adhesive proteins, growth factors, and other bioactive molecules that play a pivotal role in physiological processes, such as wound healing and tissue regeneration. 34 Although the proteins undergo a chemical modification process, the degree of modification is about 14%, allowing a sustained release of proteins from the PLMA hydrogel matrices, as previously reported. 33 Hence, protein releasing from PLMA hydrogels at 10, 15, and 20% (w/v) was herein performed in order to evaluate the release of VEGF‐A, TGF‐β1, and EGF. These growth factors have been associated with the induction of EMT and, by this way, related with an increased cell invasiveness and metastatic dissemination in primary tumor sites. 85, 86, 87, 88 In this sense, the assessment of their presence and release in the PLMA‐based invasion models is an important step toward the validation of this human‐derived matrix for 3D invasion models application. The release profile of total protein showed a fast release during the first 24 h, followed by a sustained release until the 336 h (14 d) ( Figure 7 A ). This tendency is in accordance with the previously reported analysis. 33 To address the release of the aforementioned growth factors, ELISA assays were performed and a sustained release of VEGF‐A and TGF‐β1 was verified (Figure 7B, C ). As expected, VEGF‐A and TGF‐β1 ELISA quantification assays suggest a dependency of the protein released concentration relatively to the PLMA hydrogels concentration. Comparing with the availability of growth factors reported for standard Corning Matrigel, VEGF‐A and TGF‐β1 release from PLMA hydrogels is until 100‐fold lower and 2‐fold higher, respectively. EGF ELISA quantification revealed that this growth factor was not released from PLMA hydrogels in its ELISA‐detectable form, while in standard Corning Matrigel its presence is reported. One of the reasons that may justify the nondetection of this growth factor is its loss during the dialysis process, since its molecular weight (≈6. 2 kDa) is slightly higher than the cutoff (3. 5–5. 0 kDa) of the dialysis membranes used. On the other hand, EGF may have undergone chemical modification at the ELISA recognition site; however, its release in a biologically active form may be occurring. Furthermore, matrix‐immobilized growth factors can also display available biologically active domains to directly interact with encapsulated cells and induce in vivo‐like cell behaviors. In this sense, for any of the growth factors here quantified, it can be hypothesized that their availability is higher than that detected in the ELISA assays. From this standpoint, the protein content availability of PLMA hydrogels reinforces the hypothesis that their biochemical composition recreates the protein accessibility in the native tissues ECM, promoting the exceptional cell invasiveness behavior herein reported. Figure 7 Protein and growth factor release from PLMA hydrogels. A) Total protein quantification and B, C) ELISA quantification of TGF‐β1 and VEGF‐A release from PLMA hydrogels at 10, 15, and 20% (w/v). Data are presented as mean ± SD ( n ≥ 3). 3 Conclusions The demand for more pathophysiologically relevant 3D models able to recapitulate fundamental features of tumor metastasis is ever‐rising in order to further improve the discovery of new effective therapies. PLMA‐based hydrogels, recently proposed as a new biomaterial for 3D cell culture, were herein explored as a human‐derived platform for the development of spheroid invasion models. This biomaterial demonstrated to support cellular invasion and formation of a necrotic core in spheroids of hBM‐MSC and distinct tumor cell lines, recreating the phenotypic heterogeneity of solid tumors which is undoubtedly involved in therapy resistance in vivo. Furthermore, the cells also acquired an in vivo‐like cell polarity and were able to invade the matrix, forming complex interconnective cellular networks. Regarding the invasiveness ability evaluation of each cell spheroid type, PLMA‐based hydrogels not only promoted an increased invasion kinetics comparing with the positive control, Matrigel, but also demonstrated that the invasion kinetics can be controlled through PLMA matrix stiffness. We hypothesized that this invasion kinetics can be associated with the release or cell–ECM interaction of crucial growth factors (e. g. , VEGF, TGF‐β1, EGF, IGF) involved into the stimulation of cellular invasion and migration during tumor growth and metastasis development. Overall, these results clearly suggest that PLMA‐based hydrogels can be a great alternative to the animal‐derived Matrigel. The proposed 3D invasion models may be used to study the biological mechanisms involved in metastatic cascade and as a platform for screening and validation of new therapeutic agents in a more realistic microenvironment. The evidenced ability to support different cell population survival also opens up the possibility to develop more complex disease models, integrating and studying the role of stromal cells in proliferation and drug resistance of tumor cells. Finally, the possibility of combining patient‐derived PL and cells can allow the application of this platform in personalized drug screening. 4 Experimental Section Synthesis of Methacryloyl Platelet Lysates PLMA were synthesized based on a protocol previously reported. 33 Briefly, PL (STEMCELL Technologies, Canada) were thawed in a water bath at 37 °C. Then, PLMA of low‐degree modification (PLMA100) were synthesized by reaction with methacrylic anhydride (MA) 94% (Sigma‐Aldrich, USA) in a ratio of 100:1. The reaction was performed at room temperature, under constant stirring, and pH was maintained in a range of 6–8 using sodium hydroxide (NaOH, 5 m ) (AkzoNobel, USA) solution. After the reaction, synthetized PLMA were purified by dialysis with Float‐A‐Lyzer G2 Dialysis Device 3. 5–5 kDa (Spectrum, USA) against deionized water for about 24 h. The PLMA solution was sterilized with low protein retention 0. 2 µm filter (Enzymatic S. A. , Portugal), frozen with liquid nitrogen, lyophilized (LyoQuest Plus Eco, Telstar, Spain) and stored at 4 °C until further use. Compressive Mechanical Testing PLMA hydrogels at 10, 15, and 20% (w/v), PEGDA hydrogels at 10% (w/v), and Matrigel were mechanically characterized by compression testing using the Instron 3340 Series Universal Testing System (Instron, USA) equipped with a 50 N load cell. The assays were performed on freshly prepared hydrogels with a cylindrical form (6 mm of diameter and 2. 5 mm of height), at room temperature, except for Matrigel with which the assays were performed at 37 °C in order to maintain protein polymerization. The Young's modulus was defined as the slope of the linear region (0–5% of strain) of the strain–stress curve. Cell Culture hBM‐MSCs (ATCC, USA) and MG‐63 cell line (ECACC, Sigma‐Aldrich, USA) were cultured in minimum essential medium alpha (α‐MEM) (Thermo Fisher Scientific, USA) supplemented with sodium bicarbonate (2. 2 g L −1, Sigma‐Aldrich, USA), 10% heat‐inactivated fetal bovine serum (FBS) (Thermo Fisher Scientific, USA), and 1% antibiotic/antimycotic (Thermo Fisher Scientific, USA). SaOS‐2 cell line (ECACC, Sigma‐Aldrich, USA) was cultured in Dulbecco's modified Eagle's medium (DMEM) low glucose (Sigma‐Aldrich, USA) supplemented with sodium bicarbonate (3. 7 g L −1 ), 10% heat‐inactivated FBS, and 1% antibiotic/antimycotic. A549 cell line (ATCC, USA) was cultured with Nutrient Mixture F‐12 Ham (Sigma‐Aldrich, USA) supplemented with sodium bicarbonate (2. 5 g L −1 ), 10% heat‐inactivated FBS, and 1% antibiotic/antimycotic. All cells were cultured in T‐flasks, maintained under 5% CO 2 atmosphere at 37 °C (standard culture conditions) and passaged at about 80% confluence. The medium was replaced every 2 to 3 d. Generation of Multicellular Spheroids Each of the four cells types (hBM‐MSC, MG‐63, SaOS‐2, and A549) was detached with 0. 25% trypsin/EDTA (Gibco, Thermo Fisher Scientific, USA) and resuspended in its culture medium. A density of 12 000 cells in culture medium (150 µL) was seeded onto 96‐well round‐bottom ultralow attachment plates (Corning, Thermo Fisher Scientific, USA) in order to generate spheroids. In the case of SaOS‐2 and A549 cell lines, the culture mediums were supplemented with 0. 5% (w/v) of Methocel A4M (Sigma‐Aldrich, USA). Seeded cells were centrifuged at 500 × g for 10 min and incubated for 72 h at standard culture conditions. Generated spheroids were imaged by optical contrast light microscopy (Primostar, Carl Zeiss, Germany) using ZEN Imaging software. 3D Invasion Assay After 72 h of culture, generated spheroids were embedded into different hydrogel solutions: 10, 15, and 20% (w/v) PLMA100, 10% (w/v) PEGDA ( M n = 10 000 g mol −1, Sigma‐Aldrich, USA), and Matrigel Matrix (Corning, Thermo Fisher Scientific, USA). Embedding and culture of spheroids were performed into µ‐Slide Angiogenesis (ibidi, Germany) with one spheroid per well. PLMA hydrogel precursor solutions of 10, 15, and 20% (w/v) were prepared dissolving lyophilized PLMA in a sterilized solution of 0. 5% (w/v) 2‐hydroxy‐4′‐(2‐hydroxyethoxy)‐2‐methylpropiophenone (Sigma‐Aldrich, USA), also known as Irgacure, in phosphate buffered saline (PBS) (Sigma‐Aldrich, USA). PEGDA hydrogel solution of 10% (w/v) was prepared following the same procedure and then sterilized with 0. 2 µm filter. Matrigel Matrix solution (8–12 mg mL −1 ) was used as received from the provider. For PLMA and PEGDA‐based 3D invasion assays, a first layer of hydrogel solution (5 µL) was made by photopolymerization using ultraviolet (UV) irradiation (0. 095 W cm −2 ) and a collimator during 15s, performing a semi‐crosslinking. The spheroid was deposited above, as much as possible in the center of the well, and a second layer was formed at the same conditions during 60 s for complete crosslinking. For Matrigel‐based 3D invasion assay, two layers were also performed, incubating the slides during 5 and 30 min at 37 °C for first and second layer polymerization, respectively. Thereafter, the appropriate culture medium (depending on cell type) was added to each well. The invasion systems were incubated for 14 d and the culture medium was replaced every 2 to 3 d. a) Cell viability analysis : After 24 h and 14 d of spheroids embedding, a Live/Dead cell assay was performed for viability assessment. The hydrogels were incubated in a solution of 1:100 of Calcein AM solution in DMSO (4 × 10 −3 m, Life Technologies, Thermo Fisher Scientific, USA) and 1:200 of propidium iodide (PI) (Thermo Fisher Scientific, USA) in PBS at standard culture conditions (5% CO 2 at 37 °C) for 2 h. After washing with PBS, the 3D invasion models were observed under a fluorescence microscope (Fluorescence Microscope Zeiss, Axio Imager 2, Carl Zeiss, Germany). b) Cell morphology analysis : Cell morphology assessment of invasive models was performed at 14 d of culture using a DAPI/phalloidin staining. At the aforementioned time‐point, hydrogels were washed with PBS and fixed in a 4% formaldehyde (Sigma‐Aldrich, USA) and 1% glutaraldehyde (Sigma‐Aldrich, USA) solution in PBS during at least 2 h. Before staining, samples were permeabilized with 0. 5% Triton X‐100 for 30 min and blocked with 5% FBS in PBS for 1 h at room temperature. The hydrogels were firstly incubated in a phalloidin solution (Flash Phalloidin Red 594, Biolegend, USA) diluted 1:8 in PBS at room temperature for 90 min. After washing with PBS, a DAPI (5 µg mL −1, Thermo Fisher Scientific, USA) solution diluted 1:200 in PBS was prepared and used to incubate the hydrogels during 30 min at room temperature. After several PBS washes, the hydrogels were observed under a confocal fluorescence microscope (Zeiss LSM 510 META confocal laser scanning microscope, Carl Zeiss, Germany). c) Cell proliferation quantification : Cell proliferation was assessed by ATP quantification using the CellTiter‐Glo 3D Cell Viability Assay (Promega, Madison, USA). First, all samples were washed with PBS, frozen at 80 °C in distilled water and thawed at 37 °C. PLMA hydrogels were then incubated with 1% trypsin/EDTA in distilled water (Gibco, Thermo Fisher Scientific, USA) for 1 h at 37 °C in order to expose the spheroid and invading cells. PEGDA hydrogels were mechanically dissociated in an ultrasound bath for about 10 min. PEGDA hydrogels and Matrigel were also incubated in 1% trypsin/EDTA in distilled water. Afterward, CellTiter‐Glo assay was performed in accordance with the manufacturer instructions. Briefly, CellTiter‐Glo reagent was added at a 1:1 ratio, the samples were vigorously mixed for 5 min and then incubated for 25 min at RT. Luminescence was measured in 96‐well flat‐bottom opaque white plates using Synergy HTX microplate reader (BioTek Instruments, Winooski, USA). Quantification of Spheroid Size and Invasion Length After spheroids embedding, cell spheroid invasion was monitored and imaged using an inverted optical contrast light microscope (Primostar, Carl Zeiss, Germany) with ZEN Imaging software. Image acquisition was performed every day until the third day and then at 7, 10, and 14 d of culture. The day 0 correspond to the pictures taken 1 h after spheroid embedding. Raw images were processed using the Image Processing tool of ZEN Image software. Quantification of spheroid area was performed using SpheroidSizer, a MATLAB‐based and open‐source high‐throughput image analysis software that applies an adapted active contour algorithm suitable to accurately measure spheroid size. 89 The robustness of the algorithm allows an automatic or manual delimitation of the spheroid area, excluding the invasive branches. Quantification of spheroid invasion length was performed using ImageJ software. The spheroid center was estimated and the length from that point to the longer branch of each sample—invasion radius—was measured. The invasion length was normalized for each time‐point to the spheroid radius measured at day 0 by using Equation (1) (1) Invasion length = Invasion radius − Spheroid radius Day 0 where Invasion radius is the length from the center of the spheroid to the end of the longer branch and the Spheroid radius Day 0 is the obtained value for the radius from the spheroid area measured at day 0, considering the spheroid delimitation a circumference. In both software, the sizes were converted from the acquired pixels (px), taking into account the objective amplification used, and expressed in micrometers (µm). Samples displaying post‐embedding abnormalities related with their location inside the hydrogel or lack of invasion at the end of the experiment were excluded. Quantification of Protein and Growth Factors Release The protein release assays were performed in PLMA hydrogels at 10, 15, and 20% (w/v) without encapsulated spheroids. The samples ( n = 6) were placed into falcons with PBS (5 mL, Thermo Fischer Scientific, USA) and incubated with constant agitation (60 rpm) in a water bath at 37 °C. Over 14 d, an aliquot (1 mL) was taken at each time‐point and fresh PBS (1 mL) was added. The collected aliquots were stored at −20 °C. For total protein quantification, Micro BCA Protein Assay Kit (Thermo Fisher Scientific, USA) was used. ELISA assays were performed to quantify the release of vascular endothelial growth factor (VEGF Human ELISA Kit, Invitrogen, ThermoFisher Scientific, USA), transforming growth factor β1 (TGF‐β1 Human ELISA Kit, Invitrogen, ThermoFisher Scientific, USA), and epidermal growth factor (Human EGF Quantikine ELISA Kit, R&D systems, Minneapolis, USA). Statistical Analysis All data were statistically analyzed using GraphPad Prism 7 Software and are expressed as mean ± standard deviation (SD) or mean ± standard error of the mean (SEM) of at least three independent experiments. The distribution of the initial spheroid sizes was analyzed with the D'Agostino–Pearson normality test. For invasion speed data, statistical significance between the different groups was identified using two‐way ANOVA analysis of variance combined with Tukey's multiple comparisons test, and the differences were considered significant when p < 0. 05. Statistical significance for Young's modulus and ATP quantification data was evaluated by one‐way ANOVA analysis using Tukey's multiple comparisons test. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902403
| 2,020
|
Advanced Science
|
4D Self‐Morphing Culture Substrate for Modulating Cell Differentiation
|
Abstract As the most versatile and promising cell source, stem cells have been studied in regenerative medicine for two decades. Currently available culturing techniques utilize a 2D or 3D microenvironment for supporting the growth and proliferation of stem cells. However, these culture systems fail to fully reflect the supportive biological environment in which stem cells reside in vivo, which contain dynamic biophysical growth cues. Herein, a 4D programmable culture substrate with a self‐morphing capability is presented as a means to enhance dynamic cell growth and induce differentiation of stem cells. To function as a model system, a 4D neural culture substrate is fabricated using a combination of printing and imprinting techniques keyed to the different biological features of neural stem cells (NSCs) at different differentiation stages. Results show the 4D culture substrate demonstrates a time‐dependent self‐morphing process that plays an essential role in regulating NSC behaviors in a spatiotemporal manner and enhances neural differentiation of NSCs along with significant axonal alignment. This study of a customized, dynamic substrate revolutionizes current stem cell therapies, and can further have a far‐reaching impact on improving tissue regeneration and mimicking specific disease progression, as well as other impacts on materials and life science research.
|
1 Introduction The dynamic nature of the physiological environment in vivo plays an essential role in modulating cellular behaviors and functions, which is closely related to embryonic development, tissue self‐renewal or regeneration, wound healing, and disease progression over time. [ [qv: 1–3] ] Stem cell therapy has provided a new paradigm for treating various diseases and tissue regeneration; however, these therapies have limited clinical application due to poor cell survival and differentiation potentials of the stem cells. [ [qv: 1, 4] ] Dynamic biochemical, physicochemical, and mechanostructural changes within the cellular microenvironment regulate many important biological functions of stem cells, including cell adhesion, growth, migration, and differentiation. [ [qv: 2, 3, 5] ] Biochemical (bioactive molecules), mechanobiological (elasticity or force loads), and topographic cues have been extensively studied in vitro for their influence on stem cell behaviors. However, most efforts are centered on static patterns. [ [qv: 6, 7] ] Inspired by dynamic, tissue‐specific microenvironments in vivo, the use of dynamic cell culture platforms to create synthetic microenvironments in vitro has become an attractive means to direct active changes in cellular functionality and to replicate the dynamic complexity of native tissues. [ [qv: 8, 9, 10] ] Several stimuli‐responsive methods, including applied electrical and magnetic fields, changes in pH and temperature, introduction of light and enzymes, have been applied to modulate the physicochemical properties of culture substrates. [ [qv: 2, 3, 9, 11, 12] ] Specifically, these methods can be used to modify culture substrate surface properties (e. g. , hydrophilicity/hydrophobicity), mechanical properties (e. g. , elasticity), and scaffold topographies. [ [qv: 2, 9, 10, 12, 13] ] However, there are few examples that present controlled, simultaneous regulation of these biologically relevant cues in order to influence stem cells at the more complex tissue level. Namely, those systems are simplified into single differentiation behaviors without cell–cell interactions, as compared to the complicated progression of stem cell development. [ [qv: 2, 10] ] As such, new developments in the dynamic regulation of the complex behaviors of stem cells with spatiotemporal controllability will be essential in developing novel dynamic biological tissue systems, and crucial for the full realization of their diverse applications. Characteristically, stem cell‐based neural tissue engineering is an advantageous therapeutic strategy for treating neurodegenerative diseases and injuries. [ [qv: 14] ] As such, neural tissue constructs can not only replace lost cells and secrete neurotrophic factors to stimulate axonal growth, but can also be genetically manipulated preimplantation, and can thereby repair severed axons and effectively restore locomotor functionality. [ [qv: 15] ] In the developing central nervous system, neural stem cells (NSCs) dynamically mature following predetermined spatiotemporal developmental programs, and their biological characteristics greatly vary depending on the developmental stage considered. [ [qv: 16, 17] ] It has been observed that the in vitro differentiation of NSCs typically undergoes a critical morphological changing process (NSCs monolayer → aggregation → outmigration → differentiated NSCs with neurites) over a two‐week differentiation period. [ [qv: 16] ] 3D cell aggregates can closely resemble the native configuration of NSCs in vivo, which allows for direct cell–cell signaling and cell–matrix interactions when compared to 2D culture substrates or 3D scaffolds. [ [qv: 7, 18] ] Several studies have induced NSC aggregation using a cell spheroid technique. However, the generation of conventional cell spheroids typically involves a complicated culturing process, which has been criticized for being sensitive to cell culture conditions, and often results in inconsistently differentiated cells, even between experimental trials. [ [qv: 19, 20] ] Moreover, the cell aggregation achieved using these methods largely leads to ineffective induction of differentiation, which fails to replicate the morphological and physiological features of NSCs at different developmental stages. [ [qv: 20, 21] ] To address this challenge, we propose an integrated strategy which combines printing and imprinting to create a more complex change in the cellular microenvironment, where 4D spatiotemporal cues are able to replicate the topographical and mechanobiological environment of native NSCs. 4D fabrication is a highly innovative, next generational additive manufacturing process which can be used to fabricate predesigned self‐assembly structures with a time‐dependent dynamic shape change. [ [qv: 22, 23] ] Within the diverse 4D mechanisms, shape memory polymers (SMPs) have attracted particular attention owing to their reversible “temporary‐permanent” thermomechanical reprogramming characteristics. [ [qv: 24, 25] ] Based on our previous experience in 4D fabrication, [ [qv: 24, 26, 27] ] it was expected that the addition of the 4th dimension (time) would benefit the field of neural tissue engineering. This is because the dynamic 4D effect may better mimic the unique differentiation microenvironments of neural tissue and provide a potential method for recreating the different neurodevelopmental stages undergone by NSCs. With the objective to control and manipulate the differentiation behaviors and functions of stem cells, we designed and fabricated a novel 4D substrate as a new culture platform for providing desirable dynamic extracellular microenvironments. As a model, a 4D neural culture substrate was developed to replicate NSC and neural cell development at different stages of differentiation (from NSC aggregation at an early stage to highly aligned micropatterns), as illustrated in Figure 1. To the best of our knowledge, we are the first to use a smart 4D culture substrate with a time‐dependent complex topographic transformation to replicate the dynamic process of neural development. The 4D culture system was not only able to create a unique, dynamic 3D pattern to precisely control cell–cell interactions and differentiation behaviors for stem cell biological study, but also could be integrated with biomanufacturing techniques to generate a dynamic physiological environment to induce stem cell‐based tissue regeneration. Figure 1 Schematic illustration of the fabrication procedure of a novel smart 4D neural substrate or scaffold with a time‐dependent topographic transformation, which is used to provide a new platform for modulating desired extracellular microenvironments for NSCs development at different differentiation stages (from NSC aggregation at an early stage to highly aligned micropatterns). 2 Results and Discussion 2. 1 Preparation and Optimization of Shape Memory Polymers As illustrated in Figure 1, the substrate was expected to perform a programmable‐morphing process from microwells toward an aligned pattern to facilitate the differentiation behavior of NSCs. The requirement for the material to undergo a large change in structural conformation ultimately inspired us to use SMPs as the primary 4D component of the substrate. The shape memory effect provides a programmed control over both dynamic topography and dynamic internal stress, which is released automatically from the polymer as the object is transformed between different geometrical structures. [ [qv: 22] ] The incorporation of topographical features and mechanical force to form a dynamic microenvironment may be the most effective approach to precisely guide stem cell differentiation toward the desired cell lineage. Figure 2 a shows a typical shape change process of thermoset SMPs. Chemical crosslinks in the SMPs are usually utilized to set the permanent shape, while the transition temperature ( T trans ), typically referring to the glass transition temperature ( T g ), is used to control the molecular switching segments for achieving the temporary shape. As the SMPs are heated above their T trans, the molecular switching segments are “softened” and deformation can be exerted to set the temporary shape; when the temperature is decreased below the T trans, the molecular switching segments will “freeze” to immobilize the predesigned temporary shape. The SMPs will recover their permanent shape upon returning to a temperature over T trans since the molecular switching segments are softened again, allowing the crosslink networks to revert the structure to its original shape. During this shape‐changing process of the SMPs, the chemical structure of the polymer determines the T trans and the transformation rate ( T rate ). Specifically, an appropriate ratio of stiff segments and flexible segments, and moderate crosslinking density are desired to meet various demands for different T trans and T rate. To mimic the temporal conditions of native neural development, a biomimetic neural tissue culture substrate should perform a two‐week self‐morphing transformation at 37 °C (physiological temperature or cell culture temperature) to dynamically replicate the NSC differentiation behavior. [ [qv: 16] ] Thus, by simply changing the environmental temperature, the topography can be changed dynamically and in turn a time‐dependent local force can be generated automatically on these substrates, which can be further used to direct stem cell behavior and function. Figure 2 Shape memory properties of the synthesized polymeric materials. a) Illustration of the process of shape memory effect. I, increase the temperature over transition temperature; II, exert a U shape change with enforced restriction; III, fix a temporary U shape at a lower temperature; IV, remove externally enforced restriction; V, increase the temperature to recover the original shape. b) Shape recovery of the synthesized materials. ① BP300D400, ② BP200D600, and ③ BP100D800. i) Illustration of the immobilization of the three samples which are treated with the same conditions. ii) Shape recovery process of V recorded from a side view, displaying the different recovery speeds of the three materials. The images were taken by a camera, and processed with a glowing edge effect. All the samples were bent to U shape at 60 °C and fixed a temporary shape at 23 °C. The shape recovery performed at 60 °C. Initially, we explored the shape memory properties of our synthesized materials by manipulating the components and the processing parameters of the SMPs. Five samples were synthesized with the formulations listed in the first five rows of Table S1 and Figure S1 (Supporting Information) shows the corresponding DSC results. The T g s of all samples were observed, as they underwent physical transformations from the glassy state to the rubbery state. The endothermic process of our curves illustrates that heat flows into the samples as a result of glass transition. In our design, the amount of the stiff epoxy monomer (bisphenol A diglycidyl ether, B) was kept constant, while the ratio of flexible aliphatic crosslinker ((poly(propylene glycol) bis(2‐aminopropyl) ether, P) or crosslinking modulator (decylamine, D) was changed to adjust the crosslinking density. To ensure the completion of the reaction of the epoxy, the concentration of the monoamine crosslinking modulator was controlled to adjust the crosslinking degree and to keep the total amine content equal to the epoxy content. To perform the shape transformation at 37 °C, a sample with a T g around 37 °C was designed, so that the fixing temperature and the recovery temperature were set below and above 37 °C, respectively. Among these samples, BP300D400, BP200D600, and BP100D800 exhibited T g s at 46, 40, and 25 °C, respectively. Here, BP300D400 represents that P is 0. 00300 mole and D is 0. 00400 mole when B content is kept at a constant 0. 01 moles (the details are shown in Table S1 in the Supporting Information). All of these samples were able to be softened and deformed into a U shape at ≈60 °C, and could be fixed in a temporary shape at room temperature (average ≈23 °C) (Figure 2 b‐i). As shown in Figure 2 b‐ii, the samples displayed different T rates at 60 °C in their recovery processes, due to their different T g and crosslinking density. Generally, the recovery speed of these materials is determined by the switching phase. Thus, samples with a lower T g and a higher crosslinking density contribute to a faster recovery speed ( Figure 3 a ). It can be demonstrated that the switching phase of the sample which has a lower T g is softer at the recovery temperature, which leads to faster shape recovery. Additionally, a higher crosslinking density increased the responsive efficiency of the switching phase. Figure 3 Shape memory behavior of the synthesized polymeric materials when performing the shape fix at different temperatures and the shape recovery at 37 °C. a) Illustration of the recovery speed of shape memory polymers when varying the fixing temperature and crosslinking density. b) Shape recovery process of shape memory polymers: ① sample BP300D400; ② sample BP200D600; ③ sample BP100D800, displaying the different recovery speeds at different fixing temperatures, where I) fixing the temporary shape at 60 °C and recovering at 37 °C, II) fixing at 50 °C and recovering at 37 °C, and III) fixing at 40 °C and recovering at 37 °C. c) Shape recovery process of shape memory polymers: ④ sample BP275D450; ⑤ sample BP250D500; ⑥ sample BP225D550, which were fixed at 60 °C and recovered at 37 °C. Although these samples demonstrated a typical shape recovery process, they cannot be directly used as cell culture matrices to modulate NSCs differentiation behaviors, due to their high recovery temperature and the fast recovery speeds. Therefore, we further investigated the recovery behavior of these samples at 37 °C (Figure 3b ). As expected, all samples exhibited a significantly slower recovery speed than that at 60 °C (Figure 3 b‐I). BP200D600 and BP100D800 fully recovered their original shape at three days, while BP300D400 could not completely recover at 15 days. For the neural development study, BP300D400 is too slow while BP200D600 is too fast. Additionally, another factor that influences recovery speed is the deformation temperature. As illustrated in Figure 3 a, the switching segments in the SMP are much stiffer at a lower temperature, so more energy is stored when deformed, leading to faster shape recovery. When the samples were fixed at 50 °C, the recovery speeds of samples were faster (Figure 3 b‐II). Moreover, when the recovery temperature was set at 40 °C, only BP100D800 fully recovered within 30 min, while BP300D400 and BP200D600 were unable to recover their original shapes due to their high T g s (Figure 3 b‐III). We can conclude that the effects of SMP formulation or chemical structure on shape change are much higher than that of the deformation temperature. Thus, we further developed three more samples as listed in the last three rows in Table S1 (Supporting Information), where the ratio between stiff and soft monomers was ranged from 3:4 (BP300D400) to 1:3 (BP200D600) based on the previous results in Figure 3 b. Considering that the deformation temperature does not have a remarkable effect on the shape recovery, we performed the deformation temperature at 60 °C, and investigated the recovery process at 37 °C. As shown in Figure 3 c, BP275D450 showed a recovery duration up to 2 weeks while BP250D500 and BP225D550 fully recovered their original conformations within two days. Therefore, we expected that the shape change achieved by BP275D450 could provide an appropriate timeframe for NSC aggregation and the subsequent neural cell outmigration. Moreover, the actual transformation time of BP275D450 can also be controlled by changing the deformation degree or pattern to obtain a faster recovery speed (from 3 days to 15 days). 2. 2 Fabrication of Microwell Arrays and Dynamic Cell Aggregation To investigate the transformative ability of patterning to promote the NSC aggregation, we determined the shape recovery properties of BP275D450 in the transformation from microwells to a flat surface. As shown in Figure 4 a, b, we created microwell arrays on SMP samples by imprinting stereolithography (STL) printed micropillar poly(methyl methacrylate) (PMMA) arrays via a thermomechanical reprogramming process. Since PMMA has a much higher T g (≈105 °C) than BP275D450, the microwell arrays were successfully generated after imprinting at 60 °C, and the temporary microwell arrays were immediately fixed after cooling to room temperature. The shape transformation process from microwells to a flat surface is illustrated in Figure 4c with the sample having an 800 × 800 µm well size. When incubated at 37 °C, the microwell arrays were gradually transformed, and were still visible after three days. By day 7, the SMP samples had completely restored their original flat surfaces. The dynamic recovery process of the U shape strip sample and microwell arrays are compared in Figure 4 d. The microwell arrays showed a faster recovery speed than the strip samples, which might be explained by the higher energy storage or lower deformation degree during the microwell deformation. Taking the sample fidelity and operation accessibility into consideration, we prepared all the 4D culture substrates with a thickness of 6 mm to generate the microwell arrays with a depth up to a maximum of 800 µm. Figure 4 Fabrication and characterization of microwell arrays. a) Illustration of the fabrication of microwell arrays. b) Photoimages of STL printed micropillar arrays with 800 × 800 µm pillar size (scale bar, 2. 5 mm), SMP sample (scale bar, 5 mm), and 4D microwell arrays with 800 × 800 µm well size (scale bar, 5 mm). c) Shape transformation process from microwell to flat with the sample having an 800 × 800 µm well size. When incubated in water bath at 37 °C, the microwell arrays were gradually transformed, which were still visible at three days, and almost restored its original flat surface after seven days. d) Dynamic recovery process of the strip sample and microwell array sample demonstrated by the relationship between recovery degree and incubation time. In order to evaluate the cytocompatibility of our SMPs, we performed a standard cell proliferation assay (CCK‐8 assay). The results showed that our SMP substrate without any surface treatment or modification (protein coating) had excellent cell growth after two weeks of culture, similar to the glass substrate (control) ( Figure 5 a ). Additionally, the cell morphology of the seeded NSCs was also investigated using F‐actin staining. Fluorescence images illustrated that the morphology of NSC adhesion on the nontreated SMP substrate was comparable with the glass substrate (Figure 5b ). Having established that the SMP substrates could promote cell survival and growth, we sought to explore time‐dependent neural development and the effect of 4D transformation on the neural differentiation of NSCs. To this end, we first investigated the NSC aggregating behavior in the microwell arrays during the first three days of culture. After 3 days of culture, the aggregation of GFP‐NSCs was studied by 3D confocal scanning (Figure 5c ). We further analyzed the images to better observe the aggregation feature of the NSCs. The 3D spectrum showed the fluorescent intensity distribution of the aggregates when the NSCs grew in the microwell arrays. After the 3D images were split into 20 slides of 2D images, the ortho (10/20) and (15/20) images displayed the aggregation morphology of the NSCs at the different Z positions of the microwell arrays. Figure 5 Biocompatibility of the synthesized polymeric materials and NSC aggregation behavior. a) NSC proliferation on SMP and glass samples after 1, 3, and 7 days of culture. There is no significant difference among these samples (N. S. ), suggesting our SMPs exhibited an excellent cytocompatibility in vitro. b) NSC morphology on SMP and glass samples after 1, 3, and 7 days of culture, where F‐actin is colored red and the nucleus is colored blue. Scale bar is 200 µm. c) 3D fluorescent images and image analysis of GFP‐NSCs when cells were seeded in the 4D microwell arrays. The 3D spectrum showed the fluorescent intensity distribution of aggregates when NSC grew in the microwell arrays. The ortho (10/20) and (15/20) images demonstrated the aggregation morphology of NSCs at the different Z position of the microwell arrays. The scale bar is 1 mm. d) Fluorescent images of NSCs with different cell densities on different culture substrates (flat, 400 µm, and 800 µm microwells) after 1, 3, and 7 days of culture. Dotted circles indicate the NSC aggregation in the microwell region. The scale bar is 600 µm. e) Quantitative analysis of NSC size distribution by fluorescence aggregation at 3 days. After which, we further explored the effects of cell number and microwell size on the aggregating behaviors of NSCs. The microwell arrays with two feature diameters (400 and 800 µm) were imprinted using STL‐printed micropillar PMMA arrays, and a flat sample served as a control. The NSCs with different cell densities (1 × 10 4, 3. 3 × 10 4, 1 × 10 5 cells mL −1 ) were seeded into the patterned microwells and were cultured to confluence after seven days of 4D transformation, in order to generate optimal size‐controlled aggregation. The fluorescence images showed the formation of NSC aggregates on different culture substrates, as is illustrated in Figure 5 d. The conventional aggregate formation on the flat substrate resulted in a heterogeneous distribution of cell clumps of various sizes. By the first day of culture, the 400 µm (small size) microwells yielded much smaller cell aggregates, however, the NSC density in these small wells was unable to form the uniform and dense aggregates during the 4D transformation of the substrate. In contrast, the 800 µm (large size) microwells generated uniform, compact NSC aggregates of controlled size, as we designed. Moreover, after seven days of culture, higher cell densities (both 3. 3 × 10 4 and 1 × 10 5 cells mL −1 ) exhibited excellent cell confluence or monolayers around loose aggregates. The quantitative analysis of NSC size distribution was also performed by fluorescence aggregation to determine the optimal cell seeding number at 3 days of culture (Figure 5e ). Results showed that when the cell density was higher than 3. 3 × 10 4 cells mL −1, uniform NSC aggregates were formed at three days, and further continue to confluence at seven days. This typical process fully replicated the NSC behaviors (morphology and biological features) of the first four stages of in vitro neural differentiation including noninduced, initial changes, aggregation, and outmigration. [ [qv: 16] ] The data suggested that both cell density and microwell size influenced NSC aggregation behaviors, which is expected to affect the subsequent differentiation pathways. 2. 3 Fabrication of 4D Culture Neural Substrate and Dynamic Axon Guidance Several studies have demonstrated that the specific response of neuronal cells to topographical cues can result in axon guidance and the growth of transplanted neurons to enhance therapeutic effects for nerve injury repair. [ [qv: 28, 29] ] Therefore, in order to precisely control the directional growth and influence the neuronal differentiation of NSCs, the aligned topographical features were further created to direct the formation of highly aligned axons from the differentiating NSCs. Based on previous studies, it has been demonstrated that surfaces with micro‐ or nanoscale topographical patterns were able to control NSC alignment and improve neurogenesis. [ [qv: 29, 30] ] As shown in Figure 6 a, highly aligned micropatterns with a width of 100 µm were prepared on the surface of the samples. As such, a “key‐lock copied” template strategy that incorporates a 3D fused deposition modeling (FDM) printed poly(vinyl alcohol) (PVA) sacrificial mold and polydimethylsiloxane (PDMS) microgrooves was utilized to create an aligned micropattern with a width of 100 µm, which is further described in the Experimental Section. After which, the microwell arrays were created on the aligned micropatterned sample using the aforementioned “thermomechanical imprint” method. The resultant 4D culture substrate that integrated “temporary” microwell arrays and “programmable” aligned patterns were used to facilitate the time‐dependent differentiation behavior of NSCs (Figure 6b ). As shown in Figure 6 c, optical images and 3D surface plots displayed the 4D transformation behavior of the culture substrates with time. The images confirmed that the substrate surface pattern changed from “temporary” microwell arrays to “programmable” aligned patterns. Figure 6 Fabrication and characterization of 4D aligned/microwell arrays on SMPs. a) Illustration of the fabrication of microwell arrays and their time‐dependent 4D transformation. b) Illustration of time‐dependent cell patterning process during 7 days of culture. I) Reprogrammed microwell arrays and cell seeding at day 1. II) 4D shape recovery and NSC aggregation at day 3. III) Original aligned patterns and axonal alignment of the differentiating NSCs at day 7. c) Optical images and 3D surface plots of 4D SMP substrates changing from microwell arrays to aligned pattern (microgrooves) at 1, 3, and 7 days. The scale bar is 800 µm. d) Illustration of time‐dependent differentiation behavior of NSCs, demonstrating the different stages of RA induced in vitro neural differentiation of NSCs (microscopic images), and the biological features (the expression of neurogenic markers) in the differentiated NSCs after 14 days of culture. Neuronal marker: TuJ1, mature neuronal markers MAP2, and astrocytes marker GFAP. Scale bar is 200 µm. Figure 6d shows the different stages of retinoic acid (RA) induced in vitro neural differentiation of NSCs, and the biological features (the expression of neurogenic markers) in the differentiated NSCs. As mentioned above, NSCs tend to form aggregates early in the differentiation process, while neurite extension and elongation tends to occur in the later stages of development. [ [qv: 16] ] We thus hypothesized that gradually changing 4D SMP substrates could be an excellent platform to enhance the differentiation of NSCs into neurons and glial cells, and further control axonal alignment of the differentiating NSCs. Our immunostaining data demonstrated that the developing NSC aggregates became clearly out‐migrated while the well‐array gradually disappeared after seven days of differentiation. Therein, the majority of the aligned axons from the differentiated NSCs were identified by the expression of the neuron‐specific marker class‐III β‐tubulin (TuJ1) on the 4D aligned/well‐array substrates ( Figure 7 a ). The presence of the mature neuronal marker microtubule‐associated protein 2 (MAP2) and astrocytes marker glial fibrillary acidic protein (GFAP) were characterized to confirm neuronal maturation and astroglia genesis, respectively. Compared to that of the flat and 3D aligned substrates (microgrooves), the elevated expression levels of these specific neuronal markers demonstrated that the 4D aligned/well‐array substrates accelerated the differentiation of NSCs. It can be deduced that the well‐arrays improved the formation of NSC aggregates, and thus led to earlier neural differentiation. Interestingly, significant interconnected neuronal aggregates and astroglia genesis were also observed after 14 days of culture. Moreover, the quantification of total neurite length of neural differentiation on the 4D aligned/microwell arrays was also evaluated after 7 days and 14 days of culture (Figure 7b ). The aligned/microwell arrays exhibited the longest neurite length when compared to the other groups. After two weeks of differentiation (with the continuous recovery of originally aligned patterning), it was observed that the differentiating NSCs showed excellent axonal alignment on the 4D aligned/well‐array substrates (Figure 7c ). Results demonstrated that the engineered microenvironment consisting of 4D dynamic transformation and microtopographical feature provides instructive physical cues that lead to enhanced neural differentiation of NSCs along with significant axonal alignment. Figure 7 NSC differentiation studies on 4D aligned/microwell arrays on SMPs. a) Immunofluorescent images of NSC differentiation on 4D Aligned/microwell arrays compared to flat samples and samples with alignment only, after culturing in differentiation medium for 2 weeks. GFAP and MAP2 are colored green, TuJ1 is colored red, and nuclei are colored blue. The scale bar is 500 µm. The dotted squares in TuJ1 images indicate the enlarged area, which is used to generate the images of neurite tracer (scale bar is 200 µm). The dotted circles indicate the NSC aggregation in microwell. The yellow dotted arrows show the direction of aligned patterns (microgrooves). b) Quantification of total neurite length of neural differentiation 4D aligned/microwell arrays when compared to other corresponding groups at 7 days and 14 days. Data are mean ± standard deviation; n = 9; * p < 0. 05. c) The neurite direction of the microgrooves was set as an angle of 0° (horizontal direction was set as 0° for the flat control samples). The neurite major axis with respect to the direction of the microgrooves (or horizontal direction for flat control) was defined as the neurite orientation. The neurites were considered to be aligned if their angles fell into ±20° from the original benchmark. Next, in order to quantify the expression levels of these neuronal markers, we further performed real‐time quantitative polymerase chain reaction (rt‐qPCR) analyses of neural differentiation of NSCs on 4D aligned/well‐array substrates and compared them to NSCs differentiated on only aligned, flat, and tissue culture plate (TCP) substrates. Generally, the gene expression of TuJ1, MAP2, and GFAP within differentiating NSCs increased with time, whereas the undifferentiated NSC marker Nestin decreased for all culture substrates. The expression levels of neuronal and axonal markers were upregulated on aligned and 4D aligned/well‐array substrates compared to the control TCP and flat substrates. Moreover, we found that the differentiation of NSCs on the 4D aligned/well‐array substrates showed the highest expression levels for the neuronal markers (TuJ1 and MAP2) and the astrocyte marker (GFAP) ( Figure 8 ). It can be concluded that the combined effect of having microwells and an aligned pattern on a single substrate yields significantly enhanced neural differentiation and remarkable alignment of differentiated NSCs. Therefore, it has been demonstrated the dynamic 4D effect may better mimic the special growth microenvironments of neural tissue, and provide a potential method for modulating different developmental stages of NSCs, and can thus hasten the functional recovery of injured neural tissues. Figure 8 Gene expressions of NSCs on 4D aligned/microwell arrays when compared to TCPs, flat samples, and samples with aligned only for 2 weeks of culture. The data were normalized to the expression levels of cells on TCPs. Data are reported as mean ± standard deviation, n = 6, * p < 0. 05, ** p < 0. 01. In this study, the results indicate that our 4D self‐morphing substrate provides a cutting‐edge technology to manipulate cell functions and fates, and can thus serve as the platform by which to develop a new generation of commercially available scaffolds and devices. This concept of dynamic culture can be extended to other bioengineering systems. These novel 4D tissue culturing substrates can be used to mimic the progression of specific diseases, which can thereby advance life science research as a whole. In addition to functioning as a platform by which to make dynamic cell culture substrates, the 4D SMPs also have utility for application in tissue regeneration, biomechanics research, bio‐robotics, and cancer therapy. 3 Conclusion A 4D dynamic culture substrate was developed for enhancing the growth and differentiation of neural stem cells. The smart 4D shape transformation was able to facilitate neural tissue development from cellular aggregates to highly aligned axons, which replicates the physiological characteristics of NSC‐derived neural development. It is expected that this project will have a far‐reaching impact on not only neural tissue regeneration, but also other bioengineered tissues. 4 Experimental Section Preparation and Characterization of Shape Memory Polymers Poly(propylene glycol) bis(2‐aminopropyl ether), bisphenol A diglycidyl ether, and decylamine were mixed homogeneously in a glass beaker at room temperature. The mixture was then centrifuged at 1500 rpm for 3 min to remove bubbles. The ratio of the components was varied to determine the properties of the final constructs. The 4D ink was precured at 100 °C for 1. 5 h, and finally cured at 135 °C for another 1. 5 h. The shape memory properties of the solidified materials were characterized according to our previous methods. [ [qv: 26, 31] ] Briefly, the sample strips were folded 180° into a “U” shape at different temperatures with an inner radius of 10 mm, and were kept in that conformation for 10 min. The strips were then immediately cooled to room temperature for an additional 10 min in order to obtain a temporary shape. The “U” shaped strips were then immersed in the water bath at 37 °C to recover the permanent shape. The dynamic shape change (recovery speed) of different models was recorded with a PowerShot ELPH 360HS Cannon camera. The images were processed using glowing edges filter to enhance the object's border. Fabrication of Programmable Culture Substrates Three different printing techniques, including FDM, extrusion, and STL, were used to fabricate the programmable culture matrices. First, a cylinder sacrificial mold with a diameter of 12 mm and a height of 8 mm was designed with the software Autodesk123D (Autodesk Inc, USA), and the stl. formatted file was then loaded into the software Slic3r which is licensed under the GNU Affero General Public License, version 3. The infill density, the printing speed, and the layer height were assigned in Slic3r. The predesigned structures were then printed via a Solidoodle 3D FDM printer platform with a nozzle size of 100 µm. Open‐source software (Prontrface) was utilized to control the three stepper motors with an effective resolution of 100 µm in the x ‐ and y ‐axis, and a minimum layer height of 50 µm in the z ‐axis. A PVA filament used for 3D printing was obtained from Matter Hackers (USA). The PVA filament with a diameter of 1. 75 mm was used as the sacrificial mold material, and the printing temperature was set at 190 °C. The PVA sacrificial mold was printed with a layer height of 100 µm to generate the aligned surface structure. The mold was then coated with PDMS and cured at 70 °C for 2 h in order to obtain PDMS microgrooves with a width of 100 µm. Next, preset amounts of the 4D ink materials were then extruded into the PDMS mold, and were then cured using the process described above. After the samples had cooled to room temperature, the aligned micropatterned scaffolds were taken out of the oven and were ready for additional operations. Finally, microwells were generated on our flat and aligned scaffold to improve the NSC spheroid formation at an early stage of neural development. A cubic mold 9 mm ( L ) × 6 mm ( W ) × 2 mm ( H ) with micropillars (800 and 400 µm in diameter) was designed with the software Autodesk 123D, and saved as an stl. format file. The models were uploaded into a desktop STL printer (Formlabs, USA) to slice the digital model into layers for printing. Two different diameters of micropillar templates, 800 and 400 µm, were printed with PMMA resin for the fabrication of hard molds. The speed of operation was set to 25 mm h −1, the XY resolution was 50 × 50 µm, and the layer height was 50 µm. After printing, a thermomechanical reprogramming process was performed to create the temporary microwells on our 4D scaffolds as follows: the 4D scaffolds were imprinted by a micropillar mold with a clamp at 60 °C for 10 min, and then moved to ice water for 10 min. Additionally, the 4D transformation behavior of the culture substrate as a function of time was captured by optical microscopy, and 3D surface plots were created with AmScope 3. 7 software. Cell Culture, Proliferation, and Morphology NSCs cloned from mouse neuroectoderm (NE‐4C) were purchased from American Type Culture Collection (ATCC). NSCs (passage no. 3–6) were cultured in Eagle's minimum essential medium (ATCC) supplemented with 5% fetal bovine serum (FBS), 1% (v/v) l ‐glutamine, and 1% penicillin/streptomycin solution, under standard cell culture conditions (37 °C, a humidified, 5% CO 2 /95% air environment). NSCs were seeded on glass and SMPs (without any surface treatment or modification) at a density of 5 × 10 4 cells mL −1 and continuously cultured for 1, 3, and 7 days. At the predetermined time interval, culture medium containing 10% CCK‐8 solution (Dojindo, Japan) was added and incubated for 2 h. 200 µL of the medium was transferred into a 96‐well plate, and the absorbance of the incubated solution at a wavelength of 450 nm was quantified by a spectrophotometer (Thermo, USA). The NSC morphology was evaluated by F‐actin staining. At each predetermined time, all samples were fixed with 10% formalin for 15 min and then permeabilized with 0. 2% Triton‐100 for 10 min. The samples were then stained with a Texas Red‐X phalloidin solution (1:100) to stain the cells' cytoskeleton for 30 min, followed by 4′, 6‐diamidino‐2‐phenylindole (DAPI) (1:1000) solution to stain the cells' nuclei for another 5 min. The images were observed using laser confocal microscopy (Carl Zeiss LSM 710). Cell Aggregation Evaluation The green fluorescent protein transfected NSCs (GFP‐NSCs, NE‐GFP‐4C, ATCC) were seeded on the microwell arrays of the 4D substrate at a density of 5 × 10 4 cells mL −1 and cultured for 3 days. The 3D fluorescent images were taken by confocal microscopy and analyzed by Zen software (Zeiss). Additionally, in order to optimize the formation of NSC spheroids, NSCs were seeded on the scaffolds at different densities of 1 × 10 4, 3. 3 × 10 4, 1 × 10 5 cells cm −2, and cultured for 7 days under standard cell culture conditions. At each predetermined time, the cells were fixed with 10% formalin for 15 min and permeabilized in 0. 1% Triton X‐100 for 10 min. The cells were stained with Texas red fluorescent dye for 30 min and then DAPI blue fluorescent dye for 5 min to observe NSC aggregation on the scaffolds. The double‐stained samples were imaged on the confocal microscope. The aggregation distribution of NSCs at 3 days was quantified by Image J analysis software (National Institutes of Health). After setting a scale bar, the distribution area was adjusted using threshold settings, and then was measured. Six visible areas were randomly selected for statistical analysis on each sample; there were five samples in each group. Neurogenic Differentiation The neurogenic differentiation of NSCs was performed using our reported method with modifications. The NSCs were seeded on various scaffolds at 3 × 10 4 cells cm −2 and maintained in the growth medium for 24 h. To induce the neurogenic differentiation, the samples were cultured in the neurogenic medium, which consisted of the growth medium supplemented 10 −6 m retinoic acid (RA), and the medium was exchanged every other day. Immunofluorescence Staining Neurogenic differentiation of NSCs was identified using immunofluorescence staining. After incubation with neurogenic differentiation medium for 7 and 14 days, the cells were fixed with 10% formalin for 15 min, and were then treated with 0. 1% Triton X‐100 for 10 min. Then the samples were incubated with a blocking solution (containing 1% bovine serum albumin (BSA), 0. 1% Tween 20 and 0. 3 m glycine in PBS) for 2 h. The first primary antibodies of mouse anti‐TuJ1 (1:1000), rabbit anti‐GFAP antibody (1:500) and rabbit anti‐MAP2 antibody (1:500) were gently mixed with samples overnight at 4 °C. Next, the secondary antibodies of goat antimouse Alexa Fluor 594 (1:1000) and goat anti‐rabbit Alexa Fluor 488 (1:1000) were incubated with samples in the dark for 2 h at room temperature, followed by DAPI (1:1000) solution incubation for 5 min. The immunofluorescence images were taken using confocal microscopy. The average total neurite length and orientation were quantified by Image J analysis software (NeuriteTracer). Three visible areas were randomly selected for quantifying statistical analysis on each sample; there were three samples in each group. The direction of the microgrooves was set as an angle of 0° (horizontal direction was set as 0° for the flat control samples). The neurite major axis with respect to the direction of the microgrooves (or horizontal direction for flat control) was defined as the neurite orientation. The neurites were considered to be aligned if their angles fell into ±20° from the original benchmark. Real‐Time Quantitative Polymerase Chain Reaction The neurogenic gene expression of all samples, including neuron‐specific class III β‐tubulin (TuJ1), MAP2, GFAP, and Nestin, were analyzed by the rt‐qPCR assay. Briefly, the total RNA content was extracted from the samples using Trizol reagent (Life Technologies). The RNA quality and concentration were determined from the absorbance at 260 and 280 nm with a microplate reader. RNA samples were reverse‐transcribed to cDNA using a Prime Script RT reagent Kit (TaKaRa). RT‐PCR was then performed on a CFX384 Real‐Time System (BIORAD) by using SYBR Premix Ex Taq (TaKaRa) according to the manufacturer's protocol. The gene expression level of target genes was normalized against the housekeeping gene glyceraldehyde 3‐phosphate dehydrogenase (GAPDH). The relative gene expression of the samples was normalized against the control group to obtain relative gene expression fold values and calculated via the 2‐delta delta (2 −ΔΔCt ) cycle‐threshold method. Primer sequences are as follows: TUJ1, forward primer 5′‐AGCTGTTCAAACGCATCTCG‐3′ and reverse primer 5′‐GACACCAGGTCATTCATGTTGC‐3′; MAP2, forward primer 5′‐TTCTCCACTGTGGCTGTTTG‐3′ and reverse primer 5′‐GAGCCTGTTTGTAGACTGGAAGA‐3′; GFAP, forward primer 5′‐CCTTCCTTCCCTGGTTTTCT‐3′ and reverse primer 5′‐TGCTCATCTTTCCTCTTCCC‐3′; Nestin, forward primer 5′‐GTGGCCTCTGGGATGATG‐3′ and reverse primer 5′‐TTGACCTTCCTCCCCCTC‐3′; GAPDH, forward primer 5′‐GTGGCCTCTGGGATGATG‐3′ and reverse primer 5′‐ACTCCTCAGCAACTGAGGG‐3′. Statistical Analysis The mean and standard deviation was plotted for each sample group ( n = 6). Then, a one‐way analysis of variance (ANOVA) ( p < 0. 05) with a posthoc Tukey honestly significant difference test was performed on each set of data. The statistical significance was indicated with an asterisk. That is, samples connected with an asterisk were significantly different. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902420
| 2,019
|
Advanced Science
|
Artificial Inclusion Bodies for Clinical Development
|
Abstract Bacterial inclusion bodies (IBs) are mechanically stable protein particles in the microscale, which behave as robust, slow‐protein‐releasing amyloids. Upon exposure to cultured cells or upon subcutaneous or intratumor injection, these protein materials secrete functional IB polypeptides, functionally mimicking the endocrine release of peptide hormones from secretory amyloid granules. Being appealing as delivery systems for prolonged protein drug release, the development of IBs toward clinical applications is, however, severely constrained by their bacterial origin and by the undefined and protein‐to‐protein, batch‐to‐batch variable composition. In this context, the de novo fabrication of artificial IBs (ArtIBs) by simple, cell‐free physicochemical methods, using pure components at defined amounts is proposed here. By this, the resulting functional protein microparticles are intriguing, chemically defined biomimetic materials that replicate relevant functionalities of natural IBs, including mammalian cell penetration and local or remote release of functional ArtIB‐forming protein. In default of severe regulatory issues, the concept of ArtIBs is proposed as a novel exploitable category of biomaterials for biotechnological and biomedical applications, resulting from simple fabrication and envisaging soft developmental routes to clinics.
