Hybrid sol–gel inorganic/gelatin porous fibres via solution blow spinning

For the production of neat sol–gel silica fibres a sol was prepared using formulations based on the work of Poologasundarampillai et al. [31]. Briefly, TEOS, EtOH (mutual solvent), water and 1 M HCl (acid catalyst for hydrolysis) were added in that order to a final molar ratio of 1:2:2:0.01. Typical volumes used to make all the sol-precursors are given in Table S1. The solution was mixed for 24 h at room temperature in a sealed 50-mL polypropylene container (Falcon^® tube), using a roller mixer (Stuart SRT6, Bibby Scientific, USA). This solution was then transferred to a water bath and then heated to 70 °C with the lid-removed to assist evaporation, until ~65% of the mass remained and a suitable viscosity for solution blow spinning (SBS) obtained, SBS is described further below. Evaporation of EtOH has been shown to increase solution viscosity by catalysing the polycondensation of the hydrolysed TEOS [31].

Sol–gel silica–gelatin class I hybrid solutions for conventional SBS were prepared in two parts. Aqueous gelatin solutions at 10% w/v were prepared by dissolving gelatin (typically 1 g gelatin in 10 mL deionised water) at 40 °C under magnetic stirring in a sealed container, using a temperature controlled water bath. Separately a sol was prepared as described above to give a molar ratio of 1:0.7:1.6:0.03 (TEOS:EtOH:H₂O:HCl) and mixed until the solution was clear. Hybrid solutions were typically made by mixing 6 mL of gelatin solution to 7 mL silica sol in a 100-mL sealed container under continuous magnetic stirring and the pH adjusted to 3.6 by drop-wise addition of 1 M HCl in 50 µL increments. The hybrid solution was then heated in the sealed container under continuous stirring in a water bath at 60 °C. At intervals of 5 min, samples of the stock solution were taken for solution blow spinning, in order to identify the reaction time/window to enable fibre production (determined later to be between 30 and 40 min, discussed below) as opposed to droplet spray or blocking.

The sol formulation described above was modified to enable effective freezing and lyophilisation of the hybrids fibres, by the use of tert-butyl alcohol (TBA) in place of ethanol. TBA was selected as it has a melt point of 25 °C, has been used to form silica cryogels [59], is relatively facile to lyophilise, and forms azeotropes with water and ethanol. TEOS, TBA and 1 M HCl were added in that order to give a molar ratio of 1:5:2.7:0.05 (TEOS:TBA:H₂O:HCl), typical volumes shown in Table S1. The sol was then roller mixed as described above for 24 h. Class I hybrid solutions were then prepared by mixing a 10% w/v solution of gelatin in PBS (prepared as described above) with the hydrolysed sol. Class II hybrid precursor solutions were made by functionalising gelatin (already dissolved at 10% w/v in PBS) by addition of GPTMS to 92 µL g⁻¹ of gelatin, and subsequent incubation at 40 °C for 13 h under continuous orbital shaking (150 rpm, using a Stuart Orbital incubator SI50, Bibby Scientific Limited, UK), following a method adapted from [60]. Class I and II hybrid solutions were typically made by mixing 7.26 mL of the hydrolysed sol to 8 mL of the gelatin solutions, to achieve a target theoretical inorganic content of ~40% (based on assumed SiO₂ content). Hybrid solutions were sealed in Falcon^® tubes, then placed in a water bath at 40 °C. The gelatin was initially observed to precipitate upon mixing and then solubilise within 30–40 min. The solutions were held for a further 20 min prior to cryo-SBS and rheological characterisation, described below.

Conventional and cryogenic SBS was conducted based on set-ups previously described for the production of bioactive smooth and porous nanobioactive glass-filled PLA composite fibres [9]. For conventional SBS, neat sol–gel and hybrid precursor solutions (described in “Sol–gel and hybrid fibre precursor formulations for conventional SBS” section) were injected to the inner coax of the SBS head at a rate of 200 μL min⁻¹ using a precision syringe driver (N-300, New Era Pump Systems, Inc., Farmingdale, NY, USA). The inner nozzle was surrounded by a high-pressure air sheath, or outer nozzle, through which compressed air was delivered at 0.207 MPa (30 psi) via a Bambi VT150D oil-free and water-free air compressor (Bambi, Italy) and in-line regulator 5 cm from the SBS head air inlet. The internal diameter (ID) of the SBS inner nozzle was 0.8 mm, outer diameter (OD) 1.8 mm, the outer nozzle had an ID of 2 mm and OD of 3 mm. The inner nozzle had a tapered end (truncated cone) and set to protrude 2 mm from the outer nozzle. Fibres were collected at a distance of 30 cm from the SBS head on a round (~15 cm diameter), static stainless steel collector with spokes 2 mm diameter (adapted from a computer case fan grill).

