A Study of Physical and Covalent Hydrogels Containing pH-Responsive Microgel Particles and Graphene Oxide

: In this study we mixed low concentrations of graphene oxide (GO) with microgel (MG) particles and formed composite doubly cross-linked microgels (DX MG/ GO) gels. The MG particles comprised poly(ethyl acrylate- co methacrylic acid- co -1,4-butanediol diacrylate) with pendant glycidyl methacrylate units. The MG/GO mixed dispersions formed physical gels of singly cross-linked MGs (termed SX MG/GO), which were subsequently heated to produce DX MG/GO gels by free-radical reaction. The in ﬂ uence of the GO concentration on the mechanical properties of the SX MG/GO and DX MG/GO gels was investigated using dynamic rheology and static compression measurements. The SX MG/GO physical gels were injectable and moldable. The moduli for the DX MG/GO gels increased by a factor of 4 − 6 when only ca. 1.0 wt % of GO was included. The isostrain model was used to describe the variation of modulus with DX MG/GO composition. Inclusion of GO dramatically altered the stress dissipation and yielding mechanisms for the gels. GO acted as a high surface area, high modulus ﬁ ller and played an increasing role in load distribution as the GO concentration increased. It is proposed that MG domains were dispersed within a percolated GO network. Comparison of the modulus data with those published for GO-free DX MGs showed that inclusion of GO provided an unprecedented rate of modulus increase with network volume fraction for this family of colloid gels. Furthermore, the DX MG/GO gels were biocompatible and the results imply that there may be future applications of these new systems as injectable load supporting gels for soft tissue repair. diameter sterile glass coverslips. After 24 h in culture, sterile toroid-shaped composites (see inset of Figure 7a) were introduced into wells and cultured for a further 10 days. After this time the coverslips were removed and cell viability assessed by live/dead assay (Invitrogen, U.K.). Images were taken with an Olympus BX51 ﬂ uorescence microscope and Leitz Diavert inverted light microscope.


■ INTRODUCTION
The properties of hydrogels continue to evolve along with their structural complexity. 1 Remarkable improvements in gel modulus, 2 ductility, 3 swelling ratios, 4 and toughness 5 have been achieved. While the majority of gels investigated have been constructed from small molecules, a new type of gel has recently emerged that is constructed solely from interlinked microgel (MG) particles. 1,6 MGs are cross-linked polymer particles that swell in a good solvent 7,8 or when the pH approaches the pK a of the polyacid chains comprising the particles. 8−10 Liu et al. 6 showed that concentrated dispersions of pH-responsive MG particles containing pendant vinyl groups could be interlinked to form hydrogels, which are termed doubly cross-linked microgels (DX MGs). These hydrogels were constructed using sub-micrometer sized MG particles as the building blocks. They differ from other hydrogels containing MGs 11,12 because MG particles act as both crosslinker and matrix element. If the MG particles were removed from our DX MGs, the gel would no longer exist, in contrast to hydrogels containing MGs. In this study we investigate DX MGs prepared in the presence of low concentrations of graphene oxide (GO). GO, which is a high aspect ratio waterdispersible colloid, was used because it offered good potential for mechanical property enhancement. In this work we investigate and compare physical and covalent gels containing MG particles and GO. The aim of this study was to address two fundamentally important questions concerning the mechanical properties of the physical and covalent MG/GO composite gels. First, what is the stress transfer mechanism? Second, what is the strain dissipation mechanism?
The DX MGs studied here were prepared using pHresponsive poly(ethyl acrylate-co-methacrylic acid-co-1,4-butanediol diacrylate) (poly(EA/MAA/BDDA)) MG particles functionalized with glycidyl methacrylate (GMA). These MG particles are abbreviated as MG-GMA. The MG particles swell from their collapsed, as-made, latex form with pH increase (Scheme 1), and this process transforms a low viscosity fluid into a shear-thinning physical gel. The MG-GMA particles were then subjected to free-radical coupling via pendant vinyl groups (from GMA). 6 The term DX is used to identify that two types of cross-linking were present. Besides the covalent intraparticle cross-linking (within singly cross-linked (SX) MGs), we introduced a second (doubly) covalent interparticle crosslinking that interlinked the MG particles. The latter transformed the shear-thinning (injectable) gel into an elastic, permanently cross-linked, hydrogel of interlinked MG particles. In Scheme 1 the interlinking is identified with red interfaces. DX MGs have potential application as injectable gels for the restoration of the mechanical properties of degenerated intervertebral discs (IVDs). 13 Many of the design rules governing the mechanical properties of DX MG gels have become clear since their first report in 2011. 6 The modulus of DX MG gels has contributions from both intraparticle and interparticle cross-links. 6 Because the precursor is a physical gel of swollen MG particles, close contact between vinyl groups on neighboring MG particles is essential for interparticle cross-linking and DX MG formation. Highly cross-linked MG particles show low pH-triggered swelling 14 which can lead to the paradoxical situation where high modulus MG particles lead to low modulus DX MGs because of poor inter-MG cross-linking. The modulus of DX MGs increases with MG particle concentration, 6 which provides a simple mechanical property tuning method. At fixed MG particle concentration the modulus can also be increased by increasing the extent of pendant vinyl functionalization of the MG particles. 6,15 Here, we investigated the effect of GO concentration on the mechanical properties of DX MG/GO composites at constant MG particle concentration and extent of vinyl functionalization.