|
Bacterial inclusion bodies (IBs) are water‐insoluble proteinaceous inclusions generated in the cytoplasm of recombinant bacteria, 1, 2 stabilized by an amyloid fibril architecture that confers mechanical robustness. 3 They are formed by the transgene protein product plus a diversity of residual macromolecules from bacterial cells, including nucleic acids, carbohydrates, proteins, and cell wall components. Their morphology is defined by mechanical limitations imposed by the bacterial cell wall, usually at the edge of sub‐micrometer and micrometer sizes. The biological activity associated to IB proteins, 4 together with the mechanical stability and high porosity of these protein particles has pushed to re‐evaluate them as unconventional functional materials with a wide spectrum of applications in biotechnology and biomedicine. 5 Any functional polypeptide suitable for production in bacteria can be engineered to be packaged in form of IBs. 6 This is because IBs show a complex structure with a dual organization of the forming polypeptide chains; while around 40% of the IB protein generates a mechanically stable fibril network, the remaining fraction represents functional or quasi‐functional species embedded in such stable structure. 7 This protein population is properly folded, nearly soluble, releasable under physiological conditions, and responsible for the biological activity of IBs. 1 As self‐immobilized catalysts, enzyme‐based IBs show high level of operational stability and reusability. 8 In tissue engineering, they have been adapted as nontoxic topologies that provide a combination of mechanical and biological stimuli for controlled cell proliferation. 9 In a different context, IBs have been tailored as unexpected drug delivery systems or “nanopills, ” 10 that mimic the functionalities of the secretory granules of the mammalian endocrine system 11 for the intracellular, local or remote delivery of functional (either untargeted or receptor‐targeted) IB protein. 12 Therefore, IBs, as functional biomaterials, show promise in protein replacement therapies and in any clinical uses aimed to systemic, local or precision protein delivery. The enormous clinical potential of bacterial IBs is, however, darkened by their heterogeneous and undefined composition. 6 Trapping an indeterminate catalogue of bacterial cell materials, clinically oriented IBs would hardly overcome the severe regulatory constraints imposed by medicament agencies. In this context, we wondered whether chemically pure IB mimetic particles (artificial IBs, ArtIBs) could be fabricated in vitro. This would imply the production of particulate microscale protein materials, made of pure and chemically controlled components, which would replicate those functionalities of natural IBs that are relevant to protein delivery. These properties are mechanical stability, absence of intrinsic cytotoxicity, mammalian cell penetrability, and the ability to release functional IB protein upon physiological conditions. Pure protein composition would be added to such functionally rich profile. To explore the fabrication of synthetic IBs, common laboratory enzymes that form functional IBs during recombinant production, namely, β‐galactosidase (β‐Gal) and alkaline phosphatase (AP), 13 were selected as models for two alternative approaches to ArtIB fabrication. In one ( Figure 1 a), soluble protein was salted out plus thermally aggregated to generate amyloidal networks, further used as IB‐like seeds to recruit and entrap homologous soluble protein versions. This multiple step procedure (ms), was developed to imitate the dual, sponge‐like networks in natural IBs. 7 Lipids, common component of bacterial IBs were also incorporated. In a simpler single‐step (ss) approach, divalent cations (Zn, in form of ZnCl 2 ), involved in amyloid formation 14 and generically, in protein–protein contacts, 15 were added to a protein solution (Figure 1 a). The application of these procedures resulted in mechanically stable, discrete, and moderately disperse protein particles sizing around 1–2 µm (AP) and 2–6 µm (β‐Gal) (Figure 1 b, c), whose surface rugosity, size variability, and amorphous appearance remembered those of natural IBs (Figure 1 b). Both enzymes, in this packaged form, were enzymatically active (Figure 1 d). On the other hand, cross‐β‐sheet amyloid like structure (ALS) was detected by attenuated total reflectance (ATR) in AP ArtIBs at proportions (31–37%, ss and ms, respectively) matching those found in IBs. 7 Figure 1 Fabrication of ArtIBs. a) Multiple (ms) and single (ss) step procedures for ArtIB fabrication from soluble pure protein are summarized, indicating the main operational steps (arrows). OS is organic solvent. Precise details can be found in the Experimental Section. Final products are framed. b) Representative field‐emission scanning electron microscopy (FESEM) images of AP and β‐Gal ArtIBs. Magnifications are equivalent in all images. c) Dynamic light scattering (DLS) size analyses of ArtIBs, indicating the mode (in nm) and the polydispersion index (pdi). d) Specific activity of both AP and β‐Gal ArtIBs, compared to that of commercial soluble protein counterpart. Asterisks indicate statistically different from the specific activity of the soluble protein ( p < 0. 001, Holme–Sidak test). Further, we constructed new ArtIBs ( Figure 2 a) formed by the self‐assembling modular proteins T22‐GFP‐H6 16 and T22‐PE24‐H6, 17 that are targeted to the cell‐surface cytokine receptor CXCR4 18 through the N‐terminal tumor homing peptide T22. 19 The size of these materials was smaller than that of those formed by the previously tested enzymes (Figure 1 ), a fact that, considering potential medical applications, could generally favor molecular and cell interactivity and protein release, by increasing the surface/volume ratio. When exposed to cultured CXCR4 + Hela cells, T22‐GFP‐H6 ArtIBs internalized very efficiently as in the case of IB‐based nanopills, 10 by a CXCR4‐dependent route that is inhibited by the CXCR4 antagonist AMD3100 20 (Figure 2 b). Cell viability was not affected by T22‐GFP‐H6 ArtIBs (Figure 2 c), but it was instead dramatically compromised, in a CXCR4‐dependent fashion, by the Pseudomonas aeruginosa exotoxin (PE24) contained in T22‐PE24‐H6 ArtIBs. As in the case of IBs, ArtIBs steadily released a fraction of the forming protein in soluble form when incubated in physiological buffer, at least for 7 days (Figure 2 d). T22‐GFP‐H6 solubilized in vitro from ArtIBs was fluorescent (1039. 83 AU mg −1 ), it assembled as 13 nm nanoparticles indistinguishable in size from soluble T22‐GFP‐H6 (Figure 2 e), and this material was equally able to penetrate cultured HeLa cells in a CXCR4‐dependent way (Figure 2 f). This fact unveiled a potential of ArtIBs as chemically homogenous protein reservoirs for prolonged in vivo delivery of tumor‐targeted, nanostructured protein drugs. Figure 2 Characterization of ArtIBs formed by modular proteins. a) FESEM images of CXCR4‐targeted ArtIBs, all recorded at the same magnification. At the bottom of each image, specific fluorescence decay (SFD), hydrodynamic size peak (pdi ± s. e. m. ) and percentage of ALS are shown. b) Internalization of T22‐GFP‐H6 ArtIBs in cultured HeLa cells, recorded at different times after exposure through intracellular green fluorescent protein (GFP)fluorescence (top). Bottom: AMD3100‐mediated inhibition of ArtIB internalization. c) Viability of cultured HeLa cells upon 96 h of T22‐GFP‐H6 and T22‐PE24‐H6 ArtIB exposure in presence or absence of AMD3100. d) Stain‐free protein detection of released soluble protein (r) from ArtIBs, 7 days after incubation in physiological buffer. In the plot, kinetics of soluble protein release from T22‐GFP‐H6 ssArtIBs. e) Hydrodynamic mode size peak of T22‐GFP‐H6 nanoparticles released from ssArtIBs, compared to equivalent soluble nanoparticles after purification from recombinant bacteria (those used for ArtIB fabrication). In the inset, an FESEM image of those nanoparticles released from ArtIBs. f) AMD3100‐mediated inhibition of HeLa cell internalization of recombinant soluble and ArtIBs‐released nanoparticles. Symbols indicate significant differences to the control (*, p < 0. 05, Tukey test) and between samples with or without AMD3100 ( —, p < 0. 05, two tail, t ‐test). In this context, different categories of T22‐GFP‐H6 ArtIBs were implanted subcutaneously (SC) in a CXCR4 + colorectal cancer mouse model, releasing fluorescent material from the implantation point, followed by selective uptake by a remote CXCR4 + tumor, with specific kinetics for each ArtIB type. A preliminary screening of T22‐GFP‐H6 msArtIBs and T22‐GFP‐H6 ssArtIBs (Zn 2+, at 100:1 ratio of zinc to protein) showed slow release and negligible or a small amount of material accumulated in the tumor by 21 days ( Figure 3 a). Lowering the proportion of Zn 2+ (50:1) or using alternative cations to induce ssArtIBs formation, improved protein release and tumor uptake. In particular, ss ArtIBs formed by Ca +2 were more efficient than ArtIBs (Zn +2 50:1) in maintaining a faster and progressive protein release from the SC injection site leading to a higher accumulation and longer residence time (starting at day 3 and at least until day 10) in the remote CXCR4 + tumor (Figure 3 b). Further, T22‐PE24‐H6 ArtIBs Ca 2+, containing the CXCR4‐targeted cytotoxic polypeptide PE, induced an ( p = 0. 083) inhibition of tumor growth ( n = 3, 1. 0 ± 0. 2 × 10 8 ) stronger than T22‐GFP‐H6 Ca 2+ ArtIBs ( n = 3, 1. 5 ± 0. 7 × 10 8 ), as compared to the control buffer‐treated group ( n = 2, 2. 6 ± 1. 0 × 10 8 ) (Figure 3 c). This fact occurred in absence of systemic toxicity (lack of histopathological alterations in hematoxylin and eosin (H&E)‐stained liver and kidney at the end of the experiment, not shown). These observations fully confirmed both the secretion‐like prolonged protein release and the precise cell targeting of functional materials through the blood stream, from a remote location. Figure 3 ArtIBs material release, tumor uptake and antitumor activity in a CXCR4 + colorectal cancer model. a) Preliminary screening of released material and tumor uptake after subcutaneous implantation of T22‐GFP‐H6 msArtIBs, T22‐GFP‐H6 ssArtIBs (Zn 2+ 100:1) or PBS. b) Representative FLI images obtained at the injection point (IP) and at the remote tumor (T), along time (day 0, 3, 6, and 10) after T22‐GFP‐H6 ssArtIBs Ca 2+, T22‐GFP‐H6 ssArtIBs Zn 2+ (1:50) or buffer SC administration. c) Antitumor effect, measured as bioluminiscence emission by cancer cells along time, in the CXCR4 + SW1417‐luci tumor model, after SC injection of 1 mg dose per mouse of T22‐GFP‐H6 Ca 2+ ArtIBs, T22‐PE24‐H6 Ca 2+ ArtIBs or control PBS (*, p < 0. 05, Tukey test). Fluorescence in (a) and (b), or bioluminescence in (c) intensity were measured using IVIS Spectrum and expressed as x‐ ± SE of average radiant efficiency. In summary, ArtIBs can be fabricated in vitro as a new type of biomimetic material, from pure protein and by simple physicochemical methods. These protein particles reproduce IB properties that are relevant to potential uses in biomedicine, especially protein release, but their potential use would fully prevent the immunotoxic reactions potentially associated to the administration of IBs, that contain bacterial debries at variable but significant proportions. On the other hand, since protein drugs used in clinics have a human origin, 21 their administration in an ArtIBs format is not expected to pose significant immune concerns, or to enhance any putative immune reaction over those associated to the repeated, conventional administration regimens of soluble protein drugs (such as insulin, interferons, and many others). In particular, the simpler ss fabrication method allows engineering the strength of protein–protein interactions in the material by means of the stoichiometric control of metal or nonmetal divalent cations. In contrast to other excellent and biocompatible materials developed as micro‐ or nanoparticles for the slow and sustained drug release, 22 such as those based on PLA or PLGA, 23 ArtIBs are chemically homogeneous and show no chemical distinction between carrier and cargo, thus acting as self‐contained, self‐released drug materials. Then, the protein drug itself acts, in addition, as a scaffold material, what results in intriguing, totally novel and chemically homogenous drug delivery systems with simpler fabrication processes as opposite to hybrid platforms. ArtIBs might not only replace IBs as functional protein reservoirs and offer homogeneous materials for drug‐oriented development, but they will enable, in addition, the packaging of glycosylated proteins of mammalian cell origin as IB‐like materials. Since these proteins would be never produced in bacteria in functional forms, ArtIBs will then expand, as a universal platform, the catalogue of enzymes or protein drugs that could be formulated as pure microscale biocatalysts or as secretory protein granules. Experimental Section Fabrication of ArtIBs : To produce msArtIBs, 1 mg of pure soluble protein was denatured and concomitantly precipitated by heating at 100 °C in NaCl 2 (500 × 10 −3 m ), ZnCl 2 (26. 4 × 10 −3 m ), and MgCl 2 (18. 4 × 10 −3 m ) in distilled H 2 O. The precipitate was centrifuged at 15 000 × g for 15 min at 4 °C, isolated from the soluble fraction and resuspended with 1 mg of phosphatidylcholine with chloroform/methanol 2:1 v/v in a final volume of 300 µL. The excess of organic solvent (OS) was removed by a continuous N 2 flow inducing the formation of a protein–lipid film phase that acted as scaffold. The scaffold was afterwards resuspended in 1 mg mL −1 of previous soluble protein diluted in phosphate buffered saline (PBS) at 4 °C overnight. Finally, the sample was centrifuged, and soluble fraction discarded. The manufacturing of ssArtIBs was approached by diluting pure soluble protein in distilled H 2 O at a final concentration of 2 mg mL −1 and final volume of 200 µL. Protein samples (0. 196 × 10 −3 m ) were subsequently mixed with ZnCl 2, at a 100:1 ratio of zinc to protein. After 10 min of incubation at room temperature, samples were centrifuged at 15 000 × g for 15 min and soluble fraction discarded to obtain the final product. Alternatively, zinc at a ratio 50:1 and calcium at a ratio 300:1 (in form of CaCl 2 ) were used for the in vivo experimental. β‐Gal [EC3. 2. 1. 23] and AP [EC3. 1. 3. 1], both from Escherichia coli, were purchased from Sigma‐Aldrich. T22‐GFP‐H6 and T22‐PE24‐H6 were produced as recombinant proteins and purified by single step chromatography as reported. 17 Determination of Enzymatic Activity : Between 3. 7 and 21. 3 ng of pure soluble β‐Gal protein or β‐Gal ArtIBs were mixed with 5 × 10 −3 m of ortho ‐nitrophenylgalactopiranoside in a final volume of 500 µL of PBS. The mixture was incubated for 15 min at 37 °C, the reaction stopped by adding 200 µL of Na 2 CO 3 (2. 8 m ) and the product amount determined by measuring absorbance at 420 nm (ε 420 = 4530 m −1 cm −1 ) in a UV–vis spectrophotometer (Ultrospec 1000E, Pharmacia Biotech). On the other hand, between 3. 9 and 92 ng of pure soluble AP protein or AP ArtIBs were mixed with 20 × 10 −3 m of para ‐nitrophenylphosphate (pNPP) in a final volume of 500 µL of PBS. The mixture was incubated for 15 min at 37 °C, the reaction stopped by adding 200 µL of NaOH (1 m ) and activity determined by measuring p‐nitrophenyl phosphate absorbance at 405 nm (ε 405 = 18 000 m −1 cm −1 ) in a UV–vis spectrophotometer (Ultrospec 1000E, Pharmacia Biotech). Determination of Specific Fluorescence : Pure soluble T22‐GFP‐H6 or ArtIB versions were diluted in PBS at concentrations ranging from 0. 2 to 1 mg mL −1. The excitation wavelength (λ ex ) was set at 488 nm and the emission (λ em ) at 510 nm, meanwhile the excitation slit was set at 2. 5 nm and the emission slit at 5 nm. Fluorescence was measured in a Cary Eclipse Fluorescence Spectrophotometer (Agilent Technologies) by using a quartz cell with a 10 mm path of light. The intrinsic fluorescence of each sample was then represented referred to protein concentration, defining the SFD mathematically represented as a slope. The % of SFD (%SFD) represents the relationship of the parameter with the SFD of soluble T22‐GFP‐H6 protein. Size Distribution Analysis : Volume size distribution of all nanostructures was determined at 633 nm and 25 °C in a Zetasizer Nano ZS (Malvern Instruments Limited) by using ZEN2112 3 mm quartz batch cuvettes. Protein samples dissolved in PBS from 0. 2 to 1 mg mL −1 were measured in triplicate and mode size peak and polydispersion index (pdi ± s. e. m. ) obtained. Electron Microscopy : Ultrastructural morphometry (size and shape) of ArtIBs was characterized at nearly native state with field emission scanning electron microscopy (FESEM). Drops of 20 µL of each sample diluted at 0. 3 mg mL −1 in their respective buffers were directly deposited on silicon wafers (Ted Pella Inc. ) for 30 s and immediately observed without coating with an FESEM Zeiss Merlin (Zeiss) operating at 1 kV and equipped with a high resolution secondary electron detector. Representative images of a general fields and nanoparticle detail were captured at magnifications ranging between 5500× and 8500× and a working distance of 3. 5 mm. Attenuated Total Reflectance : The most suitable concentration of ArtIBs was placed and dried with a continuous N 2 flow on spectroscopic crystal surfaces. Total reflectance spectroscopy was detected 15 times as spectra by using a scan rate of 50 cm −1 min −1 and a nominal resolution of 2 cm −1 in a Tensor 27 Bruker spectrometer coupled to a Specac Golden Gate ATR accessory. All measurements were performed at 25 °C, the absorbance obtained was corrected against the background and the PBS buffer signal was subtracted. Fourier deconvolution of the spectra and the second derivative allow the identification of the different band components. Fitting of the components to the original (not deconvolved) spectrum was essentially performed according to a described procedure. 24 Peak height, band width, and peak position of the components were allowed to vary one at a time in this order. A Gaussian shape was assumed. Cell Culture : CXCR4 + cervical cancer cell lines (HeLa ATCC‐CCL‐2) were used to study the performance of ArtIBs in vitro. Cells were routinely cultured in Eagle's minimum essential medium (Gibco), supplemented with 10% fetal bovine serum (FBS, Gibco) and incubated in a humidified atmosphere at 37 °C and 5% of CO 2. Protein Internalization : HeLa CXCR4 + cells were cultured in 24‐well plates in MEM Alpha 1× GlutaMAX medium (Gibco) supplemented with foetal bovine serum (FBS) at 37 °C in a 5% CO 2 humidified atmosphere until 70% of confluence was reached. The medium was then exchanged for serum free OptiPro medium (Gibco) before the addition of the protein. Protein uptake was determined at different times ranging from 10 min to 24 h at a final concentration of 2. 5 µg. Cells were detached, and external hooked protein removed by Trypsin‐EDTA (Gibco) at 1 mg mL −1 exposure for 15 min at 37 °C. Intracellular protein fluorescence was detected by flow cytometry using a FACS‐Canto system (Becton Dickinson) with an air‐cooled argon ion laser (15 mW) exciting at 488 nm and a D detector (530/30 nm as band pass filter). In addition, the internalization specificity through CXCR4 receptor was tested by exposing cells to the CXCR4 antagonist AMD3100 25 1 h prior protein incubation at (protein/AMD3100) 1:10 ratio. Cell Viability : HeLa (ATCC‐CCL‐2) cell line was cultured in opaque‐walled 96‐well plates at a final concentration of 6000 cells per well for 24 h. MEM Alpha GlutaMAX medium (Gibco) supplemented with FBS was used at 37 °C in a 5% CO 2 humidified atmosphere, until 70% of confluence was reached. ArtIBs were incubated at 1 × 10 −6 m for 96 h using MEM Alpha GlutaMAX medium (Gibco). Cell viability was measured by CellTiter‐Glo Luminescent Cell Viability Assay (Promega) in a Multilabel Plater Reader Victor3 (Perkin Elmer). Soluble Protein Release from Artificial IBs : ArtIBs were resuspended in 1 mL of PBS 1× reaching a final concentration of 1 mg mL −1 and incubated at 37 °C without agitation. 100 µL were taken from each sample at different times ranging from 0 to 7 days and centrifuged for 15 min at 15 000 × g at 4 °C to isolate soluble and insoluble fractions. Soluble protein was then stain‐free detected by TGX (TGX FastCast Acrylamide Kit) and subsequently quantified by ImageLab software to determine the % of released protein. In Vivo Release of Fluorescent Material by Subcutaneously Implanted ArtIBs and Their Tumor Uptake : Four‐week‐old female mice of the Swiss nude strain, in the 18–20 g body weight range (Charles River, L‐Abreslle, France), maintained in pathogen‐free conditions, were used for the in vivo experiments. All experimental procedures were approved by the Hospital de Sant Pau Animal Ethics Committee and performed according to European Council directives. To generate the CXCR4 + SW1417 CRC cancer model, a 10 mg aliquot of SW1417‐luci tumor tissue was obtained from donor animals and deposited in the anterior or posterior flank subcutis of the animals. When tumors reached ≈120–200 mm 3 volume, animals were randomly allocated and implanted in the subcutis of the mouse lumbar region with a pellet of T22‐GFP‐H6 msArtIBs or T22‐GFP‐H6 ssArtIBs Zn 2+ in a preliminary study, at a single dose injection of 1 mg per mouse, suspended in a 150 µL PBS buffer, whereas in a second study, T22‐GFP‐H6 Zn 2+ ArtIBs or of T22‐GFP‐H6 Ca 2+ ArtIBs were implanted at the same dose. Control buffer injection was used as a negative control. The ArtIBs IP was selected to position it as far away as possible from the tumor in the same mouse, being located either in the anterior or posterior flanks. After ArtIBs pellet injection, the IVIS Spectrum equipment (PerkinElmer Inc. ) was used to monitor the GFP‐emitted fluorescence by the SC implants in whole‐body mouse by registering immediately (0 h) and at specific time points (3, 6, and 10 days) after the administration to determine the fluorescence remaining in the subcutaneous ArtIBs implants, as well as the fluorescent material that reached the remote tumor along time, in each mouse. Fluorescent signal was digitalized, displayed as pseudocolor overlay, and expressed as radiant efficiency. The fluorescence intensity (FLI) ratio was calculated dividing the signal from the IBs‐treated mice by the FLI autofluorescent signal of buffer‐administered control mice either in the injection point or in the tumor. In Vivo Antitumor Activity of SC Implanted ArtIBs : The CXCR4 + SW1417 CRC cancer model used to test antitumor activity was generated as described above. The expression of luciferase by cancer cells in this model allowed for the noninvasive follow‐up of tumor growth along time. A week before the deposition of the tumor aliquot in the mouse subcutis, mice were randomly allocated to be SC administered in the mouse lumbar region with 1 mg per mouse dose of T22‐GFP‐H6 Ca 2+ ArtIBs or T22‐PE24‐H6 Ca 2+ ArtIBs suspended in a 150 µL of PBS buffer or buffer‐treated control mice. After IBs administration, mouse body weight was recorded, and bioluminescent image intensity in the tumor, measured using the IVIS Spectrum equipment (PerkinElmer Inc. ), was digitalized and expressed as radiant efficiency. Tumor tissue, liver, and kidney were formalin‐fixed and paraffin‐embedded for histology. To that aim, four‐micrometer‐thick sections were stained with H&E, and analyzed for possible histological alterations by two independent observers. Representative images were taken using Cell^B software (Olympus Soft Imaging v. 3. 3). Statistical Analysis : All analyses were performed with SPSS versus version 11. 0 (IBM) software. One‐way ANOVA and t ‐tests were performed to assess differences in assays with a minimum n = 3. The Holme–Sidak test was applied for equal variance and Tukey or Mann Whitney U test for unequal variance (indicated in the figure legend). Two tail t ‐test was also used for individual comparisons. Data were presented as means ± standard error of the mean (s. e. m). Differences between the protein samples were considered significant at p ≤ 0. 05. Conflict of Interest E. V. , R. M. , and A. V. are cofounders of NANOLIGENT, devoted to develop antitumoral drugs based on proteins. J. M. S. , H. L‐L. , P. A. , E. V. , R. M. and A. V. are co‐inventors in a patent covering the use of ArtIBs.
|
10. 1002/advs. 201902701
| 2,020
|
Advanced Science
|
Crosslinking Induced Reassembly of Multiblock Polymers: Addressing the Dilemma of Stability and Responsivity
|
Abstract Physical or chemical crosslinking of polymeric micelles has emerged as a straightforward approach to overcome the intrinsic instability of assemblies. However, the crosslinking process may compromise the responsivity of nanosystems and result in inefficient release of payloads. To address this dilemma, a crosslinking induced reassembly (CIRA) strategy is reported here to simultaneously increase the kinetic and thermodynamic stability and redox‐responsivity of polymeric micelles. It is found that the click crosslinking of a model multiblock polyurethane at the micellar interface induces microphase separation between the soft and hard segments. The aggregation of hard domains gathers liable disulfide linkages around the interlayer of micelles, which could facilitate the attack of reducing agents and act as an intelligent on‐off switch for high stability and triggered release. As a result, the CIRA approach enables an enhanced tumor targeting, improved biodistribution and excellent therapeutic efficacy in vivo. This work provides a facile and versatile platform for controlled delivery applications.
|
Self‐assembled polymeric micelles have been receiving much attention for potential use as versatile drug delivery systems. [ 1, 2, 3 ] They commonly consist of hydrophobic cores for protecting and solubilizing therapeutics and hydrophilic shells for improved colloidal stability and stealthy character. [ 4 ] To date, a good variety of polymer micelle‐based nanomedicines (Genexol‐PM, [ 5 ] NC‐6004, [ 6 ] NK105, [ 7 ] NK012, [ 6 ] and NK911, [ 8 ] etc. ) have been approved by Food and Drug Administration (FDA) or entered the clinical trials. Disappointingly, their therapeutic effect does not meet the expectations, as evidenced by the limited clinical outcome. [ 9 ] A principal challenge is that most polymeric assemblies cannot withstand the massive dilution and competing interactions with blood components due to their dynamic nature and intrinsic instability, leading to premature burst release of drugs and nonspecific biodistribution in vivo. [ 10 ] To overcome this limitation, physical or chemical crosslinking of the core, [ 11, 12 ] shell [ 13, 14 ] or interlayer [ 15 ] of polymeric micelles has emerged as a straightforward approach to stabilize the assembled structures, assuring a prolonged circulation time without disassembly‐induced drug leakage. [ 16 ] However, overly stable nanocarriers are also problematic, since slow and inadequate release of therapeutics may result in an insufficient intracellular drug availability for killing cancer cells and potential induction of multidrug resistance (MDR). [ 17, 18 ] Under these circumstances, researchers have developed various cleavable linkers [ 19, 20 ] or physical interactions [ 21 ] to construct reversibly stabilized micelles. Nonetheless, these systems still cannot release drugs efficiently because a decrosslinking process should be involved before release of payloads from their cores. This two‐stage process limits their prospects and, it seems clear that increasing the stability of polymeric micelles while ensuring the efficiency of drug release are two conflicting purposes. To address the dilemma, here we propose a crosslinking induced reassembly (CIRA) strategy to simultaneously increase the stability and specific drug release rate of polymeric micelles, taking multiblock polyurethane (MPU) as a model. MPU has been established as a leading polymer for implants, tissue engineering, and drug delivery applications. [ 22 ] With excellent molecular tunability and unique phase behavior, MPU provides an outstanding platform allowing for facile control of self‐assembly properties. We and others have recently reported a series of MPUs with controllable micellization, multimodal targeting and smart response properties for on‐demand delivery of therapeutics and imaging agents. [ 23, 24, 25 ] However, to the best of our knowledge, MPU formulations with simultaneous improvement of stability and responsivity in vivo have not yet been developed. To validate our basic concept of CIRA, we first synthesized a model clickable multiblock multifunctional polyurethane. The polymer was constructed from biodegradable poly(ε‐caprolactone) (PCL), cleavable polyethylene glycol bearing a pH‐responsive benzoic‐imine linkage (BPEG), L‐lysine ethyl ester diisocyanate (LDI) as well as a reducible chain extender generated from L‐cystine (Cys‐PA) (Scheme S1, Supporting Information). The structure of MPU was confirmed by proton nuclear magnetic resonance ( 1 H NMR), Fourier transform infrared (FTIR), and gel permeation chromatography (GPC) analysis (Figures S1–S3, Supporting Information). The MPU prepared could self‐assemble into micelles with diameters around 53 nm and negative surface charges, as determined by dynamic light‐scattering (DLS) and transmission electron microscopy (TEM) (Figures S4 and S5 and Table S1, Supporting Information). The assembled structure of MPU micelles was visually clarified by computational simulation using a dissipative particle dynamics (DPD) model. The result presented a spherical core–shell structure with a hydrophobic core formed by insoluble PCL soft segments and surrounded by an acid‐detachable hydrophilic BPEG corona. The hard segments were located mainly at the subsurface, with some still distributed in the micellar core due to neighboring hydrophobic soft segments (Figures S6 and S7, Supporting Information). The alkyne sites on the interface enabled a post‐conjugation of targeting ligands [ 25, 26, 27 ] or shell‐crosslinking [ 14 ] via click chemistry after the formation of polymer micelles. To achieve reversible and click crosslinking, we designed and synthesized a reduction‐cleavable crosslinker (SS‐Az) (Schemes S2 and S3, Supporting Information). The obtained crosslinker contains a disulfide linkage and two azide sites, allowing for an efficient crosslinking of MPU micelles using a copper catalyzed alkyne‐azide cycloaddition (CuAAC) in aqueous solution. The success of crosslinking was first verified by 1 H NMR spectra, where the characteristic peaks of 1, 2, 3‐triazole ring (8. 0 ppm) and the methylene protons near the ring (4. 5, 5. 5 ppm) were observed for crosslinked MPU micelles (CMPU) (Figure S11, Supporting Information). TEM imaging indicated that the micelles remained their well‐dispersed spherical structures with click reaction at interface (Figure S5, Supporting Information). However, the micellar size increased from 53 to 111 nm after crosslinking (Figure S4, Supporting Information), which may be due to the change of self‐assembled structure or the existence of reaggregation. To understand this phenomenon, we measured the mass‐average molecular weight of micelles using static light scattering (SLS). The calculated aggregation number ( N agg ) of CMPU micelles was almost twice as large as that of MPU micelles (Figure S12 and Table S1, Supporting Information). This result implies possible inter‐micelle crosslinking that leads to reaggregation of micelles. To further prove the success of crosslinking, the MPU and CMPU micelles were treated with 10‐fold volume of N, N ‐Dimethylformamide (DMF) and analyzed with DLS. It was found that the size of CMPU micelles increased nearly twofold due to the swelling of the hydrophobic segments in the presence of DMF, [ 12 ] while the structures of MPU micelles were completely disrupted owning to the dissolution of polymers in DMF (Figure S13, Supporting Information). The results indicate that click crosslinking enables CMPU micelles to withstand dissolution in good solvent. The crosslinking of MPU micelles was also confirmed using fluorescence resonance energy transfer (FRET), a facile and powerful tool to detect the molecular interactions within the range of 10 nm and monitor the process and dynamics of self‐assembly in real time. [ 28, 29 ] As a pair of FRET dyes, doxorubicin (DOX, donor) and 3, 3′‐diethylthiadicarbocyanine iodide (Cy5, acceptor) were encapsulated into MPU and CMPU micelles separately, followed by mixing the fluorescent‐loaded micelles. As shown in Figure S14A (Supporting Information), with mixing of DOX@MPU and Cy5@MPU micelles, an increase of fluorescence intensity at 695 nm was observed over time, which means that the two dyes were exchanged between the micelles and in a close proximity. [ 29 ] By contrast, the mixture of DOX@CMPU and Cy5@CMPU did not generate evident FRET signal (Figure S14B, Supporting Information), suggesting an inhibited movement of polymeric chains and enhanced kinetic stability of micelles after crosslinking. To investigate whether crosslinking improves the thermodynamic stability of micelles, the particle sizes under different dilution times with water or phosphate buffered saline (PBS) were measured by DLS. We found that the diameters and size distributions of MPU micelles increased greatly with water or buffer addition, while those of CMPU were almost unchanged even when diluted more than 700 times ( Figure 1 A – C and Figures S15 and S16, Supporting Information). Such a stability of CMPU is sufficient for application in the body environment. [ 30 ] Further, the FRET pair DOX and Cy5 were coloaded in MPU and CMPU micelles, and the fluorescence spectra of micelles upon dilution were collected (Figure S17, Supporting Information). It can be noticed that the FRET efficiency of DOX+Cy5@MPU decreased significantly (Figure 1D ), while that of DOX+Cy5@CMPU changed slightly (Figure 1E ). On the other hand, once administered into the bloodstream, micelles are immediately mixed with blood cells, plasma proteins, surfactants, and many other components. [ 31 ] Therefore, they should realize their high stability against the existence of blood components to ensure longevity. [ 32 ] To this end, the stability of micelles was monitored in the presence of sodium dodecyl sulfate (SDS) surfactant, bovine serum albumin (BSA) protein and fetal bovine serum. As seen in Figure S18 (Supporting Information), the size distributions of MPU micelles changed rapidly under simulated physiological conditions, while those of CMPU micelles remained basically constant after various treatments. These results demonstrate a high stability of MPU micelles after click crosslinking, which is helpful to avoid micellar disassembly and premature drug release in vivo. Figure 1 Enhanced stability and promoted redox‐responsivity of crosslinked micelles. A, B) Size distributions of MPU micelles A) and CMPU micelles B) diluted with water for a) 1, b) 3, c) 10, d) 33, e) 100 and f) 333 times. C) Normalized increase in size of MPU micelles before and after crosslinking upon dilution with water. D, E) Normalized change in FRET efficiency of MPU micelles D) and CMPU micelles E) encapsulated with DOX and Cy5 upon dilution with water for different times. The insets in (D, E) show schematic illustrations of DOX+Cy5@MPU and DOX+Cy5@CMPU micelles under dilution, respectively. F) Cumulative release of DOX from MPU and CMPU micelles in phosphate buffered saline (PBS, 10 × 10 −3 m, pH 7. 4) solutions. MPU+GSH and CMPU+GSH indicate release media containing 10 × 10 −3 m GSH, and CMPU/GSH shows release media with GSH addition at 24 h. G, H) Fluorescence emission spectra (λ ex = 480 nm) of MPU micelles G) and CMPU micelles H) encapsulated with DOX and Cy5 in the presence of 10 × 10 −3 m GSH for different times. The insets in (G, H) show schematic illustrations of DOX+Cy5@MPU and DOX+Cy5@CMPU micelles after GSH treatment, respectively. I) Normalized decrease in FRET efficiency of DOX+Cy5@MPU and DOX+Cy5@CMPU micelles incubated with 10 × 10 −3 m GSH for different times. To verify whether reversible crosslinking of MPU micelles enables controlled release of payloads in tumor microenvironment, a model drug DOX was encapsulated into the micelles followed by a click crosslinking. Evidently, DOX@MPU micelles displayed a typical burst drug release in PBS solution (pH 7. 4) (Figure 1F ), while the release of DOX was much slower for DOX@CMPU under both neutral and weak acidic conditions (pH 6. 5) (Figure S19, Supporting Information). The result verifies a high stability of crosslinked micelles even after detachment of PEG corona. Moreover, a limited acceleration of drug release was achieved in the presence of reducing agent (10 × 10 −3 m GSH) for uncrosslinked micelles (Figure 1F ). This phenomenon is in agreement with our previous findings, [ 25 ] which may arise from the shielding of disulfide bonds by the soft segments leading to a steric repulsion against the penetration of GSH. Interestingly, DOX@CMPU micelles exhibited a much higher drug release rate than uncrosslinked formulation under 10 × 10 −3 m of GSH, imparting a sensitive “on‐off” switch for controlled release (Figure 1F ). This result seems counterintuitive and contradictory to other studies, [ 33 ] as it is generally believed that CMPU micelles need to break their crosslinking first for subsequent penetration of GSH into the interior core to attack the disulfide bonds linked with hydrophobic segments. Such a two‐stage degradation of crosslinked polymers is in principle slower than that of uncrosslinked ones. To address this issue, we postulated that the enhanced responsivity might be associated with the change of hierarchical architecture of assemblies during crosslinking, as it has been shown that crosslinkers could chemically induce or kinetically trap the morphological transition of polymeric assemblies. [ 34 ] To confirm this hypothesis, we performed 1 H‐ 1 H nuclear Overhauser enhancement spectroscopy (NOESY) experiment on the polymers before and after crosslinking. Obviously, a new correlation peak between the crosslinker protons (2. 96 ppm) and the methylene next to disulfide bond of Cys‐PA (2. 65 ppm) was observed for CMPU (Figure S20, Supporting Information), further confirming the success of click reaction between Cys‐PA and SS‐Az. Moreover, it is interesting to notice that the correlation between PCL (2. 32–2. 36, 4. 08 ppm) and Cys‐PA (2. 20, 3. 83 ppm) were diminished ( Figure 2 A, B and Figure S20, Supporting Information), while the NOE signals between PEG protons (3. 61, 3. 31 ppm) and Cys‐PA (3. 36, 3. 83 ppm) as well as that among Cys‐PA groups (3. 48, 3. 83 ppm) were clearly observed after crosslinking (Figure 2C, D ). These results reveal a possible migration and aggregation of Cys‐PA moieties from the core to the subsurface layer of micelles due to the azide/alkyne click reaction occurring mainly at the micellar interface, which may induce a reassembly process (CIRA) and microphase separation between the soft and hard segments. Figure 2 Crosslinking induced reassembly of MPU micelles. 1 H− 1 H NOESY spectra of MPU micelles in CD 4 O A, C) before and B, D) after cross‐linking. FTIR spectra of lyophilized MPU micelles before and after cross‐linking in the E) N–H and F) C=O stretching regions. G) Schematic representation of CIRA process of multiblock polyurethane micelles. The CIRA was also supported by FTIR analysis. As shown in Figure 2E, the N−H stretching vibration at 3300–3550 cm −1 was shifted to lower frequencies after crosslinking due to an enhanced hydrogen bonding strength. [ 35 ] More importantly, a blue shift of C=O stretching band centered at 1730 cm −1 corresponding to free carbonyl of urethane groups, [ 36 ] and a red shift of hydrogen‐bonded ordered urea carbonyl around 1640 cm −1 were observed (Figure 2F ). [ 37 ] The results verify an enhanced microphase separation between the soft and hard segments and strengthened hydrogen bonding within urea‐bearing hard domains after crosslinking. The aggregation of hard domains can lead to the gathering of disulfide linkages around the micellar interface, thus promoting the reduction responsivity and drug release rate in the presence of GSH (Figure 2G ). To further justify the improved responsivity, the micelles co‐loading with DOX and Cy5 were treated with 10 × 10 −3 m GSH. It was found that the FRET efficiency of DOX+Cy5@MPU declined moderately under reductive environment (Figure 1G, I ), while that of DOX+Cy5@CMPU decreased much faster, with an emission at 690 nm quickly disappearing over time (Figure 1H, I ). The result agrees well with NOESY, FTIR and drug release experiments, indicating an unusual improvement of redox‐responsivity by crosslinking. To our knowledge, this is the first example of a polymeric assembly with stability and responsivity increased simultaneously by crosslinking, which provides a new strategy to address the dilemma of drug retention and release for controlled delivery. The promising properties granted by CIRA inspired us to further evaluate the capacity of CMPU for intracellular drug delivery. The MCF‐7 cells were cultured with DOX‐ and Cy5‐coloaded micelles for different times and observed by confocal laser scanning microscope (CLSM). As seen in Figure 3 A, there was no FRET fluorescence observed in the cells incubated with DOX+Cy5@MPU, and the distribution of intracellular DOX fluorescence was similar to that of free drugs (Figures S22 and S23, Supporting Information). The result indicates that MPU micelles were dissociated upon contacting with cells, resulting in premature drug leakage outside the cells. In contrast, the cells treated with DOX+Cy5@CMPU showed remarkable FRET signal in the cytoplasm in 1 h, suggesting structural integrity of crosslinked micelles during cellular uptake. After 4 h of incubation, the FRET signal diminished and strong DOX fluorescence was observed inside the nucleus, owning to triggered intracellular release of drugs in response of GSH. It is worth noting that DOX+Cy5@CMPU manifested stronger intracellular fluorescence than DOX+Cy5@MPU in both DOX and Cy5 imaging channels, which may be related to the higher responsivity and faster release property of crosslinked micelles. Besides, the cell entry efficiency of micelles with different stability during uptake should also be taken into consideration. [ 38 ] With this in mind, we further assessed the mechanism of cell internalization using flow cytometry. As shown in Figure 3B, CMPU micelles showed much higher intracellular fluorescence intensity than MPU micelles even in the presence of buthionine sulphoximine (BSO) that inhibits the production of GSH in cells, [ 39 ] revealing a greater efficiency of cellular uptake after crosslinking. Moreover, as shown in Figure 3C, the internalization rates of both MPU and CMPU were greatly inhibited at 4 °C, revealing energy‐dependent cellular uptake processes. In particular, the cell entry of DOX@MPU was reduced by chlorpromazine and colchicine, indicating clathrin‐mediated endocytosis and macropinocytosis. [ 40 ] The uptake of DOX@CMPU could be associated with macropinocytosis due to the lowest internalization rate in the presence of colchicine. [ 41 ] The different endocytic pathways may account for the enhanced cell internalization of CMPU micelles. [ 42 ] This phenomenon is of great interest and worth further investigation. The improved cell trafficking and intracellular DOX release would result in higher drug efficacy. However, we found that DOX@CMPU was less effective in killing MCF‐7 tumor cells than MPU formulation (Figure S24, Supporting Information), possibly because the prematurely released DOX from uncrosslinked micelles could rapidly diffuse into the cell nucleus to cause cell death (Figure 3A ). [ 43 ] With this in mind, we used drug‐resistant MCF‐7 cancer cells that can pump free chemotherapeutics out of cells for cytotoxicity assay. [ 44 ] As expected, DOX@CMPU micelles exhibited much greater therapeutic effect against drug‐resistant tumor cells, with a median inhibitory concentration (IC 50 ) 2. 5 times lower than that of DOX@MPU (Figure S25, Supporting Information). On the other hand, the drug‐free micelles did not show any toxic effect toward L929 mouse fibroblasts (Figure S26, Supporting Information), suggesting a good cytocompatibility of MPU and CMPUs. Figure 3 Intracellular drug delivery and in vivo biodistribution. A) CLSM images of MCF‐7 cells incubated with DOX+Cy5@CMPU and DOX+Cy5@MPU for 1 and 4 h. Nuclei of cells were stained with 2‐(4‐amidinophenyl)‐6‐indolecarbamidine dihydrochloride (DAPI, blue). For Cy5 channel, λ ex = 620 nm; for others, λ ex = 480 nm. The scale bars are 10 µm. B) Flow cytometry of BSO‐ or GSH‐OEt‐pretreated MCF‐7 cells incubated with DOX@MPU and DOX@CMPU micelles for 1, 2, and 4 h. C) Flow cytometry of MCF‐7 cells incubated with DOX@MPU and DOX@CMPU micelles for 4 h in the presence of various inhibitors at different temperatures. Cell cultured at 37 °C without inhibitor were set as control. D, E) In vivo imaging of MCF‐7 tumor‐bearing mice at different times after intravenous injection of DOX+Cy5@MPU and DOX+Cy5@CMPU micelles. Mice receiving saline were set as control. D) donor fluorescence channel, λ ex = 480 nm, λ em = 600 nm. E) FRET fluorescence channel, λ ex = 480 nm, λ em = 700 nm. F) The ex vivo imaging of major organs and tumors of nude mice bearing MCF‐7 tumors at 24 h post‐injection of DOX+Cy5@MPU and DOX+Cy5@CMPU micelles (λ ex = 480 nm, λ em = 600 nm), where a, b, c, d, e, f represent heart, liver, spleen, lung, kidney, and tumor, respectively. G) Semi‐quantitative analysis of DOX fluorescence in major organs and tumors. H) CLSM images of tumor tissue slices of nude mice at 24 h post‐injection of DOX+Cy5@MPU, DOX+Cy5@CMPU micelles and saline. The scale bars are 100 µm. I) The fluorescence intensity of tumor slices was quantified and plotted. Statistical significance: (*) P < 0. 05; (**) P < 0. 01; (***) P < 0. 005. To further explore the benefit of CIRA strategy for in vivo applications, the micelles coloading with DOX and Cy5 were intravenously injected into MCF‐7 tumor‐bearing nude mice via the tail vein, and tracked using an in vivo imaging system. It was found that DOX fluorescence spreaded widely through the abdomen and the FRET effect disappeared within 2 h for DOX+Cy5@MPU group. The result indicates the disassembly of uncrosslinked micelles and premature release of payloads leading to nonspecific biodistribution (Figure 3D, E and Figure S27, Supporting Information). In contrast, CMPU formulation showed both remarkable donor fluorescence and FRET signal rapidly gathering around the tumor tissue, and the FRET emission decreased over time (Figure 3D, E and Figure S27, Supporting Information), suggesting a higher stability, superior targeting capacity and specific intratumor drug release of CMPU micelles granted by CIRA. The ex vivo fluorescent imaging of the anatomized organs of the mouse sacrificed at 24 h evidenced an improved biodistribution of DOX for CMPU formulation, where the fluorescence in tumor was greatly enhanced while those in liver, spleen, and kidney were significantly minimized (Figure 3F and Figure S28, Supporting Information). CLSM imaging of tumor slices further confirmed that the CMPU group showed remarkably stronger DOX fluorescence in tumors than MPU group (Figure 3H, I ). The superior targeting effect of CMPU could also be observed in 4T1 tumor‐bearing mice (Figures S29 and S30, Supporting Information). Next, we evaluated the therapeutic efficacy of MPU and CMPU micelles taking MCF‐7 tumor‐bearing nude mice as a model. As shown in Figure 4 A, the tumor volumes in control mice receiving saline administration increased rapidly, while the growth of tumors was remarkably suppressed by the treatment of various DOX formulations. In particular, DOX@CMPU exhibited superior tumor inhibition effect, with mean tumor weight 1. 7‐fold and 3‐fold lower than those for DOX@MPU and control groups, respectively (Figure 4B ). Furthermore, histological analysis with hematoxylin and eosin (H&E) staining revealed a greater extent of cell remission and necrosis for DOX@CMPU group compared with uncrosslinked micelles and free DOX (Figure 4C ). The percentages of apoptotic tumor cells of DOX@CMPU group (≈90%) obtained from nuclear‐associated antigen (Ki‐67) and terminal deoxynucleotidyl transferased dUTP nick end labeling (TUNEL) assays were much higher than those of other groups (Figures S31 and S32, Supporting Information). In addition, although no mice died during the treatment period due to the low dose (Figure S33, Supporting Information), apparent weight loss was detected in mice treated with free DOX (Figure S34, Supporting Information) and, particularly, small amount of cell necrosis of kidney was noticed for mice treated with free DOX and DOX@MPU (Figure S35, Supporting Information). In contrast, no significant decrease in body weights and abnormality of major organs was detected in CMPU formulations (Figures S34 and S35, Supporting Information), demonstrating that CIRA provides an effective strategy for the development of stable and smart nanoplatform for safe and specific drug delivery in vivo. Further work is ongoing to demonstrate the versatility of CIRA strategy using different kinds of polymeric systems, stimuli‐sensitive crosslinkers and other disease models. Figure 4 In vivo anticancer efficacy. A) Change of tumor volume after intravenous administration of saline, DOX, DOX@MPU and DOX@CPU micelles in MCF‐7 tumor‐bearing nude mice. B) Mean weights of tumors separated from animals with different treatments. C) Ex vivo histological analyses of tumor sections. H&E staining, Ki67 and TUNEL immunofluorescence staining analyses of tumor sections. In TUNEL analysis, the apoptotic cells were stained green. The scale bars are 100 µm. Statistical significance: (*) P < 0. 05; (**) P < 0. 01; (***) P < 0. 005. In summary, we have developed a model multiblock polyurethane bearing disulfide linkages in the backbone and clickable active sites in the side chains. The polymer self‐assembled into core–shell micelles in an aqueous solution, and underwent a crosslinking induced reassembly and microphase separation between the soft and hard segments. The CIRA drove the migration of disulfide moieties from the inner core to the subsurface of micelles due to a facile click reaction occurring at the micellar interface. As a result, the thermodynamic stability and redox‐responsivity of micelles could be improved simultaneously, leading to an enhanced tumor targeting, specific intracellular drug delivery and excellent therapeutic efficacy both in vitro and in vivo. Our work provides a new insight into the self‐assembly of macromolecules and a promising nanoplatform for theranostics applications. Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902740
| 2,020
|
Advanced Science
|
An Alkaline Based Method for Generating Crystalline, Strong, and Shape Memory Polyvinyl Alcohol Biomaterials
|
Abstract Strong, stretchable, and durable biomaterials with shape memory properties can be useful in different biomedical devices, tissue engineering, and soft robotics. However, it is challenging to combine these features. Semi‐crystalline polyvinyl alcohol (PVA) has been used to make hydrogels by conventional methods such as freeze–thaw and chemical crosslinking, but it is formidable to produce strong materials with adjustable properties. Herein, a method to induce crystallinity and produce physically crosslinked PVA hydrogels via applying high‐concentration sodium hydroxide into dense PVA polymer is introduced. Such a strategy enables the production of physically crosslinked PVA biomaterial with high mechanical properties, low water content, resistance to injury, and shape memory properties. It is also found that the developed PVA hydrogel can recover 90% of plastic deformation due to extension upon supplying water, providing a strong contraction force sufficiently to lift objects 1100 times more than their weight. Cytocompatibility, antifouling property, hemocompatibility, and biocompatibility are also demonstrated in vitro and in vivo. The fabrication methods of PVA‐based catheters, injectable electronics, and microfluidic devices are demonstrated. This gelation approach enables both layer‐by‐layer and 3D printing fabrications.