Cryo-SBS [9] was conducted using class I and II hybrid formulations described in “Class I and Class II hybrid fibre precursors for cryogenic SBS” section. The solutions were injected separately into the inner nozzle of the SBS head as described above at a rate of 230 μL min⁻¹, using outer nozzle pressure supply of 0.138 MPa (20 psi). Fibres were sprayed directly into a bath of liquid N₂ at a working distance of 10 cm and at an angle of ~45° to the bath. The spraying angle and sheath pressure were selected to limit liquid N₂ loss. The bath container was made from an expanded-polystyrene foam box. During cryo-SBS, the level of liquid N₂ in the bath was constantly topped-up to ensure that the working distance remained ±2 cm of the target value (10 cm). The temperature of the solution was measured at the SBS nozzle, by direct contact with a thermostatic probe, to be 22 °C. Frozen fibre membranes were collected in 50 mL Falcon^® tubes, transferred to lyophilisation flasks (pre-cooled in liquid N₂) and subsequently lyophilised using a VirTis BenchTop Pro Freeze Dryer (NY, USA). The external glass portion of the lyophilisation flask was kept under liquid N₂ using the polystyrene box until a vacuum of ~50 mTorr had been reached, after which the outside of the flask was left exposed to ambient conditions. Lyophilisation was continued for 72 h, after which the fibres were collected for characterisation. Additionally, porous class I and class II hybrid spheres were formed based on a method previously described by Blaker et al. [9, 61], and briefly solutions were directly dripped into liquid nitrogen and lyophilised as described above.

The morphologies of fibres and porous spheres were assessed using scanning electron microscopy on a FEI Sirion FEG-SEM (USA), as well as Hitachi S300 N microscope (USA) for screening. Porous spheres were sectioned using a double-edged Wilkinson Sword^® razor blade, as previously described [9, 61]. Mounted samples were blown with compressed air to remove and loose material and sputter coated with platinum using a Gatan Model 682 Precision Etching Coating System (USA) to an average thickness of 7 nm. Fibre diameters were analysed using ImageJ (version 1.48, NIH, Bethesda, MD, USA) software, from a minimum of 70 fibres.

The solid state ²⁹Si MAS NMR spectra were obtained using a Bruker AVANCE (USA) 400 MHz spectrometer operating at 79.48 MHz. The spectra were recorded with a 20° pulse and a repetition delay of 20 s (4000 scans) using 4 mm zirconia rotors spinning at 6 kHz. The spectra were referenced against TMS (0 ppm). The degree of condensation (D _C) was the following equation [25]:where T ⁿ and Q ⁿ are the relative abundances of silicon atoms bound to a carbon and n bridging oxygens and silicon atoms bridging to n oxygens, respectively [25].

TGA was conducted using a TGA-Q500 (TA Instruments, USA) using a dynamic heating regime in air to ensure removal of the organic phase. Samples (2–30 mg) were first equilibrated at 40 °C for 60 s to remove trapped water then heated at 10 °C min⁻¹ to 75 °C, and held 15 min to perform an isotherm. The temperature was then raised using a high-resolution ramp at 50 °C min⁻¹ to 400 °C and held for 30 min, the isotherm ensured complete combustion of the organic phase. Heating was then continued in high-resolution mode at 50 °C min⁻¹ to 800 °C, at which point the residual mass is representative of the silica content as the gelatin has undergone thermal degradation.

Brunauer–Emmett–Teller (BET) surface area measurements of the fibres were made using a surface area and porosity analyzer (TriStar II, Micromeritics, Neu-purkersdorf, Austria). Prior to the gas adsorption experiments, impurities were removed via a “degassing” step using a FlowPrep 060 instrument (Micromeritics). Approximately 100 mg (about 1 cm³) of each sample was placed inside a glass sample chamber, which was purged with nitrogen at ambient temperature overnight. Afterwards, the nitrogen adsorption isotherms were measured at 77 K.

Class I and class II hybrid fibre membranes produced via cryo-SBS were mounted separately on glass microscope slides using thin strips of Scotch^®Magic™ Tape (3 M) applied at the sample edges to obtain reasonably flat surfaces for measurement. Contact angle measurements were taken on a KRÜSS DSA 100 (GmbH, Germany). Single droplets of water were dropped onto the substrates and the wetting behaviour recorded by video recorded with a built-in high-speed camera.