Graphene has outstanding mechanical, electrical, and thermal properties. 16−20 GO is rich with hydrophilic groups on the surface and at the edge of the sheets. 21 Moreover, GO is readily water-dispersible. 16 The hydrophilic groups comprise hydroxyl, carboxyl, epoxy, and ketone groups, and the Lerf−Klinowski model is currently the most widely accepted structural model for GO. 18 Furthermore, GO has received a great deal of attention because of its high aspect ratio, high modulus, 22 and ability to be blended with polymers and incorporated in gels. 23−25 The biological applications of GO have also been actively investigated, 26−28 and there is good potential for GO use in vivo. 29 Here, we investigate colloidal gel composites containing GO as a mechanical reinforcement colloid. Although the major potential application of our DX MG/GO gels is for repairing load supporting tissues, we do not focus on this application in this study. Rather, we concentrate on establishing structure−mechanical property relationships for the new gels studied.
The starting point for composite gel preparation (Scheme 1) was a mixed dispersion of vinyl-functionalized MG particles in latex form and GO. A pH increase caused pH-triggered swelling of the MG-GMA particles and formation of a physical gel, termed SX MG/GO x , where x is the total concentration of GO added in wt %. As will be shown, the physical gels were moldable and injectable. When heated in the presence of initiator (ammonium persulfate, APS), the MG-GMA particles covalently interlinked and DX MG/GO x gels were formed. Our DX MG/GOs differ considerably from other reported hydrogel/GO composites 30−36 because we used preformed MG particles to form the hydrogel. These are macroscopic composite gels constructed from colloidal scale building blocks. Uniquely, our approach does not use small monomers.
In this study the morphology and properties of SX MG/GO and DX MG/GO composite gels are reported for the first time. The gels are explored using dynamic rheology and compression measurements, and their morphology is assessed using SEM. The modulus is found to obey the isostrain biphasic model of Scheme 1. Depiction of the Preparation of DX MG/GO Gels a a The MG particles are prepared in their latex form and undergo pH-triggered swelling at pH values greater than their pK a (inset). The latex was mixed with GO, and the pH increased to form a SX MG/GO physical gel. Heating the physical gel in the presence of ammonium persulfate (APS) covalently interlinked neighboring MG particle interfaces (shown in red) to form a DX MG/GO gel. The MG phase is proposed to exist as domains within percolated, exfoliated, GO sheets. Morris, 37 which is the first report of this behavior for hydrogels containing GO to our knowledge. Further, our data point to a structural model in which the MG particles exist as domains within a percolated GO network. While the GO is not believed to be covalently linked to the DX MG phase, there is very good stress transfer via the GO sheets. The addition of 1 wt % of GO is shown to increase the modulus by a factor of 6 and dramatically change strain dissipation. Moreover, the inclusion of GO is shown to dramatically increase the modulus compared to earlier reports for pure DX MGs. The very good modulus tunability afforded by these new biocompatible injectable composite gels offers potential for future application in the context of repairing load supporting soft tissue.
Synthesis of Graphene Oxide. Preparation of GO was conducted using a modified Hummers method. 38 Graphite powder (5 g) was introduced in concentrated H 2 SO 4 (170 mL) with stirring. An ice bath was used to cool the mixture. NaNO 3 (3.75 g, 0.04 mol) was added and stirring continued for 5 h. Potassium permanganate (25 g, 0.16 mol) was gradually added at a rate that prevented the temperature of the mixture exceeding 20°C. After all of the additives were dissolved, the ice bath was removed and the temperature was increased to 35°C using heating with stirring for 2 h. The dispersion was then stirred at room temperature for 7 days. To obtain exfoliated GO, the product was slowly dispersed into 550 mL of 5 wt % H 2 SO 4 solution and stirred for 3 h. Then, aqueous H 2 O 2 (15 g, 30 vol %) was slowly added over a period of 5 min. The mixture became a gold colored dispersion, and stirring was continued for a further 2 h. The suspension was diluted with 500 mL of 3 wt % H 2 SO 4 solution containing 0.5 wt % H 2 O 2 and left overnight. The product was centrifuged at 9000 rpm for 20 min, and the clear supernatant liquid was removed. The remaining viscous liquid was washed with a further 500 mL of 3 wt % H 2 SO 4 /0.5 wt % H 2 O 2 solution. The washing process was repeated nine times, and then the mixture was further purified by dialysis for 7 days. The final pH of the purified, exfoliated, GO dispersion was 7.2.
