Novel injectable, self-gelling hydrogel–microparticle composites for bone regeneration consisting of gellan gum and calcium and magnesium carbonate microparticles

All materials, including GG (Gelzan^™ CM, Product no. G1910, 'Low-Acyl', molecular weight 200–300 kD), CaCl₂ (21 074), MgCl₂, Na₂CO₃ (57 795), were obtained from Sigma-Aldrich, unless stated otherwise.

Ca²⁺/Mg²⁺ solutions of different Ca:Mg elemental ratios and $\text{CO}_{3}^{2-}$ solution (0.333 M) were prepared as described in table 1.

Table 1. Concentrations of solutions used to form microparticles in this study. Equal volumes of Ca²⁺/Mg²⁺ solution and $\text{CO}_{3}^{2-}$ solution were mixed to form microparticles.

Microparticle sample group	Concentrations in Ca2+/Mg2+ solution		Concentration in $\text{CO}_{3}^{2-}$ solution
	CaCl2 (mol dm−3)	MgCl2 (mol dm−3)	Na2CO3 (mol dm−3)
A	0.333	0	0.333
B	0.250	0.083	0.333
C	0.167	0.167	0.333
D	0.083	0.250	0.333
E	0	0.333	0.333

4 ml of a Ca²⁺/Mg²⁺ solution and 4 ml $\text{CO}_{3}^{2-}$ solution were mixed for 30 s at room temperature using a magnetic stirrer. A particle suspension was formed which was centrifuged for 2 min at 4 °C at a speed of 13 000 rpm. The supernatant was removed, and the same volume of double-distilled water was added. The particles were resuspended using a vortexer. The particle suspension was centrifuged again for 5 min at 4 °C at a speed of 2000 rpm. The supernatant was removed and discarded and particles were dried in an oven for 2 d at 37 °C. Dried particles were then sterilized by autoclaving at 121 °C for 15 min (Astell, VWR International, Leuven, Belgium).

A FTIR spectrometer van Perkin Elmer type Spectrum BX in ATR mode (attenuated total reflectance) was used to obtain infrared spectra of the samples. The spectra were recorded over the wavenumber range 4000–550 cm⁻¹ at 32 scans with a resolution of 4 cm⁻¹.

Raman spectra were recorded using a Kaiser Optical Systems Rxn1-532 device equipped with a 532 nm (Nd:YAG) laser source as described previously (Douglas et al 2014b). Spectra were recorded after 10 accumulations of 5 s each.

XRD measurements were performed with a Thermo Scientific^™ ARL^™ X'TRA Powder Diffractometer set to 40 kV and 30 mA. The slits that were used are: 0.6 and 1 at the source side and 0.6 and 0.2 at the detector side. Scans were made in continuous scan mode over a 2θ range from 10 to 50 with a step size of 2θ  =  0.02° degrees and a scan rate of 1.2 s/step. To study the crystallographic phases of the structures the XRD technique with post processing analysis with Highscore plus software by PANalytical B.V. (Netherlands) was used. The phase content analysis was performed by the semiquantative Rietveld method. The theoretical CaCO₃ spectra were taken from the works of Sitepu (2009) and Le Bail et al (2011) for calcite and vaterite, respectively. The theoretical spectra of sodium chloride and magnesium calcite were taken from Strel'tsov et al (1987) and Althoff (1977), respectively.

SEM was performed using a JEOL JSM-5600 instrument equipped with a secondary electron detector. Samples were sputtered with a thin gold layer (15–20 nm) to make the sample surface conductive. In certain samples, the size distribution of vaterite microparticles and the percentage of particles attributable to vaterite and calcite were calculated on the basis of SEM images. Size distribution of particles and percentage quantification of the calcite (cubic-like structures) and vaterite (spherical polycrystals) were obtained by post processing and image analysis of SEM images by Image J software (NIH, http://rsb.info.nih.gov/ij/). At least 90 particles were used for statistical analysis. The average size was shown as mean  ±  standard deviation.