Ethanol in the above formulations was swapped for TBA to enable facile lyophilisation. Highly porous silica nanoparticles have previously been prepared using TBA and a basic catalyst [59]. Su et al. [59] previously obtained high surface area cryogels using a molar ratio of 13 (TBA:TEOS). In the current work, a molar ratio of 5 (TBA:TEOS) was used as higher concentrations of TBA lead to the precipitation of gelatin [63]. The gelatin solution was effectively diluted from 10 to 5.2% w/v on addition to the inorganic precursor solution, to obtain a total solute concentration of ∼9% in the hybrid spinning solution given 100% conversion of TEOS to SiO₂.

The viscosity of class I and class II cryo-SBS precursor solutions, as a function of shear rate at 20, 30 and 40 °C, is shown in Fig. 2a. These temperatures were selected as they represent the preparation, dissolution and reaction conditions, and an intermediate point cooling down to spinning at room temperature. All precursors exhibit Newtonian behaviour at shear rates above 100 s⁻¹. Orefice and Hench [32] reported initial shear thinning of bioactive sol–gels followed by Newtonian flow above shear rates of 10 s⁻¹. They indicate that a water to alkoxide molar ratio of 2 yielded a linear polymeric precursor that is amenable to the spinning process [32]. Despite the higher water to alkoxide (TEOS) molar ratio used in the current work, the incorporation of gelatin (in solution) also facilitated the spinning process by providing entanglements. Both class I and II solutions exhibit an increase in viscosity with decreasing temperatures, which is more marked for the class II solutions. Class II solutions are more viscous due to the development of covalent crosslinks, further evidenced below. The temperature-dependent viscosity was investigated at constant shear rate (500 s⁻¹), both classes exhibit a rapid increase in viscosity below 20 °C (Fig. 2b). The typical sol–gel transition for the 10% w/v gelatin solution is shown in Fig. S1. Below 23 °C the 10% w/v gelatin solution exhibits a sharp increase in viscosity (to 2949 mPa s at 15.3 °C), indicative of the thermally activated conformation change from random coil to triple helix [57, 64]. The hybrid solutions exhibit a similar rapid increase in viscosity, below 20 °C and occurring over a wider temperature range (to ~12 °C), likely due to the increased dilution. The extended range is also likely due to lower pH of the hybrid formulations and some protonation of the gelatin (class I pH 3.9 and class II pH 4.1, gelatin pH 6.2, gelatin-GPTMS pH 6.7).

The thermoreversibility of the hybrid formulations was tested by cycling the precursors between 30 and 20 °C, and measuring the viscosity at constant shear (500 s⁻¹) (Fig. 1c). The viscosity varied between 27.3 ± 1.1 (20 °C) to 18.1 ± 0.6 (30 °C) mPa s and 48.3 ± 4.8 (20 °C) to 28.4 ± 2.3 (30 °C) mPa s for class I and II hybrid precursors, respectively. The viscosity is tuneable in the 20–30 °C range. The class II solution also exhibits time-dependent behaviour due to the formation of covalent bonds between gelatin and silica network, a rapid increase in viscosity is observed for the sample held at 40 °C after 65 min (Fig. 1d), whereas the class I solution exhibited an increase in viscosity at ~65 min, then a plateau region (behaviour was exhibited in three repeats, representative profile shown). The gelatin-GPTMS solution underwent no significant change in the measurement period. The class II solution held at 20 °C exhibited a near linear increase in viscosity over time (Fig. 1d). As expected, the development of the silica network and covalent crosslinking is time- and temperature-dependent, and both can be used to control the viscosity and spinning window. The crosslinking reaction between the organic and inorganic phases via GPTMS is catalysed by heat and after extended periods of time the class II solutions gel irreversibly (images shown in Fig. 1e). After 24 h of incubation at 40 °C, the class II solution irreversibility forms a gel whereas the class I solution remains as a viscous solution. Irreversibility was also evident if the gelatin-GPTMS functionalisation reaction proceeded for an extended period of time, the hydrolysed trimethoxysilane terminal groups of GPTMS begin to polycondense and then become insoluble in the inorganic precursor phase, in agreement with Negahi Shirazi et al. [60].