Synthesis of Poly(EA/MAA/BDDA) Microgel. The synthesis of this MG was conducted using the seed−feed emulsion polymerization method according to a previously published method 39 and is briefly described here. The MG particles contained about 33 wt % of MAA with respect to total monomer. A comonomer solution (250 g) containing EA (167 g, 1.88 mol), MAA (83 g, 0.83 mol), and BDDA (2.5 g, 0.01 mol) was prepared. A total of 31.5 g of the comonomer solution was used for the seed formation. The latter solution was added to water (518 g) containing SDS (1.8 g, 0.006 mol), and then K 2 HPO 4 (3.15 g of a 7 wt % aqueous solution) and APS (10 g of a 2 wt % solution) were added. The seed was prepared at 80°C with mechanical stirring under nitrogen for 30 min. Then the remaining monomer solution was added at a uniform rate over a period of 1.5 h. The reaction was continued for further 2 h. The product was purified by extensive dialysis using water.
Synthesis of GMA-Functionalized Poly(EA/MAA/BDDA) Microgel. MG dispersion (100 g of 10 wt % dispersion) was mixed with GMA (16.6 g, 0.12 mol). The pH of the mixture was adjusted to 5.4 using aqueous NaOH solution (0.05 M). The mixture was heated at 50°C for 8 h. After reaction, chloroform was used to wash the product and then removed using rotary evaporation and the process repeated. The MG-GMA dispersion was then dialyzed extensively.
Preparation of SX MG/GO Physical Gels and DX MG/GO Covalent Gels. To prepare the composite gels a thoroughly mixed dispersion containing MG-GMA and GO was first prepared at a pH of about 4.7. As an example, the DX MG/GO 1.0 composite was prepared from mixing MG-GMA (6.67 g of 15 wt % dispersion) and GO (22.22 g of 0.45 wt % dispersion). (For the other DX MG/GO x systems the masses of GO were reduced accordingly.) The mixed dispersion was concentrated to ∼10 g using rotary evaporation at room temperature. An aqueous solution (0.2 g) of APS and NaOH was then added with stirring. The added solution contained 3.4 wt % of APS, and the concentration of NaOH was 4 M. The gel had a final pH of ∼7.2. This process triggered formation of an SX MG/GO physical gel. To form the DX MG/GO gel the SX MG/GO mixture was heated at 50°C for 8 h. (For the two demonstration gel preparations where TEMED was present, the covalent interlinking reaction was conducted at 37°C using an overall TEMED concentration of 0.08 M.) All the gels studied in this work contained a MG particle concentration of 10 wt % and the pH values were adjusted to ∼7.2. The SX MG/GO physical gels were studied in the absence of added APS.
Physical Measurements. Titration measurements were performed using a Mettler Toledo DL 15 titrator in the presence of aqueous 0.1 M NaCl. TGA measurements were performed using a TA Instruments Q500 with a heating rate of 10°C/min. Photon correlation spectroscopy (PCS) measurements were performed using a BI-9000 Brookhaven light scattering apparatus (Brookhaven Instrument Cooperation) equipped with a 20 mW HeNe laser, and the detector was set at a scattering angle of 90 o . TEM data were obtained using a JEOL JEM-2011F operating at an accelerating voltage of 200 kV. Zeta potential data were obtained using a Malvern Nano ZS using a GO concentration of 0.01 wt % and a background electrolyte concentration (NaNO 3 ) of 0.01 M. The rheological properties of the gels were measured using a TA Instruments AR-G2 temperaturecontrolled rheometer equipped with an environmental chamber. The measurement geometry used a 20 mm flat disk. Frequency-sweep and strain-sweep data were measured at 25°C. For the frequency-sweep data a strain of 1% was used. The strain-sweep data were measured using a frequency of 1 Hz. The compression tests were conducted using an Instron series 5569 load frame equipped with a 100 N load cell. The compression rate was 2 mm/min, and the average height and diameter of the gel cylinders were 10.8 mm and 9.6 mm, respectively. Modulus values were calculated from the initial gradient of stress− strain curves at strains (ε) of less than 10%. Data are also shown in terms of the extension ratio λ (= 1 − ε). SEM measurements were obtained using a Philips xl30 instrument. Before observation, gel samples were rapidly freeze-dried using liquid nitrogen. The SEM samples were coated with platinum prior to examination.