|
1 Introduction Engineered biomaterials have been increasingly recognized as a promising solution to overcome major hurdles in healthcare. [ 1 ] Development of stretchable and strong biomaterials with long‐term performance, chemical stability, and multifunctional properties can facilitate translating them into clinical and industrial applications. [ 2 ] Meanwhile, biomaterials which are adaptable to 3D printing and microfluidic technologies are highly beneficial to provide controlled synthesis, versatile properties, and accurate biomimicking solutions. [ 3 ] Hence, new methods to form biomaterials are necessary to make progress in advancing the field. Hydrogels can be made of a diverse group of synthetic and natural materials and can be categorized into physical and chemical gels. [ 4 ] The formation of crystallites is among the strongest physical crosslinking methods in forming hydrogels, [ 5, 6 ] which combined with hydrogen bonding can provide strength and elasticity. Polyvinyl alcohol (PVA) with tunable crystallinity has been largely used to improve mechanical properties of biomaterials used for tissue engineering [ 7 ] and biomedical applications such for wound dressings, [ 8 ] contact lenses, [ 9 ] vascular grafts, [ 10 ] artificial meniscus, [ 11 ] and vitreous substitute. [ 12 ] PVA hydrogels are commonly synthesized through two mechanisms comprising chemical bonding using a cross‐linker such as glutaraldehyde and/or physical crosslinking such as cyclic freeze–thaw methods. [ 13 ] PVA hydrogels with high mechanical properties and chemical stability require low water to PVA ratio [ 2b ] as well as dense PVA networks. [ 2a ] The strength and toughness of conventional PVA hydrogels with high swelling ratios are greatly diminished when they absorb water. [ 2b ] Accordingly, having dense polymer networks and higher degree of crystallinity are the major contributing factors in forming strong PVA hydrogels. In this paper, we aim to develop a method to form a strong and stretchable PVA biomaterial with crystalline and dense networks. We hypothesize that applying alkaline hydroxide in a high concentration into a highly dense PVA polymer network can rapidly induce crystallinity. According to our proposed mechanism, first, the deprotonation of hydroxyl groups of PVA is achieved via the basic attack of OH − ions. Subsequently, the complexation can be formed between O − group and the free Na +, facilitating the mobility of PVA chains to be stretched and aligned and form crystalline domains. This chain organization will be stabilized and crystalline domains will be developed by replacing the complex with hydrogen bonds as soon as they are in contact with water. Moreover, the higher amount of hydrogen bonding and crystallite formation lead to the expelling of water molecules, leading to a low swelling ratio without sacrificing the elasticity. To prove our hypothesis and demonstrate potential applications, we characterized the proposed process and comprehensively studied the thermal, physical, mechanical, chemical, and biological properties of our proposed physically crosslinked PVA hydrogel (PVA‐H). Finally, benefiting from the current approach, a water‐induced shape memory biomaterial, and artificial muscle system with the PVA‐H are designed. To reveal potential applications of our novel PVA biomaterial in biomedical applications, biocompatibility of PVA membranes are evaluated both in vitro and in vivo and their applications in fabricating stretchable and robust injectable electronics, catheters, and microfluidic devices are elucidated. 2 Results and Discussion 2. 1 Preparation and Mechanism A 100 mg mL −1 solution of PVA ( M w = 205 000 g mol −1, Aladdin) was prepared by dissolving PVA powder in deionized water at 90 °C under magnetic stirring overnight. A specified amount of the PVA solution was poured or coated on the surface of a Petri dish or mold and was allowed to dry completely. The PVA‐H was then formed by immersing the dried PVA film in a high‐concentration solution of strong alkaline hydroxide (NaOH) for 10–40 min (depending on the thickness of films), followed by water treatment. To explain the proposed mechanism, we called NaOH‐treatment as the intermediate step, where the polymer chains are organized ( Figure 1 A ). Afterward, the insoluble membrane was immersed in deionized water to remove ions and to stabilize crosslinked crystalline PVA networks (crystallites). All PVA biomaterials were prepared using 6 m solution of NaOH unless otherwise stated. The thickness of membranes can be readily tuned by changing the volume of the initial PVA solution. Figure 1 A) Schematic illustration of PVA‐H preparation and proposed mechanism (PVA‐Na + : PVA immersed in a NaOH solution; PVA‐H: crosslinked PVA hydrogel). B) 1 H‐NMR spectrum of PVA and PVA‐H in DMSO‐d 6 solvent. C) FTIR spectra of PVA, PVA‐Na + (intermediate step) and PVA‐H. D–F) Photos of PVA‐Hs made with different molarities of NaOH solution: D) 1 m forming gel with the thickness of 0. 05 mm, E) 3 m forming gel with the thickness of 0. 1 mm, and F) 6 m forming gel with the thickness of 0. 1 mm. G) Stress–strain curves and elastic moduli of PVA‐Hs made with different NaOH concentrations. The inset shows a PVA strip during tensile test. H) DSC curves of a dried PVA membrane and a PVA‐H. I) TGA curves of PVA and PVA‐H. J) Demonstration of the resistance of PVA‐H to needle puncture (Video S1, Supporting Information). K) Optical images of a medical glove and a PVA‐H membrane tighten in a customized holder for measuring the resistance against needle puncture in Figure 1L. L) The force–displacement graph when an injection needle was pushed onto the medical glove with 0. 025 mm and PVA‐H samples with 0. 08 mm and 0. 2 mm thicknesses. M–P) PVA‐Hs incorporating different nanomaterials including M) MNPs, N) graphene, O) CNT, and P) PVA‐H coated with PANI. Q) PVA solution injection in a NaOH bath to make fibers. R) PVA/CNT injection in NaOH bath to make PVA/CNT fibers. S) PVA solution can be coated on a mold and crosslinked to make different shapes such as a glove. In the intermediate step, we aimed to organize the stacked polymer chains to form crystallization and crosslinking with our new approach by using a high‐concentration solution of NaOH. The NaOH attack leads to the hydrolysis of the ester group of the acetate moieties from the polymer backbone as confirmed by nuclear magnetic resonance (NMR) spectroscopy, where the 1 H signal of the acetate methyl group proton signals named H c disappeared (Figure 1B ). [ 14 ] We suppose that applying a strong alkaline hydroxide into a dried PVA film results in two sequential events in the intermediate step. First, OH − of the alkaline hydroxide attacks the hydroxyl groups of PVA, resulting in disrupted hydrogen bonds and deprotonation of the hydroxyl groups of the PVA chain. Afterward, the newly formed O − groups in PVA interact with free Na + ions to form complex. This new complexation allows the polymer chain to move freely via new network as shown in Figure 1A. This new organization in the intermediate step, results from the conformation of polymer chains that form parallel and stretched macromolecules. These two sequential processes facilitate the PVA crystallization as the intermediate step induces the preorganization of the polymer chain as depicted in Figure 1A. Followed by immersing NaOH‐treated PVA (PVA‐Na + ) in water, crystalline domains are stabilized, and PVA‐H are formed. In this final step, adding water will remove Na + ions, and O − will be protonated and crystalline domain will be permanently stabilized, leading to a strong PVA‐H (Figure 1A ). In following we characterized each step with several experimental methods. The interaction of the hydroxyl groups in PVA with OH − moieties of the highly concentrated NaOH solution leads to the acid‐base reaction [Equation ( 1 )], and it is confirmed by the difference between the pKa values which is 1. 5 (14 for H 2 O/OH − and 15. 5 for R—OH/R—O − ; we chose to use the value generally used for CH 3 —CH 2 —OH/CH 3 —CH 2 —O −, as no data is available regarding the pKa of PVA). This confirms that two thirds of OH groups of the PVA backbone are deprotonated and they can be further complexed by Na + ions. [ 15 ] (1) R − OH + Na + OH − ⇄ R − O − Na + + H 2 O Fourier transformed infrared (FTIR) spectra are provided in Figure 1C which depicts the spectra of the PVA (pristine film), PVA‐Na + (intermediate step), and PVA‐H (crosslinked PVA). After the hydrolysis, the fingerprint peak related to acetate groups between 1750–1735 cm −1 disappeared which confirmed the full efficiency of the acid‐base reaction. Indeed, this peak is linked to the stretching C=O from acetate group which is in agreement with the NMR data (Figure 1B ). The appearance and sharpness of the peak at about 1142 cm −1 for the PVA‐H and PVA‐Na + confirmed the re‐organization of the polymer chains after the reaction with NaOH (Equation 1 ). Indeed, the intensity of this peak is directly related to the crystalline portion of the polymeric chains. [ 16 ] The increase in the degree of crystallinity is then enlightened by the increase of the intensity of this peak, which is the fingerprint of the crystallinity for the PVA. [ 16a ] Moreover, the intensity of the peak attributed to the CH 2 wagging and twisting at around 1018 cm −1 decreased after the hydrolysis. This suggests that the C—H bonds of the CH 2 groups of the backbone are less free to deform out of the plane, which indicates that the polymer chains are more confined, preventing the free deformation of the C—H bonds mainly owing to the van der Waals interactions. 2. 2 Mechanical Properties The effects of the NaOH concentration on the mechanical and gelation properties of PVA‐H are shown in Figure 1D – G. The crosslinking of PVA film with 3–6 m solution of NaOH leads to successful gelation (Figure 1D, E ). However, 1 m solution of NaOH resulted in the formation of a very weak and thin membrane (Figure 1D ) with a very low elastic modulus (0. 04 MPa), indicating unsuccessful crosslinking, since the concentration is too low to stoichiometrically deprotonate OH groups of the PVA and only strong basic media, pH >12, can deprotonate PVA to form a strong metal‐binding ligand. [ 17 ] According to the stress‐strain curves, all samples exhibited elastic behavior when they were stretched up to 100% strain. In addition, all samples demonstrated excellent ductility, exceeding 350% of their original length. Further comprehensive tensile experiments were carried out to probe the effect of dehydration on the mechanical properties of PVA‐Hs (Figure S1, Supporting Information). The results show that the continuous dehydration of a PVA‐H leads to denser PVA chains, and thus gradual increase of its elastic modulus and mechanical strength. To evaluate the effectiveness of alkaline metal hydroxide agents, tensile properties of PVA‐Hs (0. 1 mm thickness) crosslinked with NaOH, KOH, and LiOH, were studied (Figure S1A, Supporting Information). The use of LiOH led to an unstable and weak hydrogel, showing an ineffective crosslinking. On the other hand, NaOH‐ and KOH‐crosslinked membranes exhibited high mechanical strengths with ultimate strengths above 13 MPa and elastic moduli above 1. 7 MPa. This is because LiOH is a weak base due to its low dissociation constant in water. LiOH bond is described as having maximum covalent character because of Fajans’ Rule, which finally hampers an efficient reactivity of the OH −. [ 18 ] 2. 3 Crystallinity The increase in crystallinity was further confirmed by differential scanning calorimetry (DSC) to calculate the PVA crystalline fraction. Typical DSC for PVA and PVA‐H films are shown in Figure 1H. A considerable increase in crystallinity from 7. 35% to 56. 9% was observed, confirming the efficiency of our new crosslinking approach. This effect contributes to the enhancement of the mechanical properties of the PVA‐H compared to the PVA. The change in the crystallinity degree was also confirmed by X‐ray diffraction measurement shown in Figure S2, Supporting Information. Thermogravimetric analysis (TGA) was performed to evaluate PVA and PVA‐H thermal properties (Figure 1I ). The first region (around 100 °C) is related to the evaporation of water molecules entrapped in the polymer. The evaporation of water molecules is delayed for the PVA‐H sample as seen in the change of the slope due to the crystalline structure, which decreases the diffusion of H 2 O through the materials. [ 19 ] The second region corresponds to the hydrogen bonding break and the degradation of the PVA backbone. PVA shows higher stability in this region which is due to the lower amount of the crystallites. Crystallites are known to degrade earlier than entangled amorphous chains as the crystalline organization with the polymer enables first the chain separation and second their degradation. This is impossible for amorphous polymers as the entanglement of the chains hinder their free mobility and lead to degradation of the material, which require higher amounts of energy. [ 20 ] In the formal freeze–thaw method, the phase separation and crystallization are caused by water molecules during the repeated freezing and thawing. [ 21 ] 2. 4 Other Properties The scanning electron microscope (SEM) of a PVA‐H film is presented in Figure S3, Supporting Information, showing very fine PVA networks. Water content of PVA‐H is 33 ± 3% and the equilibrium swelling ratio of PVA‐H is 0. 9 ± 0. 1. Such a low water content and swelling ratio along with the transparency of this PVA‐H offer a unique opportunity for its application in designing biomedical devices such as contact lenses. The long‐term stability of PVA‐H was evaluated by storing PVA‐H strips in water for 2 years without observing any tangible changes in their physical and mechanical properties, which qualitatively demonstrates their long‐term stability. Provided by NaOH‐induced gelation method, PVA‐Hs performed excellent resistance to damage by pointed and sharp objects (Video S1, Supporting Information). For demonstration, a 21G needle was pushed onto PVA membranes and a nitrile glove (VWR International, LLC) to measure the puncture forces. The PVA membrane showed higher resistance to puncture by needle compared to nitrile glove (Figure 1J – L ). Furthermore, current gelation method allows the incorporation of different nanomaterials, such as magnetic nanoparticles (MNPs), graphene, and carbon nanotubes (CNTs), into PVA‐Hs that enables us to produce functional hybrid biomaterials with adjustable thicknesses (Figure 1M – P ). PVA‐H membranes can also be coated with conducting polymers such as polyaniline (PANI) using solution‐processing methods for flexible and stretchable energy storage or sensor applications. [ 22 ] For instance, the color of a PANI‐coated PVA‐H film immediately changed from green (emeraldine salt form of PANI) to blue/purple (emeraldine base form of PANI) when it was transferred from an acidic (1 m H 2 SO 4 ) to a basic solution (1 m NaOH), and it recovered its original color upon returning it to H 2 SO 4 solution, which can be employed in the development of optical pH sensors (Figure S5 and Video S2, Supporting Information). The decoration of PVA‐H with PANI showed negligible effect on its mechanical strength. As a demonstration of the strength of PVA‐H/PANI film, a dried PVA‐H/PANI strip (≈0. 5 g) was used to lift 9 kg weight, 18 000 times more than its own weight (Video S2, Supporting Information). Different molds can be used before the crosslinking step to form complex shapes such as a glove (Figure 1S ). The mechanical strength and water content of PVA‐Hs can also be tuned by controlling the dehydration level of PVA before the crosslinking step. For instance, PVA‐H fibers were immediately formed by injecting a PVA solution (200 mg mL −1 ) into the NaOH bath (Figure 1Q ). Due to the less dense network of PVA solution, these PVA fibers were mechanically weaker than PVA‐Hs that were dried before the gelation. Figure 1R illustrates the formation of PVA/CNT fibers with improved elastic modulus by injecting PVA/CNT solution (≈100 mg mL −1 PVA and 10 mg mL −1 CNTs) into the NaOH solution, indicating the possibility of direct solution‐to‐hydrogel preparation of PVA/CNT. Our results show that our PVA‐H introduces improved mechanical properties and low swelling ratio compared to conventional PVA hydrogels in the literature (Table S1, Supporting Information). [ 2, 23 ] Additional merits of current strategy include the ability to incorporate variety of nanomaterials and the possibility to fabricate various shapes comprising strong and stretchable films and tubes with an adjustable thickness as small as several micrometers. It should be noted that this method can be applied for PVA polymers with molecular weights higher than 31 000, since it provides longer chains and higher chain entanglements (Figure S6, Supporting Information). 2. 5 Shape Memory Effect and Artificial Muscle The as‐prepared PVA‐H films can recover more than 90% of large plastic elongation upon immersing in water. For instance, a PVA‐H strip with a length of 26 mm (width 8 mm, thickness 0. 1 mm) was stretched to 100 mm after tensile tests. Upon immersing the strip in water for 40 min, the length was decreased to 32. 6 mm, which is 91% recovery in plastic deformation, indicating its shape memory characteristic. PVA‐Hs can recover from plastic deformation by immersion in water. In addition, they can demonstrate a shape memory behavior when immersed in NaOH solution. The PVA‐H becomes soft and its ductility will be significantly increased might be because PVA becomes deprotonated and hydrogen bonds are destroyed ( Figure 2 A ). As such, the NaOH‐treated PVA‐H strip can be stretched to the plastic region with less force than a non‐treated one because of its lower stiffness. This is attributed to disruption of hydrogen bonds by the basic attack, which leads to an enhanced chain mobility in the PVA‐H. Figure 2 Demonstration of the shape memory property of the PVA‐H and its potential to develop artificial muscles. A) The PVA‐H strip was immersed in NaOH solution and elongated from 13 mm to 36 mm. Schematic shows the disruption of hydrogen bonds by the basic attack in the PVA‐H strip, enhancing the mobility of the PVA chains. B) Then, it was immersed in water, which regained its original length in a few seconds (Video S3, Supporting Information). C) An elongated PVA strip which was immersed in NaOH solution can provide a strong contraction force upon adding water which can lift a 0. 5 kg weight up to 4 cm (1, 2) Contraction force measured during the addition of water to the same PVA‐H strip (used in C1 and C2) using tensile machine (Video S4, Supporting Information) (3). D) The PVA‐H strip (1) can carry 9 kg (2), leading to the permanent elongation (3), which was regained upon immersing in water (4). A PVA‐H strip treated with NaOH solution (5) which was lifting 9 kg (6), underwent a large elongation (7) which can be regained upon adding water (8) (Video S5, Supporting Information). E, F) SEM images of PVA‐H immersed in NaOH solution. G, H) SEM images of a NaOH‐treated PVA‐H film, which was completely stretched into plastic region (Arrow shows stretching direction). I, J) SEM images of a PVA‐H film, which recovered to its original length after adding water. K, L) EDX spectra of samples presented in (E) and (I) respectively. Supplying water to the PVA‐H will remove Na + ions and O − will be re‐protonated, leading to the return of previously disrupted hydrogen bonds. This process results in the recovery of the plastic elongation, representing the proposed mechanism of the shape memory (Figure 2B and Video S3, Supporting Information). For example, a NaOH‐treated PVA‐H strip with an initial length of 13 cm was stretched to 36 cm (176. 9% strain; Figure 2A ). Upon immersing the strip in water for 40 s, its length returned to 14 cm, revealing 95. 7% recovery of its large plastic deformation or shape memory behavior (Figure 2B and Video S3‐part 1, Supporting Information). In a similar manner, another NaOH‐treated PVA strip with an initial length of 13 cm was stretched to 27 cm (107. 7% strain). The length of the strip decreased to 13 cm after immersing it in water for 70 s, indicating full recovery of its plastic deformation (Video S3‐part 2, Supporting Information). The quick recovery of large strain in an elongated strip by supplying water provides a strong contraction force which can be utilized in artificial muscle applications. [ 24 ] Water‐assisted recovery of previously disrupted hydrogen bonds in the NaOH‐treated PVA‐H strip provides energy to roll back the plastic elongation. So, when a weight is attached to an elongated strip, this recovered energy provides contraction force to lift a weight. For instance, a NaOH‐treated PVA‐H strip (width: 10 mm, thickness: 0. 3 mm and length: 13 cm) was manually stretched and was attached to a 0. 1 kg weight reaching the final length of ≈32 cm (167% strain). The contraction force generated by supplying water to the strip was sufficient to lift the 0. 1 kg weight up to 10 cm in ≈45 s with an average speed of ≈2 mm s −1 (Figure S7 and Video S4, Supporting Information). The same strip was used to lift the 0. 5 kg weight up to 4 cm in ≈65 s (Figure 2C ). It should be noted that the generated contraction force in the strip was slightly larger than the weight of attached objects ( Mg ), i. e. , 0. 98 N and 4. 91 N for the 0. 1 kg and 0. 5 kg, respectively. As such, the work done on the 0. 1 kg and 0. 5 kg weights by the strip during their hoist were ≈98. 1 mJ and ≈196. 2 mJ, respectively. To further measure the peak of the contraction force in the strip, the NaOH‐treated and pre‐stretched strip was installed on the tensile test grips with a 4. 91 N preloading and without further stretching. Upon supplying water to the strip, the force was gradually increased up to 10. 4 N in about 50 s (Figure 2C (3)), demonstrating that the strip can lift even heavier weights. Figure 2D and Video S5, Supporting Information further illustrate the ability of PVA‐H (treated and not treated with NaOH) in carrying 9 kg weight which led to permanent elongation in PVA‐H. We also showed that these deformations can be reversed upon adding water. SEM images of a NaOH‐treated PVA‐H are shown in Figure 2E, F. For comparison, SEM images of a NaOH‐treated PVA‐H after it was stretched to its plastic region are provided in Figure 2G, H. It can be observed that the surface morphology of the NaOH‐treated PVA‐H is more flat with cracks perpendicular to the stretching direction. According to the SEM images in Figure 2I, J, the surface morphology of the PVA‐H film after it was immersed in water is like those of freshly prepared PVA‐Hs (Figure S3, Supporting Information). This observation further supports the shape memory effect of the PVA‐H that when treated with NaOH, it can recover from large plastic elongations by adding water. The EDX spectrum of the NaOH‐treated PVA‐H reveals the embedment of NaOH ions in PVA network (Figure 2K ), whereas the EDX spectrum of the PVA‐H confirm the removal of NaOH ions after immersing in water (Figure 2L ). 2. 6 Biocompatibility To expand the potential applications of this material for biomedical applications, biocompatibility of PVA‐Hs was investigated. Two cell types, human keratinocytes (HaCaT) and fibroblasts (3T3), were cultured with medium extracted from PVA‐Hs. CCK‐8 kit cell proliferation‐toxicity assay and cell viability assay were performed to test PVA‐H's potential toxicity. Figure S8A, C, Supporting Information show live/dead staining for HaCaT and 3T3 cells on day three post‐seeding, respectively. Green fluorescence indicates living cells by cell‐permeant calcein AM and red fluorescence indicates dead cells by EthD‐1. Red fluorescence labeled cells were almost invisible for all images. According to Figure S8B, D, Supporting Information, PVA‐H films were cytocompatible after a period of 5 days and exhibited significant cell proliferation levels comparable to the control group. Protein and bacterial anti‐fouling performance was evaluated with bovine serum albumin (BSA, life, USA) as protein model and E. coli as bacteria model. Figure S8E, F, Supporting Information, represent FITC‐BSA protein adsorption of the PVA‐H films and cover glasses (control). Results suggested that developed PVA‐H membranes can resist protein adsorption better than the control group. Furthermore, according to Figure S8G, H, Supporting Information, the of number adhered E. coli on PVA‐H membranes was significantly lower than that in the control group ( p < 0. 001). Red blood cell‐induced in vitro hemolysis is a reliable and important indicator to evaluate the hemocompatibility. The hemolysis rate (HR) related to controls and PVA‐H membranes are summarized in Table S2, Supporting Information. HR below 5% were considered to be non‐hemolytic according to ISO 10993–41999. The HR of the PVA‐H group was far less than 5%, which suggest they do not induce hemolysis when in contact with red blood cells (Table S2, Supporting Information). Implantation of biomaterials will normally induce a foreign body reaction. [ 25 ] This may persist as a chronic inflammatory response [ 26 ] to degradation products, such as seen with biodegradable materials. [ 27 ] To assess the occurrence of foreign body reaction, PVA‐H membranes were subcutaneously implanted in mice. Figure S9, Supporting Information, represents hematoxylin and Eosin staining of the tissue slices of these membrane‐containing tissue explants, 4 weeks post‐operatively. With such stable PVA‐Hs, we did not observe any increased inflammatory reaction and giant cells. Only mild inflammatory reaction were observed in the form of the presence of some macrophages in the near vicinity of implants. Accordingly, no evident mature fibrous tissue capsule was formed around the PVA‐H at this time, and no tissue or cell necrosis occurred. In addition, in our 1‐month in vivo study by implantation in the subcutis of mice, no evidence of materials degradation was observed. 2. 7 Layer‐by‐Layer Fabrication and Applications This crosslinking approach allows layer‐by‐layer assembly and incorporation of various nanomaterials with PVA into a single film with adjustable thickness. For example, a triple‐layered composite membrane of PVA/silver nanoparticles (AgNPs), pristine PVA, and PVA/CNT were successfully fabricated. SEM image of the membrane cross‐section shows the effective attachment and assembly of all layers with micro‐sized thickness ( Figure 3 and Table S3, Supporting Information). Figure 3 A) Cross‐sectional SEM images of a triple‐layered composite membrane of PVA‐H/CNTs, pristine PVA‐H and PVA‐H/AgNPs (left to right). High resolution SEM images of B) PVA‐H/CNTs, and C) PVA‐H/AgNPs. D) EDX spectra of three selected areas in the composite membrane (A 1, A 2, and A 3 ). E) Images (1) and (2) show PVA‐H tubes with (9. 5, 0. 3) and (1. 5, 0. 2) (diameter, thickness [mm]), respectively. Images (3) and (4) show a PVA‐H tube (1, 9. 5) connected to compressor and blown to four times larger than its original diameter before bursting. (5) The effect of tube diameter on the burst pressure for tubes made with 0. 3 mm thickness (black); the effect of applied pressure on the diameter of balloon catheter (red); the effect of tube thickness on the burst pressure of tubes made with 9. 5 mm diameter (blue). F) A small tube (1. 5, 0. 2) (1) was pumped with water using a syringe, and it was able to store up to 220 μL of water before bursting. (2) Demonstration of the elasticity of a PVA‐H tube (1. 5, 0. 2). Given the excellent mechanical properties and facile fabrication method, developed PVA biomaterials can potentially be used to make catheters and implants. Catheters are important tools in minimally invasive delivery of therapeutics [ 2f ] promoting quick recovery time after procedures such as coronary artery bypass. [ 28 ] Catheters have been developed for delivery of new shear thinning occluding material for the treatment of brain aneurysms. [ 29 ] Catheters were used for delivering drugs in the brain and [ 30 ] liver, [ 31 ] as well as delivering pancreatic islets via the portal vein. [ 32 ] The most common cause for the failure of long‐term indwelling catheters is the formation of biofilm and crystalline deposits. Colonized bacteria on the catheter surface trigger serious infections in the urinary tract which can also lead to blockage in catheters, resulting in trauma and pain. [ 33 ] Meanwhile, there is extensive research on biofilm prevention with methods such as coating of conventional catheters with amphiphilic and antifouling polymers [ 34 ] or using hydrogels. [ 35 ] Other critical properties required in designing catheters include strong elastic properties, protein and bacterial antifouling, biocompatibility, and durability. Tubes with tunable mechanical properties, diameters and thicknesses can be built by coating PVA solution on a metal rod followed by drying and gelation with a high‐concentration solution of NaOH (Figure 3E ). To ensure a uniform coating of PVA solution, a metal rod was immersed in a PVA solution and rotated using an electric motor during the drying step, which can be repeated to attain a desired tube thickness. The diameter and thickness of PVA tubes can be adjusted such that they can be blown with fluid or gas, revealing their potential application as balloon catheters. For instance, Figure 3E (3, 4) displays a PVA tube with an outer diameter of 9. 5 mm and a thickness of 1 mm which was blown to four times bigger than its original diameter. Figure 3E (5) shows the input pressure required to increase the diameter of a tube with an initial diameter of 2. 6 mm to a balloon with a specific size. Since the resistance against internal pressure is vital in designing catheters, we also investigated the effect of tube diameter and thickness on the burst pressure (Figure 3E (5)). Increasing the tube thickness leads to higher tolerance against internal pressure, whereas increasing the tube diameter deceases the burst pressure. For example, a stretchable tube with an outer diameter of 1. 5 mm and thickness of 0. 3 mm can resist internal pressure up to 689 kPa, which is also comparable with Triple‐Lumen central venous catheters used in healthcare, [ 36 ] human saphenous vein [ 37 ] and human artery. [ 38 ] Figure 3F shows a small tube with a diameter of 1. 5 mm that was pre‐filled with around ∼40 µL of water and was able to store up to an additional 220 µL of water before bursting. Accordingly, we are able to design catheters with compliant and non‐compliant balloons by controlling the diameter and thickness of PVA tubes, using a low swelling and low water content hydrogel. 2. 8 3D Printing Fabrication and Applications Another application of the proposed crosslinking approach is in designing hydrogel‐based injectable electronics and implants capable of tolerating high shear forces during injection while maintaining their physical structure and function after injection. Developing injectable electronics can pave the way for controlled drug delivery, continuous monitoring, manipulation, and follow up of a therapy. [ 2, 39 ] Mechanical strength and the ability to resist large shear forces during injection as well as seamless integration of electronic with tissues are crucial properties for injectable electronics. [ 40 ] By combining layer‐by‐layer assembly and printing, it is also possible to build miniature electronics having required injectability and stretchability. To fabricate a conductive PVA‐H/CNT mesh, PVA/CNT ink (Experimental Section in Supporting Information) was printed on a Petri dish, followed by physical crosslinking using NaOH solution ( Figure 4 A (1, 2)). The as‐prepared mesh was readily peeled off from the printing substrate to form a strong and stretchable conductive mesh, offering numerous applications in wearable sensors and printable electronics. This mesh can preserve its integrity even after being pushed through the tiny hole of a pipette tip (Figure 4A (3, 4), Video S6, Supporting Information), revealing its injectability. Furthermore, PVA can be used as a substrate for printing conductive materials to produce stretchable electronics (Figure 4B (1)). Figure 4B (2) displays a conductive mesh effectively assembled on a PVA film (Video S7, Supporting Information shows the performance of the printed mesh on a PVA film). Figure 4 A) PVA/CNT solution while printing (1). 3D printed hydrogels after gelation (2). 3D printed hydrogel (1 cm2) was injected into water through a small pipette tip with 400 µm diameter (Video S6, Supporting Information) (3 and 4). B) Schematic of PVA/CNT conductive ink printed on a thin PVA membrane (1). Image of a stretchable electronic (2) (Video S7, Supporting Information). C) Synthesis of microfluidic channels inside a membrane wall (the inset shows the microfluidic channels built in a PVA membrane is immersed in NaOH solution) (1). Photos of a microfluidic channel under UV radiation before (top) and after (bottom) injection of water‐containing dye to the channel (2) (Video S8, Supporting Information). D) Photos of a microfluidic channel with a diameter of 300 μm that was built inside the wall of a PVA‐H tube (9. 5 mm diameter). Water‐containing dye was injected to the channel (1 and 2). A microfluidic channel inside a PVA‐H tube was injected with a dye solution and was placed in water under UV radiation (3). Observation of water and trapped air inside the channels of microfluidic arrays built within a PVA‐H film under microscope (microchannels with 20 µm diameter) (4) (Video S8, Supporting Information). E) Contact angles (CA) measured for PDMS; CA = 92. 7° (1). PVA (before gelation) dried sample; CA = 56. 9° (2). Dried PVA‐H; CA = 16. 6° (3) (contact angles were captured 5 s after the droplet touched the films). Other applications that can benefit from employing our strong and biocompatible materials include developing functional organ‐on‐a‐chip devices that can be used as alternative disease models and drug testing platforms [ 41 ] or implantable microfluidic devices. [ 42 ] Considering that only limited number of materials are being used in designing microfluidic devices, such as polydimethylsiloxane (PDMS), which is associated with the problem of absorption of hydrophobic drugs, [ 43 ] and polymethylmethacrylate (PMMA), which is rigid, strong hydrogels with stretchability, low water content, and high transparency can change the trajectory of progress in this field. PVA‐based microfluidics with adjustable channel sizes can be fabricated with the aid of layer‐by‐layer assembly and 3D printing. Here, a high‐concentration solution of alginate was employed as a sacrificial material to create microfluidic channels. According to Figure 4C, the alginate solution was printed on a dried PVA film and then was covered with another layer of PVA solution, followed by the drying step. Upon treating the composite film with NaOH solution to crosslink PVA and subsequent water treatment, the printed alginate was readily removed by gently pressing the film and passing water through the channels. To demonstrate the successful fabrication of channels inside a PVA film, a solution of fluorescein sodium salt was injected into the channels under UV light (Figure 4C (2), Video S8, Supporting Information). Microfluidic channels can also be engineered in nonplanar PVA surfaces using this approach (Figure 4D (1–3), Video S8, Supporting Information). The optical microscope images in Figure 4D (4) show arrays of microfluidic channels (≈20 µm diameter) inside a PVA‐H film produced using human hair as a sacrificial material in a similar layer‐by‐layer procedure. To provide a clear exhibition of microchannels inside a PVA‐H film, it was placed in water so that water can be perfused into channels, during which air was also trapped inside channels. Capillary‐induced contact angle between the trapped air and water inside microchannels is the indication of their hydrophilic surface. The water contact angle (Figure 4E ) for PDMS, dried PVA film before gelation, and dried PVA‐H were found to be 92. 7°, 56. 9°, and 16. 6 °, respectively. In contrast to PDMS, the low contact angle of dried PVA‐H reconfirms its high hydrophilicity, which is a critical property for microfluidic systems. The conventional PDMS‐based microfluidic systems typically require further surface treatments to improve their hydrophilicity and cell compatibility. The high elastic modulus, high hydrophilicity, biocompatibility, and biofilm antifouling of the proposed PVA‐H could open up new opportunities for the lab‐on‐a‐chip systems, microfluidic bioprinting, [ 44 ] and implantable microfluidic devices. [ 42, 45 ] 3 Conclusion Here, we developed a facile crosslinking method to fabricate strong and stretchable PVA biomaterials useful for several biomedical applications. Treating dense stack of PVA polymer with a high‐concentration solution of alkaline metal hydroxide can physically crosslink the polymer and increase the crystallinity to form an elastic material with a low water content and swelling ratio. Furthermore, by using this strategy, we can incorporate different nanomaterials to prepare functional hybrid biomaterials having different shapes with the aid of 3D printing and layer‐by‐layer assembly. Developed PVA‐Hs were characterized by shape memory property with the capability to recover 90% of plastic deformation with a contraction force sufficient to lift objects 1100 times more than their own weights which can be used for actuator and artificial muscle applications. This material is cytocompatible, hemocompatible, and biocompatible with antifouling properties. Given their strong mechanical properties, low water content, chemical stability, capability of incorporating different nanomaterials, and 3D printability, this new material can be used to develop catheters, vascular grafts, articular cartilage, corneal replacement materials, and contact lenses. Beyond medical applications, the material can also potentially used for food packaging. Conflict of Interest The authors declare no conflict of interest. Author Contributions M. A. D. and A. Kho contributed equally to this work. M. A. D. developed the material and designed the project. M. A. D. and A. Kho. developed the idea, designed and implemented the experiments, and wrote and revised the manuscript. Y. W conducted in vitro and in vivo tests. N. A. , H. A. , and A. E. assisted in chemical characterizations, writing and revising the manuscript. Q. C, K. X. and Y. L. assisted in the puncture resistant experiment and prepared PVA‐H glove. G. L. and M. X. supervised in vivo tests. A. K. and M. X. supervised the project and edited the manuscript. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file. Supplemental Video 2 Click here for additional data file. Supplemental Video 3 Click here for additional data file. Supplemental Video 4 Click here for additional data file. Supplemental Video 5 Click here for additional data file. Supplemental Video 6 Click here for additional data file. Supplemental Video 7 Click here for additional data file. Supplemental Video 8 Click here for additional data file.
|
10. 1002/advs. 201902872
| 2,020
|
Advanced Science
|
Bioresorbable Materials on the Rise: From Electronic Components and Physical Sensors to In Vivo Monitoring Systems
|
Abstract Over the last decade, scientists have dreamed about the development of a bioresorbable technology that exploits a new class of electrical, optical, and sensing components able to operate in physiological conditions for a prescribed time and then disappear, being made of materials that fully dissolve in vivo with biologically benign byproducts upon external stimulation. The final goal is to engineer these components into transient implantable systems that directly interact with organs, tissues, and biofluids in real‐time, retrieve clinical parameters, and provide therapeutic actions tailored to the disease and patient clinical evolution, and then biodegrade without the need for device‐retrieving surgery that may cause tissue lesion or infection. Here, the major results achieved in bioresorbable technology are critically reviewed, with a bottom‐up approach that starts from a rational analysis of dissolution chemistry and kinetics, and biocompatibility of bioresorbable materials, then moves to in vivo performance and stability of electrical and optical bioresorbable components, and eventually focuses on the integration of such components into bioresorbable systems for clinically relevant applications. Finally, the technology readiness levels (TRLs) achieved for the different bioresorbable devices and systems are assessed, hence the open challenges are analyzed and future directions for advancing the technology are envisaged.