The surface tension of the gelatin, functionalised gelatin, silica precursors, class I, and class II solutions are shown in Table 1. The gelatin and gelatin-GPTMS solution had a similar surface tension of 51.2 ± 2.5 and 47.4 ± 6.3 mN m⁻¹, respectively, at 30 °C. Upon addition of the silica precursor, the surface tension was reduced to 24.8 ± 1 and 25.7 ± 0.5 mN m⁻¹ for class I and class II at room temperature, respectively. The TBA contributed to reduce the surface tension, with its hydrophobic methyl groups. The decrease in surface tension of the gelatin solutions after mixing with the silica precursor indicates that the surface tension is governed by the inorganic component. The surface tension of the class I and II solutions corresponds to polymer solutions that have been successfully spun into fibres via SBS. PLA solutions (in chloroform, dichloromethane and chloroform/acetone mixtures) with surface tensions in the range 26–34 mN m⁻¹ have all been successfully spun into fibres via SBS [65‐67].

Whilst the rheological properties can give an indication of spinnability, an important consideration is that the shear, solvent evaporation and degree of polycondensation are influenced during fibre spinning as the precursor is drawn into a fibre and the solution is exposed to rapidly moving air at the nozzle exit.

In order to test the ability of the solutions to be frozen and lyophilised to form porous structures, class I and class II solutions containing TBA were manually dripped via a syringe with a 25-g needle directly into liquid nitrogen. The frozen spheres were collected and lyophilised. The lyophilised spheres exhibit a porous skin region (Fig. 3a) and a highly porous interior (sectioned sphere, Fig. 3b, and inset). Class I and class II spheres exhibited very similar internal and external morphologies. The spheres produced were ~2 mm in diameter. Porous, composite poly(D,L-lactide) (PDLLA)/nanobioactive glass-filled spheres have been produced via the TIPS process [9, 61], and the size of neat porous PDLLA spheres can be reduced to <1 μm using vibrating needles or SBS of low viscosity solutions [61]. It should therefore be possible by extension to produce hybrid porous spheres of similarly small diameters by vibrating needle/atomisation techniques, whilst porous sol–gel/hybrid foam monoliths have previously been formed by freezing vials containing aged gels into liquid nitrogen and lyophilised here are issues with cell infiltration [25, 29, 38]. Whereas the inter-sphere void size can be controlled by sphere size as well as packing efficiencies, such spheres may also have application in bioreactors for cell expansion.

Both class I and class II lyophilised hybrid fibres exhibited cotton–wool like structures, as shown in Fig. 4a, b, respectively. There is some evidence of bead formation (porous beads/spheres), both in line with the fibres and entangled within the fibre membrane, more prevalent for class II fibres. Droplet/bead formation is a result of the high driving force at the SBS nozzle causing Rayleigh instability [68]. SEM analysis revealed both class I and II fibres rough surface morphologies, shown at low magnification in Fig. 4c, d, respectively. In both cases, the fibres are seen to exhibit a more roughened surface topography, which is distinct in comparison with fibres produced by conventional SBS [69] and electrospinning [31]. The surface roughness is hypothesised to result from thermally induced phase separation occurring as the fibres contact the liquid nitrogen. The excess solvent is removed by the lyophilisation process leaving behind the texture that is seen throughout the fibres. High-magnification SEM images of the fracture surfaces of the fibres (class I and II, Fig. 4e, f, respectively) reveal a highly porous interior and a thin skin region, similar in appearance to the porous spheres described above. Class I and II porous fibres exhibit mean diameters 4.4 ± 1.7 and 5.8 ± 2.1 µm, respectively (fibre size distributions are shown in Fig. 4g, h, respectively). The higher diameter observed for the class II fibres is likely due to the higher viscosity of the precursor solution, stemming from organic–inorganic covalent crosslinking. The fibre diameters for both porous class I and II hybrids are larger in comparison with those produced by conventional SBS in the work presented here, and electrospun hybrid fibres were produced by others. As previously mentioned, authors achieved fibres diameters of 192.6 ± 8 and 239 ± 13 nm by electrospinning gelatin/silica and chitosan-PEO/silica hybrids, respectively [47, 49]. Despite the relatively large fibre size distribution, class II fibres in the range of 700 nm were obtained. It should be possible to adjust the formulation and processing parameters to obtain more consistent fibres of smaller diameters.

The ²⁹Si MAS NMR spectra evidence the structure the silica network has within the hybrid fibres, shown in Fig. 5a, b. Both class I and class II fibres show strong peaks corresponding to a Q ³ species, silicon bonded to three other silicon atoms by bridging oxygens. Q ⁿ species are standard in networks formed from TEOS as a precursor, and the smaller Q ⁴ peak is a result of the network formation occurring under non-ideal conditions and is seen commonly in the literature [44, 62, 70]. In the class II fibres, peaks can also be seen around −56.3 and −64.3 ppm which correspond to the T ² and T ³ species, a silicon atom bonded to 2 and 3 carbon atoms, respectively, via bridging oxygen atoms [71]. The signal form the T ⁿ species was very low in comparison with the Q ⁿ structure due to the concentration differences between the two species (molar ratio of 1:30 GPTMS to TEOS).