Assessment of Cytotoxicity. Human nucleus pulposus (NP) cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotic/antimycotic at 37°C in a humidified 5% CO 2 atmosphere. Cells were harvested by trypsinization and seeded at a density of 2 × 10 4 cells/well into 24-well culture plates containing 13 mm diameter sterile glass coverslips. After 24 h in culture, sterile toroid-shaped composites (see inset of Figure  7a) were introduced into wells and cultured for a further 10 days. After this time the coverslips were removed and cell viability assessed by live/dead assay (Invitrogen, U.K.). Images were taken with an Olympus BX51 fluorescence microscope and Leitz Diavert inverted light microscope.

■ RESULTS AND DISCUSSION
Microgel and Graphene Oxide Characterization. The GMA-functionalized poly(EA/MAA/BDDA) MG (abbreviated as MG-GMA) was prepared by seed−feed emulsion polymerization and subsequently vinyl-functionalized by reaction with GMA (see Experimental Details). The extent of GMA functionalization and the apparent pK a were determined to be 3.0 mol % and 6.6, respectively, from titration data (see Figure S1 and Table S1  Supporting Information) with a number-average diameter (d n(SEM) ) of 108 nm (coefficient of variation = 16%). The MG-GMA particles were pH-responsive ( Figure S2b in the Supporting Information) and had a hydrodynamic diameter (d h ) of 120 nm at pH = 4. We assume that the particles at this pH were fully collapsed because the latter diameter was close to d n(SEM) . With increasing pH the MG-GMA particles exhibited pH-dependent swelling with a d h value of ∼307 nm at pH = 7.2 (d h (7.2) ). At higher pH values the value for d h was constant. A MG-GMA particle volume swelling ratio (Q = (d h(7.2)/ d h(4.0) ) 3 ) of 16.7 was calculated from the data. The MG-GMA particles swelled considerably in the vicinity of pH = 7.4. The latter pH is a common target for biomaterial applications.
The GO used in this study was prepared by a modified Hummers method (see Experimental Details). A representative TEM image for GO is shown in Figure S2c in the Supporting Information, and a folded sheet is evident. A representative SEM image ( Figure S2d in the Supporting Information) indicated an average sheet diameter of about 2 μm, which agrees with the approximate diameter of the equivalent disk. The Raman spectrum for GO differs considerably from that of graphite ( Figure S3 in the Supporting Information) and is typical of that reported for GO 40 with the G band at ∼1600 cm −1 and the D band at ∼1350 cm −1 . Both of the bands had similar intensities. The zeta potential measured for GO at pH = 7.2 was −63.5 mV. The combination of these properties is consistent with those reported for GO in the literature. 20,41,42 Probing Stress Transfer and Dissipation of Gels Containing Graphene Oxide. DX MG/GO gels were prepared by mixing MG-GMA and GO dispersions, increasing the pH, and then heating (see Scheme 1). At pH = 4.3, which is less than the pK a for the MG-GMA particles (Table S1 in the Supporting Information), the SX MG/GO dispersions were free-flowing (see Figure 1a). A fluid-to-gel transition for the mixed SX MG/GO dispersion was triggered by increasing the pH to 7.2 ( Figure 1a). In this state the swollen MG particles occupied most of the volume and prevented translation of nearest-neighbor MG particles (and GO) due to excluded volume effects. The SX MG/GO dispersions formed gels at pH of 7.2 that were shear-thinning and could be injected through a narrow gauge syringe needle (Figure 1b). Injectability is a potentially useful property for a load-bearing gel in the context of soft tissue repair. 13,43 While colloidal gels have previously been reported that formed moldable macroscopic gels due to heteroaggregation, 44 the present gels did not require hetero-aggregation, but simply a pH change and the presence of a minor proportion of GO. It can be seen from Figure 1b that these physical gels could be molded into self-supporting shapes. The ability to form SX MG/GO gels into self-supporting shapes and cure them using covalent interlinking provides the first example of "shape and set" GO-based colloidal composite gels to our knowledge. We note that pure GO dispersions are also known to form gels 45 and images of tubes containing different GO concentrations (C GO ) for this study are shown in Figure S4 in the Supporting Information. We show below that inclusion of MG particles increased gel modulus.