TEM was performed using a Jeol (S)TEM JEM-2200FS instrument operated under 200 kV as described in a previous work (Douglas et al 2014b).

Total concentrations of Ca and Mg in particles were determined using flame atomic absorption spectrophotometry (AAS) as described previously (Liskova et al 2015). Briefly, the weight of each sample was determined to the nearest 0.1 mg by weighing. Samples were dissolved in 14 M HNO₃, and diluted by a factor of 20 to reach a final concentration of 5% HNO₃ (v/v). Concentrations of Ca and Mg were determined using flame AAS (SpectrAA100, Varian, Mulgrave, Australia), after calibration with 4 standards in the range 0–1.00 mg l⁻¹. Calibration was verified with an independent reference material with a maximum certified deviation of 7.5%. Analysis of a procedural blank was performed and the value obtained was below the quantification limit (0.1 µg mg⁻¹ for a 10 mg sample). Concentrations of Ca and Mg were calculated as µg Ca and µg Mg/mg dry weight of the sample.

Dynamic light scattering (DLS) was employed to characterize the size and surface charge of the particles by measuring the hydrodynamic diameter, polydispersity index (PDI) and zeta potential on a Zetasizer Nano Series (Malvern Instruments, Hoeilaart, Belgium). The particle suspensions were added into the disposable folded capillary cells DTS 1070. Measurements were carried out at temperature of 25 °C and a scattering angle of 173°.

Release of Ca and Mg from particles was studied using inductively coupled plasma optical emission spectroscopy (ICP-OES). Particles were suspended in 1 ml Milli-Q water at a concentration of 2.5% in 1.5 ml Eppendorf microcentrifuge tubes. After a certain time point (5, 12.5, 20 min), the tubes were centrifuged for 1 min at 10 000 rpm. 0.5 ml of the supernatant was retained for ICP-OES analysis which was conducted as described previously (Gassling et al 2013), using standard Ca and Mg solutions in the concentration range 0–5 µg ml⁻¹. For all microparticle groups, n  =  4.

Sterile GG solution was prepared as described previously (Douglas et al 2014a). Briefly, 16 ml aqueous 0.875% (w/v) GG solution was sterilized by autoclaving as described in section 2.2. To generate hydrogel–microparticle composites, microparticles were resuspended in ddH₂O at 10% (w/v). 0.2 ml of this microparticle suspension was mixed with 0.6 ml autoclaved 0.875% (w/v) GG solution. Gelation took place at 37 °C. The final concentrations of microparticles and GG in the resulting composites were 2.5% (w/v) and 0.66% (w/v), respectively. Images of composites were taken using a IX81 light microscope (Olympus) after allowing composites of volume 0.8 ml to form in 24-well plates.

Gelation speed was investigated by performing rheometrical measurements with an AR1000N Rheometer (TA Instruments). All experiments were performed at 37 °C using a plate-cone setup with a stainless steel plate and an acrylic cone of 4 cm diameter. First, a time sweep was performed for 1200 s at strain 0.1% and frequency 1 Hz. Immediately afterwards, a frequency sweep was performed from 1 to 100 Hz, which corresponds to angular frequencies of 6.3–627.7 rad s⁻¹.

The distribution of microparticles inside hydrogel–microparticle composites was examined using μCT. The samples were scanned using the HECTOR μCT setup (Masschaele et al 2013) at the Centre for x-ray Tomography of Ghent University (UGCT), which is a custom built system developed in collaboration with the company XRE (www.xre.be). The acquisition parameters for each of the hydrogel–microparticle composites were identical and the total acquisition time was 43 min/sample. The microfocus x-ray source was operated at 120 kV, with a power of 10 W and 2401 radiographs per scan were acquired with an exposure time of 1000 ms using a flat panel detector with 2000  ×  2000 pixels. The acquired radiographs were reconstructed using the software package Octopus (Vlassenbroeck et al 2007) resulting in a 3D volume of 2000  ×  2000  ×  1680 voxels with a voxel size of 5 µm.