The degree of condensation was lower in the hybrids formed here in comparison with those reported in the literature due to the high Q ³ abundance. The higher Q ³ abundancy is a result of the procedure used to make the silica precursor, i.e., reaction time, acid concentration, temperature, and TEOS concentration. The degree of condensation (D _c) for the class I fibres was 83.6%, which was slightly higher than the D _c of the class II fibres (78.0%), as calculated with Eq. (1). The lower D _c in the class II fibres was likely due to the incorporation of GPTMS that leads to the relative abundance of the T ⁿ species skewing the D _c. ²⁹Si MAS NMR provides an insight to the structure of the silica network; it does not confirm crosslinking between the organic and inorganic phases in the form of C–Si–O–Si–O. The T ⁿ species can react with either the hydroxyl groups of the silica network or another hydrolysed trimethoxysilane terminal. Coincidentally, the T ⁿ species may just represent crosslinked gelatin. Further, the class II fibres do not undergo dissolution when immersed in water, due in part to bridging between gelatin chains rather than the inorganic phase [72]. However, over the period of an hour, the viscosity of the gelatin-GPTMS solution remains constant (Fig. 2d), whereas the viscosity of the class II solution increases due to organic–inorganic bridging (Fig. 2d) and therefore suggests that the dissolution properties of the class II fibres are due to organic–inorganic bridging.

The reaction between GPTMS and gelatin has been extensively covered and characterised in the literature [25, 38, 60, 73, 74]. The methodology reported by Negahi Shirazi et al. [60] was used in the current work to functionalise gelatin with GPTMS, buffering the pH to 7.4 to reduce hydrolysis and polycondensation of the trimethoxysilane terminal and diol formation of the epoxy ring. Although Gabrielli et al. [73] reported that at a lower pH the rate of epoxide ring opening and hydrolysis increased, it was not determined for basic conditions due to precipitation. At a neutral pH (in PBS), it has been reported that GPTMS is very stable; however, the presence of nucleophiles catalyses the ring opening reaction [75, 76]. In the current work, nucleophiles on the gelatin chains (carboxylic acid and amine groups) as well as the elevated temperature (40 °C) acted as the catalyst for the ring opening reaction, according to [60].

Theoretically, given complete hydrolysis and 100% efficiency of the polycondensation reaction, the fibres should have 42 and 43 wt% inorganic for the class I and II fibres, respectively. TGA results show that the inorganic content in the class I and II fibres was 38.4 ± 1.8 wt% (Fig. 6a) and 38.6 ± 1.6%wt (Fig. 6b), indicating a polycondensation mean reaction efficiently of, 91.43 ± 4.29 and 89.77 ± 3.72%, both, respectively. Several reports have highlighted an organic phase is in the region of 50–70% and is beneficial to impart flexibility to the hybrid [28, 29, 47, 77, 78].

BET surface areas of class I and class II fibres produced via cryo-SBS were measured to be 3.4 ± 0.2 and 9.5 ± 0.3 m² g⁻¹, respectively. Medeiros et al. [9] reported surface areas of 43.5 m² g⁻¹ when applying the cryo-SBS to PLA fibres although McCann et al. [79] reported surface areas of ~9.5 m² g⁻¹ for poly(acrylonitrile) (PAN) fibres with the diameter of 1 µm using cryogenic electrospinning. Water droplets rapidly wicked into both the class I and class II cryo-SBS porous hybrid fibre mats in the range of 0.6–4 s. The wettability is in contrast to porous fibres produced by cryo-SBS based on nanobioactive glass-filled/PDLLA, which have been shown to exhibit high apparent water-in-air contact angles, ~140° and may require surface treatment.

It remains a further challenge to truly incorporate calcium into the silica network of sol–gel hybrid fibres to allow the controlled release of therapeutic ions. Calcium has been incorporated using calcium chloride and calcium nitrate tetrahydrate; however, it has been highlighted that the calcium does not enter the silica network without thermal treatment and therefore is not appropriate for the hybrid materials developed here [35, 36, 48, 49, 80]. Calcium alkoxide precursors appear a more promising route and have been used to obtain monoliths and fibres with homogeneous silica and calcium networks without the need for thermal treatment [46, 72, 77, 81]. The incorporation of calcium into the porous fibres using calcium alkoxides in the formulations and processing is the focus of current work.