Injectable scaffolds for bone regeneration

Clinical treatments of significant bone defects involve invasive procedures such as the application of auto- and allo-grafts. These procedures present many limitations including the potential for infection and rejection. There is therefore a need to develop novel therapeutic strategies able to exploit the natural regenerative potential of bone and that can be delivered in a

less invasive manner. Amongst the materials studied for the development of novel scaffolds, stimuli-responsive gels containing hydroxyapatite and carbon nanotubes as nanofillers have generated great interest. In the present work, chitosan gels containing chitosan grafted CNTs and chitosan-hydroxyapatite complex have been formed by crosslinking with glycerol phosphate.
The addition of the nanofillers afforded hydrogels with a faster sol/gel transition at 37°C and enhanced mechanical properties. The thermosensitive composite gels also showed a good bioactivity profile associated with potential for the prolonged delivery of protein drugs. The inclusion of chemically crosslinked CNTs and HA in thermosensitive gels afforded injectable composite materials with enhanced properties, including reduction of gelation time, improved mechanical properties, good bioactivity and prolonged drug release.

Introduction
Standard clinical treatments of significant bone defects involve the application of auto-and allo-grafts which present serious limitations and constitute an invasive procedure 1, 2 . Novel therapeutic strategies aim at exploiting the natural regenerative potential of bone by providing biocompatible scaffolds that sustain cell migration, division and proliferation. Amongst the materials studied for the development of novel scaffolds, hydrogels are very prominent 3 . Their biocompatibility and ability to provide controlled release of drugs and growth factors are however counteracted by their weak mechanical properties and low bioactivity. Recently, the addition of nanofillers has been investigated as a viable strategy to address these downfalls.
Carbon nanotubes have been proposed for their ability to increase the mechanical strength of hydrogels and hydroxyapatite (HA) for its bioactive role 4,5 . Furthermore, the use of thermosensitive hydrogels could overcome the limitations of invasive procedures, substituting solid implants with materials that gel in situ after injection. new composites also showed good cell adherence and spreading 2 . All of the materials described above provide solid scaffolds that require surgery in order to be placed on the required site, a way to overcome this limitation is the development of stimuli-responsive systems that gel in situ.
Tang et al. evaluated the effects of the addition of HA to thermosensitive chitosan/PVA hydrogels. Both ex situ and in situ strategies have been used for the introduction of the HA into the gels, with the latter method affording a material with better chemical interactions between its components and a better performance both in terms of mechanical properties and of control over drug release rate 8 . The disadvantage of this system is however that the drug, BSA in this study, was added simultaneously to HA precursors so this may lead to instability, above all when more sensitive protein drugs, such as growth factors, are loaded. Gao et al. also developed HA loaded thermosensitive chitosan gels, in this case the drug was loaded onto preformed HA nanoparticles.
They did not study the release profile of the encapsulated drugs but a positive effect on the proliferation of mesenchymal stem cells was reported 9 .
In the present study we aim at developing a thermosensitive system with enhanced interaction between the three constituent components. The gel is based on electrostatically crosslinked chitosan 10 , the integration with HA and CNTs is favoured by introduction of grafted chitosan on the surface of carboxylated CNTs and by wet precipitation of HA in the presence of chitosan, in this way all three components will interact with the crosslinker, glycerol phosphate. The obtained gels underwent physico-chemical characterisation and their ability to provide a platform for the controlled release of protein drugs is evaluated. Nylon filter membrane (0.22 μm) was purchased from Millipore, UK. Troclosene sodium dihydrate was obtained from Guest Medical (UK).

Synthesis of chitosan grafted carbon nanotubes
Chitosan was covalently grafted on the surface of carboxylated SWNTs and MWNTs according to the method described by Yadav et al. 11 § . Grafted CNTs were analysed by ATR and TGA. ATR spectra were obtained with a Varian 640-IR spectrophotometer (Palo Alto, CA, USA). TGA measurements were carried out with a TGA Q50 from TA Instruments, in the temperature range between room temperature and 600°C, with a heating rate of 10°C/min. The degree of grafting was calculated using Eq. 1.
Where x is the weight percent of chitosan in the grafted material.
The pH was adjusted to 7.0 by slow controlled addition of NH 4 OH (25% w/v). The reaction mixture was magnetically stirred at 37°C for 7 days, the pH was daily checked and adjusted as required. The precipitate was washed once with deionised water, collected by centrifugation (2 min at 2000 rpm) and further purified by dialysis against water and freeze dried (at -40°C). The materials obtained were characterised by ATR, XRD and TGA. The crystallinity index of HA was calculated for the infra red spectra according to the method described by Weiner and Bar-Yosef et al. 12 A baseline from 750 to 500 cm −1 was drawn and the heights of the two in-plane symmetrical bending bands ν 4a and v 4c (600 and 561 cm -1 , respectively) were measured. The (CI) FTIR index was then calculated using Eq. 2: Instruments.