|
1 Introduction Bioresorbable materials that fully dissolve in the body with biologically benign byproducts provide a unique opportunity to engineer new electrical, optical, and sensing components into implantable biodegradable systems that eliminate any boundary with the human body, granting direct access to organs, tissues, and biofluids without the need of secondary device‐retrieving surgery that may cause tissue lesion or infection. 1 Traditional medical devices designed to be implanted in the human body to treat either acute or chronic diseases (e. g. , cardiac pacemakers, cochlear implants, coronary stents, articular prostheses) are made of (or coated with) inert biocompatible materials and optimized to warrant the organ functionality from days to years, depending of the targeted problem/disease, before surgical retrieval/replacement (when needed). 2, 3 While for some specific cases (e. g. , heart stimulation, hearing loss, coronary heart disease) implanted devices ideally working/operating over the whole patient life are required, in many other cases (e. g. , muscle stimulation, bone growth stimulation, neuro stimulation, wound healing) the devices are required to work/operate only for a prescribed amount of time, so that surgical retrieval is eventually needed, which could be potentially dangerous for patient health. 4 Bioresorbable medical devices designed to be implanted in the human body and operate for a prescribed time, before they undergo spontaneous dissolution in patient body, would therefore represent a new paradigm in patient care. Besides, the continuous diffusion of noninvasive and minimally invasive sensor systems for personalized medicine and sport&wellness activities envisages an ever‐growing electronic waste that poses serious environmental hazards, given the pervasive presence and rapid turn‐over of electronic devices in everyday life. 5, 6, 7, 8, 9 The development of biodegradable electronic components and systems that facilitate disposal by natural dissolution in ambient or mild conditions without environmental threats could mitigate electronic waste problems. 10, 11, 12 Bioresorbable components and biodegradable electronics would greatly benefit one from each other findings, both pursuing the same goal of developing high‐quality devices that operate for a prescribed time and then vanish under well‐defined conditions. In both cases, mechanisms for triggering and tuning the dissolution of these devices by using different physico‐chemical conditions, such as, acidic 13, 14 or basic 15, 16 solutions, light exposure, 17 and low 18 or high temperatures, 19, 20 are envisaged to provide a more accurate on‐demand control over dissolution. Summarizing, an ideal bioresorbable system is required to monitor physio‐pathological parameters of clinical interest (e. g. , intracranial pressure), transmit the acquired data to an external receiver, and produce a feedback action on the relevant organs/tissues aimed at improving the clinical status of the patient. If implanted, the materials that compose the system are required to be biocompatible and degradable in physiological conditions, further resulting in biocompatible degradation byproducts. Eventually, the system operation and performance are required to be stable until programmed (or even better, externally triggered) dissolution occurs, to ensure correct monitoring and effective caring of the patient throughout the healing process, and the degradation products need to be rapidly resorbed and/or drained by the patient organism to avoid the risk of foreign body reactions both at local and systemic levels, especially for a chronic implant. For instance, imagine an infection sentinel system being implanted in a surgical site, able to continuously monitor bacterial charge, releasing antibiotics and sending alerts to physicians in case of infection risk, and that after complete surgical wound healing totally and safely dissolves within days. It would be a groundbreaking achievement for patient life quality and health safety. In spite of the huge research effort paid over the latest years on this subject, many important issues remain open and have still to be addressed toward the achievement of such a bioresorbable medical system, including lifetime of electronic and optical components, stability and lifetime of power sources, on‐demand triggering of material dissolution, development of chemical sensors operating in vivo, and long‐term material biocompatibility in vivo. A summary of the milestones achieved over the last decade on this field is given in Figure 1. Figure 1 Milestones on bioresorbable devices and systems. Milestones achieved over the last decade on bioresorbable devices and systems for in vitro and in vivo applications, highlighting the impressive outcomes of the research field in the last few years. In this Review, differently from recent reviews on transient, degradable, and bioresorbable materials and devices mainly focusing on material preparation and fabrication processes, 21, 22, 23, 24, 25, 26, 27 we will focus on functional operation and physiological dissolution of bioresorbable devices and systems, aimed at assessing the TRL achieved by the different facets of the bioresorbable technology. We will first discuss the dissolution chemistry and kinetics, as well as the biocompatibility of the main bioresorbable materials, including inorganic semiconductors, oxides, and metals, as well as organic polymers, systematically grouping material class and dissolution rate. Next, we will review the state‐of‐the‐art bioresorbable devices and systems, including electronic components, sensors and actuators, power sources, and optical devices. Eventually, we will draw the current picture and envisage future direction on bioresorbable devices and systems, with specific attention to the challenges that still need to be overcome toward real‐word applications. 2 Bioresorbable Materials and Dissolution Chemistry In this section we will review dissolution chemistry and kinetics, and biocompatibility (when available) of the main bioresorbable materials both in vitro and in vivo, namely: inorganic semiconductors (e. g. , silicon (Si), germanium (Ge), and zinc oxide (ZnO)), oxides and nitrides (e. g. , silicon oxides and silicon nitrides), metals (e. g. , magnesium (Mg), zinc (Zn), tungsten (W), and molybdenum (Mo)), and polymers and organic materials (e. g. , polyesters, waxes, cyclic poly(phtalaldehyde) (cPPA), and silk). Figure 2 summarizes the etching rates of these materials in physiological conditions (i. e. , pH 7. 4 and 37 °C), with organic and inorganic semiconductors being mostly used as active part of bioresorbable transistors, metals mainly used as electrodes and for interconnections, polymers typically employed as supporting substrate and flexible coating, and oxides/nitrides being used as encapsulation layer. Figure 2 Dissolution rates of the principal bioresorbable materials in physiological conditions. Summary of the dissolution rates of the different materials used in bioresorbable devices and systems at pH 7. 4 and 37 °C, namely, a, b) semiconductors, 30, 31, 33, 35, 36 c) metals, 43 d) oxides and nitrides, 39 and e) polymers. 52, 53, 54, 55, 56, 57, 58 2. 1 Inorganic Semiconductors Among the different bioresorbable materials, monocrystalline silicon in the form of nanomembranes (Si NMs) represents an emerging choice for the fabrication of high‐performance transient electronics and bioresorbable devices. 28, 29 Hwang et al. 30 studied the kinetics of hydrolysis of monocrystalline Si NMs with different dopants (type and concentration) in different solutions (buffers and biofluids), evaluating in vitro and in vivo cytotoxicity. The dissolution of single‐crystalline Si NMs in aqueous media is governed by Equation (1), namely (1) Si + 4 H 2 O → Si OH 4 aq + 2 H 2 Si NMs (B‐doped 10 16 cm −3, 100 nm thick) were fully and uniformly dissolved in 24 h upon immersion in 1 m phosphate buffered saline (PBS) (at pH 7. 4 and 37 °C) and in bovine serum (at pH ≈ 7. 4 and 37 °C), with a comparable dissolution rate r (by definition, thickness of dissolved material per day) of ≈100 nm day −1. By changing type (P‐ and B‐doped) and concentration (10 17 through 10 20 cm −3 ) of dopants, the dissolution rate was found to be more sensitive to concentration rather than type of dopants, at least for 70 nm thick Si NMs in PBS (0. 1 m, pH 7. 4 and 37 °C). For instance, dopant concentrations of 10 17 cm −3 and 10 19 cm −3 resulted in r ≈ 3 nm day −1 regardless of the type of dopants, whereas at concentration of 10 20 cm −3 the etching rate was ≈0. 8 and ≈0. 3 nm day −1 (roughly a factor 3) for P and B dopants, respectively. In vitro cytotoxicity tests were carried out on metastatic breast cancer cells (MDA‐MB‐231) cultured on Si NMs subjected to continuous dissolution for consecutive days. Although the Si NMs were fully dissolved in 4 days in the culture medium, the cell viability was excellent after ten days (93 ± 4%). In vivo studies were performed by subcutaneous implantation of Si NMs in mice dorsal skin. Biodegradability and biocompatibility (no cytotoxicity and no weight loss in mice) of the implant over 5 weeks were assessed using high‐density polyethylene (HDPE) as Food and Drug Administration (FDA)‐approved control material. After 5 weeks, no residues of degradation were visible at the implant sites by stereomicroscopic analysis, while immunohistochemistry of the skin sections and hematoxylin and eosin (H&E) staining proved similar levels of immune cells to those of HDPE (i. e. , no cytotoxicity). Expanded studies on dissolution kinetics of semiconductors for transient electronics were reported by Kang et al. 31 The authors correlated the dissolution rates of 100 nm thick NMs of polycrystalline silicon (p‐Si), amorphous silicon (a‐Si), silicon–germanium alloy (SiGe), and germanium (Ge) in aqueous solutions with different pH (7–10) and temperature (room temperature and 37 °C) values. Equation (2) applies to germanium dissolution in aqueous media, namely (2) Ge + O 2 aq + H 2 O → H 2 GeO 3 aq Dissolution rates at physiological temperature (37 °C) were higher than those at room temperature for all the tested materials, which is in agreement with the Arrhenius equation. 32 The increase of pH significantly increases the dissolution rate of p‐Si, a‐Si, and Ge NMs. For instance, at physiological conditions (37 °C and pH 7. 4) r was 2. 8, 4. 1, and 3. 1 nm day −1, respectively, while at pH = 10 the NMs were completely dissolved in a few hours, regardless of the material. Conversely, the SiGe alloy (Si 8 Ge 2 (100)) showed higher stability with pH than the other materials, and no significant dissolution was recorded until pH = 8 after 16 days, while only 25 nm were dissolved at pH = 10. Biocompatibility was evaluated by culturing two different cell types (L929 mouse fibroblast and whole splenocytes harvested from mouse spleen) over 72 h onto the NMs for cytotoxicity studies. Cell viability suggested the nontoxic nature of the four dissolved materials, compared to HDPE used as control. Two years later, the same research group deepened the understanding on dissolution kinetics of Si NMs in ground‐water and biofluid media. 33 The authors investigated the dissolution rate of Si NMs (B‐doped, 10 15 cm −3, 200 nm thick) in PBS (1×) spiked with different concentrations of albumin (0. 01–35 g L −1 ), Si(OH) 4 (0–300 mg L −1 ) and cations (Na +, Mg 2+ and Ca 2+, 1 × 10 −3 m ) at 37 °C. The increase of the protein concentration slowed down the dissolution rate due to augmented protein adsorption onto the NM surface; moreover, regardless of the concentration (and presence) of proteins, the dissolution rate reduced by increasing the concentration of Si(OH) 4, consistently with the chemical equilibrium reported in Equation (1). Conversely, the presence of cations in the aqueous medium (i. e. , PBS at pH 7. 4, with 35 g L −1 of proteins at 37 °C) led to an accelerated dissolution rate, which was greater for divalent cations (namely, Ca 2+ and Mg 2+ ) with respect to monovalent cations (Na + ). For instance, the presence of 1 × 10 −3 m Ca 2+ increased the dissolution rates of a factor ≈1. 5, from 35 nm day −1 (in absence of cations) to 51 nm day −1. This result was explained with the ability of divalent cations to deprotonate surface silanol groups, formed by absorption of water on Si surface, and to enhance, in turn, the reactivity of water and siloxane groups, as suggested for SiO 2 dissolution. 34 The Si NMs were eventually investigated as water barrier for encapsulation of bioresorbable electronics. Two different strategies were examined to decrease the dissolution rate in the presence of Ca 2+ : the use of heavily doped Si NMs (B‐doped, 10 20 cm −3 ), for which a decrease of the dissolution rates of ≈40‐fold was achieved; and nanometric superficial oxidation with UV ozone (UVO, 3 nm thick) or O 2 plasma (20 nm thick) of Si NMs, for which a delay of the Si dissolution of 10 and 30 days, respectively, was demonstrated. Yin et al. 35 further reported that in aqueous media the dissolution kinetics of Si NMs was affected by the presence of cations (e. g. , Na +, Ca 2+, and Mg 2+ ) and anions (Cl−, HCO 3 −, HPO 4 2 −) which are commonly available in biologic and environmental fluids at concentrations ranging from 0. 1 to 50 g L −1. The study reported that the dissolution rate of Si NMs (B‐doped, 10 17 cm −3 ) was ≈1 nm day −1 at low ionic concentrations (e. g. , 0. 05 m K 2 HPO 4 /KH 2 PO 4 ), and increased up to 65 nm day −1 at high concentrations (e. g. , 1 m K 2 HPO 4 /KH 2 PO 4 ). Density functional theory (DFT) simulations showed that the presence of anions could weaken the interaction between Si atoms and nearby Si–Si backbones, thus accelerating the dissolution kinetics. Dissolution of ZnO in physiological conditions was recently investigated by Lu et al. , 36 who used ZnO as semiconductor in the fabrication of bioresorbable light‐emitting diodes (LEDs). ZnO samples (200 µm × 200 µm squares, thickness ≈200 nm) immersed in PBS at pH = 7. 4 and 37 °C completely dissolved in 48 h, with a dissolution rate of about 4. 1 nm h −1 (≈100 nm day −1 ). The Zn 2+ concentration, measured at the end of the dissolution tests using a fluorescent chelating agent, namely, zinquin, was estimated to be 33 ng mL −1, confirming a full and direct degradation of ZnO according to hydrolysis reaction reported in Equation (3), namely (3) ZnO + H 2 O → Zn 2 + + 2 OH − without formation of insoluble precipitates of Zn(OH) 2 (solubility ranging from 3. 2–21 µg mL −1, experimentally observed at 25–50 °C, pH 7. 4). Further, possible precipitates of Zn(OH) 2, formed at concentrations exceeding Zn 2+ solubility, could dissolve via formation of elementary Zn 2+ ions and 2OH − hydroxides according to the equilibrium reaction Zn(OH) 2 ↔ Zn 2+ + 2OH − (with solubility constant K ps = 3 × 10 −17 ), driven by the continuous excretion of the dissolved ions (i. e. , Zn 2+ ) from the organism (Le Chatelier's principle). 37 The biocompatibility of ZnO and, in turn, of its byproducts was assessed by monitoring the dissolution of ZnO nanostructures in horse blood and serum. 38 2. 2 Silicon Oxides and Nitrides In the field of transient electronics, silicon oxides and nitrides have also key significance for digital and analog circuits and thin film displays, due to their use as gate and interlayer dielectrics, passivation coatings, and barriers against water penetration. 39, 40 Kang et al. 39 studied the hydrolysis kinetics of thin films of silicon oxides and nitrides in diverse aqueous solutions at different pH levels, ion concentrations, temperatures, and for different deposition methods. Silicon oxides dissolve in aqueous media according to Equation (4), namely (4) SiO 2 + 2 H 2 O → Si OH 4 aq The reaction is initiated and kinetically influenced by OH− concentration (i. e. , pH of solution). At this purpose, the authors investigated the dependence of dissolution rate on pH (i. e. , range 7. 4–12), temperature (i. e. , room temperature and 37 °C), and type of oxide (i. e. , thermal oxide, grown in O 2 (tg‐dry) or H 2 O vapor (tg‐wet), and deposited oxide, obtained through plasma‐enhanced chemical vapor deposition (PECVD) and electron‐beam evaporation (E‐beam)). The dissolution rate increased with temperature (according to Arrhenius equation) 32 and with pH for all the tested oxides. Once temperature and pH values were given, the different oxides were etched at different rates r tg‐dry < r tg‐wet < r PECVD < r E‐beam. In particular, r tg‐dry ranged from 6. 3 × 10 −4 to 1 nm day −1, r tg‐wet from 4. 7 × 10 −4 to 1. 4 nm day −1, r PECVD from 2. 2 × 10 −2 to 14 nm day −1, and r E‐beam from 3. 5 to 282 nm day −1, at pH 7. 4 and room temperature, and at pH 12 and 37 °C, respectively. The different etching rates were explained in terms of different physico‐chemical properties of oxides grown/deposited with different methods, namely: thermal oxide is uniformly dense and has a homogeneous stoichiometry; PECVD oxide has different densities and nonhomogeneous stoichiometry; E‐beam oxide can be affected by nanoscale fragmentation. Stability of thermal silicon oxide as biofluid barrier was further investigated by Lee et al. , 40 who examined the effect of various ionic species present in biofluids, revealing the dependence of the dissolution rate on both cation (Na +, K +, Mg 2+, and Ca 2+ ) and anion (Cl− and HPO 4 2 −) concentrations, near neutral pH values. The authors found that the presence of Ca 2+ accelerates the dissolution rate more effectively than Mg 2+ and Na +, regardless of pH value. For instance, at pH 7. 5 the dissolution rates were 34, 56, and 110 nm day −1 with 30 × 10 −3 m NaCl, 10 × 10 −3 m MgCl 2, and 10 × 10 −3 m CaCl 2 solubilized in the aqueous solution, respectively. This was explained with the enhanced ability of Ca 2+ in Si—O bonding polarization and, in turn, in making it more prone to hydrolysis. These findings were in agreement with former reports on the role of alkali metals in quartz dissolution. 41 Further, the dissolution rate was higher in the presence of HPO 4 2 − rather than Cl−: for instance, for 1 m ionic strength and pH = 6. 8 the dissolution rates were 146 ± 9 and 64 ± 7 nm day −1, respectively. The anionic‐specific increase of dissolution rate with phosphate was explained in terms of strong hydrogen bonding with the silica surface. Similar studies on dissolution kinetics were also carried out for silicon nitride. 39 Silicon nitride dissolves in aqueous media in two chemical steps: 1) oxidation to silicon dioxide, according to Equation (5), namely (5) Si 3 N 4 + 6 H 2 O → 3 SiO 2 + 4 NH 3 and 2) hydrolysis of silicon oxide, according to Equation (4). The overall reaction is reported in Equation (6), namely (6) Si 3 N 4 + 12 H 2 O → 3 Si OH 4 (aq) + 4 NH 3 Being the silicon dioxide an intermediate byproduct of the reaction, the dissolution rate of silicon nitride also increased with temperature and pH, as previously reported for silicon dioxide. Two different silicon nitride deposition techniques were considered: low pressure chemical vapor deposition (LPCVD) and PECVD. The dissolution rate of LPCVD nitride resulted lower than that of PECVD nitride (at pH 7. 4 and 37 °C dissolution rates were 0. 16 and 0. 83 nm day −1, respectively), due to better stoichiometry and higher density of the former. Stacked layers of SiO 2 and Si 3 N 4 were employed in the encapsulation of a magnesium serpentine resistor (300 nm thick) as a proof‐of‐concept application in transient electronics. In particular, it was shown that, by using a triple bilayer of PECVD‐deposited SiO 2 and Si 3 N 4 (single material layer thicknesses of 200, 200, and 100 nm, respectively, resulting in a total thickness about 1 µm) for the encapsulation, a lifetime of roughly 10 days was guaranteed for the resistor in deionized water (DIW) at room temperature. The biocompatibility of SiO 2 and Si 3 N 4 was demonstrated by subcutaneous implantation of both the materials in rodents using a stainless steel wire mesh cage as carrier. Exudates were sampled after 4, 7, 14, and 21 days, and inflammatory response and leukocyte concentrations (leukocytes mL −1 ) were evaluated. The inflammatory response generated by SiO 2 and Si 3 N 4 was not significantly different from that produced by the empty cage used as control, monitored over the experiment duration. 42 2. 3 Metals Metals are interesting in bioresorbable (and implantable) devices for their peculiar mechanical and electronic properties. 43 Mg was successfully used for electrical contacts and interconnections in silicon‐based transient electronics by Hwang et al. , due to its ease of processing, fast dissolution rate in aqueous media, and full biocompatibility. 28 Yin et al. 43 later reported on chemistry and mechanism of dissolution of Mg in deionized water. Mg dissolves in aqueous media according to Equation (7), namely (7) Mg + 2 H 2 O → Mg OH 2 + H 2 Although Mg dissolution macroscopically exhibited high uniformity, scanning electron microscope (SEM) analysis revealed that dissolution occurred with the formation of micropores and needle‐like structures. In addition, X‐ray photoelectron spectroscopy (XPS) analysis showed that the main product of the dissolution was Mg(OH) 2, with MgO and MgCO 3 as byproducts due to the presence of oxygen and CO 2 in ambient atmosphere, respectively. The outer surface was mainly composed of a Mg(OH) 2 layer that featured an increased thickness as the dissolution proceeded, according to experimental results carried out in DIW ( r = 7. 2 nm day −1 ) and Hanks's balanced saline solution (HBSS) ( r = 115 nm day −1 ) as a simulated body fluid. Other metals, including W, Mo, and Zn, showed similar dissolution chemistries and transient behaviors both in DIW and HBSS. 43 Zn (comparably to Mg) showed a fast transient mechanism both in DIW (1. 7 nm day −1 ) and in biofluids (7. 2 nm day −1 ), while W and Mo showed a slower and better tunable degradation both in DIW and HBSS (in the range of 10 −2 nm day −1 ). For this reason, the latter two materials should be preferred for medical devices that require metals to have direct contact with biological tissues, for instance, for physiological electrical signal sensing. The biocompatibility of Mg and its alloys, 44 Fe, 45 and Zn 46 was already largely assessed in vivo (with materials and devices implanted in animal models or humans), through evaluation of inflammatory response, and histological and immunofluorescence analysis. Mo and W, although of considerable interest for bioresorbable devices, have less comprehensive data on biocompatibility related to the concentration of byproducts released. 43 All the discussed metals were deposited via vacuum‐based technologies, such as, chemical vapor deposition and sputtering. However, a huge interest is nowadays directed toward the development of new and low‐cost methods to reliably fabricate bioresorbable metal films, such as laser printing and metal nanoparticle‐filled conductive inks. 47, 48, 49 2. 4 Polymers and Organic Materials Bioresorbable polymers are mostly employed as substrate support for degradable devices and systems. 50 Compared to all the above‐discussed materials, polymeric materials offer higher flexibility in tuning the dissolution timescale. Indeed, polymer dissolution can be triggered by using external signals, 14 and its rate can be tailored by varying molecular weight, material crystallinity, chemical structure, hydrophilic/hydrophobic character, and exploiting bulk/surface erosion mechanisms. 51 Commonly used bioresorbable polymers are poly(lactide) (PLA/PLLA) and poly(lactide‐ co ‐glycolide) (PLGA), 52 poly(caprolactone) (PCL), 53, 54 poly(octanediol‐ co ‐citrate) (POC), 55 poly(glycerol sebacate) (PGS), 56 poly(hydroxybutyrate) (PHB), 57 silk, 58 and, recently reported, candelilla wax. 49 All these polymers are polyesters (or polyamide, silk), and their depolymerization mechanism in aqueous solution is based on acid or basic hydrolysis of the ester (or amide) bond. As an example, Equation (8) reports the hydrolysis of PLGA The dissolution rates (expressed as polymer mass% dissolved per day in PBS at pH 7. 4 and 37 °C) ranged from a minimum value of 0. 22% day −1 for a blend of PCL and PLA (5/95) to a maximum value of 8. 6% day −1 for PGS copolymer (50/50). The biocompatibility of these polymers was largely assessed, as reported in former scientific studies. 59, 60, 61, 62, 63, 64 For instance, PLA and PLGA biocompatibility was demonstrated in vivo through histological and immunologic tests after implantation of microspheres in rats; 59 PCL biocompatibility was assessed through evaluation of the viability of L929 mouse fibroblasts cultured on films of the polymer; 60 biocompatibility of POC was assessed by monitoring form and phenotype of porcine chondrocytes cultured on POC scaffolds; 61 PGS biocompatibility was demonstrated using human cardiac mesenchymal stem cells and rat cardiac progenitor cells cultured on PGS membrane through evaluation of viability after staining with DAPI; 62 biocompatibility of PHB was assessed by subcutaneous implantation in rats and evaluation of the inflammatory response on tissue after 4 and 12 weeks; 63 silk biocompatibility was widely demonstrated both in vitro and in vivo for applications in wound healing and in tissue engineering of bone, cartilage, tendon, and ligament tissues. 64 Won et al. 49 reported on the use of natural waxes as materials for long‐time and hydrophobic encapsulation in bio/ecoresorbable electronics. Soy, myrtle, and candelilla waxes, derived from soybeans (via soybean oil), myrica cerifera (myrtle), and candelilla shrubs, respectively, were tested. Waxes are known to be environmentally and biologically degradable (hydrolysis of ester and anhydride bonds in aqueous solution), conversely to their petroleum‐derived counterparts. Biodegradation tests carried out on candelilla wax (800 µm thick foil) subcutaneously implanted in the dorsal region of rats revealed a thickness reduction of roughly 28 µm after three months. Further, candelilla wax was used as biodegradable matrix for the preparation of conductive pastes, e. g. , loaded with tungsten microparticles (C‐Wax, 800 µm thick foil). Biocompatibility of both candelilla and C‐wax were investigated through subcutaneous implantation in the abdominal region of mice for two months. No remarkable histological changes were recorded after staining of the skin tissue section with H&E (compared to sham surgery skin). In addition, no adverse immunological response was verified after staining with antibody CD45, a pan‐immune cell marker. In 2014, Lopez Hernandez et al. 14 reported an innovative photodegradable or photo‐triggerable transient electronic device using cPPA as substrate. cPPA is a metastable polymer that depolymerizes into monomers once triggered by acidic conditions. The authors developed a transient system based on a film of cPPA and a photoacid generator, namely, 2‐(4‐methoxystyryl)‐4, 6‐bis(trichloromethyl)‐1, 3, 5‐triazine (MBTT), linked to the backbone of cPPA. When the device was exposed to UV light (379 nm), the MBTT generated a highly reactive Cl • radical rapidly forming hydrochloric acid (HCl), that reacted with cPPA triggering its depolymerization. As a proof‐of‐concept, a free‐standing 2. 5% MBTT/cPPA film featuring an array of transistors was fabricated. The MBTT/cPPA film was initially robust under ambient conditions (in air) but the continuous exposition to a UV lamp activated the depolymerization process, transforming the film in an oily and shapeless agglomerate. The whole degradation process lasted about 4 h, though degradation rates could be tuned by modifying the photoacid generator concentration and the irradiance of the UV source. No biocompatibility studies of the material, byproducts, or device were reported by the authors. Compared to the previously reported bioresorbable polyester polymers, silk‐based natural peptide fibers (naturally produced by Bombyx mori larvae) have been reported as a more attractive alternative for the design of bioresorbable sensors and electronics because of the robust mechanical properties, the ability to tailor dissolution and biodegradation rates (from hours to years), the formation of noninflammatory amino acid degradation products, and the option to prepare the materials at ambient conditions to preserve sensitive electronic functions. 58, 65, 66 From the chemical point of view, silk consists of two main proteins, namely, sericin and fibroin, the latter being the structural center of the silk, while the former being the sticky material surrounding it. In vitro dissolution tests of raw silk and of fibroin and sericin filaments (PBS at 37 °C) showed that dissolution occurred with weight losses of 0. 12 and 0. 08% day −1, respectively. 58 3 Bioresorbable Electrical Devices In this section we will review the main electrical components made out, either in total or partially, of bioresorbable materials, namely, active (e. g. , transistors) and passive (e. g. , resistors) components, sensors (e. g. , physical and chemical), and power sources (e. g. , batteries and energy harvesters), paying specific attention to their performance versus biodegradability and biocompatibility in vitro and/or in vivo, where reported. From now on, for those components in which biodegradability and/or biocompatibility of some of the materials were not explicitly investigated with the component itself, the reader should refer to Section 2 of this Review. 3. 1 Electrical Components Several inorganic and organic bioresorbable materials have been used as active semiconductors, gate dielectrics, electrical interconnections, and supporting substrates for the fabrication of degradable electrical components, toward the achievement of more complex electronic circuits and (in perspective) small microprocessors. 3. 1. 1 Inorganic Transistors A great impulse to bioresorbable electronics came from the discovery that implanted silicon nanomembranes dissolve in vivo under physiological conditions, 28 allowing the use of already mature silicon integrated circuit (IC) technology for the fabrication of transient electronic devices. 67 In 2009, Kim et al. 65 proposed silk as a bioresorbable substrate for silicon‐based n‐channel metal‐oxide‐semiconductor field effect transistors (MOSFETs). The transistors were fabricated leveraging silicon technology, i. e. , silicon dioxide as dielectric and gold for electrical contacts deposited and patterned on crystalline silicon NMs (200 nm thick, p‐type) by using photolithography, reactive ion etching, plasma enhanced chemical vapor deposition, and electron beam evaporation. After fabrication, the transistors were transfer‐printed on a silk freestanding membrane using a poly(dimethylsiloxane) (PDMS) stamp. Silk was chosen as substrate because of its excellent biocompatibility and tunable dissolution, which can be tailored by modification of its chemical structure. Threshold voltage and electron mobility of the transistors were around 0. 2 V and 500 cm 2 V −1 s −1, respectively, which were comparable to those of traditional silicon MOSFETs. When immersed in water, the 25 µm thick silk substrate dissolved within 3 min, releasing the transistors in water. Electrical characteristics of the transistors after silk dissolution showed only small variations compared to as‐prepared transistors (threshold and electron mobility of 0. 5 V and 440 cm 2 V −1 s −1, respectively). In vivo studies were carried out by subcutaneously implanting the transistors laying on silk substrate in mice and retrieving them after 2 weeks. The results showed only partial dissolution of both silk film and transistors in this time frame, as well as the lack of any inflammation around the implant site. Although additional studies were required, these initial in vivo experiments suggested promising developments for this form of biodegradable electronics. In fact, three years later (2012) the same research group reported 28 the groundbreaking invention of transient electronic devices made of silicon, allowing facile fabrication of bioresorbable electronic components ( Figure 3 a). In particular, silicon nanomembrane‐based transistors were fabricated using silicon (300 nm thick, p‐type) as active material, SiO 2 (100 nm) as gate dielectric, Mg (300–400 nm) for electrical contacts, MgO (400 nm) and silk (70 µm) as encapsulant layers. The disintegrable n‐channel MOSFETs exhibited threshold voltage and electron mobility of about 0. 6 V and 500 cm 2 V −1 s −1, respectively. In vitro dissolution in DIW at room temperature resulted in stable operation for about 4 days, followed by rapid performance degradation in 30 min (due to Mg electrode disintegration). NAND and NOR logic ports were fabricated and tested in similar ways, as well. Figure 3 Representative examples of inorganic transistors. a) Left: Photographs of an array of completely bioresorbable silicon transistors on silk substrate. 28 Middle‐left: Optical micrograph of a transient silicon nMOSFET inverter on silk substrate, and respective circuit scheme. Middle‐Right: Electrical characteristic of a nMOSFET during in vitro degradation experiment in deionized water at room temperature; inset: optical micrograph of the device configuration used for the degradation experiment. Right: Functional degradation of the silicon transistor. Reproduced with permission. 28 Copyright 2012, American Association for the Advancement of Science. b) Left: Sketch of the ZnO‐based bioresorbable transistor array. 73 Middle: Representative optical microscope images during dissolution experiments of a ZnO trace (left) and a complete transistor (right) immersed in deionized water at room temperature. Right: Functional degradation of a MgO‐encapsulated ZnO transistor during immersion in deionized water. Reproduced with permission. 73 Copyright 2013, John Wiley and Sons. c) Left: Sketch of the foundry‐compatible bioresorbable Si transistor. 67 Right: Representative photographs during device dissolution in buffer at pH 7. 4 and 70 °C. Reproduced with permission. 67 Copyright 2017, The Authors, Published by National Academy of Sciences. One year later, Hwang et al. 68 reported on silicon‐based bioresorbable n‐type MOSFETs and nMOS inverters fabricated on a custom silicon on insulator (SOI) wafer. The transistors were fabricated with standard IC processes on a 100 nm thick p‐silicon layer, using SiO 2 as gate and interlayer dielectrics, and Mg for electrical contacts. Additional SiO 2 (100 nm, PECVD) and Si 3 N 4 (400 nm, PECVD) layers were used as encapsulating and masking materials, respectively. Etching of the silicon on the back‐side was carried out in tetramethyl ammonium hydroxide (TMAH) to release the transistors, which were then transfer‐printed on a silk substrate, before removal of the Si 3 N 4 layer. The fabricated bioresorbable nMOSFETs exhibited threshold voltage and electron mobility around 0. 6 V and 530 cm 2 V −1 s −1, respectively. In vitro experiments in PBS at pH 7. 4 and 37 °C were performed to monitor both kinetics and mechanisms of the transistor dissolution. Magnesium was completely dissolved in 36 h, silicon dioxide in 2 weeks, and silicon in 4 weeks. Monitoring of the transistor performance in PBS at pH 7. 4 and 37 °C was carried out using an additional magnesium dioxide layer (800 nm thick) for encapsulation. A stable operation was recorded for 8 h, followed by performance degradation in the subsequent 45 min. In vivo experiments were carried out with an array of transistors transfer‐printed on a silk substrate and subcutaneously implanted in mice. The results showed full transistor dissolution and no inflammation response after 2 weeks (verified through histological examination of the tissue surrounding the implant), while substrate dissolution was not yet complete. Using similar approaches, silicon transistors were successfully fabricated on different bioresorbable polymeric substrates, such as PLGA, 69 poly(octanediol‐ co ‐citrate), 70 levan polysaccharide, 71 as well as on metal foils. 72 In the same year, Dagdeviren et al. 73 reported on a fully bioresorbable ZnO FET (Figure 3 b). Magnesium oxide was used as gate dielectric, magnesium for electrical contacts, and zinc oxide as semiconductor. Silk fibroin was used both as substrate and as encapsulation layer. The threshold voltage of the transistors was 1 V, and carrier mobility was 0. 95 cm 2 V −1 s −1. Dissolution experiments were performed in deionized water at room temperature: the silk substrate immediately dissolved when immersed in water, while the other components slowly dissolved in 15 h without visible flaking or cracking. To demonstrate functionality in aqueous solution, the transistors were fabricated on glass substrate and encapsulated with magnesium oxide (500 nm thick), resulting in a stable operation over 3 h, followed by a rapid functional degeneration in the next 45 min. All the materials used in this work spontaneously dissolved in water, producing hydroxides (e. g. , Mg(OH) 2 and Zn(OH) 2 ) or polypeptides that were environmentally and biologically benign (as discussed in Section 2 ). More recently (2017), Chang et al. 67 developed foundry‐compatible bioresorbable MOSFETs, inverters, and logic ports (Figure 3 c). The transistors were fabricated on standard 6‐inch SOI wafers by an external foundry using common IC processes and materials, and then (in laboratory) they were transfer‐printed on a PLGA substrate and encapsulated in a further PLGA layer. Eventually, tungsten electrical connections were patterned between different single transistors as needed. Typical nMOSFET voltage threshold and electron mobility were reported to be about 1. 2 V and 630 cm 2 V −1 s −1, respectively. Accelerated dissolution experiments on wafer‐attached transistors were performed in PBS at pH 7. 4 and 96 °C. Encapsulation layers, interlayer dielectrics, and connection layer dissolved within 5 days, while buried thermal oxide dissolved in the subsequent 10 days. The only insoluble (yet biocompatible) components were adhesion layers (100 nm thick) made of titanium and titanium nitride, whose disintegration happened through delamination, peeling, and flake formation. Device integrity when transferred on PLGA substrate was tested in PBS at pH 7. 4 and 70 °C: transistors and interconnections fractured within a few hours, due to PLGA swelling and buckling. PLGA total degradation time was estimated to be around 20 days in the tested conditions (pH 7. 4 and 70 °C). Summarizing, bioresorbable silicon‐based transistors have been successfully fabricated with performance comparable to that of standard silicon devices, and the process has been demonstrated to be scaled up to wafer level using industrial IC processes. Although biocompatibility and dissolution of these devices have not always been assessed both in vitro and in vivo (in animal models), the biocompatibility of materials used for transistor fabrication was formerly assessed in independent studies, as discussed in Section 2 ; performance of isolated devices was only tested in vitro to date. 3. 1. 2 Organic Transistors Following the work of Kim et al. 65 (2009) on silicon‐based n‐channel MOSFETs on a silk bioresorbable substrate, organic bioresorbable transistors have been also proposed, exploiting well‐known degradable materials commonly used in food, textile, and cosmetic industry ( Figure 4 ). Figure 4 Representative examples of organic transistors. a) Left: Sketch of the organic transistor with biodegradable substrate and gate dielectric. 74 Middle: In vitro degradation experiments, performed in citrate buffer at pH 4 and 37 °C. Right: Representative photographs of in vitro device degradation; scale bar is 5 mm. Reproduced with permission. 74 Copyright 2010, John Wiley and Sons. b) Left: Sketch of the biodegradable Electric Double Layer transistor. 79 Middle: Transmittance spectrum and photograph of a fabricated transistor array, showing transparency and flexibility. Right: Representative photographs of in vitro device degradation in DIW. Reproduced with permission. 79 Copyright 2015, IEEE. c) Left: Sketch of the ultralightweight and biodegradable synthetic polymer transistor. 80 Middle: Monitoring of PDPP‐PD degradation in aqueous solution at pH 4. 6 by absorbance spectroscopy. Right: Representatives photographs of in vitro device degradation in pH 4. 6 solution containing 1 mg mL −1 cellulase enzyme. Reproduced with permission. 80 Copyright 2017, The Authors, Published by National Academy of Sciences. Bettinger and Bao in 2010 reported on a FET 74 making use of crosslinked poly(vinyl alcohol) (PVA) as gate dielectric, 5, 5'‐bis‐(7‐dodecyl‐9 H ‐fluoren‐2‐yl)‐2, 2′‐bithiophene (DDFTTF) as p‐channel semiconductor, and silver and gold for electrical contacts, both of which were not biodegradable, though having good biocompatibility (Figure 4 a); PLGA 85:15 was used as substrate. Threshold voltage and mobility of the transistor were about −15 V and 0. 2 cm 2 V −1 s −1, respectively. Upon immersion in citrate buffer (pH 4. 0) at 37 °C, transistor performance immediately started to deteriorate because of DDFTTF delamination from the dielectric, and the device ceased functioning within hours. PLGA structurally started degrading in one month, and the near‐total device resorption occurred approximately in 70 days. Capelli et al. 75 proposed silk fibroin as a gate dielectric in n‐type and p‐type FETs, making use of N, N′ ‐ditridecylperylene‐3, 4, 9, 10‐tetracarboxylic diimide (P13) as n‐type semiconductor, α, ω‐dihexyl‐quaterthiophene (DH4T) as p‐type semiconductor, and indium tin oxide (ITO) and gold for electrical contacts; glass was used as substrate. The p‐type FET had a threshold voltage of −17 V and a mobility of 1. 3 × 10 −2 cm 2 V −1 s −1, while the n‐type FET had a threshold voltage of 2 V and a mobility of 4 × 10 −2 cm 2 V −1 s −1. For n‐type FET, deep‐red (700 nm) light emission was also achieved at V gs = V ds = 90 V. Jeon et al. 76 proposed chicken albumen as potentially biodegradable gate dielectrics of n‐type FET, further using amorphous In–Ga–Zn–O (IGZO) as semiconductor, alumina as passivation layer, and ITO and aluminum as electrical contacts. Transistors were fabricated using either glass or paper as substrate. The FETs fabricated on glass substrate had threshold voltage and electron mobility of 2. 25 V and 5. 64 cm 2 V −1 s −1, respectively, while the ones on paper substrate exhibited 1. 73 V and 6. 48 cm 2 V −1 s −1, respectively. Irimia‐Vladu et al. 77 reported on both p‐ and n‐type semiconductor FETs exploiting glucose, caffeine, or adenine‐guanine bilayers as gate dielectrics (possibly in conjunction with inorganic alumina to increase performance), indanthrene yellow G and indanthrene brilliant orange RF (known in textile and food industry as Vat yellow 1 and Vat orange 3, respectively), β‐carotene, or indigo as semiconductors, and aluminum for electrical contacts. Caramelized glucose and gelatin were used as substrates. Operational voltages and electrical mobilities were in the ranges of 4–20 V and 10 −4 –10 −2 cm 2 V −1 s −1, respectively. Indigo pigment was further investigated by Irimia‐Vladu et al. 78 in a following work, using anodized aluminum passivated with tetratetracontane (a biodegradable long‐chain alkane commonly found in nature) as gate dielectric, and aluminum for electrical contacts; shellac resin was used as bioresorbable substrate. Indigo showed ambipolar semiconductor properties, enabling the fabrication of both p‐type and n‐type FETs, as well as of complementary MOSFET (CMOS) inverters. The FET threshold voltages were between −1. 5 and −3 V for p‐type, and between 4. 5 and 7 V for n‐type transistors, while carrier mobilities were around 10 −2 cm 2 V −1 s −1 in both cases. Indigo performance in n‐type FET degraded upon exposure to air, so that encapsulation with polyimide was used to avoid oxygen‐related degradation. Although the organic materials employed in the fabrication of these transistors were all potentially biodegradable, no degradation, dissolution, or biocompatibility tests were reported for the studies discussed above. Further, the inorganic materials used were not biodegradable, at least under physiological conditions. In 2015 Guo et al. 79 proposed an electric double layer transistor with aluminum‐zinc oxide electrodes directly deposited on a sodium alginate membrane acting as both bioresorbable substrate and active layer (Figure 4 b). Threshold voltage and carrier mobility were −0. 05 V and 6. 19 cm 2 V −1 s −1, respectively. Upon immersion in DIW at room temperature, sodium alginate rapidly dissolved within 5 min, followed by electrode hydrolysis in 1 h. In 2017 Lei et al. 80 reported on pseudo‐CMOS inverters, NAND, and NOR logic ports fabricated using alumina as die‐lectric, a synthetic conjugated polymer as p‐type semiconductor, namely, poly(diketopyrrolopyrrole)‐phenylenediamine (PDPP‐PD), and gold or iron (Fe) for electrical contacts (Figure 4 c). Ultrathin regenerated cellulose was used as substrate. The threshold voltage and carrier mobility with gold electrodes were about −4. 67 V and 0. 21 cm 2 V −1 s −1, whereas iron electrodes resulted in slightly poorer performance, namely, about −5. 75 V and 0. 12 cm 2 V −1 s −1, respectively. Upon immersion in acetate buffer at pH 4. 6, PDPP‐PD, alumina, and iron spontaneously dissolved; conversely, the use of cellulase enzyme (1 mg mL −1 ) was necessary to degrade the cellulose substrate. The entire device was completely dissolved within 30 days in cellulase‐containing buffer at pH 4. 6. PDPP‐PD degradation was further investigated via absorbance spectroscopy in acetic acid solution 1% v/v in water, showing complete depolymerization in 10 days, and total disintegration in 40 days. The biocompatibility of PDPP‐PD was proved through in vitro cell culture with HL‐1 cardiomyocytes on a glass slide coated with the polymer. No significant difference in cell viability was observed after 6 days, compared to cell cultured on a control surface. Summarizing, biodegradable organic transistors have shown poor electrical performance with respect to bioresorbable inorganic semiconductor FETs, and in most cases device degradation was only postulated and concerned with single elements of the whole devices, such as, semiconductor, substrate, or dielectric layers. Only in few cases degradation has been tested in vitro, and never in vivo, and stability of the performance of such transistors upon immersion in aqueous media has been mostly overlooked. 3. 1. 3 Passive Components Bioresorbable passive components (e. g. , resistors, inductors, antennas, capacitors, diodes) have also been proposed 28, 81, 82, 83, 84, 85, 86, 87, 88 either to complement active components (i. e. , transistors) toward the realization of more complex circuits (e. g. , antennas in radio frequency (RF) circuits), or to be used as transducers (e. g. , heaters) in biomedical applications (e. g. , antibacterial therapy). In 2012, Hwang et al. 28 reported on the fabrication of inductive coils, resistive serpentines, and capacitive electrodes using Mg for electrical contacts and conductive traces, MgO or SiO 2 as encapsulation layers, dielectric materials for capacitors and MOSFET gates, and interlayer insulators, Si as active semiconductor in diodes and transistors, and silk as substrate ( Figure 5 a). Single components were fabricated and tested in vitro, demonstrating functionality and lifetime tunability with encapsulation layer type and thickness: for example, a 300 nm thick Mg serpentine with stable resistance over a time interval tunable from 3 h (400 nm MgO) to 100 h (crystallized silk) in DIW at room temperature was reported. A magnesium coil connected to a silicon resistor was demonstrated as an effective in vitro thermal bactericidal device for an Escherichia coli colony in a Petri dish: namely, an external coil was fed with 2 W at 80 MHz for 30 min, causing inductively coupled current flow in the disintegrable circuit and, in turn, resistor heating up to 50 °C. The coil‐resistor circuit was implanted in mouse, resulting in a local temperature increase of about 5 °C upon activation. A metamaterial Mg (400 nm) antenna coated by MgO (600 nm) and encapsulated in silk (100 µm) was implanted in rat and used to monitor biofluid infiltration‐related biodegradation, showing continuous resonance quality factor degradation (Q decreasing from 7 to 2) along 15 days. After this time, the device was retrieved from the animal, showing faint disconnected Mg residues. Figure 5 Representative examples of electronic passive components. a) Left: Photograph of various bioresorbable discrete circuital elements on silk substrate. 28 Middle: Representative photographs of circuit dissolution in deionized water at room temperature. Right: Metamaterial magnesium antenna transmission peak measured during 15 days of implantation in rat. Reproduced with permission. 28 Copyright 2012, American Association for the Advancement of Science. b) Left: Photograph (scale bar: 5 mm) of four wirelessly actuated heaters and sketch of the device materials and structure. 82 Middle: Thermographic camera images (inset: optical micrographs) of activated devices during dissolution in PBS at pH 7. 4 and 37 °C. Right: Doxorubicin release with and without device activation, showing drug retention in the lipid bilayers for several days if device is not activated. Reproduced under the terms of the CC BY license. 82 Copyright 2015, Springer Nature. c) Left: Sketch of the double‐coil antenna with half‐wave rectifier circuit for nerve electrical stimulation. 83 Middle: Representative photographs during device degradation in PBS at pH 7. 4 and 37 °C. Right: Somatosensory evoked potentials in rat with and without nerve electrical stimulation via bioresorbable coil and electrodes. Reproduced with permission. 83 Copyright 2018, Springer Nature. In 2014, Tao et al. 81 used magnesium resistors (200 nm thick) and coils (2 µm thick) on silk substrate (50 µm) to demonstrate thermal effects for direct antibacterial activity and potential on‐demand drug release. The device was characterized in vitro to overheat a bacterial culture on a Petri dish, effectively killing Staphylococcus aureus in the heated zone. In vitro degradation of nonencapsulated devices in deionized water resulted in disintegration after 5 min and complete dissolution within 150 min. In vivo implants in rats infected with S. aureus resulted in sanitizing the wound with two sets of 10 min activations of the wireless resistor at 49 °C (500 mW input power of the external coil, at 80 MHz). After 7 days from implantation, only few traces of the magnesium coil and silk substrate were visible, which were completely resorbed at day 15. In vitro release of bactericidal drug was also demonstrated by coating the device with an ampicillin‐soaked silk layer, and accelerating its dissolution through resistor‐induced heat. One year later, Lee et al. 82 used a very similar approach to fabricate bioresorbable coils for frequency‐addressable drug releasing devices. Molybdenum coils and serpentines were deposited on PLGA substrate, and thermally responsive lipid bilayers were assembled on top, to encapsulate hydrophobic drugs such as antitumoral doxorubicin (Figure 5 b). The number of coil turns defined the circuit resonance frequency, while activation time and power (100–1300 mW) modulated the temperature increase to tune lipid bilayer destabilization and, in turn, drug release. In vitro degradation experiments were conducted in PBS at pH 7. 4 and 37 °C, resulting in stable operation for 7 days, with significant performance loss occurring on day 8. Complete molybdenum dissolution and PLGA resorption were estimated in several months. Ex vivo experiments with porcine skin showed no doxorubicin leak at 24 h from implantation, and significant drug release after 3 h from device activation at 43 °C. In vivo biocompatibility studies carried out on mice for 5 weeks resulted in immune‐inflammatory response comparable to FDA‐approved HDPE (similar levels of inflammatory cytokines and percentages of immune cells in blood), thus suggesting good biocompatibility of both materials and device. Very similar coils were also reported by Koo et al. 83 and Guo et al. 84 for wireless energy transfer via electromagnetic coupling. In particular, Koo et al. 83 used a wireless RF receiver to stimulate peripheral nerves in injured rodents (Figure 5 c). The device consisted in a magnesium double‐coil antenna (50 µm thick), a silicon RF diode (320 nm silicon with 300 nm Mg electrodes) and a magnesium/silicon oxide/magnesium capacitor (50 µm/600 nm/50 µm), connected so as to form a half‐wave rectifier on a PLGA substrate (30 µm), with magnesium electrodes (50 µm) for contacting the nerves with a wrapping cuff. In vitro dissolution tests in PBS at pH 7. 4 and 37 °C showed component disintegration within one month, while in vivo implant residuals were still visible after 8 weeks. The device was tested on living rodents to wirelessly stimulate muscular contraction and accelerate healing of peripheral nerve injury, using the wireless bioresorbable stimulator for 1 h per day, for 6 consecutive days after nerve cut and device implantation. In 2014, Huang et al. 85 reported on a transient printed circuit board (PCB) for the assembling of both active and passive components fabricated using rapidly biodegradable materials. The PCB (23 mm × 23 mm) consisted of stacked layers of transient metal traces and dielectric interlayers on a flexible support. Specifically, sodium carboxymethyl cellulose (Na‐CMC, 50 µm thick) was used as substrate, and Mg, W, and/or Zn (2 µm thick) were used for electrical traces. A layer of poly(ethylene oxide) (PEO) (1 µm thick) between the different Na‐CMC layers of the stack bonded them together. A transient metal paste, consisting of either W or Zn microparticles (4 to 12 µm in average diameter) dispersed in PEO in an organic solvent, was used to mount commercial‐off‐the‐shelf (COTS) components onto contact pads located on the PCB. Evaporation of the solvent at room temperature resulted in a solidified metal/polymer composite with good conductivity properties (up to 40 kS m −1 ), which acted as a conductive glue between the pins of the COTS components and the PCB. Dissolution tests of the PCB were carried out in water at room temperature, with polymeric materials completely dissolving in 10 min, and transient metals dissolving on timescales of hours to days. As a proof‐of‐concept application, a RF circuit that sensed the ambient temperature and transmitted the measured data through a frequency‐modulated wireless signal was assembled on a PCB with two stacked layers. A transmission antenna located 30 cm away from the circuit was used as power source. The frequency of the wireless signal transmitted by the circuit increased from 2. 47 to 2. 49 GHz as the temperature changed from 35 to 18 °C. In 2015, Hwang et al. 86 reported on the use of bioresorbable Mg resistors and antennas, and Si NM p–n diodes for the fabrication of bioresorbable passive circuits that changed functionality over time. In a first example, a serpentine Mg resistor was connected in parallel with a Si NM p–n diode, the latter of which was encapsulated with a layer of MgO whereas the Mg resistor was unprotected. At low electrical biases (up to 1 V) most of the current flowed through the resistor, being its resistance much lower than that of the diode. Upon immersion in water, dissolution of Mg occurred, thus changing the electrical response of the circuit from that of a resistor to that of a diode. In a further example, a Mg resistor was connected to a Mg antenna, where specific regions of the antenna were encapsulated with 400 (region 1) and 800 nm (region 2) thick layers of MgO. The other regions of the passive system were encapsulated with bilayers of MgO (800 nm) and SiO 2 (2 µm). Upon immersion of the system in water, regions 1 and 2 of the antennas progressively dissolved with different times, thus changing length and, in turn, resonance frequency of the antenna from 1. 8 GHz (before immersion) to 1. 9 GHz (regions 1 dissolved), and eventually to 2. 2 GHz (regions 2 dissolved). In 2017, Lee et al. 87 reported on a fully biodegradable micro‐supercapacitors (MSCs) with high electrochemical performance. Fe, Mo, or W (about 300 nm thick) were used as electrode materials and current collectors, agarose gel with NaCl salt as hydrogel electrolyte (150 µm), PLGA (about 15 µm thick film) was used as supporting substrate, and polyanhydride as encapsulating material. MSCs with Mo interdigitated electrodes offered an areal capacitance of 1. 6 mF cm −2 at a current density of 0. 15 mA cm −2, and areal energy density and power density of 0. 14 µW h cm −2 and 61 µW cm −2, respectively. These values were comparable to those of recently reported nontransient supercapacitors. The electrochemical performance of the supercapacitors was measured up to 10 000 charge/discharge cycles. The capacitance of Mo MSCs increased dramatically during the first 3000 cycles when charged at 0. 1 mA cm −2, reaching a maximum value of 4 mF cm −2 (20 times larger than the initial value), and then decreased. This effect was ascribed to additional redox reactions occurring at the metal oxide growing during the capacitor charge/discharge cycles, until the oxide thickness hindered electronic exchange. The performance of Mo MSCs encapsulated in PLGA (about 15 µm) was also studied upon immersion in PBS 10 × 10 −3 m at pH 7. 4 and 37 °C. The MSCs showed stable performance for 6 h after immersion in PBS, then quickly and completely dissolved afterward (at 10 h). To increase lifetime, an Mo MSC was encapsulated and sealed using polyanhydride (150 µm thick), thus extending stable operation time of the MSC up to 36 h. 3. 2 Sensors There is today an increasing demand for biodegradable and bioresorbable sensors to be used in sport&wellness and/or healthcare applications. In this section we will review the main achievements on physical (e. g. , pressure, force, temperature) and chemical (e. g. , ions, pH, biomolecules) bioresorbable sensors. In the aforementioned milestone work of Roger's group, 28 silicon strain gauges (gauge factor near 40) and silicon diodes and magnesium thermoresistive sensors (sensitivity of −2. 23 mV °C −1 and 0. 23% °C −1, respectively) had been proposed, along with Si photodiodes in a 2D array configuration, opening great possibilities for following developments. 3. 2. 1 Physical Sensors Pressure and temperature sensors represent the most studied bioresorbable physical sensors to date. The former exploit variable capacitors, piezoelectric crystals, and strain gauges; the latter mainly exploit thermoresistive materials ( Figure 6 ). Figure 6 Representative examples of physical sensors. a) Left: Sketch of the structure and functional principle of the wireless‐readable capacitive pressure sensor. 92 Middle: Sketch of the in vivo inductively coupled measurement setup. Right: In vivo rat heart rate measurement. Reproduced with permission. 92 Copyright 2019, Springer Nature. b) Left: Sketch of the piezoresistive Si NM pressure sensor. 94 Middle: Representative photographs of device dissolution during immersion in aqueous buffer at pH 12. Right: Calibration of device performance in artificial cerebrospinal fluid. Reproduced with permission. 94 Copyright 2016, Springer Nature. c) Left: Sketch of the piezoelectric pressure sensor. 96 Middle: Photographs of a fabricated device before degradation (left) and after (right) immersion in PBS at 74 °C. Right: In vivo measurement of diaphragm pressure in mouse. Reproduced with permission. 