The frequency-sweep dynamic rheology data for GO, SX MG/GO, and DX MG-GO gels are shown in Figure 2a−c. The GO data are included as a control. The values for the storage modulus (G′) are greater than those for the loss modulus (G″) over the whole frequency range studied for all systems, showing that each was a viscoelastic solid. The frequency dependencies for G′ and G″ are generally low for each gel. The G′ values scale with frequency (=ω/2π, ω is the oscillation frequency) according to G′ ∼ ω n ′, 46 where n′ is the frequency exponent. The value for n′ approaches zero as the relaxation time becomes longer than the experimental time and the rigidity increases. Figure 2d shows values for n′ plotted as a function of C GO . The value for n′ decreases in the order GO > SX MG/GO > DX MG/GO. It follows that the relaxation times increase in the same order. The presence of MG particles (for SX MG/ GO) and their covalent interlinking (for DX MG/GO) decreased the ability of the stress distributing elements to become disconnected within the experimental times probed by these oscillatory rheology measurements. Accordingly, we propose that the covalent interlinking of the MG particles (for DX MG/GO) restricted the freedom of the GO sheets to move in response to strain.
The value for G′ can be considered as a measure of network connectivity. According to Figure 2e the connectivity of the networks increased with C GO for all of the gels. (It will be shown in Figure 3 that the increase of G′ is linear with respect to C GO .) In the case of SX MG/GO, the increased connectivity arises from the swollen MG particles, which form a space-filling physical gel, and stress transfer occurs between the MG particles and the GO sheets. The fact that the G′ values for the DX MG/GO gels are higher than those for the respective SX MG/GO shows that the double cross-linking contributed additional elastically effective chains. It is proposed that the covalent interlinking of the MG particles more effectively trapped the GO sheets in their positions within the DX MG/ GO gels.
The values for G″ and G′ provide measures of the energy dissipated and stored, respectively, per unit strain. 47 Hence, as the value for tan δ (=G″/G′) increases, the proportion of energy dissipated by the gel under strain also increases. This is greatest for the GO gels (Figure 2f). It follows that the GO sheets were least constrained for the GO gels and they were not strongly linked together. This suggestion is consistent with the literature for GO where the principal attractive interactions are hydrophobic attraction and hydrogen bonding, 48,49 both of which are relatively weak (and reversible) compared to covalent bonds.
Compared to the GO gels, inclusion of non-interlinked MG particles (SX MG/GO) greatly decreased tan δ (Figure 2f), and this was further decreased by MG interlinking (for DX MG/ GO). The physical and covalently interlinked MG particle matrixes were both effective in decreasing dissipation and are proposed to have greatly restricted the ability of the GO sheets to move in response to strain. It is noteworthy that tan δ was much lower for the (GO-free, C GO = 0) DX MG gel (Figure 2f) compared to the DX MG/GO gels. Accordingly, the exfoliated GO sheets caused a greater degree of movement within the DX MG/GO gels (and hence dissipation) in response to strain. This trend is opposite to that expected if the GO sheets were covalently linked to the network because in that case less dissipation would be expected. Free-radical cross-linking of GO has been proposed for other GO hydrogel composites. 35,36 In the present case our data suggest that the GO sheets were not covalently linked to the interlinked DX MG particle matrix. While we cannot be certain that covalent cross-linking between the DX MG and GO phases was completely absent, substantial interphase cross-linking would be expected to result in a more rigid network compared to the parent DX MG gel, with similar or lower tan δ values. The tan δ increase with increasing C GO (Figure 2f) is further support for the absence of significant interphase cross-linking.
The DX MG/GO gels contained covalently interlinked MG particles as evidenced by the fact that they did not redisperse when placed in water for extended periods. The GO phase can be viewed as a reinforcing, high modulus filler. In this context the percolation concentration for GO is important. Because the rheology data showed that the GO dispersions were gels (Figure 2a), it is reasonable to conclude that the percolation concentration for GO was less than 0.25 wt %. This conclusion is in agreement with the study of Vasu et al. 45 Consequently, we tested a model for the DX MG/GO composites consisting of a relatively low modulus (MG) phase dispersed within a high modulus phase (GO). Under these conditions the biphasic isostrain model of Morris 37 should apply. The isostrain model assumes thar the modulus of the composite is the volume fraction weighted sum of the modulus values for each component. This model was tested using the following equations.