Characterisation of the freeze-dried hydrogels by DVS
Dynamic vapour sorption (DVS) analysis of the freeze-dried hydrogels was carried out with a Surface Measurement Systems DVS Advantage instrument, using nitrogen as the carrier gas. The mass change of samples subjected to a changing water vapour partial pressure at 25°C was recorded. The partial pressure was increased from 0 to 90 % in 10% increments; the partial pressure was increased to the next step either after equilibrium or after a maximum time of 360 min. A full adsorption/desorption cycle was performed; the data collected were used to calculate the adsorption and desorption isotherms as well as the hysteresis. The data were further analysed according to the following Eq. 3.

Eq. 3
Where w p is the weight gain; K p is the kinetic constant of water penetration into the composite material; n p is the exponent describing the mechanism of water penetration.
2.6. Characterisation of the swollen hydrogels A TA.XT plus texture analyser (Stable Micro Systems) was used to characterise the hydrogels.
Syringeability tests were performed on the composite mixtures before gelation took place. The liquid formulations were loaded onto 5 ml plastic syringes fitted with 19 G, 25 mm long needles.
The syringe was then vertically clamped to a syringe rig and the instrument probe (10 mm diameter) was lowered until in contact with the syringe plunger. The probe was used to compress the barrel of the syringe at a constant speed (1 mm / s, corresponding to 3.8 ml / min, comparable to normal injection rates) to a distance of 40 mm and the initial glide force, dynamic glide force and maximum force were measured 13 . All measurements were performed in triplicate at room temperature. Texture profile analysis was then performed by depressing a polycarbonate probe (10 mm dia) into the gels at 1 mm / s and to a depth of 5 mm, six measurements were taken at room temperature before and after the sol/gel transition took place. Force/distance curves were obtained and used to calculate the following parameters: 1) Compressibility (N · mm); work required to deform the product during the compression phase. 2) Adhesiveness (N · mm); work necessary to overcome the attractive forces between the sample and the probe. Gels prepared by  The disappearance of the band at 644 cm −1 (OH) and the shift of the CS signals from 1375 to 1413 cm −1 suggested that HA hydroxyl groups might interact with the hydroxyl groups of chitosan via the formation of hydrogen bonds 16,17 . Also the characteristic amino group band of the polysaccharide, at 1650cm −1 , almost disappeared while the amide I band at 1589cm −1 shifted to 1633 cm −1 and became less intense, indicating electrostatic interactions and hydrogen bonding is probably taking place between the PO 4 3− groups of HA and the NH 3 + groups of chitosan 18 .
From the IR spectra it is possible to calculate the Weiner and Bar-Yosef's Crystallinity Index -(CI FTIR-heights ), this is related to the molecular and atomic bonds within the crystal as opposed to the CI obtained by XRD data that is related to the structural arrangement and the size of the crystal 12 . The CI index obtained from the IR data showed a higher value for the composite compared to the HA alone, further confirming the core role of the polysaccharide in the structure of the composite (Table 1). suggested that the CS peaks would not be seen because of the difference in the intensity of the diffraction peaks between HA and CS. TGA results also supported this observation as they indicated that chitosan accounted only for 17% of the composite weight (Fig. 1C). Chitosan degraded completely when heated to 600°C, after a first water loss at low temperature, it underwent a 2 step degradation process involving depolymerisation, degradation of pyranose rings through dehydration and deamination and lastly ring-opening reactions. 1 Error! Bookmark not defined. The residue mass observed at 600°C was assigned to the presence of inorganic material. The poorly crystalline apatite structure obtained is expected to be more bioactive than a fully developed crystalline hydroxyapatite 20 .

Characterisation of the grafted nanotubes
Homogeneous aqueous dispersions of CNTs are difficult to achieve as they form bundles thermodynamically stabilised by numerous π-π interactions. Techniques such as ultrasonication and chemical modification of the CNT surface have been used to perturb the extended delocalised π system enabling better dispersion of CNTs in aqueous media 21,22 . In the present study chitosan was covalently grafted to carboxylated CNTs to afford stable CNTs aqueous dispersions and to guarantee efficient inclusion of the nanomaterial in the hydrogels, preventing CNTs leaching out of the site of application 23 . A simple coupling procedure involving the use of carbodiimide was used, as this could be carried out at room temperature and in aqueous media, preserving the molecular structures of both chitosan and CNTs (zero-linker) and using nonhazardous reagents. The chemical modification of the composites was confirmed by IR (Fig S1  § ) and TGA (Fig 2).
When chitosan was grafted onto CNTs, the peak at 1648 cm -1 of the secondary amide bond disappeared and a noticeable peak near 1633 cm -1 appeared in the spectrum of CS-CNTs, indicating the formation of a new amide bond. The successful amidation reaction for both CS-CNTs was also confirmed by the appearance of new peaks between 550 and 1680 cm -1 ; our results correlate well with previous studies 24 . TGA showed that SWNTs had a higher degree of modification with 84% of their total weight formed by chitosan as compared to 75% in CS-MWNTs (Fig. 2).