96 Copyright 2018, The Authors, Published by National Academy of Sciences. Luo et al. 89 fabricated fully bioresorbable capacitors using a PLLA substrate sealed with PCL, with metal connections and electrodes made of iron (adhesion promoter) and zinc. Namely, iron (5–10 µm thick) and zinc (50 µm) were electroplated through a mask in order to fabricate a coil (inductor) connected to two disks (capacitor electrodes); the obtained metal traces were then laminated on a poly( l ‐lactide) (PLLA) substrate (200–300 µm), and the assembly was folded so as to obtain a plane capacitor with the metal disks separated by a 30 µm PLLA O‐ring, and sealed with 40 µm PCL spacer. The capacitor was thus enclosed in a passive inductor, capacitor, and resistor (LCR) circuit, and the shift of the resonant frequency of the circuit induced by changes of the capacitance value upon application of a pressure was monitored with an external antenna. The system was characterized in the pressure range 0–20 kPa, showing a sensitivity of about −39 MHz kPa −1 (relative capacitance sensitivity 0. 05 kPa −1 ) in air. In vitro stability tests were performed on the LCR sensing circuit immersed in aqueous saline solution (NaCl 0. 9%) without an external applied pressure, resulting in an increased performance during the first 107 h (sensitivity of −54 MHz kPa −1 ), then followed by rapid performance degradation until hour 170. In vitro degradation of the metal traces was tested using freestanding iron/zinc electroplated electrodes immersed in saline solution at room temperature, resulting in disintegration within 24 h, and almost complete dissolution in roughly 300 h. Boutry et al. 90 fabricated a capacitive pressure sensor using iron (2 nm thick, used as adhesion promoter) and magnesium (100 µm) as capacitor electrodes, printed on polyhydroxybutyrate/polyhydroxyvalerate (PHB/PHV, 25 µm) substrates, separated by a PGS (150 µm) dielectric spacer. To improve device recovery time after pressure application and reduce hysteresis, the PGS surface was patterned with micropyramids. The relative capacitive sensitivity was 0. 76 kPa −1 until 2 kPa, and 0. 11 kPa −1 from 2 to 10 kPa. A 5 × 4 bidimensional grid of sensors was obtained using orthogonal electrical connections in the top and bottom layers. Blood pulse wave from radial, femoral, and carotid arteries was successfully measured on human volunteers by application of the device on the skin. Material degradation was tested by immersing the device in PBS at pH 7. 4 and 37 °C for seven weeks, showing rapid metal dissolution, and rather slow polymeric dissolution, which lasted a few months (after seven weeks, the device retained about 85% of its initial weight). Further, mechanical characteristics of PGS cylinders did not vary significantly during seven weeks of incubation in PBS, due to slow material surface erosion mechanism (instead of rapid bulk erosion, typical in many water‐swelling polymers, like PLGA), suggesting device functional characteristic stability over time. The same group recently reported a hybrid sensor for simultaneous strain and pressure measurement. 91 It was composed by a capacitive pressure sensor (similar to that previously described 90 with Mg electrodes separated by PGS micropyramids) and a capacitive strain sensor (two comb Mg electrodes (100 nm thick) patterned on top of PLLA substrates (50 µm thick) and stacked so that a 50 µm thick PLLA acted as dielectrics). The entire device was encapsulated in poly(octamethylene maleate (anhydride) citrate) (POMaC) and PGS. In‐plane strain caused changes in Mg comb alignment, and, consequently, capacitance variations. The capacitive strain sensor was tested in laboratory environment, showing a sensitivity of about −0. 23 pF % −1, with a capacitance percentage change of 50% at strain of 15%, with fast response time (within milliseconds), good stability to cycling (capacitance drift about 10% after 20 000 cycles between 5% and 10% strains), and no crosstalk between pressure and strain measurements. The device was tested in vitro by immersion in PBS at pH 7. 4 and 37 °C, showing stable performance for 2–3 weeks, with much longer degradation time (after 7 weeks most of the device was still not degraded). In vivo subcutaneous implantation of the sensor in the back of rats resulted in good pressure and strain signal detection after 3. 5 weeks, thanks to POMaC and PGS slow surface erosion mechanism. In vivo biocompatibility studies (both immunohistochemistry and H&E staining) carried out over 8 weeks from implantation showed inflammatory response comparable to that of silicone patch used as control. In another recent work of the same group, 92 a miniaturized capacitive pressure sensor with slightly different geometry was implanted in rat around the femoral artery, and heart rate was wirelessly measured on‐demand, exploiting RF coupling to monitor LCR resonant circuit frequency shift. 89, 93 Namely, PHB/PHV and POMaC (thickness of 10 µm each) were used as bottom and top substrate, respectively, laser‐cut magnesium foil (50 µm) was used for inductive coil antenna and fringe‐field capacitor metal traces, PLLA (40 µm) was inserted as coil spacer, and micropatterned PGS (40 µm) acted as dielectric (Figure 6 a). The sensor was characterized in vitro with both contact and noncontact experiments: for contact measurements, a PDMS slab was pressed (up to 20 kPa) on the device, while in noncontact measurements the PDMS slab was only approached to the device without touching it. In vitro sensor characterization was carried out using an LCR meter directly wired to the device, resulting in a relative sensor sensitivity of about 2. 2% kPa −1 for contact pressure experiments (in the 0–6 kPa range) and 2. 5% nm −1 for noncontact experiments. Noncontact mode was of chief importance because a tight wrapping of the device around blood vessel could hinder blood flow inside it. The device was implanted in rat, wrapping femoral artery with softer POMaC layer facing blood vessel wall, and harder PHB/PHV facing outside, in order to reduce pressure interferences due to respiration movements. Rat heart rate was successfully measured by inductive coupling between the internal coil and an external antenna connected to a vector network analyzer, monitoring real‐time resonance frequency shifts. One week after implantation, the signal had a poorer quality, but it was still recognizable. Partial bioresorption (only PHB/PHV left) was observed after 12 weeks without severe inflammation signs, through immunohistochemistry analysis with CD68, a surface marker of macrophages. Besides variable capacitors, piezoresistive materials that experience resistance changes when subjected to mechanical strain (piezoresistive effect) have been used for bioresorbable pressure sensors. As piezoresistive materials are also (often) sensitive to temperature changes, the same piezoresistive serpentine has been used as temperature sensor if deposited on a rigid surface, and as strain sensor if constrained on a flexible membrane. In 2016, Kang et al. 94 fabricated silicon serpentines (for both pressure and temperature monitoring) by patterning Si NMs (300 nm thick) on a SOI wafer and subsequent transfer‐printing on a flexible PLGA membrane (30 µm thick). For pressure sensing, the latter was suspended over a square cavity (about 30–40 µm deep) etched into a nanoporous silicon layer (about 60–80 µm thick, 71% porosity) used as supporting substrate (Figure 6 b). Magnesium and molybdenum were used for metal traces, and SiO 2 was used for electrical insulation and as encapsulating material. Silicon serpentine fabricated outside the cavity area served as temperature sensor. The pressure sensor was calibrated in vitro using artificial cerebrospinal fluid (ACSF), resulting in strain gauge resistance around 258 kΩ, and sensitivity around 0. 6 kΩ kPa −1, corresponding to a gauge factor of 30. Similarly, temperature sensor in ACSF showed a sensitivity around 0. 1 kΩ °C −1, with a resistance of 74 kΩ at 35 °C, corresponding to a thermal coefficient about 0. 14% °C −1. Device dissolution studies were performed in vitro in buffer solution at pH 12 and room temperature, leading to complete dissolution in about 30 h. To increase the device functional time, a polyanhydride encapsulation was also used, at the cost of a decrease in pressure sensor sensitivity (0. 38 kΩ kPa −1 ). The bioresorbable pressure and temperature sensors were implanted in living rat brain and wired to an external non‐bioresorbable power supply and communication unit fixed on the scalp, successfully proving intracranial pressure and temperature measurement for 3 and 6 days, respectively, and showing no adverse effects in the animal. Biocompatibility studies of polyanhydride were performed 2, 4, and 8 weeks after implantation through H&E staining, showing no overt adverse immune response from rat brain cells, compared to HDPE used as control. An alternative to the external wiring was proposed by using a commercial non‐biodegradable NFC chip subcutaneously fixed on a PLGA substrate with magnesium electrical connection. Besides pressure and temperature, also acceleration, thermal conductivity, and flow rate sensors were fabricated by Kang et al. 94 using serpentine sensors based on patterned Si NMs. Full dissolution, characterization, and calibration of these sensors were not detailed, though their output signals were reported to be comparable to those of commercial non‐bioresorbable control sensors. For instance, an acceleration sensor was fabricated with a piezoresistive strain gauge transfer‐printed on a PLGA cantilever suspended on a cavity etched in the porous silicon substrate. Deflection of the PLGA cantilever, upon an acceleration solicitation, resulted in change of the strain gauge resistance, that was monitored as output signal. Further, a thermal conductivity sensor was achieved using a silicon serpentine resistor. Upon injection of a current through the silicon resistor, its temperature rapidly increased; on the other hand, when the current flow ceased, the temperature of the silicon resistor returned to its initial value with a dynamic that depended on the thermal conductivity of the surrounding medium. Proof‐of‐concept demonstration was given in water, ethylene glycol, toluene, and hexane, though no calibration was reported. Eventually, a flow rate sensor was achieved using two temperature sensors with a silicon resistor in between, spatially arranged in a linear fashion. When the resistor was heated through injection of a current, a temperature gradient arose between the two sensors aligned with the flow direction of the liquid, so that the difference between the temperature sensors was directly related to the fluid flow rate. The sensor was tested in a water bath up to 3 mL s −1 showing a sensitivity around 0. 15 °C s mL −1. Shin et al. 95 reported on a piezoresistive pressure sensor fabricated using silicon strain gauges on a suspended amorphous silica membrane. Two SOI wafers with a 100 nm thick top silicon layer (Si NM) were bonded together using a 3 µm thick PDMS adhesion interlayer. Before bonding, four silicon meander‐shaped resistors (defining the pressure and temperature gauge sensors) were defined on the Si NM of one the SOI wafers, and a square trench with size 200 × 200 × 10 µm (defining the pressure sensitive region) was etched on the Si NM of the other one. Two of the silicon resistors, the ones designed to work as pressure‐sensitive strain gauges, were placed in correspondence of the pressure sensitive region; the other ones, designed to work as temperature‐sensitive gauges, were placed on rigid region of the wafer. Thermal treatment at 550 °C for 2 h calcined the PDMS adhesion layer yielding an amorphous silica layer (about 200 nm thick), which acted as a flexible membrane in the pressure sensitive region, where the silica layer was suspended on the square trench. Chemical removal of the back‐side silicon of SOI wafers resulted in ultrathin, bioresorbable electronic devices encapsulated in thermal SiO 2 layers. The two pressure‐sensitive strain gauges (on the suspended silica membrane) were connected with the two silicon pressure‐sensitive gauges in a Wheatstone bridge configuration, so as to obtain a bioresorbable pressure sensor chip with automatic rejection of thermal effects on pressure measurements. The pressure sensor chip was calibrated in ACSF at pH 7. 4 and 37 °C, showing a pressure sensitivity around 0. 13 Ω mmHg −1 (corresponding to 0. 98 Ω kPa −1 ), and good signal stability (variation in pressure sensitivity within 1. 5%) after 22 days of continuous operation. Temperature gauges were calibrated in a similar way in ACSF, resulting in a temperature coefficient of resistance (relative sensitivity) of 0. 12% °C −1. Accelerated dissolution tests were performed in vitro in PBS at pH 7. 4 and 95 °C, showing that the pressure sensor chip was completely dissolved within 80 h, with only the thermal SiO 2 layer of the bottom SOI wafer left after this time; dissolution time in physiological conditions was estimated to be about 400 days. The pressure/temperature sensor chip was implanted in living rat skull and connected to an external digital multimeter to record voltage data. Commercial clinical pressure and temperature sensors were also implanted and used as control. In vivo pressure (while compressing and releasing the rat flank) and temperature (while applying heating blanket or ice pack to the rat) measurements recorded over a period of 25 days were consistent with clinical control sensors, though sensor accuracy degraded with time, and signals disappeared after day 25 (possibly due to dissolution of metal Mo electrodes). The authors further demonstrated wireless pressure data transmission to a personal computer (PC) from a miniaturized potentiostat wired to the implanted sensor chip and externally secured on rat skull. Insights into the physiological reactions to the proposed pressure/temperature sensor chip were achieved through implantation in rats of a miniaturized version (750 µm × 750 µm × 11 µm vs 1. 3 mm × 1. 3 mm × 16 µm) of the chip without thermal SiO 2 encapsulation. Evaluation of the biodistribution of dissolved silicon resulted in silicon accumulation mostly in kidney, liver, and spleen after one week, and normal level recovery within 5 weeks; no adverse effects were observed in blood analysis and in histological analysis of brain, spleen, heart, and kidney tissues, by comparison between mice with and without implanted sensor. In 2018, Curry et al. 96 fabricated a bioresorbable piezoelectric pressure sensor using PLLA as piezoelectric material. The sensor structure (5 mm × 5 mm in size, 200 µm thick) consisted of two layers of piezoelectric PLLA (27 µm) sandwiched between Mo or Mg electrodes and encapsulated in normal PLA (Figure 6 c). To make PLLA piezoelectric, a PLLA film was stretched at an annealing temperature of 90 °C, so as to simultaneously induce polymer chain orientation and improve material crystallinity. The PLLA film was eventually cut with an angle of 45° with respect to the stretching direction to maximize piezoelectric response. PLLA films with different draw ratios (between final stretched length and initial length) were investigated to assess piezoelectric response (through both impact and vibration testing), showing optimal piezoelectrical response of PLLA sensors with a draw ratio around 5 (for which crystallinity is about 55%). The PLLA pressure sensor was then connected to a charge amplifier circuit to convert force‐induced charge to voltage, and calibrated with predefined weights in the range 0–18 kPa. The calibration curve showed two linear regions, namely, 0–2 kPa with sensitivity of 75 mV kPa −1 and 3–18 kPa with sensitivity of 14 mV kPa −1. Performance degradation of the PLLA pressure sensor was studied both in vitro and in vivo by placing the device in PBS at pH 7. 4 and 37 °C and subcutaneously implanting the device in the back of mice, respectively. In both cases, sensor output signals were comparable before and after 4 days of tests, though after 8 days from implantation there was no detectable signal. Accelerated degradation tests were then performed in PBS at pH 7. 4 and 74 °C, still showing visible device fragments after 56 days. In vivo subcutaneous implantation in the back of mice showed mild immune response after 2 weeks, which reduced to normal levels after 4 weeks. Histological staining with H&E was performed to observe inflammatory cell response, while Masson's Trichrome blue staining was carried out to detect fibrosis; immunohistochemical staining with CD64 antibody was performed to reveal macrophages. As a proof‐of‐concept application, the PLLA pressure sensor was fixed to mouse abdomen and used to measure diaphragmatic contraction of the living animal during anesthesia and subsequent euthanasia via anesthetic overdose. In 2017, Salvatore et al. 97 reported on a biodegradable thermoresistive sensor fabricated by using a magnesium serpentine resistor encapsulated in a compostable, water‐resistant biopolymer, namely, Ecoflex. The magnesium serpentine (10 µm width, 250 nm thick), which was patterned in a fractal design that ensured stretchability of the sensor structure, was sandwiched between silicon dioxide (100 nm) and silicon nitride (100 nm) layers that provided electrical insulation, and eventually encapsulated between two thin films (about 16 µm in total) of Ecoflex. The sensor was calibrated on a hotplate, exhibiting a linear behavior in the temperature range 20–50 °C without significant hysteresis and with absolute sensitivity of about 70 Ω K −1 (relative sensitivity about 0. 2% K −1 ). Sensor resistance was almost insensitive to folding, crumpling, and stretching (variations less than 10%), envisaging wearable and implantable applications. The sensors was tested in vitro in 150 × 10 −3 m saline solution at 25 °C, resulting in reliable continuous temperature measurement for 1 day with a resistance variation of 0. 6%; electrical connectivity was lost after 28 days, and complete dissolution of the SiO 2 /Mg/SiN 3 stack occurred in 67 days. The above‐described design was extended to the fabrication of an array of sensors to demonstrate scalability of the approach and broadening of applications. An array of temperature sensors (9 in total) was then integrated onto an Ecoflex fluidic device (overall size 2. 8 × 2. 5 cm 2 ) to perform a 2D mapping of the temperature distribution. Injection of a warm liquid (35 °C) into the channels and simultaneous monitoring of the sensor array response proved that temperature distribution and flow directions matched those measured with a handheld infrared camera. The possibility of wireless monitoring was also demonstrated using a standard non‐biodegradable microcontroller unit with Bluetooth connectivity, wired to the sensor with Ecoflex‐encapsulated zinc wires. 3. 2. 2 Chemical Sensors Although the concept of chemical sensors operating in vivo is very appealing, bioresorbable chemical sensors have been limited so far to the detection of ions or small organic molecules, probably because of the challenge of enabling reliable and specific sensing of larger and more complex (bio)molecules in vivo. In 2015, Hwang et al. 70 proposed a bioresorbable pH sensor by using FET‐like structures based on silicon nanoribbons (Si NRs). Si NRs (length 500 µm, width 40 µm, thickness 300 nm) with p‐type and n‐type doping (about 10 20 cm −3 ) were defined on a SOI wafer, provided with Mg (300 nm) source and drain electrodes (and electrical interconnections), and coated with SiO 2 (300 nm) acting as interlayer dielectric in the Si NR sensing regions. A further thin layer of SiO 2 encapsulated the devices, except for Si NR gate and electrode connection pads. The FET‐like Si NRs were then transfer‐printed on a POC flexible substrate ( Figure 7 a). The SiO 2 surface of the nanoribbons was functionalized with 3‐aminopropyltriethoxysilane (APTES), which acted as pH‐sensitive probe. In fact, at low pH values, the −NH 2 (amine) groups of the APTES underwent protonation (−NH 3 + ), while at high pH values the −SiOH (silanol) groups of the silicon dioxide surface underwent deprotonation (−SiO − ). The resulting changes in the electrostatic charge of the SiO 2 layer with pH modulated the conductance of the Si NRs by depleting or gathering charge carriers: in p‐doped Si NRs (boron, 10 20 cm −3 ) silicon conductance increased with pH with a sensitivity of 0. 3 µS pH −1, whereas in n‐doped Si NRs (phosphorus, 10 20 cm −3 ) silicon conductance decreased with pH with a sensitivity of −0. 1 µS pH −1, when immersed in PBS 0. 1 m at pH ranging from 4 to 10. Operation lifetime of these devices for pH measurements was investigated with boron‐doped Si NRs during immersion in PBS 0. 1 m at pH from 4 to 10 over 5 days at 37 °C. At pH 7. 4 the conductance of the Si NRs changed by less than 1% during immersion for 5 days. Conversely, unencapsulated Si NR pH sensors exhibited similar changes (less than 1%) after just one day in such conditions. At higher pH (i. e. , 10), a linear decrease in conductance (less than 10%) during this period occurred, as expected. Degradation experiments of unencapsulated Si NR pH sensors were performed by immersion in PBS at pH 10 at room temperature: in 12 h magnesium traces disappeared completely, while POC, Si, and SiO 2 slowly dissolved in several weeks. The same approach was used by the same group to fabricate a pH sensor by functionalizing phosphorus‐doped silicon nanoribbons on a PLGA substrate, obtaining a sensitivity around −0. 3 µS pH −1. 94 Figure 7 Representative examples of chemical sensors and biopotential electrodes. a) Left: Sketch and photographs of the conductometric pH sensor. 70 Right Top: Calibration of p‐doped (left) and n‐doped (middle) Si NR sensors, and sensor stability when immersed in aqueous solutions at different pH values (right). Right Bottom: Representative photographs of device dissolution during immersion in PBS at pH 10. Reproduced with permission. 70 Copyright 2015, American Chemical Society. b) Left: Sketch of the fabrication process of the silk‐based electrochemical electrode. 98 Middle: Photograph of a fabricated sensor on flexible silk substrate. Right: Calibration of the glucose biosensor in ferrocene‐containing solution. Reproduced with permission. 98 Copyright 2016, Elsevier. c) Left: Sketch of the actively multiplexed transistor array. 101 Middle‐Left: Electrical scheme of a single sensing electrode cell. Middle‐Right: Photograph of an 8 × 8 array of multiplexed electrodes. Right: Spatiotemporal mapping of somatosensory evoked potentials in anesthetized rat. Reproduced with permission. 101 Copyright 2016, Springer Nature. An organic, bioresorbable (yet not completely) electrochemical biosensor for specific detection of glucose was reported by Pal et al. 98 (Figure 7 b). A poly(3, 4‐ethylenedioxythiophene)‐poly(styrene sulfonate) (PEDOT:PSS) copolymer (conductive material) was dispersed (19%) in silk sericin‐based photoresist (SPP) to obtain a conductive photocurable ink. By adding 200 U of glucose oxidase (GOx) enzyme to the SPP‐PEDOT:PSS conductive ink, glucose‐selective electrodes were patterned on flexible and biodegradable fibroin protein photoresist (FPP) films. The analytical performance of the electrochemical biosensor was tested in PBS 100 × 10 −3 m at pH 7. 4, with ferrocenyl methyl trimethyl ammonium iodide (0. 1 × 10 −3 m in solution) as redox mediator, at different glucose concentrations. The calibration curve measured for glucose concentrations in the range 0–8 × 10 −3 m showed a linear behavior with sensitivity of 7. 57 µA m m −1 cm −2 and limit of detection of 1. 16 × 10 −3 m. Good selectivity was achieved against other sugars, namely, fructose, galactose, and sucrose (100 × 10 −3 m in solution). GOx‐functionalized sensors were claimed to be stable over time and to operate for 2 weeks after preparation. Degradation experiments on bare SPP‐PEDOT:PSS electrodes (i. e. , without GOx enzyme) on FPP were performed in PBS 10 × 10 −3 m containing 10 U of protease enzyme, resulting in SPP and FPP complete dissolution in 4 weeks. The PEDOT:PSS copolymer was not biodegradable (though biocompatible), and remained as fibrous strands in the solution. Recently, Kim et al. 99 proposed a flexible, bioresorbable electrochemical sensor that employed Si NMs coated with iron‐containing nanoparticles as catalyst to enable the detection of neurotransmitters (i. e. , dopamine (DA)). The sensor consisted of highly doped (boron, 10 20 cm −3 ) crystalline Si NMs (300 nm thick) uniformly decorated with hybrid nanoparticles (NPs) of iron (5 nm) and carboxylated polypyrrole (CPPy, 100 nm) as active sensing elements. Chromium/magnesium metal traces (10/200 nm thick) were employed for electrical contacts and interconnections, and a layer of SiO 2 (100 nm thick) isolated the conducting Mg traces from biofluids and adjacent tissues, apart from the sensing area. As a supporting element, a thin and flexible sheet of PCL was used. Investigation of the detection mechanism highlighted that: dopamine molecules were adsorbed onto the CPPy surface of the hybrid NPs (Fe 3+ _CPPy NPs) via π–π interactions; Fe‐based NPs catalyzed oxidation of DA accumulated on CPPy to form DA‐derived quinone; eventually, electrons generated from the oxidation reaction were transferred to highly p‐doped Si NM‐based electrodes producing a change in their electrical characteristics. The sensor was tested in vitro in PBS 10 × 10 −3 m, showing good sensitivity toward dopamine (a concentration as low as 1 × 10 −12 m was clearly detected), with high selectivity against interferents and other neurotransmitters, e. g. , ascorbic acid, uric acid, epinephrine and norepinephrine (no appreciable responses were recorded at concentration of 1 × 10 −3 m each). Selectivity was explained in terms of π–π interactions between polypyrrole and DA, not likely to happen for the other molecules tested. The electrochemical device spontaneously and fully dissolved in 15 h when immersed in PBS 10 × 10 −3 m at pH 11 and 37 °C (equivalent to 5 days at pH 7). An open challenge on bioresorbable chemical sensors and biosensors, with respect to physical sensors, is related to the sensor protection from dissolution during operation lifetime. In fact, differently from physical sensors, in chemical sensors the sensing material is required to have direct contact with the fluid of interest containing the target analytes, so that they cannot be fully encapsulated with a barrier material. The research on new bioresorbable polymers to be used as sensing material in chemical sensors with optimal characteristics in terms of tunable dissolution is still ongoing. 51 3. 2. 3 Electrodes for Biopotentials Bioresorbable materials, either metallic or semiconducting, have been also used as biopotential sensors with the aim of revealing abnormalities in bioelectric signals related to severe neurological diseases. Campana et al. 100 demonstrated a biocompatible, organic electrochemical transistor able to measure electrocardiographic signals. PLGA (20 µm thick, 75:15 lactide:glycolide) was used as a flexible substrate, on top of which drain and source gold electrodes (30 nm thick, length 30 µm, and width 1000 µm) were patterned. PEDOT:PSS (200 nm thick) was patterned on top of electrodes and employed as organic active material of the transistor, for which biopotential was used as gate signal by directly contacting skin surface with the polymer. Characterization of the electrochemical transistor was carried out (also in the presence of mechanical stress, i. e. , bending the transistor down to a curvature radius of 80 µm) by casting in the channel region a drop of PBS 0. 1 m at pH 7 and room temperature, in which a Pt or Ag wire was immersed and used as the gate electrode. Output curves and transfer characteristic agreed with that expected for a thin film transistor, with low sensitivity to mechanical stress. Assessment of the potentiometric recording capabilities of the bioresorbable electrochemical transistor in a medically relevant setting was achieved by recording a human electrocardiogram (ECG) by attaching the exposed PEDOT:PSS transistor channel directly to skin, using commercial electroconductive gel to promote adhesion and to reduce impedance with skin. In this configuration, the skin replaced the metal gate electrode, so that skin potential changes with respect to a ground contact led to transient fluctuations of the drain current. When operated at V SG = 0. 5 V, with respect to the grounded body, and V SD = −0. 3 V (in order to work with high transconductance and low leakage current), the recorded current trace contained the typical spikes of the heart beat with an amplitude of ≈100 nA, which agreed well with spikes of 500 µV measured with traditional potentiometric recording. The results highlighted the feasibility of realizing simple organic bioelectric interfaces on implantable bioscaffolds, which would allow the recording of signals from muscular or nervous tissue to monitor health state, or which could provide electrical stimulation to influence the tissue activity. Hwang et al. 70 used stretchable Mg (300 nm) serpentines, with SiO 2 as interlayer dielectric and encapsulation layer, transfer‐printed on POC flexible substrate, as capacitive electrodes for precordial ECG and forearm electromyogram (EMG) recording in human volunteers. The Mg electrodes were coupled to the skin through the thin sheet of POC, and the sensing mechanism relied on displacement currents generated by capacitive coupling through the SiO 2 interlayer. The resulting ECG and EMG data were quantitatively comparable to those obtained with conventional gel electrodes. Degradation tests performed in PBS at pH 10 and room temperature showed that Mg and SiO 2 completely dissolved within hours (Mg) or days/weeks (SiO 2 ). One year later, Yu et al. 101 reported on the use of bioresorbable Si NMs with high doping levels for the fabrication of multiplexed passive and active electrodes for neural signal recording. An array of thin, flexible, passive electrodes was fabricated by patterning phosphorus‐doped (10 20 cm −3 ) Si NMs (thickness 300 nm), which were then transfer‐printed on a flexible sheet of PLGA (thickness 30 µm) used as substrate. A layer of SiO 2 (thickness 100 nm) coated the Si connection traces to isolate them from biofluids and adjacent tissue, except for the terminal pads where Si was exposed and used as direct neural interface. Bioresorbable neural electrode arrays with several measurement channels, namely, 4 and 256 (in 16 × 16 configuration, overall area 3 cm × 3. 5 cm), were reported. In vitro characterization of passive electrodes was carried out by immersing the electrodes (placed on a hydrogel substrate) in PBS at pH 7. 4 and room temperature, and monitoring the electrode impedance versus operation frequency (through impedance spectroscopy measurements, with a gold electrode as standard reference). Experimental results showed that the performance of Si NM electrodes was comparable with that of the Au electrode. In vitro degradation tests highlighted that complete dissolution of the passive electrodes was achieved in 4–5 weeks in PBS at pH 7. 4 and 37 °C. Passive electrodes were used for in vivo neural recording experiments in anesthetized adult rat. A craniotomy exposed a 4 × 8 mm 2 region of cortex in the left hemisphere and the Si NM electrodes (four channels) were positioned on the cortical surface for recording physiological oscillations under isoflurane anesthesia, next to a standard stainless‐steel microelectrode used as a control. Experimental results showed that epileptiform spiking activity induced by application of crystals of bicuculline methoxide was successfully recorded by the bioresorbable electrodes, in agreement with the control electrode, also with high signal‐to‐noise ratio (42 for bioresorbable, 32 for control electrode). The bioresorbable electrodes were also implanted on the periosteum and used for high‐fidelity recording of electroencephalogram (EEG) and evoked potentials, successfully capturing theta waves (highlighted in power spectral analysis of the recorded traces) and sleep spindles. Further chronic tests of electrocorticogram (ECoG) recording indicated long‐term stability of operation for more than one month without significant performance change, using electrodes with increased thicknesses of SiO 2 (300 nm) and Si NMs (1000 nm). Sudden failure occurred at day 33, probably due to an open circuit state of interconnections. Studies of tissue reactions to bioresorbable electrode arrays involved chronic implants and were carried out in 14 animals, using Pt electrodes with similar geometries as control. Immunohistochemical analyses showed no significant tissue inflammation at the implantation site, for both Si NM and control electrodes, when compared to the control contralateral hemisphere. Similar passive electrodes were fabricated by and Lee et al. 33 using Mo traces for electrical connection, instead of a single long Si nanomembrane; SiO 2 was used as encapsulation layer, leaving openings for Si recording regions. Yu et al. 101 also reported on arrays (8 × 8 configuration) of active electrodes fabricated by using a similar technology (Figure 7 c). Namely, 128 bioresorbable transistors were fabricated using Si NMs transfer‐printed on a PLGA sheet, employing SiO 2 (100 nm thick) as gate dielectric, Mo traces (thickness 300 nm each) for electrical connections and sensing gate electrodes, and a stack of SiO 2 /Si 3 N 4 /SiO 2 (thickness 300/400/300 nm) as interlayer dielectric and encapsulation layer, with openings for Mo sensing gate electrodes. Every electrode cell was composed of two nMOSFETs, namely, a buffer transistor and a selection transistor, in order to obtain a densely packed electrode array with reduced external connections enabling multiplexed reading. In vitro accelerated degradation of the multiplexed electrode array was investigated in PBS at pH 12 and 37 °C, and allowed to estimate a degradation time in physiological conditions (pH 7. 4) of about 4–6 weeks for PLGA, Si, and Mo, and 6 months for SiO 2 and Si 3 N 4. Eventually, the multiplexed electrode array was used in vivo to measure 2D ECoG in anesthetized rats. Drug‐induced epileptic spikes were successfully recorded, allowing spatiotemporal mapping of different neural wave propagation, highlighting spiral, diagonal, and lateral sweeps. In addition to pathological condition, physiological somatosensory evoked potentials were also recorded on barrel cortex under whisker stimulation, giving amplitude spatial distributions consistent with the activation zone of the solicited whisker. No in vivo degradation tests were reported for multiplexed active electrode arrays. 3. 3 Power Supply Nowadays, an increasing interest toward organic, harmlessly degradable, and even ingestible batteries is rising. 102 Bioresorbable power sources, either batteries or energy harvesters, represent indeed an essential component of implanted bioresorbable devices, regardless of the targeted applications ( Figure 8 ). Figure 8 Representative examples of batteries. a) Left: Sketch of the structure of the Mg–Mo battery. 103 Middle‐left: Photograph of the 4 cell Mg–Mo battery used for powering a LED. Middle‐right: Photographs of the battery degradation during immersion in PBS at 37 °C and 85 °C. Right: Battery discharge under a constant current density of 100 µA cm −2. Reproduced with permission. 103 Copyright 2014, John Wiley and Sons. b) Left: Photograph of the PCL‐encapsulated liquid‐harvesting battery. 104 Middle: Representative photographs of battery degradation during immersion in PBS 1× at 37 °C. Right: Battery discharge in different aqueous solutions under a constant current density of 230 µA cm −2. Reproduced under the terms of the CC BY license. 104 Copyright 2015, Springer Nature. c) Left: Sketch of the structure of the layered‐cathode battery. 108 Middle‐left: Photograph of continuous powering of a LED after 16 h of immersion in PBS. Middle‐right: Representative photographs of battery degradation during immersion in PBS at 37 °C and 85 °C. Right: Battery discharge under a constant current density of 25 µA cm −2. Reproduced with permission. 108 Copyright 2018, John Wiley and Sons. 3. 3. 1 Batteries In 2014, Yin et al. 103 developed a biodegradable battery making use of biodegradable metals for electrodes, and filled with PBS as electrolyte (Figure 8 a). Mg was chosen as anode material because of its high energy density and long shelf‐life (when not immersed in water), while Fe, W, or Mo were used as cathode materials, thus achieving Mg‐X single cell batteries (X = Fe, W, or Mo). The operating voltages were about 0. 75, 0. 65, and 0. 45 V for Fe, W, and Mo, respectively, at discharging current of 0. 1 mA cm −2, and were stable for at least 24 h. A 1 cm 2 Mg–Mo single cell battery containing 8. 7 mg of magnesium (50 µm thick) and 8. 2 mg of molybdenum (8 µm thick) showed an energy capacity of about 2. 4 mA h, when discharged at 0. 1 mA cm −2 for 24 h. Due to the corrosion of Mg foils during operation, the measured capacity was significantly lower than the theoretical capacity of Mg (2. 2 A h g −1 ). In fact, at the magnesium anode two oxidation reactions occurred simultaneously, namely, current‐driven oxidation (Mg→ Mg 2+ +2e − ) and spontaneous water‐driven oxidation (Mg + 2H 2 O→ Mg(OH) 2 + H 2 ). Spontaneous dissolution of magnesium in water reduced lifetime of the battery when filled with PBS, compared to that of nonactivated battery (i. e. , Mg in air), leading to rough halving of the battery energy capacity in 1–2 days after PBS filling. Four Mg–Mo cells were stacked together and encapsulated in a polyanhydride layer, in order to achieve a stable (up to 6 h) output voltage of about 1. 6 V, at a discharging current of 0. 1 mA cm −2. As proof‐of‐concept applications, the stacked battery was used to power a RF circuit and a conventional LED. Battery dissolution experiments carried out by immersion in PBS at 37 °C showed that the polyanhydride degraded after 11 days, leaving partially dissolved Mg and Mo foils; accelerating the dissolution by increasing the temperature to 85 °C resulted in a complete battery dissolution after further 8 days. One year later, Tsang et al. 104 proposed a bioresorbable battery based on the employment of electrolytes usually available in the human body, namely, MgCl 2, NaCl, and PBS, in combination with PCL polymer as separator (Figure 8 b). Magnesium was used as anode and iron as cathode, and the electrodes were separated by and encapsulated in PCL (5 µm thick). Battery performance in various body fluids was investigated by immersion in MgCl 2 (0. 1 m ), NaCl (0. 9% by weight), and PBS solutions (1× at pH 7. 4) at a discharging current density of 230 µA cm −2. Output voltages around 0. 5, 0. 4, and 0. 7 V, and lifetimes of 49, 90, and 99 h, respectively, were achieved, corresponding to specific capacities of 509, 1100, and 1060 mA h g −1. Dissolution experiments were carried out in PBS at 37 °C, showing almost complete magnesium dissolution and device delamination after 20 days, while iron electrode dissolved at a much slower rate, and could last from months to years. To tackle relatively short lifetimes and low power densities of batteries using liquid electrolytes and operating in physiological conditions, Jia et al. 105 proposed a battery that exploited biocompatible (yet not totally bioresorbable) ionic liquid polymer electrolytes. A magnesium‐aluminum‐zinc alloy (AZ31, 200 µm thick) was used as anode material, poly(pyrrole)‐ para (toluene sulfonic acid) (PPy‐ p TS, 40 µm tick) was used as cathode material, and chitosan‐choline nitrate (CS−[Ch][NO 3 ]) was used as polymer electrolyte. Eventually, a thin layer of CS−[Ch][NO 3 ] solution was dropped onto the assembled battery to hold the device components together (like a glue) and improve mechanical integrity. A 1 cm 2 (60 µm thick) single cell battery exhibited a high output voltage (1. 33 V) and long stability (160 h) when discharged at 10 µA cm −2, though performance rapidly degraded at higher discharge currents due to the low ion mobilities, and no stable operation was achieved above 1 V at 0. 1 mA cm −2. No dissolution tests were performed on this battery. The same group next developed biodegradable batteries that used an AZ31 foil anode and either PPy‐ p TS 106 or gold nanoparticles 107 deposited on a silk fibroin substrate as cathode. In the latter work, silk fibroin−choline nitrate (SF‐[Ch][NO 3 ]) was used as polymer electrolyte. 107 When discharged at 5 µA cm −2, a 1 cm 2 single cell battery exhibited a capacity of 2. 2 mA h cm −2 with a plateau voltage of around 1 V. To optimize the battery for implantation, 107 AZ31 alloy was sputtered on crystallized silk fibroin (instead of using a rigid AZ31 alloy foil) and further encapsulated in crystallized silk by thermal processing. The encapsulated battery showed a specific capacity of 0. 06 mA h cm −2, when discharged at 10 µA cm −2, which was lower than that of the unsealed battery (1. 43 mA h cm −2 ), due to both rapid depletion of AZ31 and limited oxygen availability. A stable open‐circuit voltage (OCV) of 1. 21 V for about 180 min was achieved for the encapsulated battery in air, likely due to the corrosion of the AZ31 layer at the interface with the polymer electrolyte. When exposed to PBS, a two‐stage transient behavior was observed, with a stable OCV above 1. 21 V for 64 min (with same discharge profile as in air), followed by a rapid functional degradation in 22 min, explained by a failure in the crystallized silk protection film due to swelling upon water adsorption. The overall encapsulated battery (170 µm thickness) almost totally degraded after 45 days, with AZ31 thin film (500 nm thickness) completely disappearing in 4 h. The SF‐Au film 0. 09 mg Au in 230 mg total device) was physically fragmented in PBS solution because of the degradation of the silk substrate, and Au nanoparticles (being not biodegradable, though considered nontoxic) were possibly cleared by renal excretion in a potential implantation in living humans. More recently (2018), Huang et al. 108 developed a magnesium‐molybdenum oxide (Mg–MoO 3 ) bioresorbable battery with extended lifetime using calcium crosslinked alginate in PBS as electrolyte‐retaining polymer (Figure 8 c). The anode was a magnesium foil (200 µm) and the cathode consisted of MoO 3 nanoparticles mixed with PLGA and cast (150 µm) on Mo foil (30 µm). A polyanhydride layer on top of a PLGA layer was used to encapsulate the battery after fabrication. The battery featured a lifetime of about 13 days and a total energy capacity of 6. 5 mW h cm −2, when discharged at a current density of 25 µA cm −2, with an output voltage around 1. 5 V for 50 h, during which molybdenum oxide was reduced and dissolved, and 0. 6 V during the next 250 h, during which molybdenum foil acted as cathode active material. Remarkably, the high voltage output during the first phase was maintained up to a discharge current density of 150 µA cm −2. As proof‐of‐concept applications, the battery was used to power a digital calculator and a low‐power ECG signal detector. Powering of a standard LED over 16 h upon immersion in PBS was also demonstrated. Degradation of the battery (with magnesium and molybdenum foil thicknesses reduced to 50 and 5 µm, respectively) was investigated in PBS: at 37 °C, the polyanhydride and PLGA encapsulation degraded first, followed by the simultaneous dissolution of Mg, sodium alginate hydrogel, MoO 3 /PLGA layer, and Mo. Most materials completely dissolved within 9 days, except Mo that needed further 10 days at 85 °C to fully dissolve. In vitro cytotoxicity tests of MoO 3 paste electrode encapsulated by a PLGA layer showed a good biocompatibility after co‐incubation with L‐929 mouse fibroblast cells. Further, in vivo subcutaneous implants (using a battery with reduced foil thickness) performed on a rat model demonstrated full degradability of the entire battery in 4 weeks, without apparent inflammatory response on skin tissues and on different organs. All the bioresorbable batteries discussed above are rather bulky compared to other electrical (and optical) bioresorbable components; further, their performance quickly and continuously degrade when operated in biofluids, so that energy provided should be used in a short time to limit losses, while complete dissolution of battery materials require a much longer time. 3. 3. 2 Energy Harvesters Energy harvester devices, such as piezoelectric 109 and triboelectric 110 generators, able to directly and continuously produce energy once implanted in human body, represent an appealing alternative to energy storage devices, such as, batteries ( Figure 9 ). Figure 9 Representative examples of mechanical energy harvesters. a) Left: Sketch of the ZnO piezoelectric energy harvester. 73 Middle: Photograph of a fabricated device, demonstrating flexibility on a silk substrate. Right: Device electrical output during periodic bending. Reproduced with permission. 73 Copyright 2013, John Wiley and Sons. b) Left: Sketch and photograph (inset) of the polymeric triboelectric energy harvester. 111 Middle: Representative photographs of device dissolution during immersion in PBS at 37 °C. Right: Electrical output degradation during in vivo implantation in rat. Reproduced with permission. 111 Copyright 2016, The Authors, Published by American Association for the Advancement of Science. c) Left: Sketch of the optically monitorable silk triboelectric energy harvester. 112 Middle: Photograph of in vitro device degradation during immersion in DIW at room temperature. Right: In vivo demonstration of device degradation after subcutaneous injection of physiological saline solution. Reproduced with permission. 112 Copyright 2018, John Wiley and Sons. Dagdeviren et al. 73 reported a mechanical energy harvester using zinc oxide (ZnO) as piezoelectric material deposited on a silk substrate (Figure 9 a). ZnO strips (500 nm thick) were provided with top and bottom magnesium electrodes (500 and 300 nm thick, respectively) defining an active area of 50 µm × 2 mm. The energy harvester consisted of 60 single Mg/ZnO elements connected in series and parallel, namely, six series groups of ten parallel elements. Application and release of buckling stimuli to the silk substrate led to bending of the ZnO strips, resulting in positive and negative variations of voltage and current outputs. Peak values of about 1. 14 V and 0. 55 nA were achieved, respectively, with a peak output power density of 10 nW cm −2. Dissolution tests were carried out in deionized water, PBS, and serum at room temperature. The silk substrate (about 25 µm thick for this case study) quickly dissolved in water, causing disintegration of the device physical structure. Afterward, all electronic materials, i. e. , Mg, MgO, and ZnO, completely dissolved in 15 h in a controlled manner, without cracking, flaking, or delamination. Triboelectric effect, differently from piezoelectric effect, relies on different charge affinity of two (usually dielectric) materials. When in contact (or even rubbed against each other), electron clouds overlap, and there is an electron redistribution at the two material interfaces. After separation, most of the electrons remain trapped in the new potential wells, resulting in surface polarization of the two materials, that now have an excess and a vacancy of electrons, causing negative and positive interface charge, respectively. This surface charges can produce current flow when the materials (contacted to an electric load) are displaced one respect to each other, being the system equivalent to a capacitor with varying capacitance. 110 Zheng et al. 111 fabricated biodegradable triboelectric nanogenerators using PLGA and PCL films featuring a nanostructured surface (Figure 9 b). The nanogenerator had a multilayered structure. PLGA and PCL slabs (2 cm × 3 cm, 50–100 µm thick) were patterned at the nanoscale on one surface to increase effective contact area, and the two polymer layers were assembled facing each other as friction parts, with a 200 µm thick spacer set between them. A magnesium film (50 µm thick) was then deposited on the back flat side of each friction layer as electrode, and the whole structure was eventually encapsulated in PLGA (75:25 lactic:glycolic ratio, 100 µm). Upon application of a periodic compression force (and subsequent release) at a frequency of 1 Hz to simulate low‐frequency biomechanical motion, a periodic output signal with peak voltage of about 40 V and peak current of about 1 µA was achieved. By connecting the nanogenerator to a load resistance of 80 MΩ, a power density of 3. 26 µW cm −2 was achieved. As a proof‐of‐concept application, primary neurons were repeatedly (at 1 Hz, 10 V mm −1 ) stimulated with the nanogenerator for 5 days, after 24 h of initial culture, showing that the nerve cell growth was successfully orientated, which was crucial for neural repair. In vitro degradation experiments were performed in PBS at pH 7. 4 and 37 °C, showing slow water uptake of PLGA in the first month, followed by a rather rapid swelling and mass loss that led to structural failure of PLGA after 50 days, and, in turn, to complete dissolution of the whole device after 90 days. The biocompatibility of the nanogenerator was assessed by culturing endotheliocytes (ECs) on both PLGA and PCL films (≈100 µm). After 7 days, most of the ECs were viable and showed no significant difference with the reference control group (standard cell culture). In vivo testing was carried out by subcutaneous implantation of the nanogenerator in the dorsal region of a rat, and monitoring the output voltage of the implanted nanogenerator. The output voltage (generated upon finger‐tapping stimulation through skin) was stable for one week, then slowly decayed until complete failure after one month. On the other hand, after 9 weeks from implant, the wound was healed with no infection or inflammation signs observed, and the integrity of the device structure had been destroyed, indicating that most of the materials were biodegraded in the animal body. Recently, Zhang et al. 112 reported a silk‐based triboelectric generator in which the degradation state can be optically monitored (Figure 9 c). Magnesium and silk, which was longitudinally patterned at the microscale, were used as triboelectric materials. The Mg film also functioned as electrode. Silk with different thickness and crystallinity was eventually used as encapsulation material. This silk micropattern had the twofold aim of increasing the triboelectric response and creating a diffractive grating operating in the near infrared (NIR) spectral region (at 1040 nm). Illumination of the triboelectric generator with a laser at 1040 nm produced a diffraction pattern that smoothened and eventually vanished together with the triboelectric response of the generator, as the micropattern structural integrity degraded due to silk dissolution. In vitro characterization of the generator was carried out through application of mechanical impulses at a frequency of 2 Hz to mimic low‐frequency biomechanical motions. An open‐circuit voltage up to about 60 V and a short‐circuit current (absolute value) as high as 1 µA were achieved, depending on the load resistance. A power density of about 3. 85 µW cm −2 was recorded using a load resistance of 100 MΩ. Remarkably, by using different thicknesses and crystallinity degrees for the encapsulating silk, the device lifetime in water was tuned from tens of minutes (unencapsulated) to over 10 h (encapsulated in a double layer of 50 µm thick crystallized silk), with failure being caused by magnesium dissolution after destruction of the encapsulating material. The triboelectric signals and diffractive optical readouts (first‐order diffraction intensities of the silk micrograting monitored at 405, 532, and 650 nm) showed highly positive correlation with the device degradation (in vitro). In vivo experiments were carried out on encapsulated generators implanted in the subdermal region of mice. The implanted generator showed a stable output of about 6 V for 6 h, when a constant external force was repeatedly applied onto the skin of the implanted region. Injection of 10 mL of physiological saline solution at the implant region was used to trigger degradation, with triboelectric function being lost within 30 min. The implant region was reexamined after three weeks, showing gradual reintegration and apparent revascularization in the subdermal layers with no obvious inflammatory reactions detected, so proving good in vivo compatibility of the device. Tests carried out with silk components crystallized to a lower degree (water vapor annealing for 2 h) completely disappeared after three weeks in vivo, whereas for a similar device with a higher crystallization degree (water vapor annealing for 12 h) residues were still observed in the implant region after three weeks. Finally, as a proof‐of‐concept application, the authors loaded penicillin or phenobarbital within the silk encapsulation and/or triboelectric layer and demonstrated in vivo drug release in an animal model. Penicillin release was tested in rats infected at the surgical site with Staphylococcus aureus. Drug loading had negligible effect on the electrical properties of the generator and did not affect silk degradation. Analysis of the infected tissue and bacterial culture on infected samples revealed that infection was healed at day 7 from implantation. The device was also used as epilepsy sensor/medicament by analyzing (with an external unit) the output amplitude and frequency of a phenobarbital‐loaded device. When epileptic symptoms were induced in anesthetized mouse with Penicillin G Sodium injection, the control unit recognized the pathologic state and activated a resistor in order to accelerate silk dissolution and thus increase phenobarbital release: after a 10 min treatment, the epileptic convulsion was remarkably alleviated. Photovoltaic (PV) cells 113 represent an alternative choice to piezoelectric and triboelectric generators for in vivo energy harvesting. Lu et al. 114 reported on a bioresorbable array of photovoltaic microcells that were fabricated using biocompatible and biodegradable materials. The single element of the array of microcells consisted of a thin p–n Si photodiode (thickness 1. 5 µm, dimensions 390 µm × 410 µm, B and P concentrations about 10 20 cm −3, back‐side surface coated with a SiO 2 layer), with patterned Mo (thickness 1. 5 µm) serving as electrodes and for electrical interconnections; PLGA (200 µm thick) was used as both bioresorbable substrate and encapsulation coating. The single element (with a fill factor of about 73%) open‐circuit voltage and short‐circuit current were 0. 40 V and 4. 34 mA cm −2, respectively, when exposed to 1 sun illumination at room temperature (solar simulator with AM 1. 5G filter and 100 mW cm −2 power density). The whole array of photovoltaic microcells (72 in total), connected as 12 columns in series and 6 rows in parallel, resulted in open‐circuit voltage and short‐circuit current of 4. 84 V and 34. 45 mA cm −2, respectively, for a total power of 122 µW (enough to power a commercial blue LED). In vitro dissolution tests of the array of microcells carried out in PBS at pH 7. 4 and 37 °C showed device failure after 5 days from immersion, mainly due to dissolution and water penetration through the PLGA encapsulation layer, whereas complete dissolution of the device required 1–2 months. Ex vivo functional tests were performed with porcine skin (with and without fat), showing degraded performance of the array of microcells due to significant skin absorption in the UV–vis range (transmission without fat about 20% up to 500 nm, and about 60% beyond 800 nm). Placing fat under the skin contributed to an additional transmittance loss up to 25% for wavelengths between 400 and 1100 nm. For instance, upon illumination with a NIR LED (780 nm, 200 mW cm −2 ) under 2 mm thick porcine skin with 2 mm thick fat underneath, a total power of 64. 4 µW was reported. In vitro cytotoxicity was investigated (immunofluorescence live/dead assay) with human umbilical vein endothelial cells (HUVECs) seeded and cultured for 7 days on photovoltaic microcell array as substrate (dimension 2000 µm × 4000 µm × 1. 5 µm), revealing good biocompatibility. Eventually, in vivo tests were carried out by implanting a photovoltaic microcell array (used to power an external blue LED) in the infrascapular region of an adult Sprague–Dawley rat, showing operation (under NIR irradiation) of the implanted device for 3 days. During 4 months from implantation, the rat showed no signs of disease or debilitation, and little/no device residues were apparent under visual and optical microscope (H&E staining) evaluation after this time. 4 Bioresorbable Optical Devices Absorption and scattering of light in tissues result from fundamental light–matter interactions and have enabled a variety of stimulation and monitoring techniques for biomedical therapy and imaging. 115 Very recently, bioresorbable optical waveguides, photonic structures, and optical sensors working in the vis–NIR spectral region and able to operate in vivo in physiological conditions have also attracted attention for photomedicine applications 116, 117, 118 ( Figure 10 ). Figure 10 Representative examples of optical devices. a) Left: Photograph and sketch of a bioresorbable silicon waveguide. 123 Middle‐Left: Representative photographs of a spiral waveguide dissolution during immersion in PBS at 70 °C. Middle‐Right: In vitro calibration of the optical transmittance for oxygen concentration measurement. Right: Comparison between the measured oxygen saturation and the environmental concentration. Reproduced with permission. 123 Copyright 2018, John Wiley and Sons. b) Left: Sketch of the structure and functioning principle of the optical pressure sensor. 124 Middle‐Left: Sketch of the implanted device configuration. Middle‐Right: In vivo calibration of the temperature sensor, against commercial sensor measurements. Right: In vivo calibration of the pressure sensor, against commercial sensor measurements. Reproduced under the terms of the CC BY license. 124 Copyright 2019, The Authors, Published by American Association for the Advancement of Science. c) Left: Sketch and optical microscope images of the bioresorbable ZnO LED. 36 Middle: Optical microscope images during dissolution of the bioresorbable LED in PBS at pH 7. 4 and 37 °C. Right: Electrical characteristic of the ZnO LED; insets: optical microscope images showing device electroluminescence. Reproduced with permission. 36 Copyright 2019, John Wiley and Sons. In 2016, Nizamoglu et al. 119 reported on comb‐shaped bioresorbable waveguides fabricated using silk, poly(vinylpyrrolidone) (PVP), PLLA, and PLGA, which were able to deliver light into living systems down to a depth >10 mm. The thickness of the films was typically adjusted in the range 200–800 µm to achieve a suitable waveguide cross‐section so as to optimize mechanical rigidity/flexibility and optical extraction efficiency, depending on the targeted depth and, in turn, specific application. For instance, PLLA waveguides with cross‐section of 240 × 650 µm 2 featured loss coefficients of 0. 16 dB mm −1 in air, 0. 76 dB mm −1 in water, and 2 dB mm −1 in oil, which were adequate for applications where the majority (90%) of input optical energy has to be delivered through a tissue section down to depth of 10 mm. By increasing the cross‐section to 440 × 580 µm 2, loss coefficients reduced to 0. 