The 3D volume was visualized and analyzed using the 3D analysis software Avizo (FEI). The more attenuation microparticles were segmented from the hydrogel matrix using grey-value based single thresholding on the 3D images, resulting in a 3D distribution of the microparticles shown in figure 8.

2.6.1. Cell culture in eluates from hydrogel–microparticle composites

Cytocompatibility was evaluated by determining the viability of MG-63 cells after culture in eluate from composite samples. Eluate was produced by incubating 1 g of composite samples in 5 ml of medium (EMEM, PAN Biotech, Germany) for 24 h at 37 °C. The eluate was diluted in EMEM by factors of 1 (undiluted), 2, 4, 8 and 16. MG-63 cells (7000/well of a 48-well plate) were cultured for 24 h and then subsequently incubated in the eluate (1 ml) at the abovementioned dilutions for the next 24 h, 3 and 7 d.

Cell viability was then evaluated using Alamar Blue reagent (In vitro Toxicology Assay Kit, Resazurin based). 0.05 ml of Alamar Blue reagent was added and the cells were incubated for 4 h at room temperature. Reduction of Alamar Blue was measured fluorescently (excitation wavelength 530 nm, emission wavelength 590 nm) (FLUOstar Omega, BMG labtech) and calculated according to the following formula:

$$\begin{eqnarray}&&\%\,\text{Reduction} ~\text{of} ~\text{Alamar} ~\text{Blue}=\frac{{{S}^{x}}-{{S}^{\text{control}}}}{{{S}^{100\%\text{reduced}}}-{{S}^{\text{control}}}}\cdot 100\%\end{eqnarray}$$

where S^x is the fluorescence of samples, S^control is the fluorescence of medium without cells and S^100%reduced is the fluorescence of reagent reduced in 100% (reagent with medium was placed in autoclave for 15 min at 121 °C).

The result of this measurement is the reduction ratio of the reagent (the higher the reduction, the more cells). This test was conducted three times, after 24 h, 3 and 7 d. Measurements were performed in triplicate.

Cell attachment, distribution and viability were evaluated 24 h, 3 d and 7 d post-seeding by fluorescence microscopy using an Axiovert 40 microscope (Carl Zeiss, Germany) after live/dead staining (Calcein AM/propidium iodide).

2.6.2. Cell culture directly on hydrogel–microparticle composites

FTIR spectra of microparticles are displayed in figure 1. Differences were observed between different microparticle groups. Groups A, B, C and D displayed bands at 712 and 871 cm⁻¹, corresponding to ʋ₄ and ʋ₂ symmetric and antisymmetric bending, respectively. These bands are typical for calcite, as is the band at approximately 1400 cm⁻¹, corresponds to ʋ₃ antisymmetric stretching (Andersen and Brecevic 1991). However sample E, which contained only magnesium and no calcium, displayed a markedly different absorption spectra. The double band at 1472 and 1416 cm⁻¹, which corresponds to ʋ₃ antisymmetric stretching of carbonate groups, and the band at 792 cm⁻¹, which corresponds to ʋ₂ symmetric bending of carbonate groups, both demonstrate the presence of the magnesium carbonate hydromagnesite (Mg₅(CO₃)₄(OH)₂·4H₂O) (White 1971, Frost 2011).