For the above equations G′ GO and G′ DX MG are the storage moduli for GO and DX MG, respectively. We used Φ GO and Φ DX MG for the volume fractions of GO and DX MG gel networks on a dry weight basis as we could not be certain of the actual volume fractions occupied by each phase in the gels. Equation 2 was obtained using the assumption Φ GO + Φ DX MG = 1. We further assumed that the density of the dry MG particles 6 and GO 22 was 1.2 and 1.8 g cm −3 , respectively. Figure  3 shows that the G′ vs Φ GO data were linear and could be fitted using eq 2 and a G′ DX MG value of 36.3 kPa (from Figure 2e). The data imply that there is good potential to dial up the modulus of these composite GO-based gels using composite formulation. Using the gradient and eq 2, a value of G′ GO of 14.6 MPa was calculated. Dikin et al. 22 reported a tensile modulus of 32 GPa for "GO paper", which is four orders of  magnitude higher than G′ GO . The reason why G′ GO is much lower than the modulus value from Dikin et al. is that the GO sheets were dispersed within a gel and had some freedom to move under strain. The interactions that transferred stress between the GO sheets and DX MG domains most likely involved water-mediated hydrogen bonding and also reversible hydrophobic interactions. It is not surprising that these interactions, which are reversible and dissipative, are less efficient for load transfer compared to those present for GO paper. 22 Strain-sweep dynamic rheology measurements are well suited to probe strain-induced failure mechanisms for gels. The GO gels (Figure 4a) had low ductility in that G′ decreased at strains greater than about 0.2%. This tendency for G′ to decrease at low strain for GO gels was reported by Vasu et al. 45 The GO gels showed only weak G″ maxima, and both the G′ and G″ data decreased strongly with increasing strain. This behavior contrasts to SX MG/GO (Figure 4b) and DX MG/GO gels (Figure 4c) where the G′ values remained constant until much higher strain values were reached and strong maxima for G″ are evident. The maxima for the MG-containing gels are attributed to cage breaking 50 and are largest for the DX MG/GO gels. This trend can be seen more clearly in Figure 4d, where ΔG″ n (= (G″ − G″ min )/G″ min ) is plotted as a function of C GO . (G″ min is the minimum G″ value in the linear viscoelastic region, i.e., at the lowest strain.) The use of ΔG″ n normalizes the G″ maxima 51 and show that inclusion of GO progressively decreased the relative magnitude of the G″ maximum. The extent of cage breaking during dissipation is proposed to be diminished by GO inclusion. This conclusion would follow from a percolated structure where MG domains are increasingly separated by GO sheets (see Scheme 1) and the behavior occurred when the GO concentration was as low as 0.25 wt %.
Applying the general approach used elsewhere for brittle viscoelastic solids, 52 we define a yield strain (γ*) corresponding to the γ value where the initial modulus had decreased significantly. In the present study a 10% modulus decrease was used. The value of γ* corresponds to a point where 90% of the network connectivity was intact. (Higher static yield strains were measured using compression measurements; Figure 6.) The data from Figure 4e show that the GO gels were brittle, which is due to the percolated network of stiff GO sheets. Inclusion of GO within the SX MG/GO gels increased γ* substantially compared to the GO gels and therefore improved the ductility of the GO network. We suggest that the MG particles separated GO sheets and opposed breaking of network connections under strain due to their swollen state, which restricted excessive GO sheet movement.
For the DX MG/GO gels covalent interlinking of the MG particles decreased the γ* values and hence the ductility ( Figure  4e) compared to the SX MG/GO gels. This is expected due to a decreased ability of the MG particles to move under an imposed strain due to covalent inter-MG linking. Overall, the γ* data reveal that inclusion of GO dramatically changed the dissipation and yielding behavior of the composite gels compared to the GO-free (C GO = 0) SX MG and DX MG gels. Generally, the SX MG/GO and DX MG/OGO became more like the GO gels in terms of stress dissipation and yielding behaviors. These data are congruent with our proposal that the gels consist of a percolated GO network containing MG particle domains.
DX MG/GO Morphology, Static Mechanical Properties, and Biocompatibility. In this part of the study we focus on the properties of DX MG/GO gels because they have the most promising properties for future biomaterial application, e.g., as load supporting injectable gels for IVD repair. 13 A detailed study of the potential for application of the SX MG/ GO and DX MG/GO gels will be presented in future work. Here, we focus on the fundamental aspects of structure− property relationships and also present preliminary biocompatibility data. Because the gels investigated here were prepared at 50°C (which is too high for use in vivo), they could only have potential future use in humans after being injected into molds. However, we note that introduction of an accelerator enabled the DX MG interlinking temperature at 37°C, 13 and this approach is expected to be suitable for our DX MG/GO gels. To test this proposal two demonstrator DX MG/GO gels were prepared at 37°C in the presence of TEMED and the mechanical properties probed by dynamic rheology. The DX MG/GO 0.5 and DX MG/GO 1.0 gels cured within about 20 min, and the strain-sweep data were similar to those for the respective gels prepared at 50°C (see Figure S5 in the Supporting Information). The remainder of the study focuses on gels prepared at 50°C in the absence of added TEMED.