15 dB mm −1 in air, 0. 62 dB mm −1 in water, and 1. 5 dB mm −1 in oil. In vitro dissolution experiments of different polymer waveguides were carried out in PBS over 24 h: PVP waveguides dissolved within minutes; silk deformed after several minutes and swelled within hours, but did not fully dissolve; PLGA and PLLA largely retained their physical structure, with modest degradation of the optical transparency over 24 h. To assess biodegradation of polymer waveguides in vivo, transparent pieces of PLGA 50:50 (1 × 5 × 0. 5 mm) were subcutaneously implanted in living mice for 35 days. Shape and transparency of the implant were largely intact at day 6, though significant degradation of both shape and transparency was apparent at day 17, and on day 35 the implant was not visible to naked eye. Visual inspection and histology indicated no signs of inflammation at the implantation site, proving good biocompatibility. In situ waveguide‐assisted photochemical tissue bonding (PTB), which is a dye‐assisted photochemical technique that induces crosslinking between wound surfaces for skin wound enclosure, was demonstrated in animals, using a device with three PLA waveguides to deliver light to porcine skin tissues. Each waveguide was tapered, with a uniform region of 1 mm in width, 440 µm in thickness, and 10 mm in length, and had corrugated edges for optimal extraction of optical energy over the entire length of 10 mm. A full‐thickness incision was made on the excised dorsal skin of a pig immediately after being killed, and Rose Bengal dye was applied to the wound. The polymer waveguides were inserted into the wound, with tissue sides brought in physical contact with the waveguide surface, and a 532 nm laser light at 1 W was launched for 15 min. Eventually, the protruding polymer was trimmed, leaving the comb teeth inside the wound to freely degrade. Tensiometer‐based shear tensile strength measurements showed that the wound treated with the bioresorbable waveguides had a shear strength six times higher than wounds treated with conventional surface‐illuminated PTB (i. e. , 1. 94 kPa vs 0. 33 kPa). In 2018, Fu et al. 120 reported PLLA optical fibers to be used as a bioresorbable optical neural interconnection between biological matter and optical instrumentation, thanks to negligible optical losses and high transmission coefficient (about 95%) of amorphous PLLA in the visible range. 121, 122 The fiber diameter was tuned to about 220 µm, like that of standard silica optical fibers, to facilitate interconnection with commercial optical components. PLLA fibers exhibited a bending stiffness of about 1. 5 × 10 4 N m −1, which is about ten times smaller than that of conventional silica fibers with same geometry (stiffness of 2. 4 × 10 5 N m −1 ) and thus ensured higher flexibility for in vivo biomedical applications. In vitro dissolution experiments were performed in PBS at pH 7. 4 and room temperature for 42 days, during which fiber propagation losses were evaluated. Surface erosion of PLLA fibers occurred with soaking in PBS, resulting in increased light scattering and, in turn, augmented propagation losses (loss coefficient of 1. 64 dB cm −1 at day 0, and 4. 9 dB cm −1 at day 42). Accelerated dissolution experiments were performed in vivo using fibers made of PLGA 50:50 (with shorter degradation time, compared to pure PLLA that required years for complete dissolution). The fibers were implanted in the brain of different three‐month‐old mice at a depth of 2. 5 mm (covering the whole depth of cortex), and perfusion and brain section were performed after different days. On the day of implantation, the lesion created by the PLGA fiber was almost equivalent to that of a silica fiber with similar geometry. As the PLGA fiber gradually and fully dissolved in the brain over the two weeks following implantation, the lesion in the brain region was reduced, and almost disappeared on day 60. Further PLLA fibers were applied for in vivo brain function investigation, including neural signal sensing and interrogation. The PLLA fibers were implanted in the brain of mice and connected to standard optical setups for photometric and optogenetic experiments, namely, enhanced green fluorescence protein recording in deep‐brain hypothalamus, and optogenetical stimulation of hippocampus neurons with a blue laser (at 473 nm) to induce seizures. In both cases the fibers were fully functional at day 0 and gradually decreased their performance until functionality ceased around day 10, though PLLA fibers were not fully dissolved after 10–15 days. However, tissue damage induced at the implantation region was fully recovered after 60 days, suggesting a promising pathway for neural activity research. The latter was evaluated through immunohistochemical staining of neurons using NeuN, a red biomarker for neurons. Bai et al. 123 fabricated flexible infrared waveguides using thin filaments of Si NMs to tackle major drawbacks of PLGA and PLLA waveguides, such as, susceptibility to swelling as a result of water uptake, and limited refraction index contrast, resulting in increased propagation losses and poor optical mode confinement in tissues and biofluids (Figure 10 a). Flexible silicon filaments were obtained from SOI wafers by patterning silicon nanomembranes to form planar coiled structures, for instance, zigzag and spiral waveguides, which were then transfer‐printed on and coated with a PLGA layer (10 µm thick). Eventually, laser milling was used to define strips of PLGA acting as cladding, surrounding the Si filaments, acting as core. The coiled silicon waveguides were designed to withstand an unfurling strain up to 99% without breaking. Optical losses associated with unfurling of the waveguides were reported to be negligible (around 0. 05 dB cm −1 ), compared with optical losses associated with absorption in the PLGA cladding and scattering from imperfections in the silicon filamentary structures (around 0. 7 dB cm −1 in total). Accelerated dissolution experiments of the waveguides (Si core with thickness of 1500 nm and width of 50 µm, PLGA cladding with thickness of 10 µm) were performed in PBS at pH 7. 4 and 70 °C. Transmitted power steadily decreased over time, until (at day 6) losses prevented waveguiding, in agreement with dramatic decrease in transparency of PLGA by day 3, which was consistent with swelling, water uptake, and initial stages of hydrolysis, and consequent dissolution of the silicon core, which was fully degraded by day 10. By selectively removing the PLGA cladding in a local region of the waveguide, the silicon core was exposed to the surrounding medium, thus acting as a transient optical sensor. The Si/PLGA transient optical sensor was used for glucose sensing in mouse blood in vitro, monitoring the return losses at 1160 and 1330 nm in the concentration range 120–180 mg dL −1 (glucose blood concentration of interest in humans is 70–180 mg dL −1 ). Calibration curves exhibited a good linearity in the investigated glucose concentration range, though interfering effects associated with absorption by other biological species were not considered. The transient optical sensor was further used in vitro and in vivo to measure blood oxygen saturation by injecting light in the waveguides (at one end) at 1050 and 1200 nm, and measuring light transmitted through the waveguide (at the other end) at those wavelengths (corresponding to absorption at 1050 and 1200 nm). In vitro tests of hemoglobin in PBS at various oxygenation levels showed calibration curves with a linear response from 5% to 100% oxygen saturation. In vivo experiments were carried out by inserting the transient optical sensor into the subcutaneous region of mice near the thoracic spine and connecting the fibers to external light sources and optical power meters. The experiments involved changing the concentration of oxygen in the experiment chamber to trigger corresponding changes in the oxygen saturation of the blood hemoglobin (SO 2 ). The oxygen saturation measured by the transient optical sensor was consistent with the measurement of a commercial oximeter positioned at the paw of the mice, with uncertainty of about 4% SO 2, mostly arising from motion artifacts, tissue heterogeneities, and venous pulsations. The implanted transient optical sensor was stained with tungsten nanoparticles in order to monitor its dissolution in vivo through microcomputed X‐ray tomography (microCT) imaging, highlighting complete degradation after 15 days. In 2019, Shin et al. 124 reported on bioresorbable optical sensors based on resonant effects, for instance, a Fabry–Pérot interferometer (FPI) and a photonic crystal cavity, for in vivo pressure and temperature measurements (Figure 10 b). Concerning the FPI, starting with SOI wafers, an air cavity was realized as a square through‐hole (250 × 250 µm size) in a Si slab (10 µm thick), which was sandwiched between two Si NM (250 nm thick) diaphragms, using amorphous silica (200 nm thick, obtained from PDMS calcination) as an adhesion layer; thermal SiO 2 (10 nm thick) was used as encapsulation layer. Bioresorbable PLGA (75:25 lactic:glycolic ratio) optical fibers with diameter of 200 nm were used to couple the FPI with commercial optical sources/spectrometers for both pressure and temperature measurements, by monitoring the infrared spectral region around 1500 nm. For pressure sensing, the optical fiber was coupled with the air cavity region of the FPI, whereas for temperature sensing it was coupled with the noncavity (solid) region of the FPI. Pressure‐induced deflections of Si NM diaphragms changed the optical thickness of the air cavity, causing shifts (toward shorter wavelengths) of the position of Fabry–Pérot resonant peaks in the reflection spectrum; temperature‐dependent changes of the refractive index of Si (positive thermo‐optic coefficient) changed the optical path of light traveling within solid Si layer of the FPI and induced, again, shifts of the position of Fabry–Pérot resonant peaks in the reflection spectrum. Experimental results achieved on FPI sensors resulted in a sensitivity of −3. 8 nm mmHg −1 and an accuracy of ± 0. 40 mmHg in the range 0–15 mmHg for pressure sensing, and a sensitivity of 0. 090 nm °C −1 in the range 27–46 °C for temperature sensing. In vitro degradation tests of FPI in PBS at pH 7. 4 and 95 °C showed a complete dissolution of the device within 80 h (corresponding to about 195 days at 37 °C). The bioresorbable fiber performance (i. e. , transmission efficiency) reduced in 3 days when immersed in PBS at 37 °C, due to polymer swelling upon water uptake, and were fully degraded over a period of 3 weeks. Operation lifetime of the FPI was tested in PBS at 37 °C over a period of 8 days using a device with thicker thermal SiO 2 encapsulation layer (thickness 300–1000 nm) and air cavity (thickness 100 µm). A highly stable pressure response throughout the test period (variation within ± 6%) was observed, while that obtained from a device without SiO 2 layer (air cavity thickness 10 µm) under similar test conditions exhibited a gradual increase in sensitivity and baseline over time. In vivo experiments were carried out with FPI sensors implanted in rat skull, successfully measuring intracranial pressures in the range 3–13 mmHg (sensitivity −3. 1 nm mmHg −1, around 1518 nm) and temperatures in the range 34. 9–38. 8 °C (sensitivity 0. 089 nm °C −1, around 1571. 7 nm), with sensing performance that was in good agreement with that achieved in vitro. Eventually, histology of brain, heart, kidney, liver, lung, and spleen tissues collected from animals with an FPI device implanted in the brain after 5 weeks was conducted; the results were comparable with those achieved from a control mouse without any implant. Histopathological evaluation of the acquired images revealed no signs of inflammation, necrosis, or structural abnormality in any organ. Very recently, Lu et al. 36 demonstrated a fully bioresorbable LED based on a II–VI semiconductor, namely, ZnO, and an ultrathin transparent electrode of Mo (Figure 10 c), advancing the state‐of‐the‐art research on this subject, with respect to partially biodegradable riboflavin‐ and peptide‐based organic LEDs, 125, 126 and demonstrating full biodegradability of the two key elements of a LED, namely, direct bandgap semiconductor that allows recombination of injected electrons and holes for light emission, and transparent electrode that allows escaping of the emitted light. Specifically, a ZnO layer (200 nm thick) used as n‐type semiconductor was deposited on top of a Si (12 µm thick) membrane used as p‐type semiconductor, whereas W (100 nm) and Mo (8 nm) were used for electrical contacts. The LED emitted light starting from a threshold voltage of about 5 V, with a rather broad emission spectrum ranging from 420 to 650 nm and a maximum optical power density of 0. 7 mW cm −2, at 9 V (maximum voltage before device damage). Although the threshold voltage was comparable to that of standard ZnO p–n junction LEDs, 127, 128 the intensity of the emitted light was significantly lower. Addition of bioresorbable Fabry–Pérot optical filters based on silicon nanomembranes was proposed to select specific wavelength bands in the visible region and narrow, in turn, the emission spectrum. Dissolution tests were performed in vitro in PBS at pH 7. 4 and 37° C, showing dissolution of thin Mo within 1 day, whereas ZnO and SiO 2 disappeared in 8 and 30 days, respectively, and W degradation occurred in 80–200 days. Full degradation of the Si substrate (12 µm tick) under physiological conditions was estimated to occur in about 5 years, though this time can be reduced by decreasing the Si thickness. No in vivo tests were reported for the proposed LED. 5 Implanted Bioresorbable Systems The ultimate goal of bioresorbable devices is their synergistic integration into an autonomous system to be implanted in the human body, able to monitor parameters of clinical interest, and then degrade on‐demand once no more needed. In spite of the significant number of works in which bioresorbable components have been tested in vivo, only in a few cases two or more bioresorbable components belonging to different functional subsystems (sensor/transducer, electrical/optical readout, power/driving) were interconnected into an implantable and bioresorbable system. In this section, we will review the effort paid over the last decade on the development of implantable bioresorbable opto‐electronic systems, intended as interconnection of two (or more) different bioresorbable functional subsystems. A summary of the bioresorbable devices and systems tested in vivo is reported in Table 1. Table 1 Summary of principal in vivo experiments with bioresorbable devices and systems In vivo experiments carried out with bioresorbable components and systems Bioresorbable device In vivo test Ref. Si‐based transistors with Au electrical contacts on silk substrate. Subcutaneous implantation in mouse for biodegradation and biocompatibility 65 Discrete Mg and Si‐based circuitry elements, i. e. , capacitors, inductors, resistors, diodes, and transistors. Mg antenna with MgO protection on silk substrate and with silk encapsulation. Mg coil and Si serpentine with MgO insulation on silk substrate and with silk encapsulation. Subcutaneous implantation in rodents for biodegradation and biocompatibility. Subcutaneous implantation in rat, with power transfer functionality tests for 15 days. Subcutaneous implantation in rat and wireless power transfer for thermal therapy. 28 Silicon transistors with Mg electrical contacts on silk substrate. Subcutaneous implantation in mouse for biodegradation and biocompatibility 68 Mg resistor on levan polysaccharide. Subcutaneous implantation in mouse back, and irradiation with NIR laser to accelerate degradation. 71 Mg coil and resistor on silk fibroin, with possible antibiotic loading. Subcutaneous implantation in rats for wireless power transfer for thermal therapy and drug delivery. 81 Mo coil and resistor on PLGA substrate, with doxorubicin‐loaded lipid membrane. Subcutaneous implantation in rat for biodegradation and biocompatibility. Subcutaneous implantation in porcine model and on‐demand drug release. 82 Mg coil and capacitor, Mg/Mo electrodes, and Si diode on PLGA substrate. Implantation in rat and electrical nerve stimulation for 10 min every day for 6 days. 83 Si NMs on PLGA sealed on nanoporous Si substrate, with SiO 2 passivation and polyanhydride encapsulation. Implantation in rat brain, for intracranial pressure (3 days) and temperature (6 days) measurement. 94 Si NMs on amorphous SiO 2 suspended membrane on Si substrate; encapsulation with SiO 2. Implantation in rat brain, for intracranial pressure and temperature measurements for 18 days, with complete signal loss after 25 days. 95 Thermo‐mechanically treated PLLA piezoelectric sensor with Mg and Mo electrical connections. Implantation in rat: subcutaneous for bioresorption, under diaphragm to demonstrate force measurement. 96 Mg electrodes on PLLA substrate, with PGS spacer, and POMaC + PGS encapsulation. Subcutaneous implantation in rat, for pressure and strain measurements over 24 days. 91 Mg traces with PGS and PLGA spacers on POMaC and PHB/PHV substrates. Implantation around rat femoral artery for wireless heart rate measurement for one week. 92 Si NMs with SiO 2 encapsulation on PLGA substrate. Similar Si transistors‐multiplexed Mo electrode array. Implantation in rat brain for ECoG measurement over one month. Spatiotemporal mapping of evoked potentials in rat brain. 101 Si NMs with Mo connections with SiO 2 encapsulation on PLGA substrate. Implantation in rat brain for ECoG and evoked potential recording. 33 Battery with Mg anode, Mo/MoO 3 in PLGA paste cathode, polyanhydride encapsulation. Subcutaneous implantation in mouse for biodegradation and biocompatibility. 108 Triboelectric generator in PLGA and PCL with sputtered Mg, PLGA encapsulation. Subcutaneous implant in rat, finger‐tapped every day for one month for performance monitoring. 111 Silk and Mg triboelectric generator with silk separator. Subcutaneous implantation in rat, performance measurement for 6 h and stimulated degradation. 112 Si NM photodiodes with SiO 2 encapsulation and Mo interconnections on PLGA substrate. Subcutaneous implantation in rat, with photovoltaic cells functioning for 3 days supplying a blue LED. 114 PLGA and PVP waveguides. Subcutaneous implantation in rat for biodegradation. 119 PLGA and PLLA optical fibers. Implantation in rat brain for biodegradation, fluorescence measurement and optogenetics. 120 Si waveguide on PLGA substrate. Subcutaneous implantation in rat for oxygen saturation measured by NIR absorption. 123 Si NMs on amorphous SiO 2 suspended membranes on Si substrate, SiO 2 encapsulation and PLGA optical fiber. Implantation in rat brain, for acute intracranial pressure and temperature measurements. 124 PLGA fiber on Si photodiodes with Zn connections on PLGA substrate; encapsulation with SiO2. Implantation in mouse brain, for intracranial temperature and oxygenation measurements; proof‐of‐concept of brain calcium concentration measurement. 132 Mg‐Zn‐Mn alloy (ZM21) stent loaded with CeO 2 and Au nanoparticles. Insertion in dog carotid, demonstrating digital data transfer, reduced immune response and NIR and RF‐induced thermal therapy. 129 John Wiley & Sons, Ltd. A bioresorbable multifunctional stent was proposed by Son et al. , 129 capable of measuring flow and temperature, store and wirelessly transmit recorded data, and release drugs upon stimuli ( Figure 11 a). A standard bioresorbable magnesium‐zinc‐manganese alloy (ZM21) was used as stent mesh material and RF antenna, and acted as substrate for all the others components, namely a resistive RAM (RRAM), 130, 131 a resistive temperature sensor, and a thermoresistive flow sensor. Therapeutic effects were also designed with gold nanoparticles for RF‐ and NIR‐triggered drug release, and cerium oxide nanoparticles as anti‐inflammatory agent. A magnesium serpentine (100 nm thick, with 400 nm MgO encapsulation and 15 µm PLA) served as thermoresistive sensor, with a thermal coefficient of about 0. 08% °C −1 (calibrated in vitro), and as flow sensor, by monitoring serpentine resistance change with time. An MgO (12 nm thick) active layer sandwiched between two Mg electrodes (60 nm thick) was used for the fabrication of the RRAM, with set and reset voltages of −0. 8 and +0. 7 V, respectively, and an estimated retention time of several years. Figure 11 In vivo bioresorbable implanted systems. a) Left: Sketch of the multifunctional bioresorbable stent. 129 Right‐Top: Wireless data transmission between an external antenna and the implanted stent. Right‐Bottom: Thermal camera image of the Au@MSN‐covered implanted stent during NIR irradiation (left) and in vivo thermal‐assisted Nile Red diffusion (right). Reproduced with permission. 129 Copyright 2015, American Chemical Society. b) Top‐Left: Sketch of the structure of the bioresorbable fiber‐photodetector system. 132 Top‐Middle: Optical microscope image of a fabricated device with RGB detector. Top‐Right: In vivo response to brain oxygenation change of the red photodiode, and comparison with tail blood oxygenation measurements. Bottom‐Left: In vivo response of the bioresorbable intracranial temperature sensor, and comparison with thermographic camera measurements. Bottom‐Right: Thermographic camera images of freely moving rat during food searching and eating. Reproduced with permission. 132 Copyright 2019, Springer Nature. Ex vivo testing on canine aorta of a two‐bits RRAM coupled to the flow sensor demonstrated the possibility of measuring and storing data on the same device (data analysis and clustering, power supplying and memory driving were performed by an external PC). CeO 2 nanoparticles for reactive oxygen species (ROS) scavenging were englobed in PLA encapsulating layer, to act as catalyst and reduce the concentration of ROS that promote apoptosis ant restenosis. CeO 2 nanoparticles were tested in vitro with HUVECs and cardiac muscle cells (HL‐1), showing cell viability preservation already after 15 min exposure to 50 × 10 −6 m H 2 O 2, while in control experiments without CeO 2 nanoparticles cell viability dropped to 55% and nearly 0% for HUVECs and HL‐1 cells, respectively. Gold nanoparticles with mesoporous silica shell (Au@MSN) were used to thermally trigger drug release via RF signal or NIR laser exposure. In vivo tests were performed by deploying the bioresorbable stent with CeO 2 and Au@MSN nanoparticles‐soaked PLA encapsulating layer in canine aorta: Nile Red dye was used as drug model to demonstrate increased tissue permeability enabled by Au@MSN heating induced by NIR laser and/or RF coil stimulation, while CeO 2 nanoparticles inhibited macrophage migration and inflammatory response. Moreover, data transmission at 1 Mbps using the stent structure as an RF antenna was successfully demonstrated in vivo (though the stent was externally wired to the network analyzer), with a power transfer efficiency of −21. 8 dB at 900 MHz, at a working distance of 1 cm. Although Au@MSN and CeO 2 nanoparticles were not bioresorbable, injection into an 8‐weeks‐old mouse showed complete clearance after 50 h, with no adverse effects on the animal. No biodegradation experiments were reported for the device. Very recently, a bioresorbable optoelectronic system was reported by Bai et al. , 132 consisting of an optical fiber used to deliver light, a photodetector for generating electrical signals in response to transmitted light, and electrodes for electrical interconnection of the photodetector to external measurement setups (Figure 11 b). The system was assembled into a shape that resembled hypodermic needles (600 µm wide, 160 µm thick, and several mm long) to facilitate minimally invasive implantation. The optical fiber consisted of a 150 µm diameter PLGA (75:25 lactide:glycolide ratio) core coated with an alginate hydrogel serving as a cladding. The photodetector was made from crystalline Si NMs transfer‐printed on a PLGA substrate (10 µm thick) in two different configurations: interdigitated p‐i‐n junctions (1500 nm thickness), and tri‐color stacked junctions (four highly doped silicon layers with different thicknesses and alternating doping, namely, 200 nm (n + ‐doped), 400 nm (p + ‐doped), 1400 nm (n + ‐doped), and 4500 nm (p + ‐doped)). Metal electrodes for readout of the photodetector were made of zinc (thickness 400 nm). Eventually, the photodetector provided with electrodes was encapsulated with sputtered SiO 2 (50 nm). The optical fiber (connected to a commercial laser source) was then placed on top of the 10 µm thick PLGA substrate containing the photodetector/electrode, through the 50 nm thick SiO 2 encapsulating layer. An additional bioresorbable optical filter (distributed Bragg reflectors) made of a multistack of alternating layers of SiO x and SiN x with controlled thickness and periodicity was also reported, to be placed on top of the photodetector with the aim of filtering out laser excitation and/or collect fluorescence emission. A simple bioresorbable spectrometer was demonstrated using the tri‐color photodetector coupled with a PLGA fiber for light delivery, and the four Zn metal electrodes for electric readout. I – V curves (in a dark environment) for the three junctions of the photodetector indicated excellent rectifying behaviors, with dark currents of about 10 −1 –10 −2 µA and responsivities of about 0. 15 A W −1. In vitro biodegradation experiments of the system encapsulated with a 200 nm thick SiO 2 layer were performed in PBS at 37 °C, resulting in stable operation for 10 days, followed by performance degradation (in terms of I – V curves and responsivity of the photodetectors), until at day 25 the system ceased functioning. Computed X‐ray tomography images of the system implanted in the subcutaneous region near the flank of a mouse model showed that different materials and components were bioresorbed at different times, with complete bioresorption achieved on day 45. Further hematology and biochemistry studies were carried out to infer into the biodistribution of elements (Si and Zn, associated with dissolution of the biodegradable spectrometers implanted in mouse models) in blood, brain, heart, kidney, liver, lung, muscle, and spleen tissues explanted from mice at 1, 3, 5, and 7 weeks after implantation. The results in all the measured organs of mice with an implanted device, compared with those in the control group with no implantation, showed no abnormal accumulation of dissolved Si and Zn in the tissues during the 7‐week‐long implantation period. Histological analysis of key organ tissues (i. e. , heart, kidney, lung, and spleen) showed no damages to the tissues and no identifiable immune cells related to implantation. Analysis of complete blood counts and blood chemistry tests also indicated no signs of organ damage or injury, and no changes in the electrolyte and enzyme balance. The system was then implanted in rat brain, showing the possibility to continuously sense cerebral temperature, cerebral oxygen saturation, and neural activity (using a DBR filter coupled with the photodetector and injecting a calcium‐sensitive fluorescent dye, namely, Oregon Green 488 BAPTA 2‐AM) in living animals. Implantation induced minimal inflammatory glial responses (at day 1), which gradually decreased over time to a level comparable to the control group. The implanted bioresorbable systems discussed in this section clearly point out that the missing tile in bioresorbable opto‐electronics appears nowadays to be efficient and stable implantable power supply sources, which are still in their infancy. In fact, though fully integrated circuits for sensor/transducer driving and readout entirely made of bioresorbable materials were not reported yet, their realization has been plainly envisaged by numerous works on bioresorbable electrical and optical, passive and active components. Eventually, a further challenge to be addressed is related to the operation time of the implanted systems, which is rather short with respect to full dissolution time. 6 Open Challenges and Future Directions The analysis of the state‐of‐the‐art literature points out that although bioresorbable technology is still in its infancy, a number of important discoveries and applications have been made since the first report on (partially) bioresorbable transistors, in 2009. Remarkably, biocompatibility (though not taking long‐term effects into account), operation, and biodegradability of a variety of both electrical and optical components have been successfully demonstrated in vivo in animal models, which were also employed to provide several proof‐of‐concept demonstrations of potential clinical applications, through the interconnection of a few components into basic sensing systems ( Figure 12 a and Table 1 ). In spite of these important achievements, there are challenges that still need to be addressed toward real‐world (and commercial, possibly) applications of such a groundbreaking technology, as clearly highlighted by the technology readiness levels (TRLs) 133 of bioresorbable components and systems achieved to date, which range from 2 (technology concept and/or application formulated) to 4 (component and/or system validation in laboratory environment). Radar plots in Figure 12 b, c summarize (to the best of our knowledge) the TRLs of the different bioresorbable components and systems reported in the scientific literature to date. Figure 12 Current picture of bioresorbable devices, systems, and applications. a) Examples of in vivo applications of bioresorbable devices and systems, in animal models. TRL of bioresorbable b) electrical and c) optical devices and systems. Considerations on optical devices are limited to a much lower number of works with respect to electrical devices. The general approach in bioresorbable technology has been based, so far, on the fabrication of a bioresorbable device with a specific functionality, followed by the encapsulation of the device with a protection material. The encapsulation material mainly defines the operation lifetime of the device, by preventing direct contact of the device with biofluids for a prescribed time. Active and passive, electronic (e. g. , transistor, inductor) and optical (e. g. , waveguide, photodiode) components and physical sensors (e. g. , pressure, temperature) have been advanced at a fast pace, driven by the synergistic interaction of well‐established silicon technology and rapid‐prototyping polymer technologies, which has led to bioresorbable electrical and optical devices of high quality, with good performance, tunable lifetime, and wafer‐scale fabrication (in some cases, at least). Nonetheless, only a few, preliminary attempts to push the technology from components to system level have been reported, the most advanced ones being limited to basic interconnection of sensors with driving or readout components (e. g. , use of transistors as switch for multiplexing electrode biopotential measurements, or LCR resonant circuits for wireless transmission with antennas). An open challenge, here, is the fabrication of complex bioresorbable electronic circuits able to promote in situ and in vivo signal amplification, filtering of unwanted signals, signal‐to‐noise ratio improvement, signal modulation for RF transmission. On‐chip fabrication of a bioresorbable operational amplifier (OP‐AMP) composed of tens of transistors connected to resistors and capacitors with simultaneous operation is envisaged to advance the technology, OP‐AMPs being the building‐block of driving, readout, and power management of advanced analog electronic circuits. Further, wireless transmission of signals monitored in situ has been mostly limited to shift of the position of resonance frequency and wavelength of electrical and optical resonant devices, respectively. Electronic circuits enabling analog and/or digital modulation of physicochemical signals recorded in vivo (and demodulation of external analog/digital signals transmitted to impart precise commands/actions to implanted devices) are required to enable an effective wireless data communication and avoid, in turn, wire cables exiting the patient body through skin breaches that are prone to infections. Besides, chemical sensors and power sources, both electrical and optical, which are essential components of any bioresorbable system designed to operate in vivo, have fallen significantly behind other components. As to bioresorbable chemical sensors, these have been mostly limited to pH sensing in vitro and oxygen and calcium sensing in vivo. An open challenge, here, is related to stable in vivo operation of these sensors from medium to long time. In fact, differently from active/passive components and physical sensors, in chemical sensors the sensing material must be in direct contact with the biofluid containing the target analytes, so that protection by full encapsulation of these class of sensors with a barrier material is intrinsically not achievable. Research on novel sensing materials and/or functionalization techniques to be used in chemical sensors, with designed specificity to target analytes, stable operation in contact with biofluids, and tunable dissolution on request, is envisaged to advance the field of bioresorbable chemical sensors. Concerning bioresorbable power sources, all the (primary) batteries reported so far share a low specific energy, with short operation lifetime and rather long full‐degradation time. Water‐based (biocompatible) electrolytes of these bioresorbable batteries lead to enhanced degradation of electrodes (e. g. , magnesium‐based anodes), thus limiting the effective energy and the operation lifetime. A challenge, here, is the fabrication of medium to long lifetime high‐performance primary batteries able to operate in vivo (not yet demonstrated) to be used as power supply of implanted devices and systems. In fact, a conventional, external power supply has been commonly used with bioresorbable devices and systems, so far. On the other hand, piezoelectric and triboelectric generators and photovoltaic cells, whose functionality has been already demonstrated in vivo, are unlikely to provide enough continuous energy to power devices and systems over medium to long operation time, being bioresorbable energy storage devices to be connected with not yet available. In this case, the development of a rechargeable (secondary) battery to be used with harvesters would enable energy accumulation and storage, then allowing powering of more complex circuits (e. g. , OP‐AMPs) from medium to long times. Despite optical fibers and photovoltaic and photodetector cells have been assessed in vivo, the development of bioresorbable optical components has been overlooked with respect to the electrical counterpart, with fully bioresorbable LED only recently reported, in fall 2019. Nonetheless, the LED operation has been only reported in vitro with poor performance, both from electrical and optical point of views. Here, an open challenge is connected with the development of light sources, such as LEDs and lasers, able to operate in vivo, to allow bypassing drawbacks related with absorption of tissues and organs in the UV–vis region. In situ light stimulation with implanted LEDs and lasers locally connected with wavelength‐selective photodiodes would allow to avoid connection with externally tethered light sources and signal analyzers. Eventually, lifetime of all bioresorbable device reported to date ultimately depends on the spontaneous dissolution rate and mechanism of the encapsulating materials in contact with the biofluid, in physiological conditions. Hydrophobic organic polymers (e. g. , polyanhydrides) and inorganic materials (i. e. , silicon oxides/nitrides) have shown to be excellent encapsulation/barrier layers against water infiltration and to guarantee slow surface erosion for tunable lifetimes. Nonetheless, full degradation of bioresorbable components, once operation lifetime has expired, usually takes a much longer time with respect to operation time, lasting from months to years compared to a few days of device operation. Such a discordance between rapid degradation of performance and slow dissolution of materials in bioresorbable components reported to date represents an open challenge to be addressed in future research toward healthcare commercial applications. Further, being dissolution rate of bioresorbable materials depending on biofluid characteristics (e. g. , temperature, pH, and ionic strength), operation lifetime is only roughly preset, especially in vivo. A smart encapsulation layer featuring triggered dissolution upon an external or internal stimulus is envisaged, which allows setting the material lifetime ad‐hoc, depending on medical application needs and patient clinical evolution, by triggering device dissolution. The trigger is required to be specific, to avoid accidental dissolution of the device, and reliable, to ensure dissolution of the device on‐demand. Trigger mechanisms reported so far mostly rely on acceleration of the material dissolution in harsh environments, such as, acidic or basic solutions and high temperatures, and photoinduced depolymerization using light stimuli of suitable wavelength, which requires photosensitizer molecules and photoacid generators. However, none of these approaches has been evaluated in vivo, either in terms of specificity and reliability or biocompatibility. An interesting strategy to be explored is the use of thermoresponsive polymers that change molecular configuration (and hence solubility) when cooled below a tunable threshold temperature (namely, lower critical solution temperature (LCST)), e. g. , poly( N ‐isopropylacrylamide) (PNIPAAM), which has shown good biocompatibility. 134 7 Conclusions In the last decade an increased research effort has been directed toward the use of biodegradable materials for the fabrication of electrical and optical components designed to operate in vivo for a prescribed time and then dissolve within the body in harmless byproducts, namely, bioresorbable technology. This new trend has germinated from former efforts on biocompatibility and biodegradability investigation of materials used in conventional implantable devices for chronic diseases, perfectly complementing them by providing a means to address short to medium terms medical applications. The challenge is the realization of bioresorbable systems designed to be implanted in the human body and interact with tissues and organs to monitor physiological parameters and/or delivery therapeutic agents, and then dissolved upon request in the body itself generating safe byproducts, without needs for surgical retrieval. Over the last decade several milestones have been achieved, which have brought such a bioresorbable technology to TRL 4, by demonstrating in vivo operation of a number of devices and systems for some specific clinical applications (e. g. , neural activity monitoring, intracranial pressure and temperature recording). On the other hand, many challenges are yet to be addressed toward commercial, real‐world applications of such a technology, such as in vivo operation of bioresorbable batteries, and on‐demand bioresorbability of encapsulation coatings. Although, in our opinion, real‐world applications on humans will require medium to long terms (5–10 years) research, we do believe that the potential of such a bioresorbable technology is enormous and it can be envisaged that the efforts on such a research topic will increase in the next few years, driven by both clinical and market requests for resorbable devices and systems. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201902931
| 2,020
|
Advanced Science
|
BMP‐2 Signaling and Mechanotransduction Synergize to Drive Osteogenic Differentiation via YAP/TAZ
|
Abstract Growth factors and mechanical cues synergistically affect cellular functions, triggering a variety of signaling pathways. The molecular levels of such cooperative interactions are not fully understood. Due to its role in osteogenesis, the growth factor bone morphogenetic protein 2 (BMP‐2) is of tremendous interest for bone regenerative medicine, osteoporosis therapeutics, and beyond. Here, contribution of BMP‐2 signaling and extracellular mechanical cues to the osteogenic commitment of C2C12 cells is investigated. It is revealed that these two distinct pathways are integrated at the transcriptional level to provide multifactorial control of cell differentiation. The activation of osteogenic genes requires the cooperation of BMP‐2 pathway‐associated Smad1/5/8 heteromeric complexes and mechanosensitive YAP/TAZ translocation. It is further demonstrated that the Smad complexes remain bound onto and active on target genes, even after BMP‐2 removal, suggesting that they act as a “molecular memory unit. ” Thus, synergistic stimulation with BMP‐2 and mechanical cues drives osteogenic differentiation in a programmable fashion.
|
1 Introduction Chemical and physical cues act upon cells to mediate a wide range of behaviors, including growth, differentiation, and survival. [ 1, 2 ] Mechanotransduction enables cells to sense and respond to physical cues like extracellular matrix (ECM) viscoelasticity, ligand density, and topography. [ 3, 4, 5, 6 ] These mechanical cues are translated into biochemical signals, ultimately activating nuclear transcription factors that control gene transcription for downstream molecular outputs and phenotypic maintenance or diversity. [ 7, 8 ] Separately, biochemical cues provided by soluble growth factors regulate a variety of cellular processes, including the induction of cell fate during development. Growth factors initiate numerous signaling cascades by binding to and activating complementary receptors. [ 9 ] While the activity of several growth factors has been shown to be dependent on mechanical cues, [ 10 ] the crosstalk between mechanotransduction pathways and growth factor signaling is not yet well‐understood. Bone morphogenetic protein 2 (BMP‐2) belongs to the transforming growth factor β (TGF‐β) superfamily. Beyond its canonical role in initiating the differentiation of osteoprogenitor cells into mature osteoblasts, it also is capable of stimulating the transdifferentiation of non‐osteogenic cells into osteoblasts. [ 11 ] The BMP‐2 signaling cascade is activated via interactions between BMP‐2 and its heteromeric transmembrane receptor composed of types I and II serine/threonine kinase receptors. This activation induces recruitment and phosphorylation of Smad family signal transducing proteins Smad1/5/8. Heteromeric complexes are subsequently formed between phosphorylated Smad1/5/8 and the common mediator Smad4. These complexes then accumulate in the nucleus [ 12 ] to regulate target gene transcription. [ 13, 14 ] Several previous studies have noted that the efficiency of BMP‐2‐induced osteogenic differentiation is highly dependent on cell shape, cytoskeletal tension, cell–ligand interactions, and matrix stiffness. [ 15, 16, 17, 18 ] In other words, BMP‐2 signaling and mechanotransduction pathways appear tightly interconnected. [ 14, 19 ] A biological example of crosstalk related to cell shape regulation exists in the soft articular joints or soft bone marrow, where chondrogenic progenitor or adipogenic precursor cells, respectively, exhibit a spherical phenotype and resist being forced into an osteogenic fate when exposed to local BMP gradients. [ 15 ] Here, we examine how mechanical cues, that is, matrix stiffness and cytoskeletal tension, affect biochemical signaling in a key step of the BMP‐2‐induced Smad1/5/8 signaling cascade. We then identify which step of osteogenesis towards an osteogenesis lineage is suppressed when cells lack cytoskeletal tension. Hippo pathway effectors YAP (Yes‐associated protein) and TAZ (transcriptional coactivator with PDZ‐binding motif, also known as WWTR1), which are well‐known mechanosensitive mediators of mechanical cues, [ 8 ] are highlighted as key transcriptional regulators that coactivate osteogenic genes with Smad1/4/5/8 heteromeric complexes. Finally, we provide evidence that Smad signaling activity persists even after the removal of BMP‐2 in a process that serves as a “molecular memory unit” for BMP‐2 signaling. Ultimately, programmable BMP‐2 stimulation and mechanical cues were synergistically utilized to induce and enhance osteogenic differentiation. 2 Results 2. 1 Cytoskeletal Tension Mediates BMP‐2‐Induced Osteogenic Differentiation To investigate the role of mechanotransduction in BMP‐2 signaling, we monitored BMP‐2‐induced osteogenic differentiation of C2C12 mouse myoblasts on polyethylene glycol (PEG) hydrogels, which were functionalized with RGD peptides to promote cell adhesion. C2C12 cells are myoblasts that are able to transdifferentiate into osteoblasts upon stimulation with BMPs and represent an established in vitro model system to study BMP‐2 signaling. [ 20, 21 ] C2C12s are a compelling cell line for untangling the independent contributions of both substrate stiffness and BMP‐2 signaling pathways precisely because their osteogenic differentiation is dependent upon both, rather than just one (as is the case with mesenchymal stem cells (MSCs)). [ 22 ] In C2C12s, osteogenic differentiation is only observed when cells are treated with BMP‐2 and are exposed to high levels of substrate stiffness. By altering these two variables independently, we can better understand the dynamic role each modality plays in the onset of osteogenesis. In MSCs, the activation of osteogenetic genes can be provoked by either BMP‐2 signaling or substrate stiffness. Thus, perturbing one pathway will not result in a relative loss of osteogenesis, preventing us from understanding the interplay between the two modalities. Therefore, the C2C12 model is a powerful tool for investigating the synergistic role of matrix mechanics and BMP‐2 signaling in osteogenic differentiation. Hydrogel stiffness was tuned by varying the concentration of PEG diacrylate, modulating the number of backbone polymers within the hydrogel network. The elastic moduli of the hydrogels, measured by rheometry, ranged from 0. 8 to 77 kPa ( Figure 1 a ), within the range of physiological elasticity of natural tissues. [ 22 ] PEG hydrogels prevent the nonspecific adsorption of proteins, [ 23 ] allowing cells to directly sense matrix stiffness via integrin–RGD ligand interactions. [ 24, 25 ] Figure 1 BMP‐2‐induced osteogenic differentiation of C2C12 cells. a) Average elastic modulus of PEG hydrogels (error bars are SD, n = 3, two technical replicates, Welch's t ‐test). b) Quantitative assay of ALP activity in C2C12 cells cultured in different conditions for 7 days ( n = 2–3, two technical replicates, one‐way ANOVA followed by post hoc Tukey's multiple comparisons test). c) Representative images of ALP staining for C2C12 cells on PEG hydrogels of different stiffness for 5 days. d) Representative images of ALP staining for C2C12 cells on TCPS with or without BMP‐2 and inhibitor treatment for 7 days. e) Representative images of F‐actin staining of C2C12 cells cultured in different conditions for 24 h. f) The number of attached C2C12 cells related to cells on TCPS without BMP‐2 treatment ( n = 5, two technical replicates, Welch's t ‐test). g) Representative high‐magnification images of F‐actin staining of C2C12 cells cultured in different conditions for 24 h. h) Spread area of C2C12 cells cultured in different conditions for 24 h ( n = 50, two technical replicates, Welch's t ‐test). TCPS: cells were cultured on TCPS; Blebb: cells were treated with blebbistatin; Calyc: cells were treated with calyculin A; X kPa: cells were cultured on hydrogels with stiffness X kPa. Osteogenic differentiation of C2C12 cells on the hydrogels was evaluated by quantifying alkaline phosphatase (ALP) activity, as well as staining for ALP expression (Figure 1b – d ). Following addition of BMP‐2 to the culture media, cells on the stiffest (77 kPa) hydrogels expressed the highest ALP activity, comparable to cells on tissue culture polystyrene (TCPS). ALP activity decreased as stiffness decreased, with the lowest expression observed on the softest (0. 8 kPa) hydrogels. These results confirm previous reports that BMP‐2‐induced osteogenic differentiation in C2C12s and MSCs is mediated by matrix stiffness. [ 17, 18 ] To gain more insight into the potential role of mechanotransduction in BMP‐2 signaling, C2C12 cell adhesion on hydrogels with varying stiffness was investigated. As anticipated from previous studies, [ 21, 26 ] a higher amount of cells adhered on stiff hydrogels than on soft hydrogels (Figure 1e, f ). Cell spread area was also greater on stiffer hydrogels (Figure 1g, h ). BMP‐2 has been reported to increase cytoskeletal tension and alter spread area; [ 27 ] however, we did not observe significant differences in spread area as a result of BMP‐2 treatment, suggesting that substrate stiffness overrides the effect of the growth factor. F‐actin staining showed that stress fibers were more pronounced and parallel to the cell's major axis on 35 and 77 kPa hydrogels, while cells on 8 kPa hydrogels were smaller and displayed decreased stress fiber bundles. On 0. 8 kPa hydrogels, in which cells were the smallest, stress fiber formation was greatly hindered (Figure 1g ). As stress fiber formation contributes to the generation of intrinsic cytoskeletal tension and is thus a key step in mechanotransduction, we can conclude that mechanotransduction may be linked to BMP‐2 signaling via cytoskeletal tension. To confirm this, pharmacological agents that interfere with cellular tension and actin dynamics were utilized in cells on TCPS. As myosin is one of the key components regulating cytoskeletal tension, we employed blebbistatin (Blebb), an inhibitor of myosin II, and calyculin A (Calyc), which inhibits myosin light‐chain phosphatase from dephosphorylating myosin, resulting in increased myosin II activation. Perturbation of myosin II activity by Blebb dramatically decreased stress fiber formation and cell spread area (Figure 1g, h ), as well as ALP activity (Figure 1b, d ). In contrast, increased myosin II activation resulted in enhanced organization of stress fibers and cell spread area, as well as ALP activity. Surprisingly, stimulation by Calyc alone in the absence of BMP‐2 was sufficient to induce an increase in ALP expression in C2C12 cells. This is likely due to the largely enhanced cytoskeletal tension. Of all culture conditions, cells treated with both Calyc and BMP‐2 exhibited the highest ALP activity. Together, the results of the inhibitor experiments indicate that cytoskeletal tension is the mediator of BMP‐2‐induced osteogenic differentiation of C2C12 cells. This finding further raises the question of whether cytoskeletal tension is critical for either the initiation or the subsequent propagation of the BMP‐2 signaling cascade. Elasticity directs cellular mechanotransduction through a set of extra‐ and intracellular signaling pathways involving integrins, focal adhesion kinase (FAK), Rho/ROCK, lamin‐A/C, and YAP/TAZ, to name a few. [ 28, 29 ] As mechanosensitive signaling pathways have been heavily investigated in recent years, we chose here to focus specifically on the relationship between mechanosensitive signaling and BMP‐2 signaling. 2. 2 BMP‐2 Signaling Initiation Is Relatively Independent of Mechanical Cues Fusion and differentiation of C2C12 cells into multinucleated myotubes was monitored by staining for the marker myosin heavy chain (MHC). [ 11 ] In the presence of BMP‐2, C2C12 cells failed to form MHC positive myotubes on TCPS or hydrogels, with or without pharmacological agent treatment (Figure S1, Supporting Information). Conversely, in the absence of BMP‐2, MHC‐positive myotubes were clearly observed when cells were confluent in all experimental conditions except Calyc treatment. This indicates that BMP‐2 signaling is initiated (as proven by inhibition of myogenesis), but later blocked (as proven by lack of osteogenesis), on soft hydrogels where ALP activity is at a minimum. To determine how mechanical cues affect BMP‐2 signaling, we assayed the BMP‐2‐induced Smad signaling cascade in a stepwise fashion. BMP‐2 binds to and stabilizes membrane complexes consisting of types I and type II receptors. The type II receptor then phosphorylates and activates the type I receptor, recruiting and phosphorylating Smad1/5/8 at its C‐terminal SSXS‐motif to initiate the canonical Smad cascade. [ 14 ] Thus, phosphorylated Smad1/5/8 (pSmad1/5/8) was used as a reporter for BMP‐2 activity. [ 30, 31 ] To examine its phosphorylation state, cells were first seeded on TCPS or hydrogels with or without Blebb or Calyc pretreatment, then stimulated with BMP‐2 for 60 min. pSmad1/5/8 levels were determined by Western blot and normalized to β‐actin expression ( Figure 2 a, b ). In the absence of BMP‐2 stimulation, Smad1/5/8 phosphorylation was not observed in any experimental conditions. In BMP‐2 stimulated cells, the use of soft hydrogels or Blebb treatment resulted in a 20–25% decrease in Smad1/5/8 phosphorylation compared to TCPS or stiff hydrogels. Figure 2 Activation of BMP‐2‐induced Smad signaling. a) Representative Western blots for pSmad1/5/8 and the housekeeper protein β‐actin of C2C12 cells that were stimulated for 60 min in different conditions before lysis. b) Quantification of Smad1/5/8 phosphorylation based on Western blot images; n = 1–2, two technical replicates. c) Nuclear‐to‐cytoplasmic ratios of pSmad1/5/8 for C2C12 cells cultured in different conditions for 3 h followed by BMP‐2 treatment for 1 h. Quantification was based on immunofluorescence images; n = 15, two technical replicates. d) Luciferase activity of C2C12 BRE‐Luc cells cultured in different conditions for 24 h; n = 2, three technical replicates. e) Representative immunofluorescence images of C2C12 cells stained with DAPI (blue), Alexa Fluor 488 phalloidin (green), and anti‐pSmad1/5/8 (red) after culturing in different conditions for 3 h followed by BMP‐2 treatment for 1 h. High‐magnification images of anti‐pSmad1/5/8 are available in Figure S2, Supporting Information. f) Heat map of the relative expression levels (y) of the genes related to the early ( RUNX2 ), intermediate ( ALP ), and late ( OPN ) periods of osteogenesis in C2C12 cells cultured in different conditions. Data are analyzed by one‐way ANOVA followed by post hoc Tukey's multiple comparisons test for all panels. Downstream Smad signaling processes, which involve nuclear translocation of Smad complexes and further binding to target genes, were investigated by immunostaining of the Smad1/5/8 complex (Figure 2c, e ) and luciferase reporter assay (Figure 2d ), respectively. pSmad1/5/8 was highly enriched only in the nuclei of cells stimulated with BMP‐2, with Blebb treatment and the lowest hydrogel stiffness (0. 8 kPa) inducing a slight decrease in nuclear localization. C2C12 cells stably transfected with a BMP‐responsive luciferase reporter (BRE‐Luc) containing Smad binding elements derived from the Id1 promoter [ 32 ] were treated with BMP‐2. BRE‐Luc activity was fivefold to 10‐fold higher in cells stimulated with BMP‐2 compared to unstimulated cells. Consistent with Smad1/5/8 phosphorylation and Smad complex nuclear translocation, Blebb treatment or 0. 8 kPa hydrogel stiffness resulted in a 30% decrease in BRE‐Luc activity. It has been reported that cell shape regulates Smad signaling via RhoA/ROCK activity and downstream cytoskeletal tension generation. Smad signaling was decreased, but not totally blocked, when cell spreading was limited. [ 15 ] Furthermore, cell spreading on matrix‐bound BMP‐2 has been shown to not affect Smad signaling; ROCK‐dependent cytoskeletal tension is not directly required for the Smad activation. [ 21 ] On soft substrates, soluble BMP‐2 has been shown to decrease Smad activity to a certain degree. In our system, we also found that decreased cytoskeletal tension resulting from Blebb treatment or a soft substrate slightly decreased Smad signaling in cells stimulated with BMP‐2. However, Smad signaling in low tension, BMP‐2 stimulated cells was much higher than in unstimulated cells of all conditions. Thus, the initiation of the BMP‐2 signaling cascade via Smad signaling is relatively independent of mechanical cues. 2. 3 Osteogenic Gene Expression Is Regulated by Mechanical Cues As the activation of Smad signaling was found to be independent of mechanical cues, osteogenic gene expression was analyzed via quantitative real‐time reverse transcriptase polymerase chain reaction (qRT‐PCR). Three marker genes related to early (Runt‐related transcription factor 2; RUNX2 ), intermediate ( ALP ), and late (osteopontin; OPN ) stages of osteogenic differentiation were assayed. All three markers were significantly upregulated in a time‐dependent fashion in C2C12 cells grown on TCPS and stiff 35 and 77 kPa hydrogels in the presence of BMP‐2 after 7 days (Figure 2f ). The most dramatic upregulation of osteogenic markers was observed in cells cotreated with BMP‐2 and Calyc. Furthermore, Calyc treatment resulted in upregulation of these markers even in the absence of BMP‐2. In contrast, BMP‐2‐treated cells in combination with Blebb or cultured on the softest 0. 8 kPa hydrogels displayed a reduction in osteogenic marker expression. These patterns of osteogenic gene expression are consistent with the ALP activity assay and staining results described above. This observation suggests that osteogenic gene activation is mediated by cytoskeletal tension and dictated by mechanotransduction pathways. Thus, on soft matrices, the BMP‐2 signaling pathway driving osteogenic differentiation is blocked before gene expression is activated, although the Smad complex successfully binds to target genes (Figure 2d ). Therefore, other transcription factors or transcriptional regulators may be required to coactivate osteogenic genes and promote subsequent cell fate specification. 2. 4 YAP/TAZ Are Regulated by Cytoskeletal Tension A wide range of transcription factors and transcriptional regulators play a role in mechanosensitive gene expression. Specifically, YAP and TAZ have been shown to be sensitive to mechanical cues via regulation of the Ras‐related GTPase RAP2 [ 33 ] and the ARID1A‐containing SWI/SNF complex. [ 34 ] In addition, YAP/TAZ are well‐known for mediating cellular mechanoresponses. [ 35 ] Thus, we investigated the correlation between endogenous YAP/TAZ subcellular localization and the presence of cytoskeletal tension in the presence or absence of BMP‐2 stimulation (Figures S3 and S4, Supporting Information). To this end, immunostaining of both YAP and TAZ was performed in C2C12 cells cultured on surfaces with different stiffness or treated with pharmacological inhibitors. YAP and TAZ both localized to the nucleus on rigid TCPS and stiffer 35 and 77 kPa matrices. On softer 0. 8 and 8 kPa matrices, YAP and TAZ became predominantly cytoplasmic. Similarly, both YAP and TAZ were predominantly cytoplasmic when cytoskeletal tension was decreased via Blebb treatment. Calyc treatment did not alter YAP/TAZ localization for cells on TCPS, while BMP‐2 stimulation did not alter YAP/TAZ localization in any conditions. Next, we investigated whether cell spread area regulates BMP‐2 signaling and YAP/TAZ localization. Micropatterned “islands” of defined size generated on an antifouling PEG coating were utilized to induce changes in cell spread area based on the available adhesive area in the presence of BMP‐2 (Figure S5, Supporting Information). On these micropatterns, nuclear translocation of Smad complexes was not affected by cell size. In contrast, YAP exhibited strong nuclear localization on large islands and weak nuclear localization on small islands. Similarly, osteogenic differentiation, as measured by osterix expression and localization, was enhanced on large islands. These results on single‐cell micropatterns rule out the effects of cell–cell contacts on BMP‐2 signaling pathways. [ 8 ] YAP/TAZ localization in response to cell spreading and cytoskeletal tension support the notion that they are potential mediators of BMP‐2 signaling during osteogenic differentiation. 