CS-HACS-MWNTs
24.31 ± 0.27 **** 23.57 ± 0.31 **** 24.36 ± 1.00 **** The evaluation of syringe performance of formulations of human drugs and biologics is an FDA regulatory requirement, this is tested by measuring the force needed to start and maintain the movement of the plunger along the syringe barrel 25 . The composite materials were tested to verify their ease of administration via a syringe. Initial glide force (force required for the initial movement of the plunger), dynamic glide force (force required to maintain the movement of the plunger) and maximum force (highest force measured during the experiment) needed to expel the mixture though a 19 G needle were measured ( Table 2). All preparations presented force values below 30 N, considered the upper limit of reasonable injection force as defined by Burkbuchler et al. 13 .
These authors also found that the force needed to inject a liquid into subcutaneous tissue is 1.1 fold higher than that measured in air, indicating that our materials would not go over the 30 N limit even when injected in vivo. All parameters showed an increase in value when HACS was added, while the addition of HA alone did not have a significant effect on syringeability. This might be due to the fact that HA alone does not take part in the gel network and does not affect the flow of the gel subjected to the pressure of the plunger, while the presence of chitosan in the HACS composite allowed for a more intimate contact between HA and the gel structure with participation in the formation of the crosslinking. For this reason the CS-HACS gel provided a higher resistance to flow. Further increase in the syringeability parameters was observed with the addition of both SWNTs and MWNTs, these, being grafted with chitosan, provided further crosslinking points with significant effects on the flowability of the hydrogels. Once ascertained that the composite formulations were injectable, the time to achieve a sol/gel transition at 37°C was determined for each formulation.  (Table 3), which decreased significantly with the addition of HACS and CNTs. The gels containing CNTs had the most marked decrease in gelation time to 8 and 6 min for MWNTs and SWNTs, respectively. Two-tailed, paired t-test showed that there was a significant difference for all samples and all parameters before and after gelation, except when indicated (ns). One-way Anova was performed and returned p < 0.05 for all parameters after gelation. The results of Tukey's multicomparison tests are reported in the graphs (*, p < 0.05; ****, p < 0.0001). All data are reported as mean ± SD (n =6).
This showed that the crosslinking occurring between the three components has a very important role in determining the properties of the composite gels. The gels were further characterised by studying their textural properties 26 . All samples showed a significantly higher compressibility value after gelation confirming that all formulations undergo a sol/gel transition at 37°C (Fig. 3 and Fig. S2  § ). Addition of the nanofillers increased, even if not significantly, the compressibility of the materials. Adhesiveness was also significantly changed after gelation took place, with a significant decrease in all cases except for the hydrogel containing SWNTs. This sample also showed a significantly higher adhesion than the chitosan only gel, this fact is interesting as a highly adhesive gel would be more likely to remain on the site of injection for longer. Compression studies carried out on cylinder shaped gels provided more evidence of the role of CNTs in strengthening the gel structures (Table 3 and Fig S3  § ). The increased Young's modulus and compressive strength observed after addition of CNTs can be attributed to several factors: a higher crosslinking density; high mechanical property of the CNTs reinforcing the matrix; and uniform dispersion of CNTs in the polymer matrix 2 . Comparing the data with information available about the human bone, it is clear that these gels cannot be applied in load bearing joints, however they could be used in non-load-bearing bone or as initial fracture healing callus substitute (Young's modulus 1 MPa), which can then be remodelled and developed into new bone tissue 27, 28 .