The morphology of the freeze-dried DX MG-GO gels was probed by SEM. Figure 5a shows a freeze-dried image for a DX MG/GO 1.0 in the as-made state. The morphology was porous and typical of that reported for conventional hydrogel/GO composites. 31,36,53 For our DX MG/GO gels, spherical MG particles can be seen from the inset of Figure 5a (arrow). However, the GO sheets were less clear. In order to reveal the GO sheets a freeze-dried DX MG/GO 1.0 hydrogel was heated at 550°C under N 2 for 1 h. This process was conducted because GO has superior thermal stability to conventional polymers. 54 The heating process resulted in the freeze-dried mass content decreasing by 90% (see TGA data in Figure S6 in the Supporting Information). SEM images (Figure 5b) show that the GO sheets became exposed. Not all of the GO sheets were exposed because significant, degraded, DX MG remained. Nevertheless, the exposed GO sheets can be seen to have a space-filling morphology. This morphology generally supports our proposal that GO formed a percolated network throughout the DX MG/GO gels. Because the DX MG/GO gels were prepared by interlinking SX MG/GO gels (without further mixing), it is reasonable to infer that the morphology of the SX MG/GO gels was similar to that of the DX MG/GO gels ( Figure 5).
Static compression stress vs extension ratio data were measured for the DX MG/GO gels (Figure 6a). The inclusion of GO decreased the maximum strain (ε max ) (Figure 6b) when the data for the DX MG/GO gels are compared to that for DX MG (C GO = 0). For the DX MG/GO gels, the ε max values (and hence ductility) were not significantly dependent on C GO despite the major increase in the maximum true stress values that occurred (σ max(T) , Figure 6b). The fact that the ε max values for DX MG/GO were not distinguishable and were all significantly lower than for the GO-free DX MG (Figure 6b) confirms the trend for the γ* values that was apparent from strain-sweep rheology (Figure 4e). These data support our assertion that a major change in the gel architecture occurred upon inclusion of GO within DX MG/GO. The values for ε max were about 25%, and these are potentially useful in the context of an injectable gel for IVD repair because the axial regional compressive strains experienced within the IVDs of humans are much less than 10%. 55 The values for the modulus, E, were calculated from the gradients of the stress−strain curves and are shown in Figure  6c. A linear relationship between E and C GO is evident, which is the same general trend established from rheology experiments ( Figure 3). These data show that it should be straightforward to tune the modulus of the DX MG/GO gels through control of C GO , which could be useful in the context of IVD repair. 13 The maximum E value was 300 kPa, which is an impressive value for a gel containing a high water content (89 wt %). It is remarkable that E increased by a factor of 6 while the increase in network forming material (i.e., GO) accounted for only an additional 1 wt %. This major increase in the modulus achieved through addition of low GO concentration is in agreement with trends reported elsewhere for sugar-based aqueous gels, where  functionalized GO was used. 56 However, for the present gels functionalization of GO was not required.
As the DX MG/GO gels have potential future application in the context of IVD repair, we conducted a preliminary biocompatibility study. The biocompatibility of DX MG/ GO 0.5 gels was investigated by culturing human NP cells in direct contact with composites for 10 days, and cell viability was assessed by live/dead assay (Figure 7). Annular DX MG/GO 0.5 composites (inset of Figure 7a) were introduced into 24-well culture plates containing cultures of human NP cells on glass coverslips. Live cells in the assay are distinguished by the presence of calcein (green), whereas dead cells are stained with ethidium homodimer-1 (red). Figure 7a shows the adherence and morphology of NP cells in culture with the DX MG-GO 0.50 composite. The morphology was indicative of viable NP cells. Figure 7b shows live/dead assay fluorescent microscopy images of NP cells in contact with the gel, and a very high proportion of live cells were evident (greater than ca. 95% from our images). It follows that there was a high proportion of viable, adherent NP cells present after 10 days of culture in direct contact with DX MG/GO 0.50 . The DX MG/GO gel had good biocompatibility under the conditions studied. These trends match those established earlier for GO-free DX MGs. 13 Proposed Morphology for SX MG/GO and DX MG-GO Gels and Roles of Each Component. Our MG particles aided GO dispersion (Figure 1a), which is consistent with the reported amphiphilic nature of GO 16 and related MG particles. 57 The MG particles are proposed to have physically adsorbed onto the GO sheets via hydrophobic and hydrogen bonding interactions. The subsequent pH-triggered MG particle swelling increase caused physical gel formation ( Figure  1a,b); when heated, MG particle interlinking gave DX MG/GO gels (Figure 1c). The depictions for the SX MG/GO and DX MG/GO gels shown in Scheme 1 capture the key morphological features. In each case exfoliated GO sheets separate domains of MG particles which are either sterically confined (SX MG/GO) or covalently interlinked (DX MG/ GO). The GO sheets, which form a percolated network throughout the gels, occupy a much smaller volume fraction than depicted in Scheme 1. (The proportion of MG particles shown was decreased for clarity.) Because the tan δ values of the DX MG/GO gels were higher than for the parent DX MG (Figure 2f), it is proposed that cross-linking did not occur to a significant extent between the DX MG and GO phases.