2. 5 YAP/TAZ Regulate BMP‐2 Signaling YAP/TAZ subcellular localization patterns merely indicate that YAP/TAZ are molecular “readers” of cytoskeletal tension. To prove that YAP/TAZ are relevant in mediating BMP‐2 signaling, transient siRNA‐induced knockdowns of YAP, TAZ, or both YAP and TAZ in C2C12 cells were performed and verified by Western blot ( Figure 3 a ). Importantly, Smad signaling was not affected by YAP/TAZ knockdown as measured by luciferase reporter assay (Figure 3b ). Individual YAP or TAZ knockdown resulted in a reduction of BMP‐2‐induced ALP activity by nearly 50% within 1 day of siRNA treatment, and simultaneous knockdown of both YAP and TAZ resulted in an even stronger reduction in ALP activity (Figure 3c, i ). After 3 days, ALP activity increased in all samples, consistent with transient recovery of YAP/TAZ. While individual YAP and TAZ knockdown samples returned to control levels, dual YAP/TAZ knockdown samples still exhibited statistically significant reductions in ALP activity after 3 days (Figure 3d, i ). Figure 3 Upregulation and downregulation of YAP/TAZ to mediate BMP‐2 signaling. a) Representative Western blots of YAP and TAZ expression in lysates collected 1 or 3 days post‐siRNA treatment. b) Luciferase activity in C2C12 BRE‐Luc cells treated with siRNAs; n = 3. c, d) Quantification of ALP activity for C2C12 cells cultured c) 1 day or d) 3 days post‐siRNA treatment; n = 3. e) Representative Western blots of total YAP/TAZ and Flag‐tagged YAP/TAZ expression in lysates collected 1 day post‐transfection. f) Luciferase activity in C2C12 BRE‐Luc cells transfected with Flag‐tagged YAP/TAZ; n = 3. g, h) Quantitative assay of ALP activity for C2C12 cells cultured g) 1 day or h) 3 days post‐transfection; n = 2, two technical replicates. i, j) Representative images of ALP staining in BMP‐2‐stimulated C2C12 cells cultured 1 or 3 days i) post‐siRNA or j) Flag‐tagged YAP/TAZ transfection. k) Representative immunofluorescence images of C2C12 cells cultured on soft 0. 8 kPa hydrogels in the presence of BMP‐2 2 days after transfection with pEGFP‐C3‐hYAP1 and pEF‐TAZ‐N‐Flag plasmid. Blue: DAPI; magenta: anti‐TAZ; green: plasmid‐generated YAP; red: anti‐osterix. Data are analyzed by one‐way ANOVA followed by post hoc Tukey's multiple comparisons test, p ‐values referenced to NC group. We also examined YAP and TAZ overexpression in C2C12 cells via Flag‐tagged plasmid transfection, both individually and in combination (Figure 3e ). Plasmid‐induced YAP/TAZ overexpression did not affect Smad signaling (Figure 3f ). Overexpression of YAP, TAZ, and YAP/TAZ in combination for 1 or 3 days resulted in significantly higher levels of ALP activity than in control groups (Figure 3g, h, j ). TAZ overexpression resulted in higher levels of ALP activity than YAP expression, potentially due to the fact that TAZ is negatively regulated by YAP expression but not vice versa, [ 36 ] although we did not observe a decrease of TAZ levels during YAP overexpression by Western blot. We also did not observe this phenomenon in the YAP knockdown experiments, possibly because the siRNA‐induced knockdown we employed was transient as opposed to the more permanent short hairpin RNA (shRNA) or CRISPR‐Cas9 systems utilized in the literature. [ 36 ] As YAP/TAZ are involved in numerous signaling pathways in cells, [ 37 ] permanent perturbation may cause undesired cell responses including apoptotic cell death. [ 38 ] Overexpression of both YAP and TAZ not only resulted in enhanced osteogenic differentiation on TCPS, but also stimulated osteogenic differentiation on soft 0. 8 kPa matrices in the presence of BMP‐2, as measured by osterix nuclear localization (Figure 3k ). The EGFP‐YAP plasmid and the TAZ plasmid were mixed together prior to lipids preparation for transfection. Nuclear osterix localization (red) was greatly enhanced in efficiently transfected cells (green), although increased YAP and TAZ expression did not enhance nuclear accumulation (as measured by nuclear‐cytoplasmic ratio) of YAP/TAZ (Figure S6, Supporting Information) as observed via immunostaining of total TAZ (magenta) and overexpressed YAP (green). In comparison, nuclear osterix localization was not observed in BMP‐2‐treated cells without plasmid expression. Thus, YAP/TAZ overexpression in the presence of BMP‐2 is sufficient for C2C12 osteogenesis, even when cytoskeletal tension is limited due to soft substrate elasticity. The spread area and stress fiber assembly of TCPS‐adherent cells were also monitored when the expression of both YAP and TAZ was downregulated or upregulated. No obvious differences were detected (Figure S7, Supporting Information). Thus, cytoskeleton tension is not altered during siRNA or plasmid treatment. Together, this indicates that YAP and TAZ play a major role in BMP‐2‐induced osteogenesis by serving as mechanosensitive mediators of BMP‐2 signaling. 2. 6 Smad Complexes and YAP/TAZ Synergize to Activate Gene Expression To explore the interplay between Smad complexes and YAP/TAZ during mechanosensitive gene activation, C2C12 cells were treated with BMP‐2 and Blebb, followed by whole transcriptome shotgun sequencing (RNA‐Seq) (Data 1, Supporting Information). Principal component analysis (PCA) of 5813 differentially expressed genes (DEGs) inherent to specific treatment combinations ( q value < 0. 05) and hierarchical clustering analysis identified transcriptional signatures sufficient to differentiate cell populations as a function of treatment (Figure S8, Supporting Information). A total of 1536 genes were found to be significantly upregulated in response to BMP‐2 (Data 2, Supporting Information). Importantly, half of these BMP‐2 responsive genes (867 genes) were antagonized by Blebb, as revealed by unsupervised hierarchical clustering analysis (see green cluster, Figure 4 a ), suggesting that this cluster of BMP‐2 responsive genes is mechanosensitive. Similarly, a subset of our BMP‐2 upregulated genes, which we characterized as bona fide Smad1/4/5 targets based on previous ChIP‐seq data [ 39 ] (Data 3, Supporting Information), was significantly downregulated when BMP‐2 was coupled with Blebb (70% of these genes were antagonized by Blebb), indicating that the majority of Smad1/4/5 targets are also mechanosensitive (Figure 4b, c ). Figure 4 Cooperative gene activation by BMP‐2 stimulation and mechanical cues. a) Hierarchical clustering analysis of genes upregulated by BMP‐2. b) Hierarchical clustering analysis of the Smad1/4/5 targets identified from genes in (a). c) Pie chart of the Smad1/4/5 targets in (b) based on the data of Blebb&BMP‐2 versus BMP2. d) GSEA of BMP‐2 genes revealed an enrichment for YAP/TAZ‐regulated genes. e) Canonical YAP/TAZ targets were upregulated in C2C12 cells by BMP‐2 and antagonized by Blebb. f) Hierarchical clustering analysis of the YAP targets identified from genes in (b). Gene set enrichment analysis (GSEA) of BMP‐2 genes revealed an enrichment of YAP/TAZ regulated genes (Cordenonsi signature, Figure 4d ), suggesting a BMP‐2‐dependent regulation of YAP/TAZ targets. Indeed, canonical YAP/TAZ targets were upregulated in C2C12 cells by BMP‐2 and antagonized by Blebb (Figure 4e ). Interestingly, 26 of our BMP‐2‐upregulated Smad1/4/5 targets have also been previously identified as targets of YAP, [ 40 ] although in that study, BMP‐2 was not used to stimulate cells, suggesting that other Smad targets may also be YAP targets (Data 4, Supporting Information). BMP‐2‐induced upregulation of over 65% of these YAP target genes was decreased in response to Blebb treatment (Figure 4f ), further strengthening the finding that BMP‐2‐induced gene expression is highly mechanosensitive. These results suggest that 1) Smad complexes are capable of activating certain genes in the absence of mechanical stimulation, and 2) Smad complexes coactivate other genes in a concerted manner along with YAP/TAZ. The genes identified in (1) likely contribute to the inhibition of myotube formation, while those in (2) play a greater role in initiating osteogenic differentiation of C2C12 cells. In addition, osteogenic differentiation likely also includes other YAP/TAZ‐binding transcription factors. Indeed, the genes Id1, Id2, Id3, and Id4, all of which are DEGs upregulated by BMP‐2, were not affected by Blebb treatment (Figure S9, Supporting Information). These genes encode DNA‐binding protein inhibitors that have been shown to bind to and deactivate MyoD and its cofactors, resulting in an inhibition of myogenesis. [ 41 ] Together, these results indicate that Smad signaling can occur both independent of and in conjunction with mechanical stimulation. 2. 7 Nuclear Accumulation of Smad Complexes Maintains BMP‐2 Signaling after BMP‐2 Withdrawal but Additional Mechanical Stimulation Is Required for Osteogenesis Data presented, thus, support the hypothesis that mechanical cues mediate BMP‐2 signaling, as osteogenic differentiation of C2C12 cells requires both Smad complex activation and YAP/TAZ localization. Next, we wanted to determine whether we could activate these two signaling pathways independently. We generated an antifouling polymer coating on TCPS based on a biomimetic amphiphilic block copolymer ( Figure 5 a ), [ 42 ] which prevents ECM protein adsorption and cell adhesion, [ 43 ] limiting external mechanical cues and cytoskeletal tension. C2C12 cells were initially stimulated with BMP‐2 on antifouling TCPS (Anti) and then transferred onto adhesive TCPS (untreated) without further BMP‐2 stimulation (Figure 5b ). Cell adhesion was not observed on antifouling TCPS, but cells became well‐spread after transfer onto adhesive TCPS (Figure S10, Supporting Information). It is worth emphasizing that nonadherent cells were transferred via simple pipetting, allowing us to avoid trypsinization or other confounding factors present during passaging. Therefore, in this setup, Smad signaling can be activated first via BMP‐2 stimulation, followed by subsequent and separate mechanical stimulation and cytoskeletal tension induced by adhesive TCPS. Figure 5 Synergy between BMP‐2 signaling and mechanotransduction. a) Antifouling coating scheme on TCPS. b) Scheme of the experimental setup. C2C12s were cultured on antifouling TCPS (green) with BMP‐2 stimulation (blue) for 30 min to 3 days. BMP‐2 was removed (red) and C2C12s were maintained in culture on antifouling TCPS for 1 day. C2C12s were subsequently transferred onto adhesive TCPS (yellow) for 1–3 days. c) C2C12 BRE‐Luc cell luciferase activity when cultured on antifouling or adhesive TCPS for 1 day; n = 3, two technical replicates. d) C2C12 BRE‐Luc cell luciferase activity when pretreated with BMP‐2 on antifouling TCPS for 1 day and then transferred onto antifouling or adhesive TCPS in the presence (+) or absence (–) of BMP‐2; n = 3, two technical replicates. e) Representative immunofluorescence images of BMP‐2‐pretreated (1 day) C2C12 cells stained with DAPI (blue), Alexa Fluor 488 phalloidin (green), and anti‐YAP (red) after being transferred onto adhesive TCPS for 20 min or 24 h, and related nuclear‐to‐cytoplasmic ratios of YAP; n = 20, Welch's t ‐test. High‐magnification images of anti‐YAP are available in Figure S13, Supporting Information. f) Quantitative ALP activity assay for C2C12 cells pretreated with BMP‐2 on antifouling TCPS for 1 day and then transferred onto antifouling or adhesive TCPS in the presence (+) or absence (–) of BMP‐2; n = 3. g) Representative images of ALP staining for C2C12 cells pretreated with BMP‐2 on antifouling TCPS for 1 day and then transferred onto adhesive TCPS in the presence or absence of BMP‐2. h) Luciferase activity of C2C12 BRE‐Luc cells and i) ALP activity of C2C12 cells pretreated with BMP‐2 on antifouling TCPS from 30 min to 3 days, and then transferred onto adhesive TCPS without BMP‐2 for 1 day. For “1 d‐1 d” samples, cells were pretreated with BMP‐2 on antifouling TCPS for 1 day, and then cultured for an additional day on antifouling TCPS without BMP‐2, before being transferred onto adhesive TCPS without BMP‐2 for a final day; n = 4 for luciferase activity and n = 3 for ALP activity. The cell proliferation was limited by mitomycin C. Data are analyzed by one‐way ANOVA followed by post hoc Tukey's multiple comparisons test (except in (e), Welch's t ‐test). To eliminate the effects of proliferation, cells were pretreated with mitomycin C when cultured on adhesive TCPS for more than 1 day. Luciferase reporter assays indicated that C2C12 BRE‐Luc cells on antifouling TCPS exhibited slightly lower luciferase activity compared to cells on adhesive TCPS after BMP‐2 stimulation for 1 day (pretreatment), but exhibited eight times higher activity compared to cells that did not receive BMP‐2 treatment (Figure 5c ). Interestingly, luciferase activity can be maintained in pretreated cells for at least 3 days on both antifouling and adhesive TCPS after the removal of BMP‐2 (Figure 5d ). This means that BMP‐2‐induced Smad complexes continuously bind and activate target genes once initiation has occurred. Immunostaining also confirmed that nuclear pSmad1/5/8 localization was persistent in pretreated cells once they had been transferred onto adhesive TCPS in the absence of BMP‐2 (Figure S11, Supporting Information). Nuclear YAP accumulation did not occur in pretreated cells subsequently transferred onto adhesive TCPS and fixed after 20 min, indicating that the pretreatment phase did not induce YAP/TAZ translocation. Once cells were sufficiently spread with well‐organized actin stress fibers, nuclear YAP accumulation was observed (Figure 5e ). Pretreated cells did exhibit high ALP activity after both 1 and 3 days post‐transfer onto adhesive TCPS, but not when kept on antifouling TCPS (Figure 5f, g ). Osteogenic differentiation was also monitored via osterix nuclear localization. In line with nuclear accumulation of YAP, only well‐spread, pretreated cells exhibited high osterix localization (Figure S12, Supporting Information). To determine the effect of the duration of BMP‐2 pulse stimulation on Smad signaling, we provided BMP‐2 stimulation to C2C12 and C2C12 BRE‐Luc cells at various time points between 30 min and 3 days before transferring them onto adhesive TCPS. Cells were then cultured for 1 day in the absence of BMP‐2, after which luciferase reporter and ALP activity assays were performed. Luciferase activity, which was low when pretreatment only lasted 30 min, increased gradually until a pretreatment length of 1 day, at which point luciferase activity became saturated (Figure 5h ). ALP activity followed a similar pattern (Figure 5i ). Thus, higher‐level gene activation caused by Smad complexes in early signaling events only results in higher ALP activity after the application of mechanical cues. When cell proliferation is inhibited by mitomycin C, 1 day of BMP‐2 stimulation is sufficient to activate osteogenic differentiation in C2C12 cells. Moreover, luciferase activity and ALP activity did not decrease when pretreated cells were kept on antifouling TCPS for an additional day without BMP‐2 stimulation before being transferred onto adhesive TCPS (Figure 5h, i ). These results 1) highlight the role of mechanotransduction in BMP‐2‐induced osteogenic differentiation, and 2) prove that C2C12 osteogenic commitment can be initiated in a step‐wise process via independent pulsed BMP‐2 stimulation and mechanical cues. In other words, cells can “remember” BMP‐2 stimulation history via Smad complex gene targeting. Osteogenic differentiation is subsequently activated via nuclear YAP/TAZ translocation as a function of cytoskeletal tension. Human MSCs are a useful model for mechanosensitive stem cell differentiation. Similar to our findings in C2C12 cells, we observed osteogenic differentiation of MSCs in response to stimulation with BMP‐2 and mechanical cues (Figures S14–S16, Supporting Information). As MSCs are widely used in regenerative medicine and hold great therapeutic potential, these results extend our findings to more translational conditions. 3 Discussion BMPs are a potent class of growth factors that regulate the development of many organ systems in the body and play an especially important role in osteogenesis. BMP therapies are becoming promising alternatives to autografts, which are currently the gold standard for chronic bone defects, but remain limited by low availability as well as donor site pain and inflammation. [ 16, 44 ] Dysregulation of BMP signaling has been shown to contribute to a number of pathological processes, including cancer and ectopic bone formation. [ 45, 46 ] Thus, understanding how BMP‐2 regulates differentiation and manipulating BMP‐2 signaling are both critically important for both clinical regenerative medicine and the rational design of growth factor‐doped biomaterials. In this study, BMP‐2‐induced C2C12 osteogenesis was investigated from two complementary directions: 1) biochemical pathways that have been shown to play a role in BMP‐2 signaling, and 2) mechanotransduction pathways that are influenced by different matrix stiffness or the application of cytoskeletal tension‐modifying inhibitors. These synergistic modalities are summarized in Figure 6. By transiently downregulating and upregulating the expression of YAP and TAZ in C2C12s, we found that they contribute to the crosstalk between BMP‐2 signaling and mechanotransduction pathways. Additionally, the shuttling of YAP and TAZ from the cytoplasm to the nucleus is independent of BMP‐2 signaling, but this translocation enhances BMP‐2‐induced differentiation. In parallel, the initiation of Smad signaling is independent of mechanical cues. Crosstalk between these two signaling pathways synergistically enhances osteogenic gene expression. Figure 6 Synergy between cytoskeletal tension (YAP/TAZ) and the BMP‐2‐Smad pathway. Left: No activated Smad signaling or osteogenic differentiation is observed in the absence of BMP‐2, even in the presence of stiff matrix. Right: The initiation of BMP‐2‐induced Smad signaling is independent of cytoskeletal tension. Smad1/5/8 can be phosphorylated and form heteromeric complexes that translocate into the nucleus and bind to target genes. However, osteogenic gene activation requires cytoskeletal tension‐induced nuclear accumulation of YAP/TAZ. Thus, BMP‐2 signaling responds to mechanical cues by sensing nucleocytoplasmic shuttling of YAP/TAZ. The crosstalk between BMP‐2 signaling and mechanotransduction pathways is likely due to binding between Smad heteromeric complexes and YAP/TAZ, possibly via phosphorylated Smad1, [ 47, 48 ] as well as the assistance of other YAP/TAZ‐binding transcription factors. It has been suggested that TAZ coactivates RUNX2‐dependent gene transcription, driving cells towards osteogenic differentiation under BMP‐2 stimulation. [ 49 ] In other studies focusing on TGF‐β signaling, the osteogenic gene activation is suggested to be coassociated with TEAD [ 50 ] and OCT4. [ 51 ] Thus, gene activation is likely the result of synergistic cooperation between multiple transcription factors regulated by diverse stimuli, as indicated by RNA‐seq analysis. In 3D systems, YAP/TAZ may be regulated by mechanical cues in a different way than on 2D surfaces, likely due to spatial limitations on cytoskeletal assembly. [ 52 ] For example, YAP/TAZ nuclear localization has been shown to be higher in cells encapsulated in soft hydrogels than in stiff ones. [ 53, 54 ] Cells sense mechanical cues in both 2D and 3D conditions via the same mechanotransduction pathways. Thus, the key step in mechanotransduction in both 2D and 3D conditions is the generation of intracellular cytoskeletal tension. This tension is capable of inducing downstream YAP/TAZ nuclear localization and further crosstalk with BMP‐2 signaling pathways. In this work, we chose to focus on the intrinsic myosin‐based cytoskeletal tension generated on simple 2D material models, but it is reasonable to speculate that these same molecular mechanisms can be extended to 3D systems. Smad signaling can be initiated by pulse BMP‐2 stimulation, which becomes saturated within 1 day. Interestingly, Smad complexes can continue binding to and activating target genes regardless of external mechanical cues after the removal of the BMP‐2 stimulus. This maintenance enables Smad‐activated cells to either keep their phenotype without applying further mechanical cues or differentiate to osteoblasts by applying later mechanical cues. The differentiation process is therefore programmable, a fact that can be leveraged in the future to control cells and growth factors in biomedical engineering. Please note the cell proliferation can dilute the concentration of targeted Smad complexes on genes. We only focused on the molecular mechanism and used mitomycin C to limit this dilution. For practical applications, we would suggest to initially treat cells with relatively high concentration of BMP‐2 for about 1 day and to keep the BMP‐2 concentration in relatively low level to continuously stimulate the proliferated cells. Furthermore, this sequential activation of differentiation programming may play a major role in cellular differentiation in vivo, where migrating stem cells are exposed to diverse and dynamic mechanical and biochemical environments on their journey from the stem cell niche to their ultimate differentiation site. In addition to canonical Smad signaling, BMP‐2 also induces noncanonical pathways by activating MAPK cascades through Smad in a transcription‐independent fashion. [ 55, 56 ] However, the molecular basis for the activation and signal transduction of noncanonical pathways is still not fully understood. YAP/TAZ may not be the only crosstalk point between mechanotransduction pathways and BMP‐2 signaling. Integrins, which have been shown to colocalize with BMP‐2 receptors, [ 57 ] may play a role in integrating mechanical signals with BMP‐2 signaling. Indeed, BMP‐2 stimulation has been shown to upregulate the expression of multiple types of αvβ‐ integrins in human osteoblasts. [ 57 ] In addition, αvβ3 integrin mediates BMP‐2‐induced Smad signaling through the Cdc42‐Src‐FAK‐ILK cascade. [ 21 ] Moreover, other cell membrane receptors like N‐cadherin [ 58 ] and FGF receptor [ 59 ] have been found to modulate BMP‐2. ROCK activity and RhoA/ROCK‐mediated cytoskeletal tension also regulate BMP‐induced Smad signaling and osteogenic differentiation in human MSCs. [ 15 ] Although it has been demonstrated that mechanical cues affect Smad signaling through the various signaling pathways described above, here we demonstrate for the first time that cytoskeletal tension‐mediated nuclear accumulation of YAP/TAZ is critical for BMP‐2 signaling and subsequent osteogenic differentiation. 4 Conclusion Overall, we used a carefully selected model cell line to study the crosstalk between BMP‐2 signaling and mechanotransduction pathways to understand how these two disparate modalities can synergize to mediate osteogenic differentiation at a molecular level. We identified YAP/TAZ as a primary crosstalk junction and demonstrated the integration of these two distinct pathways for altered gene transcription. In addition, Smad complexes, the transcription factors involved heavily in BMP‐2 signaling, were found to maintain BMP‐2 activity and remain bound to target genes after BMP‐2 withdrawal. This observation was then leveraged to dictate programmable BMP‐2 stimulation and mechanical cues for on‐demand cell fate determination. Investigations into the role of mechanotransduction in BMP signaling may identify important mechanisms linking chemical cues from the extracellular environment to physical cues from the ECM during cell differentiation and tissue development. Understanding this link will allow for the guided design of new biomaterials for regenerative therapies, as well as provide critical information on the use of BMP as a therapeutic tool for enhancing clinical bone repair success. Ultimately, future tissue engineering strategies for bone repair must take both biochemical and mechanobiological phenomena into account. 5 Experimental Section PEG Hydrogels Fabrication and Mechanical Characterization PEG hydrogel stiffness was controlled by varying poly(ethylene glycol)‐diacrylate (PEG‐DA) macromer with Mn 700 (455008 Sigma‐Aldrich) concentration in water. Specifically, 80, 100, 200, and 400 mg mL –1 concentrations were utilized. The adhesive peptide cyclo(Arg‐Gly‐Asp‐D‐Phe‐Cys) (c(RGDfC), PCI‐3686‐PI, Peptides International) was added to the PEG‐DA solution at a constant concentration of 100 × 10 −6 m. The photoinitiator Irgacure 2959 (410896, Sigma‐Aldrich) was then mixed into the solution to achieve an initiator‐to‐acrylate ratio of 1:100. The gelation solution was pipetted onto glass coverslips modified with 3‐(trimethoxysilyl)propyl acrylate (475149, Sigma‐Aldrich) and covered with quartz slides modified with 1 H, 1 H, 2 H, 2 H ‐perfluorodecyltrichlorosilane (L1658403, Alfa Aesar). Polymerization was initiated via 365 nm UV irradiation. Fabricated hydrogels were equilibrated in phosphate‐buffered saline (PBS) buffer before use. For cell experiments, the c(RGDfC) peptide was utilized as an adhesive ligand. RGD peptide concentration was kept constant at 100 × 10 −6 m, comparable to previous biological studies, [ 60 ] and was sufficient to allow cells to adhere, but not induce spatial sensing‐induced cell adhesion. [ 61 ] Mechanical properties of the hydrogels were measured with a rotational rheometer (Kinexus Pro, Malvern) in a humidity chamber at a constant shear rate of 1 Hz. Cell Culture and Inhibitor Treatment Mouse C2C12 myoblasts (ATCC CRL‐1772) were cultured as subconfluent monolayers in growth media, consisting of α‐MEM (A1049001, Gibco) supplemented with 10% fetal bovine serum (S0115, Biochrom) and 1% penicillin/streptomycin (15140122, Gibco) at 37 °C and 5% CO 2. Cells were serum‐starved in serum‐free growth media for 3–4 h before stimulation with 20 × 10 −9 m recombinant human BMP‐2 (rhBMP‐2, Morphoplant GmbH) in growth media. For cells stimulated with BMP‐2 on antifouling surfaces, BMP‐2 was removed before cell transfer. Since cells could not adhere to the antifouling surface, their removal was possible with gentle pipetting. For this, media and cells were gently collected, centrifuged, and washed with PBS at least three times to remove residual exogenous BMP‐2 before seeding to new antifouling or cell adhesive TCPS. For inhibitor studies, 20 × 10 −6 m (–)blebbistatin (Blebb, B0560, Sigma‐Aldrich) or 1 × 10 −9 m calyculin A (Calyc, C5552, Sigma‐Aldrich) was added to cell culture media. To prevent cell proliferation during Smad complex analysis post‐BMP‐2 removal, cells were treated with 10 µg mL –1 mitomycin C for 2 h before use. ALP Assay The SensoLyte pNPP alkaline phosphatase assay kit Colorimetric (AnaSpec) was used according to the manufacturer's instructions. Cell numbers were counted prior to lysis. Final values were measured at a wavelength of 405 nm with a Tecan Infinite M200 Plate Reader. ALP staining was performed with Leukocyte Alkaline Phosphatase kit based on Naphthol AS‐BI and fast blue BB salt (86C, Sigma‐Aldrich) according to manufacturer's instructions. Protein Isolation and Western Blot Analysis Cells were lysed in RIPA buffer (89900, Thermo Fisher Scientific) freshly supplemented with protease and phosphatase inhibitor cocktails (78442, Thermo Fisher Scientific) on ice and then centrifuged at 14 000 × g for 15 min at 4 °C. Lysates were collected and stored at –20 °C. After BCA assays to determine concentration, protein extracts were separated via SDS‐PAGE and transferred to a nitrocellulose membrane. After blocking with 5% BSA in TBST for 1 h, membranes were incubated overnight with diluted primary antibodies in 5% BSA in TBST at 4 °C followed by secondary HRP‐linked antibodies at room temperature for 2 h. Chemiluminescence via Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare) was detected with a Fujifilm LAS‐3000 Imager. ImageJ was used for band quantification. For the Western blot analysis of pSmad1/5/8, more than 10 k cells were seeded on substrates in serum‐free growth media for 3 h before stimulation with 20 × 10 −9 m BMP‐2 for 1 more hour. If required, 20 × 10 −6 m Blebb or 1 × 10 −9 m Calyc was added during the whole 4 h. For the Western blot analysis of YAP/TAZ, more than 10 k cells were treated as described below (Section RNA Interference and Plasmid Transfection). Primary antibodies and corresponding concentrations used in this study were rabbit anti‐pSmad1/5/8(9) (1:1000) (13820, Cell Signaling Technology), rabbit anti‐YAP (1:1000) (4912, Cell Signaling Technology), mouse anti‐TAZ (1:1000) (560235, BD Pharmingen), mouse anti‐β‐actin (1:2000) (A1978, Sigma‐Aldrich), and mouse anti‐FLAG (1:1000) (F1804, Sigma‐Aldrich). Secondary antibodies were HRP‐linked anti‐rabbit IgG (1:2000) (7074, Cell Signaling Technology) and HRP‐linked anti‐mouse IgG (1:2000) (7076, Cell Signaling Technology). PageRuler Plus Prestained Protein Ladder, 10–250 kDa, served as a molecular weight marker (26619, Thermo Fisher Scientific). Luciferase Assay Luciferase assays were performed using the Luciferase Assay System (E1500, Promega) according to the manufacturer's instructions. C2C12 BRE‐Luc cells stably transfected with pGL3(BRE)‐luciferase reporter construct have been previously described. [ 32 ] Cell numbers were counted prior to lysis. Chemiluminescence was detected with a Tecan Infinite M200 Plate Reader and normalized to cell counts. Immunofluorescence Staining and Microscopy Cells were washed once with cell culture media and twice with PBS before fixation with 4% paraformaldehyde at room temperature for 15 min. Samples were then washed thrice with ice cold PBS. Cells were permeabilized with 0. 25% v/v Triton X‐100 in PBS for 10 min at room temperature, then washed thrice with PBS. Nonspecific antibody binding was blocked by incubating samples with 1% w/v bovine serum albumin (BSA) in PBST (0. 1% v/v Triton X‐100 in PBS) at room temperature for 45 min. Next, samples were washed briefly with PBST and incubated with primary antibodies for 60 min at room temperature. Following primary antibody incubation, samples were washed twice with PBST and thrice with PBS. Samples were then incubated with secondary antibodies and Alexa Fluor 488 phalloidin for 60 min at room temperature, followed by washing twice with PBST and thrice with PBS. Finally, samples were placed on microscope slides in Fluoromount‐G with DAPI (00‐4959‐52, eBioscience) mounting media. Immunofluorescence images were acquired on an Axiovert 200M Microscope (Carl Zeiss). Primary antibodies and corresponding concentrations used were rabbit anti‐pSmad1/5/8(9) (1:200) (13820, Cell Signaling Technology), rabbit anti‐YAP (1:100) (4912, Cell Signaling Technology), mouse anti‐TAZ (1:100) (560235, BD Pharmingen), rabbit anti‐Osterix (1:100) (ab22552, Abcam), and mouse anti‐myosin heavy chain (1:10) (MF20, Developmental Studies Hybridoma Bank, University of Iowa). Secondary antibodies used were Alexa Fluor 568‐linked anti‐rabbit IgG (1:1000) (A11011, Thermo Fisher Scientific), Alexa Fluor 647‐linked anti‐mouse IgG (1:1000) (A21235, Thermo Fisher Scientific), and Alexa Fluor 488‐linked anti‐mouse IgG (1:1000) (A11029, Thermo Fisher Scientific). RNA Extraction and qRT‐PCR Total RNA was isolated using TRIzol reagent (15596026, Invitrogen) according to the manufacturer's instructions. Collected RNA was converted to cDNA using a high‐capacity cDNA reverse transcription kit (4368814, Applied Biosystems) along with an iCycler thermal cycler gradient (Bio‐Rad) according to the manufacturer's instructions. Gene expression profiles were determined for RUNX2, ALP, and OPN with glyceraldehyde 3‐phosphate dehydrogenase used as a housekeeping gene. qRT‐PCR was carried out on a 7500 Real‐Time PCR System (Applied Biosystems). cDNA was amplified with the following conditions: 1 cycle at 50 °C for 2 min and 95 °C for 2 min, then 40 cycles at 95 °C for 15 s and 60 °C for 1 min. Amplification was monitored with SYBR Green (4309155, Applied Biosystems). Data were then normalized to the housekeeping gene as an index of cDNA content after reverse transcription and further normalized to the group on TCPS at day 1. Primer sequences [ 62 ] are listed in the Table S1, Supporting Information. RNA‐Seq Total RNA from C2C12 cells was extracted with a Quick‐RNA miniprep kit from Zymo Research (R1055) following manufacturer's instructions. RNA concentration was assessed via NanoDrop and then stored at –80 °C. For RNA‐Seq experiments, library preparation was performed with the TruSeq RNA Sample Prep Kits v2 (Illumina) following manufacturer's instruction. RNA‐Seq libraries were then run on the Agilent 2100 Bioanalyzer (Agilent high‐sensitivity DNA chip) for quantification and quality control and then sequenced on NovaSeq 6000 (Illumina). RNA Interference To knock down endogenous YAP and TAZ levels, C2C12 cells were transfected with siRNAs against YAP (TriFECTa DsiRNA Kit, design ID, mm. Ri. Yap1. 13, Integrated DNA Technologies, Inc. ) or TAZ (Stealth siRNAs MSS227747, MSS227748, MSS227749, Thermo Fisher Scientific) using Lipofectamine 2000 (11668027, Thermo Fisher Scientific) according to the manufacturer's instructions. To analyze resulting protein expression levels, cells were lysed for Western blot analysis at indicated time points after transfection. Cells were seeded in antibiotic‐free growth media for 24 h prior to transfection. Transfection solution was prepared as follows: 1 mg mL –1 Lipofectamine 2000 was diluted to 30 µg mL –1 in Opti‐MEM media (31985, Gibco) and mixed for 15 min, while 20 × 10 −6 m solutions of each siRNA were diluted 20× to 1 × 10 −6 m in a separate aliquot of Opti‐MEM media. The two solutions were then mixed at a 1:1 by volume for another 15 min at room temperature. Afterwards, the final mixture was added into cell culture medium without antibiotics to achieve a final concentration of 100 × 10 −9 m for each siRNA in 3 µg mL –1 Lipofectamine 2000. Cells were then incubated with siRNA for 24 h at 37 °C with 5% CO 2. Cells treated with Lipofectamine 2000 alone were used as a negative control group. Plasmid Transfection p2xFLAGhYAP1 and pEGFP‐C3‐hYAP1 plasmids were provided by Marius Sudol (17791 and 17843, Addgene); [ 63, 64 ] pEF‐TAZ‐N‐Flag plasmids were provided by Michael Yaffe (19025, Addgene). [ 65 ] Plasmids were isolated and purified with a Qiagen Plasmid Plus Midi Kit (12943, Qiagen) and were stored at –80 °C. Cells were seeded in antibiotic‐free growth media for 24 h prior to transfection. Transfection solution was prepared as follows: 1 mg mL –1 Lipofectamine 2000 was diluted to 60 µg mL –1 in Opti‐MEM media (31985, Gibco) and mixed for 15 min, while 200 ng µL –1 solutions of plasmid were diluted 10× to 20 ng µL –1 in a separate aliquot of Opti‐MEM media. The two solutions were then mixed at a 1:1 by volume for another 15 min at room temperature. Afterwards, the final mixture was added into cell culture media without antibiotics to achieve a final concentration of 1 ng µL –1 for each plasmid in 3 µg mL –1 Lipofectamine 2000. Cells were then incubated with plasmid for 24 h at 37 °C with 5% CO 2. Cells treated with Lipofectamine 2000 alone were used as a negative control group. Statistics Statistical analysis was performed in GraphPad Prism 7. One‐way ANOVA followed by post hoc Tukey's multiple comparisons test or Welch's t ‐test were carried out as described in figure captions. All results are displayed as mean ± standard deviation, * p < 0. 05, ** p < 0. 01, *** p < 0. 001, and **** p < 0. 0001. Significance is indicated with by *( p < 0. 05). Conflict of Interest The authors declare no conflict of interest. Supporting information Supporting Information Click here for additional data file. Supporting Information Click here for additional data file. Supporting Information Click here for additional data file. Supporting Information Click here for additional data file. Supporting Information Click here for additional data file.
|
10. 1002/advs. 201902953
| 2,020
|
Advanced Science
|
Advances in Hybrid Fabrication toward Hierarchical Tissue Constructs
|
Abstract The diversity of manufacturing processes used to fabricate 3D implants, scaffolds, and tissue constructs is continuously increasing. This growing number of different applicable fabrication technologies include electrospinning, melt electrowriting, volumetric‐, extrusion‐, and laser‐based bioprinting, the Kenzan method, and magnetic and acoustic levitational bioassembly, to name a few. Each of these fabrication technologies feature specific advantages and limitations, so that a combination of different approaches opens new and otherwise unreachable opportunities for the fabrication of hierarchical cell–material constructs. Ongoing challenges such as vascularization, limited volume, and repeatability of tissue constructs at the resolution required to mimic natural tissue is most likely greater than what one manufacturing technology can overcome. Therefore, the combination of at least two different manufacturing technologies is seen as a clear and necessary emerging trend, especially within biofabrication. This hybrid approach allows more complex mechanics and discrete biomimetic structures to address mechanotransduction and chemotactic/haptotactic cues. Pioneering milestone papers in hybrid fabrication for biomedical purposes are presented and recent trends toward future manufacturing platforms are analyzed.
|
1 Introduction Using automatized technologies for the fabrication of hierarchical structures with the aim to obtain biologically functional tissue analogues has been an essential development within biofabrication. [ 1, 2 ] Such hierarchical 3D tissue constructs have broad applicability including for drug screening, [ 3 ] disease modeling, [ 4 ] and organotypic models [ 5 ] in addition to basic biomedical science research [ 6 ] and in vivo implantation. [ 7 ] The central hypothesis of biofabrication is that defined positioning of cells in a 3D hierarchical manner aids the formation of complex tissue. [ 2 ] This includes “on‐a‐chip” technologies [ 8 ] where complexity is needed to replicate the cell microenvironment and communication pathways more reflective of the in vivo environment. Automated 3D positioning is therefore expected to improve tissue maturation into more complex, functional ones. While developments in additive manufacturing (AM) technologies have driven much biofabrication research, other fabrication technologies such as electrospinning, [ 9 ] centrifugal spinning, [ 10 ] liquid–liquid centrifugal casting, [ 11 ] uniaxial freezing, [ 12 ] micromolding, [ 13 ] and electrochemical compaction [ 14 ] provide diverse manufacturing options for biomedical researchers. These approaches are also becoming increasingly automated and, as outlined in this review, are part of a greater trend of manufacturing technology hybridization for the creation of complex, hierarchical tissue constructs for biomedical applications. There are many different automated processes—not only those based on AM principles—available to manufacture 3D tissue constructs. [ 1, 15 ] However, much of the published research for tissue engineering and regenerative medicine (TERM) relies on a single fabrication technology that is expected to achieve all the complex elements required to produce a tissue construct. [ 16 ] In reality, each manufacturing process has its strengths and weaknesses, with respect to resolution, fabrication rates and the compatibility with cell‐based maturation. [ 1 ] There are also an exclusive materials or bioink library of each of these fabrication methods. Ultimately, the complexity and resolution required to mimic natural tissue is greater than what a single technology can deliver. While extracellular matrix (ECM) mimicry is relevant, especially in connective tissue, fabricating a nano‐microhierarchical morphology at the first stage may not be required. While indeed natural tissues have nanoscale elements, there is an abundance of research [ 17, 18, 19 ] showing that microscale structures allow cells to attach and then reorganize to produce nanostructured ECM (in other words, a nanoscale scaffold structure or substrate topography is not necessarily required for cells to secrete their natural ECM components). From this perspective, the aim of bioprinting and biofabrication is not maximal ECM mimicry, but final tissue formation and maturation, with ECM ideally secreted and shaped by the cells themselves. Nanoscale mimicry while essential within the mature tissue construct, diverges from what we originally wanted to address in the review, which is hybrid fabrication technologies. Therefore, the hybridization of fabrication technologies from both AM and non‐AM origins is the next logical step. When combining complementing manufacturing technologies, it is possible that the limitations of each individual one can be mitigated. There is already a trend toward hybridization of automated technologies, [ 20, 21, 22, 23 ] which we believe will increase over the next decade. Since the term “hybrid” has different meanings depending on the context, this article uses the word to describe the combination of individual manufacturing technologies (simultaneously or in series) to provide a multimodal, [ 24, 25 ] multiphasic, [ 26, 27 ] or multimaterial structure [ 7, 28 ] from which to support tissue maturation or regeneration. As depicted in Figure 1, these hierarchical structures can be grouped as acellular and cell‐based technologies; the first instance involves the fabrication of acellular biomaterials and scaffolds/lattices, while the latter involves cellular incorporation as an integral part of the fabrication technology. For reproducibility reasons, we foresee manual aspects of fabrication such as pipetting and transfer of substrates to eventually become fully automated technologies based on robotization. Figure 1 A graphical depiction of melt electrowriting (MEW) and melt electrospinning (MES) as an example of hybrid fabrication (top) and an overview how such technologies can be combined within a biofabrication or tissue engineering and regenerative medicine (TERM) paradigm. Hybrid fabrication approaches are influenced by the level of automation, with a fully automated scenario envisioned in the future. Example inset images previously unpublished and provided by Mr. Marius Berthel. Cellular figure reproduced with permission. [ 31 ] Copyright 2017, American Chemical Society. Acellular figure reproduced with permission. [ 32 ] Copyright 2018, The Authors, Published by WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. Automated to manual right‐hand side images reproduced with permission. [ 33 ] Copyright 2019, IOP Publishing. 2 Hybrid Fabrication: Specific Case Analyses and Emerging Key Technologies There are several milestone and breakthrough articles highlighted in Figure 2 ; studies that first describe and demonstrate the convergence of different fabrication technologies to manufacture more complex 3D tissue constructs. Advantages and disadvantages of specific fabrication technologies used for TERM are summarized in Table 1. All of the following technologies are envisioned to eventually be incorporated into a fully automated biofabrication system as shown in Figure 1. This hybrid fabrication would likely be part of a greater future automated platform based on high‐throughput screening and computational modeling. [ 33 ] Figure 2 Timeline and milestone papers toward hybrid fabrication for TERM and biofabrication. Schematic of Stankus et al. reproduced with permission. [ 39 ] Copyright 2006, Elsevier. Schematic of Schuurman et al. reproduced with permission. [ 40 ] Copyright 2013, IOP Publishing. Schematic of Xu et al. reproduced with permission. [ 41 ] Copyright 2013, IOP Publishing. Schematic of Visser et al. reproduced with permission. [ 42 ] Copyright 2015, The Authors. Published by Springer‐Nature Publishing. Figure of Hrynevich et al. reproduced with permission. [ 32 ] Copyright 2018, The Authors, Published by WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. Schematic of Jungst et al. reproduced with permission. [ 21 ] Copyright 2019, The Authors, Published by Wiley. Schematic of Moroni et al. reproduced with permission. [ 43 ] Copyright 2008, Wiley. Schematic of Lang et al. reproduced with permission. [ 44 ] Copyright 2010, IEEE. Schematic of Ozbolat et al. reproduced with permission. [ 45 ] Copyright 2014, Elsevier. Schematic of Silva et al. reproduced under the terms and conditions of the Creative Commons CC BY 3. 0 License. [ 46 ] Copyright 2016, The Authors, Published by PLOS ONE. Figure of Mekhileri et al. reproduced under the terms and conditions of the Creative Commons CC BY 3. 0 License. [ 20 ] Copyright 2018, The Authors. Published by IOP Publishing. Table 1 Comparative advantages and disadvantages of different selected fabrication technologies used within TERM and biofabrication Technology 1 (T 1 ) Technology 2 (T 2 ) Advantages Disadvantages Refs. Solution Electrospinning Electrospraying Low cost; simple to establish; Lack of polymer/solvent options; generation of airborne particles; [ 39 ] Solution Electrospinning Extrusion‐based Printing Low cost; high total volume; surface area; improved in vitro Solvent use; volume fraction occupied by polymer high; [ 43, 48, 49, 50 ] Solution Electrospinning Inkjet Cell deposition control; Solvent use; electrospun substrate required on fluid; small pore size; [ 41 ] Extrusion‐based Printing Extrusion‐based Bioprinting High total volume; Volume fraction occupied by polymer high; [ 40, 54 ] 2PP Spheroids Individual building blocks have mm sizes; High establishment cost; requires better seeding efficiency; [ 46 ] Extrusion‐based Printing Spheroids Dispensing control; mm‐sized cell building blocks; Complex printer; volume fraction occupied by polymer high; [ 20, 70 ] MEW Spheroids Low polymer fraction; design variations; Time to manufacture; manual seeding; [ 32, 81 ] MEW Extrusion‐based Bioprinting Low polymer fraction reinforcement; Commercial or custom‐built printer required; [ 97 ] MEW Electrospinning Reinforcement of solution electrospun tube with minimal layers; MEW layers align cells; Custom‐built printer required; [ 21 ] Melt Electrospinning Electrospraying Decoupling scaffold mechanics from level of drug delivery; Airborne generation of small particles; ventilation and safety required for electrospraying; [ 100 ] John Wiley & Sons, Ltd. 2. 1 Solution Electrospinning Undoubtedly, solution electrospinning [ 34 ] has had a profound impact on TERM research. Well‐reviewed in depth elsewhere, electrospinning is a processing method that stretches charge polymer solutions/melts into nano‐ and ultra‐fine fibers. [ 35, 36 ] Solution electrospinning is particularly pertinent for tissue engineering, as first illustrated in 2002 when Bowlin and colleagues electrospun collagen fibers. [ 37 ] This created significant interest in using electrospinning to mimic collagen fibrils, however the solid nature of collectors tended to result in compact fibrous nonwoven sheets with minimal porosity for cell penetration. [ 38 ] In 2006, Stankus et al. circumvented this challenge of cell penetration into solution electrospun meshes by simultaneously solution electrospinning and cell electrospraying [ 39 ] (Figure 2 ). Since most volatile solvents are toxic, and water can be slow to evaporate, this study used hexafluoroisopropanol (HFIP) to dissolve the elastomeric poly(ester urethane)urea into an electrospinning solution while smooth muscle cells isolated from the rat aorta were suspended within Dulbecco's Modified Eagle Medium and electrosprayed onto either a flat collector and a tubular mandrel. One benefit of a tubular mandrel in this instance is that the positions of the two nozzles/spinnerets could be positioned diametrically opposite to each other—this configuration is one repeated in numerous other papers involving tubular collectors. The cells/fibers formed striated layers even though deposition of the two sources was simultaneous. In 2008, Ekaputra et al. simultaneously combined electrospraying of a osteoblast‐containing heparin/hyaluronic acid matrix with solution electrospun PCL fibers and showed that leachable PEO fibers only minimally improved scaffold porosity in this architecture. [ 47 ] A different approach to increasing the porosity of electrospun materials also addressed a disadvantage for extrusion‐based fabrication technologies, namely cell seeding efficiency. First to report from two almost simultaneously published papers, [ 43, 48 ] Moroni and colleagues combined extrusion‐based fabrication of molten poly(ethylene oxide‐terephthalate)/poly(butylene terephthalate) with the solution electrospinning of the same, dissolved polymer. [ 43 ] After seeding with chondrocytes, the electrospun containing group had significantly higher cell entrapment, and the glycosaminoglycan/DNA ratio significantly higher after 28 days. Furthermore, the chondrocytes were spread on the extruded scaffold and were rounded when the electrospun fibers were integrated. In a similar experiment with chondrocytes and with similar findings, Park et al. used poly(ε‐caprolactone) for the two components; one extruded as a melt and the other dissolved in HFIP and electrospun between the layers of the extrusion‐based fabrication process [ 48 ] ( Figure 3 A ). Similar studies by Mota et al. combining these two fabrication technologies to demonstrate the different morphologies that result on the different diameter surfaces when MC3T3 murine preosteoblasts are seeded Figure 3 B. [ 49 ] Centola et al. was the first to generate such multimodal scaffolds (Figure 3C ) on tubular collectors, using material extrusion upon solution electrospun membranes. [ 50 ] Recently, solution electrospun tubes were hybridized with melt electrowriting by Jungst and colleagues, (Figure 2 ), to make multiphasic tubes from PCL for vascular applications. [ 21 ] Figure 3 Hybrid fabrication technologies. Combination of material extrusion and electrospinning on A) flat surfaces and B) seeded with MC3T3 cells on a hybrid material‐extruded and electrospun scaffold, shown with white and red arrows respectively. C) A similar approach but with a cylindrical collector. D) Combination of melt electrospinning and solution electrospinning. E) Live/dead staining from a combination of melt‐extrusion and extrusion‐based 3D bioprinting of decellularized bioink containing human inferior turbinate‐tissue‐derived mesenchymal stromal cells. F) Tissue section showing the combination of solution electrospinning and inkjet printing of cells. (A) reproduced with permission. [ 48 ] Copyright 2008, Elsevier. (B) reproduced under the terms and conditions of the Creative Commons CC BY 3. 0 License. [ 49 ] Copyright 2018, The Authors. Published by MPDI Publishing. (C) reproduced with permission. [ 50 ] Copyright 2010, IOP Publishing. (D) reproduced with permission. [ 51 ] Copyright 2010, Elsevier. (E) reproduced. [ 52 ] Copyright 2018, The Authors. Published by Springer‐Nature Publishing. (F) reproduced with permission. [ 41 ] Copyright 2013, IOP Publishing. While solution electrospinning attracted extensive attention for tissue engineering, melt electrospinning typically produced larger diameter fibers [ 53 ] that were (at the time) beyond the nanoscale dimensions that drove nanotechnology research. However, this size discrepancy was used in a hybrid fabrication approach, Kim et al. combined melt electrospinning of poly(lactide‐ co ‐glycolide) (PLGA) to establish large fibers, while dissolving PLGA into HFIP and electrospinning the solution produce the sub‐micrometer diameter fibers (Figure 3D ). [ 51 ] Similar to the aforementioned articles, this configuration used two diametrically opposed nozzles/spinnerets and collected onto a tubular mandrel in the middle. Such a combined structure provides volume due to the larger diameters and enables improved cell seeding with human epidermal keratinocytes. [ 51 ] 2. 2 Bioprinting Approaches While electrospinning is often used as a second process to introduce a high surface area for cell attachment, [ 50 ] the combination of bioprinting with extrusion‐based fabrication technologies was adopted by Malda and colleagues (Figure 2 ) for a different reason. [ 40 ] When bioprinting a cell‐laden hydrogel for several layers to achieve volume, the weight of the upper layers of the bioink imparts a significant force on the underlying layers and the structure can collapse or become unstable. Therefore, a more rigid, material extruded (typically a biodegradable thermoplastic polymer such as PCL) fiber is used to minimize the forces imparted on the weaker bioink. [ 40 ] Decellularized tissue was also used as bioinks by Pati et al. within a material extruded scaffold as shown in Figure 3E. [ 54 ] In 2013, Xu et al. demonstrated a different hybrid printing approach involved electrospinning and inkjet bioprinting of cells (Figure 2 ). [ 41 ] A solution of PCL and Pluronic F‐127 was dissolved in acetone and electrospun onto a phosphate‐buffered saline‐filled Petri dish after which chondrocytes/fibrinogen/collagen were dispensed with an inkjet valve. This process was repeated until there were three electrospun layers and two bioink layers, 1 mm thick. Cell viability in vitro and their evaluation as hybrid cartilage constructs in vivo was performed (Figure 3F ). The mechanics of the hybrid constructs were superior compared to the printed hydrogel alone and the chondrocytes remained viable while producing cartilage‐specific extracellular matrix. [ 41 ] As outlined in‐depth later in this review, the melt electrowriting (MEW) of PCL fibers provided an alternative reinforcement structure for such voluminous hydrogel structures. [ 42 ] What is distinct about the research reported by Visser et al (Figure 2 ) is that significant mechanical reinforcement could be obtained under compression with 19 µm diameter MEW filaments that occupied between only 2% and 7% of the total construct volume. Individually, the PCL scaffold and hydrogel had a compressive moduli of ≈10–15 kPa, however substantially increased to 405 kPa when combined. [ 42 ] Numerous studies on such soft network composites have confirmed that small volume fractions of well‐positioned, small diameter fibers can profoundly affect the mechanical properties. [ 55, 56, 57 ] MEW has also recently been performed to generate chemically crosslinked hydrogels for biomedical applications, [ 58 ] and could be considered a secondary structure similar to extrusion‐based systems previously described by Schuurman et al. [ 40 ] A prerequisite for combining bioprinting with other fabrication technologies is the precise control over the material properties of the biomaterial inks, and in case it is a formulation of cells and materials, the bioinks. [ 59 ] Crucial material parameters such as a proper rheological behavior, the corresponding characterization, postfabrication curing and shape fidelity, and most recent material and fabrication developments have recently been reviewed in several comprehensive reviews and are thus not further discussed here. [ 60, 61, 62, 63, 64, 65, 66 ] 2. 3 Tissue Spheroids Bioassembly Tissue spheroids are densely packed cell aggregates and the idea to use tissue spheroids as building blocks in biofabrication and bioprinting were introduced almost a decade ago. [ 67 ] There are numerous advantages of using tissue spheroids in biofabrication as building blocks: i) they are 3D and can have maximal cell density for replication of natural condensation reactions observed in developmental tissue biology and tissue formation; ii) they have a spheroidal shape suitable for bioprocessing; iii) they have intrinsic capacity for tissue fusion or formation of larger size tissue engineered constructs via self‐assembly process; iv) it enables self‐assembly/spheroid formation with multiple cell types, i. e. , spheroid coculture to replicate the native condensation niche and cell‐cell communication; v) the diameter of tissue spheroids (usually 250 ± 50 µm, also up to 1 mm in diameter) allows significant (up to 20 times) reduction in the number of fabrication layers and, thus, reduce printing time required) spheroids have been shown to form almost all tissue types. Tissue fusion is a ubiquitous natural process and occurs during embryonic development. [ 68 ] In order to create 3D tissue constructs from tissue spheroids they must be positioned so that they can fuse. Over the past decade many hybrid fabrication technologies have involved tissue spheroids. Lang et al. (Figure 2 ) [ 44 ] first melt‐extruded a porous thermoplastic scaffold and sequentially assembled tissue spheroids into the pores. Mekhileri et al. used a more advanced automated approach employing a specially designed hybrid 3D bioprinter enabling microfluidic singularization and precision placement and insertion of tissue spheroids into scaffolds designed to mimic mechanical properties of the target surrounding native tissue (Figure 2 ). [ 20 ] This provides the capacity for the overall construct to be manufactured using a “modular” 3D bioassembly strategy. [ 69 ] Unlike dissociated cell seeding that is inefficient in extrusion‐based fabrication structures without an electrospun mesh to entrap individual cells, modular spheroids are of such dimensions that they allow 100% seeding efficiency without the electrospun component combined with precision 3D spatial organization or arrangement of individual tissue modules. Biofabrication of spheroids embedded within 3D bioprinted bioinks was further demonstrated with a “multiarm bioprinter. ” [ 45 ] In this instance spheroids could be placed discretely and a coaxial nozzle permitted the codelivery of an alginate solution with a calcium chloride solution to facilitate crosslinking. This approach enabled the spheroids to be placed and also fixed into position with a reinforcing structure. While such reinforcing structures are extruded to allow matrix and spheroid handling and incubation, at 400–600 µm in diameter, such filaments occupy a significant volume fraction of the overall tissue construct (below 75 vol% porosity). [ 45 ] While many biofabrication approaches adopt extrusion‐based principles to deposit biomaterial inks and bioinks, light‐based technologies have a different set of capacities. Widely appreciated as one of the best resolved AM technologies, two‐photon polymerization (2PP) can be used to design a different type of reinforcement construct. One example is the “lockyball, ” which are spherical, porous containers that have protruding hooks and structures that interlock. [ 46 ] Silva et al. showed that cells can be seeded within such lockyballs (Figure 2 ), that facilitate large volume structures through interaction and assembly of the locking structures. When discretely dispensing two different types of spheroids into a dome structure made by extrusion‐based printing Mekhileri et al. (Figure 2 ) provided a demonstration of future spheroid bioassembly for hierarchical structures. [ 20 ] Using the osteochondral defect as a target tissue, the extrusion of a thermoplastic copolymer was used to fabricate sequential layers, including the anatomic dome shape of the articular cartilage surface. This paper is the first to show discrete automated 3D bioassembly of cellular spheroids and cell‐laden hydrogel spheroids into reinforcing scaffolds, with the automated singularization and assembly of spheroids occurring in a microfluidic aqueous environment, i. e. , the nozzle is under the surface of the media when dispensed. This strategy represents an automated layer‐by‐layer hybrid biofabrication approach alternating between extrusion‐based 3D plotting of thermoplastic polymer and microfluidic bioassembly of preformed cellular spheroid modules. This research group also described the development of a 96‐well plate bioreactor showing perfused drug screening on 3D bioassembled cocultured cancer spheroid model (i. e. , 3D plotted scaffold + coculture cancer spheroids). [ 73 ] Similar hybrid approaches have been employed by Ozbolat and colleagues (Figure 2 ) using alginate hydrogels for fabricating porous scaffolds suitable for placing tissue spheroids. [ 45 ] 3D Bioprinting Solutions, a company in Russia, also developed a hybrid biofabrication technology using original multifunctional bioprinter “Fabion” [ 74 ] ( Figure 4 B ). Figure 4 Hybrid biofabrication technologies including manual pipetting (top‐seeding) of 3D tissue. A) Extrusion printing of thermoplastic scaffold combined with precision placement and fusion of tissue spheroids (chondrospheres), stained with Saf‐O. B) Extrusion bioprinting of collagen hydrogel with robotic placement of spheroids. C) Extrusion bioprinting of collagen hydrogel with placing of ovarian follicle (white arrow) using pipetting. D) Electrospinning of polyurethane matrix with robotic placing of tissue spheroids fabricated from human fibroblasts. E) Biofabrication of tissue spheroids inside synthetic microscaffold ("lockyball") fabricated by 2PP. F) Top seeding of tissue spheroids into scaffold fabricated by MEW. G) Precision bioassembly of multicellular tissue spheroids in sequential layers of a mechanically stable and anatomically shaped 3D scaffold. (A) reproduced with permission. [ 70 ] Copyright 2012, Springer‐Nature Publishing. (B) reproduced with permission. [ 71 ] Copyright 2019, Springer‐Nature Publishing. (C) reproduced under the terms and conditions of the Creative Commons CC BY 4. 