Zoom In Zoom Out Reset image size

Raman spectra of microparticles are shown in figure 2. Microparticle group A showed bands at 1087 cm⁻¹ and 283 cm⁻¹ which are consistent with the υ¹ $\text{CO}_{3}^{2-}$ symmetric stretching and vibration (L) lattice modes, respectively (Bischoff et al 1985). Microparticle group B showed similar but broader bands at 1089 cm⁻¹ and 285 cm⁻¹, as well as another band at 714 cm⁻¹, which corresponds to the υ₄ $\text{CO}_{3}^{2-}$ in-plane stretching. The shift and broadening of the υ₁ and L bands to higher values is consistent with incorporation of Mg in the calcite lattice (Bischoff et al 1985). Another band was observed at 1069 cm⁻¹, which is untypical for magnesian calcite; the reason for its appearance is unclear. Microparticle group C showed a single broader band at 1087 cm⁻¹, suggesting increasing amorphicity. Microparticle group D displayed a single band at 1097 cm⁻¹. This shift suggests that calcite is no longer present, but that a Mg-rich carbonate phase has formed (Bischoff et al 1985). Microparticle group E shows a broad band in the range 1080–1130 cm⁻¹ with a peak at 1109 cm⁻¹ which appears to correspond to the υ₁ $\text{CO}_{3}^{2-}$ symmetric stretching mode (Edwards et al 2005, Frost 2011).

Zoom In Zoom Out Reset image size

XRD results are shown in figure 3. Microparticle group A displayed peaks at 2θ values of 23.2, 29.5, 36.1, 39.5, 43.3, 47.7, 48.6 and 57.5 which are characteristic for calcite (Sitepu 2009), and peaks at 25.0, 27.2 and 32.9, which are characteristic of vaterite (Le Bail et al 2011), suggesting the presence of both calcite and vaterite. Quantitative analysis based on the Rietveld method revealed that the sample contained 60% vaterite and 40% calcite by mass. In microparticle group B peaks characteristic of vaterite were not present, however, the main peak characteristic of the (1 0 4) plane of calcite was observed at 29.9. This shift from 29.5 is characteristic of magnesian calcite (Rahman et al 2013). Further shifts typical for magnesian calcite were observed in the 2θ range 30–50 (Althoff 1977). In addition, peaks characteristic for NaCl were observed, particularly at 31.9 (Strel'tsov et al 1987). Quantative analysis based on the Rietveld method revealed that the sample contained 80% magnesian calcite and 20% NaCl by mass. Microparticle groups C and D were more amorphous because they showed very broad peaks at approximately 32.0. The broadness and low intensity of these peaks suggest the presence of an amorphous phase. Microparticle group E appeared to be largely amorphous, However, peaks at 2θ values of approximately 31.9, 32.9 and 45.5 were also observed, which can be ascribed to NaCl (Strel'tsov et al 1987).

Zoom In Zoom Out Reset image size

Results of DLS measurements of average particle diameter are presented in table 2. The average diameters ranged from approximately 2 to 8 µm. The high polydispersity index of the particles in sample groups A, D and E indicated a large variability in the particle size or homogeneity.

SEM images of microparticle groups A, B, C, D and E are shown in figure 4. The microparticle diameters observed on SEM images were consistent with the DLS data (table 2). Group A contained porous spherical deposits characteristic of vaterite, and cubic, non-porous deposits characteristic of calcite. The size distribution of the vaterite particles is shown in online supplementary figure S1 (stacks.iop.org/BMM/11/065011/mmedia). The average diameter was 1.6  ±  0.3 µm. Analysis of SEM images shown in figure S1 revealed that 83% of the particles were vaterite and 17% were calcite. Microparticles in all other groups were irregular in shape.

Zoom In Zoom Out Reset image size

Results of AAS analysis to determine Ca and Mg content and elemental ratio of microparticle groups A–E are shown in table 3. The elemental ratios in all microparticle groups were approximately equal to the elemental compositions of the starting solution.

TEM images and associated SAED diffraction patterns are presented in figure 5. In microparticle group A, porous vaterite microspheres were observed. The diffraction pattern revealed rings, indicating the presence of small crystallites, which corresponded to the (0 0 2), (1 1 0), and (2 2 0) planes of vaterite. In contrast, all other microparticle groups were composed of agglomerates of roughly spherical nanoparticles, which varied in size. Indistinct and broad rings were seen on diffraction patterns, which indicate far lower crystallinity.

Zoom In Zoom Out Reset image size

Addition of microparticle groups A and B did not induce GG hydrogel formation. In contrast, addition of microparticle groups C, D and E caused gelation.