Considering the mechanical properties of the DX MG/GO gels, the GO sheets acted as a high modulus rigid filler. The adsorbed MG particles transferred load between neighboring GO sheets via the DX MG domains that occupied inter-GO spaces, i.e., GO-to-DX MG-to-GO load transfer. Because of their space-filling nature, the MG particles effectively lock the GO sheets in place until a critical yield strain is reached. The high swelling pressure of the DX MG phase opposed the large scale strain-induced GO sheet alignment required for yielding of pure GO gels. 45 Hence, the role of the MG particles was to provide an effective load transfer mechanism between the GO sheets and an energetic barrier to their strain induced alignment. The role of the GO was to provide a high modulus continuous phase which increasingly distributed load as the total GO surface area increased. The latter increased with increasing C GO .
Mechanical Property Comparison of GO-Free and GO-Containing SX MG and DX MG Gels. Although a range of SX MGs and DX MGs have been reported, 6,13,58 like conventional gels, moderate increases in polymer concentration are required to substantially increase G′. Figure 8 shows a comparison of the reported G′ values for DX MGs and SX MGs for a range of network volume fractions (ϕ net ). The network comprises polymer from the MG particles for conventional SX MGs and DX MGs, or polymer and GO for SX MG/GO and DX MG/GO. Figure 8 shows that the G′ values for DX MG/GO and SX MG/GO dramatically increase for inclusion of a small increase of ϕ net compared to the other DX MG and SX MG gels. (Note that 1 wt % of GO is equivalent to a ϕ net increase of 0.05.) The rate of increase was greatest for the DX MG/GO gels (line of best fit shown). These data show that inclusion of GO is a far more effective means to obtain very large increases in modulus compared to increasing the network concentration by simply adding more MG particles. The rate of modulus increase for the DX MG/ GO gels is unprecedented for DX MG-based gels, which are a new class of gel. 1 These data also imply that very high G′ DX MG-GO gels could be obtained by adding higher C GO values than studied here and, if confirmed by experiment, such an effect may be potentially important in the context of high modulus injectable gels for soft tissue repair of cartilage. Future

■ CONCLUSION
This study has investigated the morphology and mechanical properties of colloidal composite gels comprising MG particles and GO. SEM data revealed that the GO sheets were well distributed throughout the matrix. The modulus values (G′ and E) for DX MG/GO were found to be proportional to C GO . The modulus values increased by a factor of 5−6 with inclusion of only 1 wt % of GO. The modulus data could be described using the isostrain model of Morris. 37 The modulus values for the SX MG/GO were sufficiently large that they could form moldable gels. However, the gels were shear-thinning and were shown to be injectable. The inclusion of GO changed the failure mechanism for the gels to one that resembled that of pure GO gels. Cage breaking was greatly diminished by GO. The ductility of the SX MG/GO and DX MG/GO gels was lower than that for the respective GO-free gels. This was explained in terms of the morphology changes caused by GO, and we propose that the morphologies depicted in Scheme 1 generally apply. The GO sheets were dispersed within MG particle domains (which probably did not involve covalent bonds from GO) for both the SX MG/GO and DX MG/GO gels. GO contributed increasingly to load distribution as the GO concentration increased and behaved as a high surface area, high modulus filler. The results of cell-challenge studies showed that the DX MG/GO 0.5 gel was not cytotoxic to NP cells. Comparison with data published for GO-free DX MG gels shows that inclusion of GO within DX MG/GO gels provided an unprecedented rate of modulus increase with network volume fraction for this family of colloid gels. The ability to form SX MG/GO gels into self-supporting shapes (Figure 1b) and then cure them using covalent interlinking (Figure 1c) provides the first example of "shape and set" GO-based colloidal composite gels. Considering the good mechanical performance of these novel gel composites, they can be considered as a promising candidate for injectable loadsupporting soft tissue repair therapies in the future. ■ ASSOCIATED CONTENT * S Supporting Information Titration data, SEM and PCS data for the MG particles, TEM, SEM, and Raman spectroscopy data for GO, digital photographs for GO gels, dynamic rheology data for composite gels prepared with and without TEMED, and TGA data for the DX MG/GO composite. This material is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION Notes
The authors declare the following competing financial interest(s): B. Saunders and T. Freemont are the founders of a university spin out company that aims to translate microgel technologies to healthcare applications.