Characterisation of the freeze-dried gels
The morphology of the freeze-dried materials was observed by SEM (Fig. 4). The chitosan gel presented a structure with wide pores, the addition of HA and HACS increased the roughness of the surface giving a scaly appearance to the composite with presence of smaller pores together with the wider gaps observed for chitosan alone. Finally, the addition of CNTs afforded composites with a more compact and homogeneous structure but still presenting pores of different sizes. The gels were further characterised by studying their water sorption and desorption profile at 25°C (Fig. 5). A freeze-dried solution of chitosan was used as the control (Fig. 5A). Chitosan showed a profile typical of bulk water sorption, characterised by limited hysteresis and a significant total amount of water absorbed. The hysteresis is due to reversible and elastic swelling deformations as a consequence of water molecules penetration between the chains 29 .
Furthermore, chitosan presented a typical sigmoidal shape isotherm (type II of the BDDT classification) indicative of a mechanism of water absorption via formation of a monolayer. The crosslinked chitosan gel presented an increase in the total amount of water absorbed (Fig. 6), as well as in the hysteresis (Fig 5B). Considering that the crosslinking with glycerol phosphate reduced the number of groups available for hydrogen bonding formation with water, the increased absorption can be explained by formation of pores in the structure, as shown by SEM ( Fig. 4) 30 .  The isotherm shape also changed to a type III, indicating that a different mechanism of water sorption applies in the case of the gel structure; a cluster mechanism. The addition of HA and HACS resulted in similar effects with increased total amount of water sorbed, type III isotherm and similar hysteresis profile. Furthermore, it was observed that chitosan gel, CS-HA gel and CS-HACS gel, not only were the materials that absorbed the highest amount of water but they all failed to reach equilibrium during the initial drying stage, i.e. after 360 min at RH 0%. This indicates that these materials all contained some bound water, that was not observed in the more hydrophobic, CNTs containing, composites. The inclusion of CNTs in the composite materials also significantly decreased the total amount of water adsorbed, phenomenon that could be explained by the increased hydrophobicity of the material or by the reduction in pore size, as a more compact structure was observed by SEM (Fig. 4). Analysis of the kinetics data (Table 4) revealed that the crosslinked chitosan behaves as a glassy polymer with a case II transport type of water penetration kinetic (n > 0.89), where the mechanical behaviour of the polymer is the determinant of the penetration kinetic 31 . This behaviour was maintained when HA and CNTs composites where added to the gels with an increased diffusional exponent (n) value possibly due to stiffening of the structure as HA and CNTs are added and an increased importance of the mechanical properties of the material in the determination of the water penetration mechanism. These findings were consistent with the fact that all samples exhibited hysteresis implying that the amount of water associated with the solid is greater for the desorption isotherm than the sorption isotherm, at a given relative humidity. This can be explained by changes in polymer chain conformation i.e. chain relaxation or irreversible swelling, molecular ordering, or a combination of these with water penetrating the polymeric structure. The penetration of water into the bulk of the polymer during sorption may be driven by the built-up surface condensed water, but not during desorption, hence resulting in a different mass gain following sorption and desorption at the same relative humidity 32 .

3.5.
In vitro calcification of the composite gels In vitro calcification studies were performed to test the bioactivity of the composite materials.
SEM was used to visualise the deposition of calcium salts on the surface of the materials after 7 and 14 days of incubation in simulated body fluid. The SEM images (Fig. 7) revealed that cluster deposition of calcium salt started to occur after 7 days in all samples containing HA but not on the chitosan only gel, indicating that HA might be an important nucleating site for deposition of further calcium salts. Additional deposition was observed following incubation for 14 days.

Drug release studies
The release profile of BSA, as a model macromolecule, was evaluated (Fig. 8). The release of the drug was studied from hydrated and formed gels. In clinical application, the gels would be hydrated but administered in the form of viscous suspensions. Since the time of gelation for the 5 formulations studied was significantly different, it would be correct to assume that the gels, that in the in vitro experiment showed very similar release profiles, would show significant differences in vivo, with a marked bust release for those gels that take longer to undergo gelation.
All gels, in the conditions used, showed the capacity of controlling drug release over a very long period of time (p > 0.05, One-way ANOVA), with a maximum release of about 30% in 14 days, consistent with similar formulations previously reported 33 . Since the hydrogels are already swollen, it is expected that the mechanism of drug release would follow Fick's law. This was confirmed by the n values obtained by the Korsmeyers Peppas plot ( Table 5). All gels presented an n < 0.45 typical of a case I or simple diffusion mechanism, as opposed to the results obtained for water penetration into the dry gel presented in Table 3 discussed above.

Conclusion
Thermosensitive injectable composite hydrogels were obtained by crosslinking chitosan with chitosan grafted CNTs and chitosan-HA complexes. The crosslinking strategy allowed for strong and homogeneous interaction between the three materials affording composites with enhanced characteristics. The composite gels underwent rapid sol/gel transition at body temperature and presented enhanced mechanical as well as bioactive properties. The materials developed in the present study are promising scaffolds for bone regeneration.

Associated content § Supporting information
Synthesis of chitosan grafted carbon nanotubes and of HA and HACS composites. ATR spectra of chitosan grafted CNTs, texture profile of composite hydrogels, hydrogels stress strain curves. 2392843586; e-mail:marta.roldo@port.ac.uk

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.