Light microscopy images of the hydrogel– microparticle composites are shown in figure 6. It can be seen that there was some variance in the size of the microparticles and/or microparticle aggregates present, which was most apparent for composites containing microparticle group E. Aggregates were larger in width than the microparticles observed using SEM (figure 4) and measured using DLS (table 2) Images of the hydrogels can be seen in online supplementary figure S2. All samples were non-transparent to the naked eye.

Zoom In Zoom Out Reset image size

Results of rheometric analysis are shown in online supplementary figure S3. Time sweeps revealed that a plateau value was reached after 1200 s (20 min) in all groups, indicating completion of gelation. Composites containing microparticle group E reached the plateau value after approximately 400 s, considerably more slowly than those containing microparticle groups C and D. The storage modulus decreased in the order D  >  C  >  E. Subsequent frequency sweeps showed that angular frequency at which the storage modulus declined increased in the order E  <  C  <  D.

The release profiles of Ca and Mg from microparticle groups C, D and E suspended in Milli-Q water are shown in online supplementary figure S4. No marked differences in Ca and Mg concentration were observed between 5 and 20 min. This suggests that most Ca and Mg release occurs in the first 5 min of incubation. Release of Ca from microparticle groups C and D was comparable. As expected, no Ca release from microparticle group E was observed. Release of Mg from microparticle groups C, D and E was markedly higher than Ca release from microparticle groups C and D. Release of Mg from microparticle group D was markedly higher than that from groups C and E.

μCT analysis results are shown in figure 7. Aggregates of varying sizes were seen in three composite types, which were obviously larger in size than the particle sizes observed by SEM (figure 4) and DLS (table 2). Aggregates in composites containing microparticle group E were smaller than those observed in those containing microparticle groups C and D.

Zoom In Zoom Out Reset image size

Fluorescence microscopy images of MG-63 cells cultured in eluates from hydrogel–microparticle composites are shown in figure 8. Growth and vitality of cells were comparable to control samples after 1, 3 and 7 d. Results of the Alamar Blue assay are shown in online supplementary figure S5. These results are in agreement with the results of the fluorescence microscopy analysis (figure 9).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Fluorescence microscopy images of MG-63 cells cultured directly on hydrogel–microparticle composites are shown in figure 9. Cells cultured on TCPS displayed a well-spread morphology. In contrast, cells grown on GG hydrogels and hydrogel–microparticle composites displayed a more rounded morphology and were mainly present as single cells or small cell clusters. After 7 d, larger cell clusters were observed on hydrogel–microparticle composites than on GG. Some dead cells were observed on composites containing microparticle group E.

Results of the Alamar Blue assay are shown in figure 10. Proliferation after 7 d was highest on composites containing microparticle group C. Proliferation after 1, 3 and 7 d was lower on composites containing microparticle groups D and E than on composites containing microparticle group C and GG samples.

Zoom In Zoom Out Reset image size

In the absence of Mg, i.e. in microparticle group A, a mixture of vaterite and calcite formed. This was demonstrated by the FTIR (figure 1) and Raman (figure 2) XRD (figure 3), SEM (figures 4 and S1), TEM and SAED (figure 5) data.

Vaterite particles have a porous polycrystalline structure with surface with porosity of approximately 40% (Volodkin et al 2004, Parakhonskiy et al 2013, Donatan et al 2016) independently of particle size and shape. For this reason the density of the vaterite spherolite was 1.6 gcm⁻³ (Volodkin et al 2004) which is 1.7 times lower than the solid calcite or vaterite crystal with a density of 2.6–2.7 g cm⁻¹. Based on this fact it is possible to explain the difference between the amount of calcite and vaterite particles calculated on the basis of SEM images (by number of particles) and by XRD spectra (by mass). Simple recalculation of the vaterite mass using spherical approximation of the particles shape yields a particle volume of 2.1 µm³, which in turn yields a particle mass of 3.4  ×  10⁻¹² g. For calcite, the average particle size is 2.1  ±  0.6 µm, which translates to a particle mass of 13  ×  10⁻¹² g.