0 License. [ 72 ] Copyright 2017, The Authors. Published by Springer‐Nature Publishing. (D) reproduced under the terms and conditions of the Creative Commons CC BY 4. 0 License. [ 30 ] Copyright 2016, The Authors, Published by Whioce Publishing Pte Ltd. (E) reproduced under the terms and conditions of the Creative Commons CC BY 4. 0 License. [ 46 ] Copyright 2016, The Authors, Published by PLOS ONE. F) reproduced with permission. [ 32 ] Copyright 2018, The Authors. Published by WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. (G) reproduced under the terms and conditions of the Creative Commons CC BY 3. 0 License. [ 20 ] Copyright 2018, The Authors. Published by IOP Publishing. In another approach, tissue spheroids were robotically placed onto a polyurethane electrospun matrix [ 30 ] (Figure 4D ). These tissue spheroids attach and spread on the surface of an electrospun matrix. Precision robotic placement and patterning of tissue spheroids on the surface of electrospun matrices allowed control over the thickness of resultant 3D tissue construct after attachment, spreading and tissue fusion. Moreover, it has been suggested (but not yet implemented) that tissue spheroids could be also bioassembled on the opposite side of an electrospun matrix enabling more complex layered 3D tissue construct from different types of tissue spheroids. 3D Bioprinting Solutions used robotic 3D Bioprinting of tissue spheroids biofabricated by rounding of mouse embryonic tissue explants on printed collagen scaffolds for biofabrication of first functional and vascularized organ construct—a mouse thyroid gland. [ 75 ] Implantation of bioprinted mouse thyroid gland construct under kidney capsular into the experimental animal model of hypothyreosis using radiation ablation allowed for restored levels of the thyroid hormone thyroxine (T4). [ 75 ] The tissue spheroid could be also formed in microscaffolds fabricated by two photon polymerization [ 46, 76, 77 ] (Figure 4E ). Additionally, the seeding and capture of tissue spheroids into specifically designed MEW scaffolds was readily performed [ 32 ] (Figure 4F ). Thus, different types of solution electrospun, melt electrospun, melt‐extrusion based 3D plotting and two‐photon polymerized scaffolds could be used for development of hybrid fabrication technologies combining tissue spheroid assembly. The automated bioassembly of tissue spheroids using fluidic handling and special hybrid 3D bioprinters [ 20, 78, 79 ] has an obvious advantage compared with manual placement in further advancing of tissue spheroids based hybrid biofabrication technologies. This also leads to the development of complex automated 3D in vitro tissue models for medium‐ or high‐throughput screening. Material‐extruded scaffolds for reinforcement often occupy considerable polymer volume; space where cells could be occupying to regulate their microenvironment until the reinforcing part degrades. Conversely, MEW readily produces scaffolds above 80 vol% accessible to the cells. [ 80 ] In 2018, Hrynevich et al. used MEW in a manner that can create a spectrum of fiber diameters, with each fiber placed in a specific position. [ 32 ] This was expanded in a separate paper to show such composites can be sectioned, stained and handled; increasing the number to many hundreds of spheroids was demonstrated into a construct that could be further manipulated (Figure 2 ). [ 81 ] The capacity to design reinforcing scaffolds at the low‐micrometer resolution provides an additional tool for future hybrid fabrication approaches. Aspects of the printing process that could cause the structure of the spheroid to be compromised during and after printing have been raised or addressed in literature that has pioneered spheroid bioassembly approaches. For example Mekhileri et al. [ 20 ] demonstrated that by adopting a fluidic spheroid singularization and insertion print head strategy, cell viability and spheroid shape was maintained and equivalent to non‐bioassembled control spheroids. Furthermore, assembly of cell‐encapsulated hydrogel beads or spheroids offers additional protection to cells from flow, shear or mechanical stress during assembly. The development of hybrid approaches that aim to combine spheroids and/or organoid tissue modules should be cognizant of the biological demands and fragility of spheroid/organoid manipulation during optimization and automation of biofabrication systems and determination of spheroid viability, composition and integrity must be validated throughout all steps of any bioassembly strategy and form a key part of any proof‐of‐concept validation studies. 2. 4 The Kenzan Method The Japanese word “Kenzan” originates from a spiky structure used for flower arrangement. From the biomedical fabrication perspective, this approach positions spheroids into 3D arrangements by skewering them onto microneedles ( Figure 5 A – D ) that are spaced sufficiently to allows spheroid fusion. Deformation forces caused by skewering the spheroids [ 82 ] during assembly are not reported to affect cell viability, ECM production or fusion. With the Kenzan method sheets, tubes and heterogeneous tissue have been fabricated by using the appropriate spheroid types and position. [ 83 ] For example, Figure 5E, F shows how smooth‐muscle forming spheroids made from human induced pluripotent stem cells form one week after removal from the microneedles. During their positioning on the microneedles, then spheroids fuse and seemed to encapsulate microvascular fragments. [ 82 ] Figure 5 Overview of the Kenzan Method. A–D) Schematic of the needle‐based substrate that allows assembly and consequent fusion of spheroids into a tubular structure. E, F) Immunofluorescent staining of one such construct after one week removed from needles prepared from human induced pluripotent stem cell‐derived smooth‐muscle forming cell spheroids. G, H) A hypothetical hybridization of Kenzan with 2PP, solving a persistent challenge of how to embed spheroids within a capillinser structure. Scale bars for (E, F) are 1 mm. (A–D) reproduced under the terms and conditions of the Creative Commons CC BY 4. 0 License. [ 84 ] Copyright 2017, The Authors. Published by Springer‐Nature Publishing. E, F) reproduced with permission. [ 82 ] Copyright 2017, Wiley Publishing. G, H) provided by Dr. Fred Pereira. While the Kenzan method is described as scaffold‐free, one can argue that the microneedles are a “temporal support” that behaves as a scaffold. Irrespective of the definition, it is a fascinating approach that could have particular applicability to hybrid fabrication approaches. Since the Kenzan method allows the physical transfer of tissue structures to another location, they can be sustained for fusion and maturation within a spectrum of environments. One can consider how other AM methods (such as 2PP, MEW; stereolithography (SLA)) can be combined with the Kenzan method to expand the manufacturing possibilities (Figure 5G, H ). In this hypothetical example of hybrid fabrication, the challenge of encapsulating spheroids within a 2PP‐fabricated capillinser could be solved using the Kenzan method. 2. 5 MEW There are upper and lower diameter limits for electrospinning and microextrusion respectively, due to inherent physical phenomenon with processing such viscoelastic fluids. From a manufacturing perspective, MEW can be considered a hybrid fabrication technology in itself between electrospinning and microextrusion. Distinct from both these two technologies, the fiber diameter is variable, ranging from 800 nm to 150 µm (typically much lower than microextrusion), while fiber placement is excellent (which remains a disadvantage in solution electrospinning). MEW is, unlike bioprinting, a cell‐free manufacturing process, although work on electrohydrodynamic (EHD) jetting with cell‐containing fluids enables this feature. [ 85, 86 ] The EHD jetting of the MEW process relies on a phenomena that stabilizes fluid columns at low flow rates with a voltage that is applied across a nozzle and collector. [ 87 ] Outlined by Taylor in 1969, [ 88 ] and shown as “floating water” bridges, [ 89 ] such applied voltages prevent fluid column break up [ 87 ] between two points. This differs from electrospinning where a high voltage is used to initiate electrical instabilities (i. e. , whipping) in a jet to produce ultra‐fine fibers. [ 35 ] MEW has been a notable technology for hybrid fabrication, and is reviewed elsewhere in this context. [ 90 ] An advantage of MEW is that low‐micrometer‐scale fibers can be accurately placed into position, [ 32, 43, 48, 49 ] first shown in 2011 by Brown et al. [ 93 ] MEW scaffolds have been used for both in vitro [ 91 ] ( Figure 6 A – C ) and in vivo [ 94 ] research. MEW scaffolds tend to have a high porosity, typically from 80 vol% and even up to 98 vol% pore volume (Figure 6B, C ). [ 92 ] This allows for both cell attachment as well as self‐organization within the scaffold pore. Jungst et al. (Figure 2 ) combined MEW with solution electrospinning, then seeded with endothelial and stromal cells make tubular vascular grafts. Interestingly, thin solution electrospun PCL layers did not affect the accuracy of MEW PCL fiber placement onto the collector, and good fiber fusion between the two regions was seen (Figure 6D ). The capacity to accurately place MEW fibers also leads to their use as a customizable support structure for spheroids (Figure 6E, F ). [ 81 ] Figure 6 Examples of MEW products and integration within biofabrication. A) Top‐seeded human dermal fibroblasts on a MEW scaffold with a 30˚ pitch deposition. Osteoprogenitor cell line seeded on a 98% porous scaffold after B) six and C) 14 days in vitro, demonstrating typical pore closure for ECM depositing and proliferating cells. D) A multiphasic tube combining a solution electrospun substrate (SES) that is subsequently MEW at a specific angle. E) An adipose‐derived spheroid sheet for transfer, including F) SEM image of two spheroids in adjacent pores. G) sinusoidally printing and intersecting patterns for highly flexible scaffolds or soft network composites. (A) reproduced with permission. [ 91 ] Copyright 2013, IOP Publishing. (B, C) reproduced with permission. [ 92 ] Copyright 2015, Mary‐Anne Liebert. (D) reproduced with permission. [ 21 ] Copyright 2019, The Authors. Published by Wiley. (E, F) reproduced with permission. [ 81 ] Copyright 2019, The Authors, Published by WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. (G) reproduced with permission. [ 55 ] Copyright 2017, American Chemical Society. The defined placement of MEW fibers has a profound effect on their mechanical properties, when embedded within a second matrix. Bending of MEW fibers is restricted by the matrix and significantly higher mechanical properties can be achieved while maintaining a low volume composite fraction. For example, Visser et al. showed that a MEW fiber‐reinforced matrix was much stronger in compression (≈ 50 times) than the individual components, despite the MEW scaffold occupying only 7% space within the composite volume. [ 42 ] This form of mechanical reinforcement also extends to shear stresses. [ 95 ] It is therefore an efficient method to reinforce matrices, including GelMA, [ 42, 55, 57 ] alginate, [ 42 ] and Matrigel. [ 96 ] Furthermore, the mechanics of the MEW composite can be tuned, by the fiber placement. Direct writing the fibers in a sinusoidal pattern (Figure 5G ) [ 55 ] affects the mechanical properties, with tensile testing demonstrating a distinct “toe region” that can be controlled with different amplitudes and the wavelength laydowns. In another example of MEW used for hybrid fabrication, it was combined with bioprinting to produce a soft network composite. [ 97 ] The capacity to significantly alter the fiber diameter during printing extends the design capability for MEW scaffolds. [ 32 ] With an appropriate calibration, MEW scaffolds can be constructed based on their intended dimensions rather than a specific outcome that is usually defined by the stable manufacturing parameters. [ 32 ] Furthermore, the desired mechanics of a target tissue—demonstrated with heart valves—can be designed into the MEW reinforcing structure so that the reinforced hydrogel displays similar mechanical behavior. [ 98 ] With comparatively less research performed on this technique compared to electrospinning, the use of such voltage‐stabilized jets offer more options in design. While use of melts is advantageous from a rapid fabrication perspective, the stabilized jet is equally applicable to polymer solutions, [ 99 ] including bioinks. [ 86 ] From a fabrication perspective, avoiding the use of organic solvents for MEW has advantages when it comes to potential toxicity and ventilation requirements. However heat can affect chemical stability, and nonwoven fabrics based on melt electrospinning have been previously coupled with solution electrospraying for drug delivery purposes. [ 100 ] The millimeter‐scale collector distance between the nozzle and the sample in MEW has also enabled monitoring of the jet to obtain useful information. In a research article introduced in 2019 to demonstrate high‐throughput screening of printing parameters (i. e. , Printomics), Wunner et al. devised a conveyor‐belt collector system that allowed digitized information to be automatically generated from a fully automated printing system. [ 101 ] This allows the rapid identification of stable processing within a multiparametric printing process such as MEW. While beyond the scope of this review, such digitized information is well‐suited toward artificial intelligence‐based manufacturing directions. There have been several publications related to machine learning within biofabrication to date. [ 102, 103, 104 ] 2. 6 Tomographic Volumetric Bioprinting Many of the afore‐mentioned fabrication technologies use AM principles, in that the scaffold or construct is made in a layer‐by‐layer approach. Tomographic manufacturing produces objects in a volumetric manner, by projecting a series of images into a rotating container so that an object is only created in the regions where sufficient photopolymerization occurs ( Figure 7 ). Described as being a “reverse tomography” technique, volumetric bioprinting is distinguished by the extremely rapid production of a 3D object—in the range of 20–30 s. This method was inspired by a medical therapy treatment procedure in which a specific dose of ionizing radiation is deposited in 3D by rotating an intensity modulated source around the patient. This was described first by Mackie et al. in 1993. [ 105 ] The implementation of this technique in the visible light range for photopolymerization was performed independently 25 years later by two groups as described in the works of Kelly et al. [ 106, 107 ] and Loterie et al. [ 108 ] The latter authors applied this approach to bioprinting, using visible light initiators and incorporated cells within the photopolymerizable resin. [ 109 ] The current printing resolution for volumetric printing is 80 µm. [ 110 ] Figure 7 Volumetric bioprinting and its potential utility in hybrid fabrication. A) Schematic showing how incoming light can start a gelation process that B) produces a specific shape in the center when the sample in rotated while the projected image is altered. C) A photograph of the final printed material, which has a smooth surface finish. D) A CAD drawing of a meniscus. E) A photograph of the volumetrically printed product. F) A live/dead stain after 28 days. G) A graph showing significant cell compatibility and improvements in mechanical strength of the tissue. H, I) A demonstration of how hybrid fabrication could be achieved by placing structures within the container prior to volumetric bioprinting. H) A PCL MEW tube with aligned fins that allows light to transmit (or project) into the center. I) A schematic of how spheroids (blue) positioned with the Kenzan method could be further entrapped with a bioresin. (A–I) reproduced with permission. [ 109 ] Copyright 2019, The Authors, Published by WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim. (H) includes a photograph of a MEW tube kindly provided by Mr. Thomas Robinson. The disruptive potential of volumetric bioprinting on TERM and biofabrication research is substantial, especially if the method can be delivered at an economical price. While the rapid production of millimeter volume cell‐containing constructs is in itself exciting, it is the capability to hybridize volumetric bioprinting with other manufacturing technologies that will drive this technology even further. For example, it is not unreasonable to predict that the photopolymerizable resin can be mixed with other AM fabricated constructs, such as those made by MEW, SLA or 2PP prior to, or after volumetric bioprinting. Furthermore, the photopolymerizable resin could be modified with another process prior to volumetric bioprinting to create discrete structures or gradients within the printed construct. 2. 7 Increasing Complexity of Living Building Blocks—The Potential for Fabrication with Organoids Organoids are stem cell derived self‐organized mini‐tissues which have certain morphological and functional characteristics of natural human organs. [ 111 ] They differ from spheroids in that organoids are formed by natural growth and differentiation (self‐organization), while spheroids are fabricated by artificial (bio)assembly from cell suspensions. [ 112, 113 ] There are already at least four published papers which suggested (but not implemented yet) the combination of bioprinting [ 112, 114 ] and biofabrication [ 3, 115 ] with rapidly emerging organoid technologies. Organoids usually have irregular shape, they are not vascularized and do not have stroma and innervation, but they have internal histological structure closely resembling the histomorphology of natural human organs. Moreover, they recapitulate certain physiological functions of human organs. In this context, human organoids are more advanced and potentially more attractive alternative to tissue spheroids in the development of hybrid biofabrication technologies. Modern biofabrication and microfluidics technologies can enable vascularization and innervation of organoids and, thus, could made them physiologically and morphologically even more relevant and similar to native human tissue and organs. [ 116 ] In particular, the main future development will be the automated 3D manufacture and increased complexity and throughout offered by hybrid biofabrication and organoid technologies that offer significant breakthroughs in developmental biology, drug and disease screening in precision medicine. [ 3 ] Engineering of organoids of desirable size and shape is already under investigation using specially designed biofabrication technologies and novel biopolymers and hydrogels. [ 112, 113, 117 ] There is a role for biomaterials in organoid research, [ 118 ] and in establishing fabrication methods to investigate/control their behavior. [ 117, 119 ] Sizing, shaping, and directed differentiation of organoids using advanced biomaterials is already started [ 112, 113 ] and could be developed further using advanced hybrid fabrication technologies. 3 Implications of Future Hybrid Fabrication Technologies The use of hybrid fabrication in this review is directed toward TERM, biofabrication and bioprinting. In the context of hybrid fabrication, there are several implications for these fields. First, a hybrid approach enables the development of more sophisticated and more biomimetic TERM scaffolds and tissue constructs. Second, the hybrid approach is able to increase scaffold porosity without compromising its biomechanical properties and viability. Third, hybrid biofabrication can increase the initial cell density and, thus, save time otherwise necessary for cell proliferation and tissue engineered construct maturation. Finally, both manual and robotic‐based hybrid technologies allow greater precision assembly of cell and tissue spheroids. 3. 1 Increasing the Available Volume for Tissue Constructs Some of the most recent contributions by MEW involve being able to reinforce a matrix while maintaining an overall low volume fraction. We believe that this is essential for voluminous tissue constructs that require both cell‐friendly matrices and handling capabilities. So far, there have been two general directions for TE; scaffold‐based and scaffold‐free. [ 77 ] One can argue that there is a middle‐ground of “high porosity” scaffolds, including those made via MEW, that have a low polymer volume fraction than most other scaffolding technologies. [ 120 ] In these cases, the volume occupied by a non‐cell‐penetrating structure is less than 10%, and currently as low as 2%. [ 92 ] The substantial mechanical reinforcing effects observed by well‐positioned, low‐micrometer dimension fibers provides another argument for such “high porosity” approaches. In a different context, increasing the overall volume of TERM constructs to the centimeter scales required for human tissue has been a long‐standing challenge in the field. [ 7 ] Such an overall volume of tissue constructs or high cell densities require a vascularization strategy depending on their intended use, [ 28, 116 ] well reviewed elsewhere. [ 121 ] 3. 2 Increasing Complexity of Scaffolds and Tissue Constructs with Biomimicry Hybrid fabrication strategies offers practically unlimited opportunities for creative combination of different technologies in order to achieve highly desirable histotypical organization of tissue constructs with maximal possible level of complexity and authenticity. Closely packed tissue spheroids can, after tissue fusion, reach a higher density than in natural tissue and organs. In certain tissues (for example, mesenchymal condensation in cartilage during embryonic chondrogenesis) this increased high density is a necessary precondition for sequential initiation of extracellular matrix synthesis and deposition. [ 69 ] Thus, it is expected that combining tissue spheroids with high packing cell density will provide more voluminous TERM products. In turn, vascularization strategies become important to maintain nutrition. This leads toward the several levels of complexity within natural tissue and organs that require addressing. The establishment of a complex vasculature, chemotactic gradients, heterogeneous cell densities biomechanical cues and mimicking the ECM distributions are just some areas where current fabrication technologies are far away from matching the delicate and complex distribution for natural tissue. Previous hybrid fabrication research has touched upon this, including the integrated tissue–organ printer that combines melt extrusion and bioprinting for the fabrication of various vascularized musculoskeletal tissue. [ 7 ] Mechanical reinforcement principles, as previously outlined, have been first performed with melt extrusion and bioprinting/spheroid printing for bone/cartilage tissues [ 20, 52 ] and recently followed by MEW for cartilage and heart valve applications. [ 42, 98 ] Approaches to vascularize tissues by Kolesky et al. , also showed how combining different extrusion based multimaterial bioinks and fugitive inks can be used for a more complex vascularized structure for liver biofabrication, [ 28 ] and for more large volume tissues. [ 122 ] In another important study on complex vasculature fabrication, the DLP of a hydrogel allowed the mechanical environment of vascular channels surrounding alveolar to be studied. [ 6 ] More recently, the potential of a multimaterial switching during bioprinting was demonstrated that could further increase distributions requiring a change in cell density and mechanical reinforcement. [ 123 ] Cell self‐organization can be somewhat relied upon to establish such biomimicry, the improved resolution of the “fabrication technology toolbox” to establish a sufficient microenvironment for tissue maturation is likely required. While replicating the complexity of natural tissue may seem insurmountable to perform with current fabrication technologies, improved resolutions with new fabrication technologies and their hybridization with others could advance our capability in this aspect. Improvements in fabrication resolution may also converge with our improved ability to map the cellular and molecular distributions within tissue. A separately developing field is the process of tissue clearing [ 124 ] which uses protocols to render entire organs/tissues such as the brain, spinal cord, eye and kidney, transparent. Combining tissue clearing with immunohistochemistry and light‐sheet microscopy, a fully 3D mapping of organs at the cellular and even molecular level can be obtained. There are several tissue‐clearing protocols such as CLARITY, [ 125 ] 3DISCO [ 126 ] and, more recently, SHANEL [ 127 ] that can provide greater volumetric information at the cellular level on intact organs. We foresee that such tissue clearing processes will converge with biofabrication through the digitization of discrete tissue at this cellular and volumetric level. Such information will converge with improved manufacturing resolutions to provide an increasingly complex biomimetic structure for printing pathways. Developing volumetric models of the 3D distribution of cells and ECM within an organ will become increasingly important for biofabrication. Interestingly, deep learning and neural networks have been applied to the analysis and understanding of such treated tissues. Excellent research articles [ 125, 126, 127 ] describing in‐depth such tissue clearing methods and reviews [ 128, 129 ] can be found elsewhere. 3. 3 Manual or Automated—More Precision Placing of Cells and Spheroids When combining cells or cell‐laden matrices to AM scaffolds, manual pipetting remains the most commonly used approach, often related to cost. For example, in a breakthrough paper, Laronda et al. used manual pipetting to place tissue spheroids (ovarian follicles) onto 3D printed gelatin constructs (Figure 4C ). [ 72 ] This implantation of an artificial printed “ovary” into sterilized mice to restore their fertile function provided a dramatic demonstration of how biofabrication can tackle significant medical challenges. [ 72 ] This trend of automation is even progressing toward in situ bioprinting [ 130 ] of cartilage, [ 131 ] skin, [ 132 ] and cranial defects. [ 133, 134 ] The variance of results and the increasing complexity of experiments demands that such processes eventually require automation. This transition is becoming economically easier with more options for low‐cost automated dispensing technologies becoming available. Since the central hypothesis of biofabrication is that hierarchical cell/spheroid placement is essential to recapitulate more in vivo like tissue constructs, then defined dispensing and direct writing is an inevitable but important evolution in this aspect. Biofabrication, and the accompanying automation could alter the long‐appreciated challenge of primary cell efficacy such as a loss of potency, change in phenotype and factors associated with rapid cell proliferation. Automation won't remove the issue of loss of phenotype but the repeatable cell handling steps can reduce variation make the most of potency available. Integration of fabrication systems with bioreactor systems may also provide opportunity for tissue maturation in this sense. [ 3 ] Automated fabrication strategies that do not require cells but deliver cues or factors that can recruit host cells or modulate repair offer new opportunities to potentially circumvent this loss in primary cell efficiency. Another manual dispensing example involves fiber‐reinforced hydrogels that have been primarily made by manual dispensing of precursors prior to gelling. Such improvements in mechanical properties are also observed with hybrid MEW and 3D bioprinting approaches. The need for more reproducible outcomes and more discrete hierarchical structures within biofabrication will drive the replacement of manual dispensing with automated systems. The fluid mechanics and variations currently observed with the seeding of cells is supporting the hypothesis that adopting a fully automated dispensing system will improve both the quality and reproducibility of experiments. 3. 4 Assembly Lines and Multifunctional Biofabrication Printers The design and implementation of automated biofabrication devices is another important trend. It opens a direct pathway for automated, scalable, standardized and cost‐effective biofabrication of complex tissue constructs which is critically important for successful clinical translation and commercialization of TERM products. Another strategic question is how hybrid fabrication technologies will be implemented together. Based on previous approaches this would be through i) the use of a multifunctional device combining a single set of different functional capabilities or ii) by development of separated mono‐functional devices integrated into one united biofabrication assembly line [ 135 ] ( Figure 8 A ). Moreover, certain fabrication technologies are not possible to integrate into one device due to certain technological restraints and limits. Figure 8 Different strategies in the development of hybrid fabrication technologies (T1, T2, T3). A) Schematic demonstration of two different strategies for the hybridization of different manufacturing technologies: strategy 1 based on construction of integrated fabrication assembly line consisted of separate biofabrication devices (T1+T2+T3) and strategy 2 based on combination of different biofabrication technologies integrated into one single multifunctional device (T1/T2/T3). B) A practical implementation of strategy 1: a multifunctional printer concept incorporating several types of heads that allow hybrid fabrication. C) A practical example of implementation of strategy 2 as a biofabrication assembly line developed to automate tissue construct preparation. (B) is previously unpublished and was kindly provided by Dr. Felix Wunner and Professor Dietmar Hutmacher. (C) is previously unpublished and was kindly provided by Tom Bollenbach, Ph. D. , ARMI | BioFabUSA. There is also the availability of multifunctional biofabrication printers that contains all of the separate technologies to be hybridized, and maintain it in a single location. [ 74 ] There are several commercial multifunctional biofabrication printers based around the technology of extrusion‐bioprinting (Figure 8B ). With such multifunctional printers, the tissue construct is fabricated in a single location, and the different technologies combine at this point. In fact many commercial 3D bioprinters have thermoplastic extrusion heads on them, so as to perform hybrid fabrication and, in turn, will drive further research on this topic. Currently there are a number of commercial and laboratory‐based multifunctional biofabrication machines that allow for hybrid construct fabrication for tissue engineering and regenerative medicine and has been recently reviewed. [ 136 ] In general these machines utilize a collection of rapidly interchangeable print head, each offering a specific function and control systems such and temperature, pressure, flow rate etc. Fabrication technologies of each print head can typically include the following: 1. Extrusion hydrogel bioprinting of cell laden bioinks. 2. Extrusion thermoplastic polymer dispensing. 3. MEW of thermoplastic polymers with associated collector. 4. Melt or solution electrospinning with associated collector. 5. Piezo or jet‐based bioprinting (or inkjet printing). 6. Fluidic or needle (Kenzan) based bioassembly of spheroids. 7. Microfluidic print head containing multiple nozzles for multimaterial/multibioink dispensing. There are several initiatives that reduce human interactions from cell culture procedures, such as the “Skin Factory” biofabrication assembly line (now disassembled) developed by Fraunhofer IGB and other modular, assembly line biofabrication systems. [ 33 ] More recently, a low‐cost and modular, assembly line approach is being taken at the Advanced Regenerative Manufacturing Institute (ARMI), in Manchester, USA. This is an enclosed system built outside a cleanroom, which saves on costs and accessibility. With the backing of industrial partners, the purpose of ARMI and its BioFabUSA program is to develop the engineering tools for automated tissue construction. The first of many planned automated assembly line prototypes (Figure 8C ) has been built to identify improvements for the next version. Ultimately, factory‐line systems should be reduced in size to a similar footprint of a multifunctional printer. In the context of the numerous diseases and injuries to organs/tissue, full automation is ultimately the level of control necessary to deliver a scalable product that is both safe and validated. 4 Conclusions The evolving field of TERM, and more recently biofabrication, will be further advanced by the hybridization of manufacturing processes. There are already several variants of hybrid technologies which have been developed based on the combination of at least two fabrication technologies, and further combinations should accelerate in the next five years. Different technologies within a biofabrication or TERM approach must complement each other and provide better outcomes through synergistic effects. In this review, recent fabrication technologies—MEW, volumetric bioprinting, Kenzan method and spheroid formation—have been illustrated as examples of technologies where the full potential remains to be investigated. Two future strategies of hybrid manufacturing, either in multifunctional devices or with a fabrication assembly line have been discussed and outlined. Conflict of Interest The authors declare no conflict of interest.
|
10. 1002/advs. 201903027
| 2,019
|
Advanced Science
|
Pressure‐Driven Two‐Input 3D Microfluidic Logic Gates
|
Abstract Microfluidics is a continuously growing field with potential not only in the fields of medical, chemical, and bioanalysis, but also in the domains of optics and information technology. Here, a pressure‐driven 3D microfluidic chip is demonstrated with multiple logic Boolean functions. The presence and absence of fluid at the output of the gates represent the binary signals 1 and 0, respectively. Therefore, the logic gates do not require a specially functionalized liquid to operate. The chip is based on a multilevel of poly(methyl methacrylate) (PMMA)‐based polymeric sheets with aligned microchannels while a flexible polyimide‐based sheet with a cantilever‐like structure is embedded to enable a one‐directional flow of the liquid. Several Boolean logic functions are realized (AND, OR, and XOR) using different fluids in addition to a half adder digital microfluidic circuit. The outputs of the logic gates are designed to be at different heights within the 3D chip to enable different pressure drops. The results show that the logic gates are operational for a specific range of flow rates, which is dependent on the microchannel dimensions, surface roughness, and fluid viscosity and therefore on their hydraulic resistance. The demonstrated approach enables simple cascading of logic gates for large‐scale microfluidic computing systems.
|
Recent developments in microfluidics 1 are tackling technological challenges in a wide range of applications including chemistry, 2 3D printing, 3 tissue engineering, 4 drugs development, 5 biomedical research, 6 and most lately organs‐on‐chip. 7, 8 Particularly, microfluidic devices which enable the handling of exceptionally low volumes of fluids in the range of micro‐ to pico‐liters are employed in biological and chemical analysis applications due to the precise manipulation of particles and liquids in a microscopic environment. 9 Additionally, microfluidic devices can provide the opportunity to analyze, isolate, concentrate, control, and identify biomolecules with an improved sensitivity and throughput, in addition to being simpler and easier than conventional techniques. For instance, microfluidics have recently shown the capability to use very small amounts of samples and chemicals to detect cancer cells and their interaction with myeloid cells with high sensitivity, high resolution, fast analysis, and low cost. 10 The future generation of microfluidic devices should be capable of performing in situ complex sample analysis and treatment. Today, integrated circuits can execute complex operations using electronic building blocks known as logic gates. This allows the system to autonomously make decisions following Boolean rules, thus eliminating the necessity for any manual intervention. Therefore, in order to automatically categorize, tag, isolate, and identify markers in complex fluidic samples such as blood, drugs, sweat, and so on, it is essential to include microfluidic logic functions in the miniaturized analysis. To date, multiple techniques have been demonstrated for logic computing including fluid flow resistance, 11 electrochemical reactions, 12 fluorescent molecular devices, 13 nonlinearity in fluid viscosity, 14 pneumatic pressure, 15 and bubbles flowing in microchannels. 16 Nonetheless, the major drawbacks of these approaches lie in the different interpretations of the input/output signals in addition to requirement for specially functionalized liquids for the logic gates to operate which makes the scaling and integration of multiple logic gates more complicated and challenging to achieve. As a result, microdroplet‐based microfluidic computation has received a growing attention in the past years due to its simple interpretation of output signals, where the presence and absence of the droplet represents the binary signals 1 and 0, respectively. However, this approach still requires the generation of microdroplets and their dispersal in another continuous liquid, in addition to the different required mechanisms for microdroplet movement such as relative flow resistance, 17 applied voltage, 18 magnetic field, 19 etc. To overcome these challenges, here, we report pressure‐driven 3D microfluidic logic gates that can operate using any fluid. The operation of the chip is based on the fluid pressure within the microchannels which is dictated by the flow rate defined using a syringe pump, a conventional tool that is used in most of the microfluidics applications. It is worth to note that the flow pressure within the channels depends on the hydraulic resistance as well which is a function of the microchannel dimensions, surface roughness, and fluid viscosity. However, these variables are usually fixed and known for a specific device and fluid. The 3D microfluidic chip is fabricated using CO 2 ablation of poly(methyl methacrylate) (PMMA) sheets and their following bonding using a thermocompression process. It is important to note that PMMA has been one of the most popular material for microfluidic devices due to its transparency, low cost, rigidity, reliability, and compatibility with different existing biomolecular techniques. 20 Therefore, this work enables easy integration of logic gates within most of the existing microfluidic devices. Several Boolean logic gates including AND, OR, and XOR are realized, in addition to a half adder digital microfluidic circuit, which is an essential component in the arithmetic logic unit (ALU) in microprocessors. In the demonstrated devices, the presence of the fluid at the output is interpreted as a binary signal 1 while the absence of the fluid is binary signal 0. The presented approach allows easy integration and cascading of microfluidic logic gates for complex logic computations. The microfluidic chip is based on the 3D integration and stacking of four layers of PMMA sheets with microchannels. CO 2 laser ablation with different power and speed recipes is used to create microchannels with different dimensions. The obtained shape of the microchannels is Gaussian due to their small widths ( Figure 1 a–c) which restricts the proper development of the well, while macrochannels with large widths (>1 mm) result in a rectangular shape. In particular, due to this effect, changing the width of the microchannel while fixing the power and speed of the laser result in channels with different depths (Figure S1, Supporting Information). Moreover, for a specific fixed width, it is known that in order to get deeper channels, the power can be increased with a constant cutting speed or the speed of the laser can be reduced for a giver power. This ensures that the laser ablation effect on a specific area is more pronounced, as a result, a deeper channel is obtained. However, it is observed that the obtained width of the microchannel is also affected as shown in Figure 1 a–c where a fixed power of 10% and a speed of 8% result in a channel width of 473 µm while reducing the speed to 4% results in a channel width of 518 µm. The same applies to the case when the laser speed is fixed and its power is modified, the width of the channel gets affected in addition to the depth. Therefore, in order to maintain the same width of the microchannel but to increase/reduce its depth, the laser recipe needs to be optimized in terms of both laser power and speed. Another observed effect during the CO 2 laser ablation of the PMMA is that the surface roughness of the channel sidewalls increases at higher speeds as shown in Figure 1 d–i. The surface roughness of the microchannel affects its hydraulic resistance and consequently the flow rate range of operation of the microfluidic logic gates. Figure 1 Microchannels fabrication in PMMA sheets and optimization using CO 2 laser. Cross‐sectional scanning electron microscopy (SEM) image of the microchannel fabricated using a) a power of 10% and speed of 8%, b) a power of 10% and speed of 6%, c) a power of 10% and speed of 4%. The width of the microchannel is shown to be dependent on the laser speed. SEM image of the sidewall of the microchannel d) of (a), e) of (b), and f) of (c). Surface roughness using a profilometer of the sidewall of the microchannel g) of (a), h) of (b), and i) of (c). Once the laser ablation recipes are optimized to get microchannels with different depths and widths, the PMMA sheets were then bonded using a thermocompression process as shown in Figure 2 a (Figure S2, Supporting Information). The microfluidic gates are pressure driven, therefore, the outputs of the two gates are designed to be located at different heights where the AND output is placed at a higher level than the OR output (Figure 2 b). As a result, a higher pressure drop is required in order to activate the AND gate. This is achieved when fluid flows in both inputs A and B resulting in fluid flowing out of both AND and OR gates ( A. B = 1, A + B = 1). However, when only one input is activated ( A ′. B = 1 or A. B ′ = 1), the pressure drop across the AND gate is not large enough to turn it on and the fluid flows only out of the OR gate. One challenge that has been observed during the operation of the microfluidic chip with one fluid input ( A ′. B = 1 or A. B ′ = 1) is the backflow of a small amount of the fluid into the other input due to capillary forces as shown in Figure 2 c. To overcome this challenge, a flexible polyimide (PI)‐based cantilever‐like structure is embedded in between the PMMA sheets at the intersection of both inputs to enable a one‐directional flow of liquids as shown in Figure 2 d. Using this approach, the AND and OR gates with two inputs are shown to work properly (Figure 2 e–i, Videos S1–S3, Supporting Information). Blue and yellow colored water are used at inputs and the output is green colored which confirms the mixing of both inputs (Figure 2 g–i). Similarly, an XOR gate is prepared using four layers of PMMA sheets and two PI layers with cantilever‐like structures to enable unidirectional flow of fluid between different PMMA levels and within the same level as well. The XOR gate is then combined with an AND gate to achieve a half adder microfluidic circuit ( Figure 3 a, b). The operation of the half adder circuit is shown in Figure 3 c–e (Videos S4–S6, Supporting Information) where the outputs in this case are the sum (sum = XOR ( A, B ) = A ⊕ B ) and carry (carry = AND ( A, B ) = A. B ). Figure 2 Microfluidic chip fabrication and design optimization. a) 3D microfluidic chip with multilevels of PMMA sheets. Microchannels are fabricated using CO 2 laser followed by thermocompression of the different layers. b) Cross‐sectional illustration showing the fluid flow in the microchannels. c) Fabricated microfluidic chip showing a challenge of fluid backflow into the inputs. d) Optimized microfluidic chip design with scaled down microchannels and a flexible polyimide sheet embedded between the PMMA sheets to enable one‐directional flow of fluid. Inset shows the stack of the rigid PMMA and flexible polyimide sheets for fluid backflow blockade. Scale bar: 1 cm. Simulated 3D microfluidic chip showing the AND/OR outputs for e) A = 1 and B = 0, and f) A = 1 and B = 1. g) Optical image of the operation of the microfluidic chip with inputs A = 1 and B = 0. Scale bar: 1 cm. h) Optical image of the operation of the microfluidic chip with inputs A = 0 and B = 1. Scale bar: 1 cm. i) Optical image of the operation of the microfluidic chip with inputs A = 1 and B = 1. Scale bar: 1 cm. Figure 3 Half adder digital microfluidic circuit fabrication and design optimization. a) 3D half adder microfluidic circuit with four levels of PMMA sheets and two PI layers for unidirectional fluid flow. b) Half adder circuit design showing a color map where level 1 corresponds to the top PMMA sheet while level 4 corresponds to the bottom PMMA sheet. c) Optical image of the operation of the microfluidic half adder circuit with inputs A = 0 and B = 0. Scale bar: 1 cm. d) Optical image of the operation of the microfluidic half adder circuit with inputs A = 0 and B = 1. Scale bar: 1 cm. e) Optical image of the operation of the microfluidic half adder circuit with inputs A = 1 and B = 0. Scale bar: 1 cm. f) Optical image of the operation of the microfluidic half adder circuit with inputs A = 1 and B = 1. Scale bar: 1 cm. To study the effect of scaling down the microchannels on the performance of the logic gates, different recipes were optimized to obtain either different depths, lengths, or widths while fixing the other dimensions as shown in Figure 4 a. For the depth study, the length and width of the microchannels are fixed at 38 mm and ≈300 µm, respectively. For the length study, the depth and width are fixed at ≈470 and ≈300 µm, respectively, while for the width study, the length and depth are fixed at 38 mm and ≈250 µm, respectively. The pressure drop between an input and an output in the chip is Δ P = R H × Q, where R H is the hydraulic resistance and Q is the flow rate. The hydraulic resistance is related to the dimensions of the channel and fluid according to R H = C geometrical × μ × L W × D 3, where L, W, and D are the length, width and depth of the microchannel, respectively, µ is the viscosity of the fluid and C geometrical is a geometrical factor that depends on the shape of the channel and its roughness. 21 Therefore, since R H is not simple to calculate especially with several variables that are not fixed for the whole channel such as the roughness and geometrical factor, the effect of the flow rate on the device performance is studied. In fact, to insert the fluids into inputs A and B, a syringe pump is used which enables the user to set the flow rate. Figure 4 Operation of the microfluidic chip with different dimensions and fluids. a) CO 2 laser recipes used for the fabrication of the different microchannels. For the depth study, the length and width are fixed at 38 mm and ≈300 µm, respectively. For the length study, the depth and width are fixed at ≈470 and ≈300 µm, respectively. For the width study, the length and depth are fixed at 38 mm and ≈250 µm, respectively. b) Operation regions of the microfluidic chip with different dimensions and DI water. c) Operation regions of the microfluidic chip with 20% glycerol and 80% DI water. Response time of the OR Boolean function for different d) lengths, e) widths, f) depths of the microchannels. The flow rate is fixed at 300 µL min −1 in the experiments. Two fluids are tested: deionized (DI) water and 20:80 glycerol:water which results in a higher viscosity. 22 It is observed that the logic gates are operational for a specific range of flow rates (Figure 4 b, c). Using flow rates beyond the upper limit, one fluidic input ( A ′. B = 1 or A. B ′ = 1) results in fluidic output in both AND and OR gates due to the high resulting fluid pressure. While using a flow rate below the lower limit results in no output from the AND gate when both inputs are turned on due to the very low pressure drop across the channel. In particular, the upper and lower flow rate limits are shown to reduce at higher lengths, smaller widths, and smaller depths. In fact, since the hydraulic resistance is directly proportional to the length and inversely proportional to the width and to the cubic depth, this will cause the increase of the hydraulic resistance and therefore increase the pressure drop across the channel. As a result, with one fluidic input, both AND and OR outputs are turned on. Therefore, as the length (depth, width) is increased (reduced, reduced), the hydraulic resistance is increased which require the reduction of the flow rate to maintain the same operational pressure drop (Figure 4 b). Moreover, using a more viscous fluid (glycerol:water), the upper and lower limits of the operational flow rate are reduced compared to the water case (Figure 4 c) due to the fact that the hydraulic resistance increases with the viscosity. The surface roughness of the microfluidic channel plays an important role as well in determining the operational flow rate. To explain this effect, two devices with different dimensions are considered: Device 1 is chosen from the width study in Figure 4 b, using W = 313 µm (while D and L are fixed at 250 µm and 38 mm, respectively), and Device 2 is chosen from the depth study, using D = 200 µm (while W and L are fixed at 300 µm and 38 mm, respectively). It can be concluded based on the dimensions of the microchannels in the two devices that Device 1 has a smaller R H (due to the larger W and D ), however, the upper and lower limits of the operational flow rates are smaller for Device 1 which is contradictory to the above explained reasoning. However, based on Figure 4 a, it can be seen that the used recipe to create the microchannel in Device 1 shows a higher power and higher speed ( P = 20%, S = 12%) than the recipe used for Device 2 speed ( P = 5%, S = 4%). As a result, the surface roughness in Device 1 is higher which overcompensates the effect of the larger D and W and causes an overall increase in hydraulic resistance compared to the case of Device 2. As a result, the operational flow rate in Device 1 is lower than in Device 2. When comparing the slope of the operational flow rate for the devices with different dimensions (Figure 4 b, c), it can be observed that the depth study shows the smallest slope, the reason is the dependence of the R H on 1/D 3. However, it is also observed that the width study shows a reduction in operational flow rate by a larger factor than the depth study. Since the hydraulic resistance is inversely proportional to the width and directly proportional to the length, it is expected that the operational flow rate of both studies should change by around the same factor when a different fluid is inserted. However, our devices are based on the laser ablation of channels followed by the 3D stacking of PMMA sheets, as a result, when the width of the microchannel is increased, the fluid will be in contact with a larger surface area of polished PMMA (top of the channel). Therefore, in this case, as the width is increased, multiple opposing mechanisms that affect the hydraulic resistance compete including 1) increase in polished surface area in contact with the fluid (top of the channel), 2) increase in amount of fluid that is not in direct contact with the sidewalls of the channel (center of the channel), and 3) increase in surface roughness of the sidewalls (due to the higher laser power and speed as shown in Figure 4 a). As a result, the overall hydraulic resistance experienced by the fluid is increased by a smaller factor when the width is reduced than in the case when the length is increased. Finally, the response time for the fluidic OR gate is studied for the devices with different dimensions and different fluids using a fixed flow rate of 300 µL min −1 (Figure 4 d–f). The results show that the response time increases linearly with length (Figure 4 d) but not with width and depth (Figure 4 e, f). This can be explained by the fact that the hydraulic resistance in the latter cases is affected by the change in the surface roughness as well due to the different optimized recipes for the width and depth studies (Figure 4 a) in addition to the R H variation with 1/D 3. In conclusion, a 3D multilevel microfluidic chip with different Boolean logic gates and a digital microfluidic circuit (AND, OR, XOR, and half adder) is demonstrated. The microfluidic logic gates operation is based on the pressure drop between the inputs and outputs of the device, which is a function of both the hydraulic resistance (fixed for a given device) and the flow rate set using a syringe pump. The presence of the fluid at the output represents a logic signal 1 while its absence is a logic signal 0. As a result, the demonstrated microfluidic logic devices can be easily used with any fluid and cascaded to achieve integrated and complex computations. The results show that the logic gates are operational for a specific range of flow rates which is dependent on the microchannel dimensions, surface roughness, and fluid viscosity. Finally, the response time of the logic gates is studied for devices with different dimensions, the results confirm its dependency on the channel hydraulic resistance. Experimental Section Fabrication of PMMA Sheets with Microchannels : A CO 2 laser tool (Universal Laser Systems) with maximum power of 75 W and speed of 223 mm s −1 was used to cut the PMMA sheets and create the microchannels. A power of 10% corresponds to 7. 5 W while a speed of 10% corresponds to 22. 3 mm s −1. PMMA sheets with 2 mm thickness were used in addition to a flexible 120‐µm‐thick PI sheet for blocking the backflow of fluids. The device consists of four PMMA sheets and an embedded PI sheet. The PI sheet was patterned using the CO 2 laser to obtain a cantilever‐like structure to enable the one‐directional flow of fluid. All the layers were aligned using metallic pins inserted at the edges of the device. Food coloring was used to color the water at the inputs of the microfluidic logic gates. Bonding of the PMMA Sheets : A thermocompression tool was used to bond the several layers of the device as shown in Figure S2 in the Supporting Information. The bonding procedure was the following: first, the PMMA sheets were aligned using metallic pins, next the microfluidic device was placed between two silicon wafers to avoid direct contact between the hot plates and the PMMA sheets. The complete sandwich of silicon wafers and PMMA sheets was then placed in between the hotplates in the thermocompression tool. The temperature of the top and bottom plates was set to 120 °C and the spacing between the plates was narrowed down to the exact thickness of the sandwich (no applied pressure) to provide heat transfer by conduction and avoid trapping air within the device which would otherwise result in air bubbles between the PMMA sheets. Once the temperature of the system reaches 120 °C (higher than the glass transition temperature of PMMA), a pressure of 20–40 lbs was applied between the hotplates to compress and bond the layers. The heaters were then turned off and the cooling valves were opened to cool down the device. The pressure was kept constant until the temperature reached a value of 90 °C. The bonding process is shown in Figure S2 in the Supporting Information. Characterization : The microfluidic devices were tested using a Harvard syringe pump. Two syringes with 60 mL capacity were installed. Deionized water and 20:80 Glycerol:DI were used during the experiments. In the apparatus, the volumetric flow set point (flow rate) was changed to study its effect on the operation of the logic gates. Conflict of Interest The authors declare no conflict of interest. Author Contributions N. E. ‐A. conceptualized the idea. M. M. H. directed the study. N. E. ‐A. designed the system, characterized the devices, and analyzed the data, J. C. C. assisted in the fabrication and characterization of the microfluidic logic gates, and all authors contributed to writing the manuscript. Supporting information Supporting Information Click here for additional data file. Supplemental Video 1 Click here for additional data file. Supplemental Video 2 Click here for additional data file. Supplemental Video 3 Click here for additional data file. Supplemental Video 4 Click here for additional data file. Supplemental Video 5 Click here for additional data file. Supplemental Video 6 Click here for additional data file.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.