Based on SEM results, numerically 83% of the particles are vaterite and 17% are calcite. This corresponds to 57% vaterite and 42% calcite by mass. This in turn corresponds well with the XRD results (60% vaterite and 40% calcite).

It is well known that when Ca and carbonate ions are mixed in solution, the polymorph vaterite forms, which undergoes conversion to calcite in aqueous media and in the presence of sodium chloride (Parakhonskiy et al 2011, 2012, 2013). The fact that a significant proportion had recrystallized to calcite was probably due to the fact that drying took place overnight at 37 °C. Residual moisture may have enabled some recrystallization to take place.

Increasing Mg content of the microparticles led to a reduction in crystallinity, as demonstrated by the decrease of the intensity and increase of width of Raman bands (figure 2), XRD peaks (figure 3) and the increasing occurrence of undefined non-crystalline mineral deposits by SEM and TEM (figures 4 and 5) and the decrease in clarity of SAED diffraction patterns (figure 5). Microparticle group B appeared to consist of magnesian calcite with some vaterite, as evidenced by Raman (figure 2) and XRD (figure 3). Mg ions are more strongly hydrated than Ca ions and can inhibit calcite formation by binding to the surface of calcite nuclei and hindering further crystal growth (Folk 1974). This explains the decrease in crystallinity in the order A  >  B  >  C  >  D  >  E.

Presumably, the increasing amorphicity in the order A  >  B  >  C  >  D  >  E results in increasing solubility, which in turn leads to higher release of Mg and Ca ions which can crosslink GG. Indeed, the amorphicity of sample D led to higher Mg release (online supplementary figure S4). This would explain why microparticle groups C, D and E induced gelation of GG (figure 6, online supplementary figures S2 and S3), whereas microparticle groups A and B did not. The decrease in mechanical strength of the hydrogel–microparticle composites in the order D  >  C  >  E (online supplementary figure S3) might be explained by the differences in Ca and Mg release (online supplementary figure S4). Microparticle group D released more Mg than group C, which in turn released more Ca than group E. It should be borne in mind that release studies were performed in water. The kinetics of ion release into water and the diffusion of ions in water may not be similar to those in GG solution and crosslinked GG hydrogels, as GG has the ability to bind Ca²⁺ and Mg²⁺ ions and crosslinking of GG polymer chains during hydrogel formation may impede diffusion.

Cytotoxicity testing of hydrogel–microparticle composites (figure 8 and online supplementary figure S5) revealed no striking differences between them and controls. This suggests that the concentration of Ca and Mg ions in eluates is not cytotoxic and does not promote cell proliferation.

Cells seeded on hydrogel–microparticle composites did not show a more spread morphology than those on GG (figure 9). Perhaps the particle concentration (2%) is too low to significantly influence cell adhesion. The reason for the increased cell clustering on hydrogel–particle composites in comparison to GG remains unclear. The reasons for the increased cell death on composites containing microparticle group E remain unclear. Ion release from microparticle group E was lower than that from groups C and D (online supplementary figure S4), which suggests that increased ion release is not the cause, but in the absence of further data, such discussion must remain speculative. It is also unclear why cell proliferation was higher on composites containing microparticle group C (figure 10). One may speculate that the differences in ion release from microparticles (online supplementary figure S4), caused by differences in crystallinity, and differences in hydrogel stiffness (online supplementary figure S3) may play a role.

MG-63 cells are considered to be a good model for studies on cell adhesion and proliferation, as performed in this study. Further work will include investigation of osteogenic differentiation and bone matrix production. For such studies, Saos-2 cells and primary osteoblasts will be selected as their expression of important osteogenic markers is generally much higher on both mRNA and protein levels (Pautke et al 2004, Shapira and Halabi 2009, Saldana et al 2011, Czekanska et al 2012, 2014).