Figure 1(a) shows SEM image of pure regenerated silk fibroin nanofibers electrospun from a regenerated silk solution dissolved in HFIP at a concentration of 8 wt %. SEM analysis indicates a broad diameter distribution, with an average diameter of 1,250 nm and standard deviation (SD) of 410 nm.

The silk/TMOS nanofibers, shown in Figure 1(b,c), were obtained by adding 5 wt % TMOS in 8 wt % regenerated silk fibroin solution within 30 min under stirring, electrospinning at a voltage of 16 kV and a TCD of 10 cm, and finally drying at 25 °C for 24 h under humidity of 20%. Interestingly, the adjacent fibers in silk/TMOS hybrid electrospun nanofibers caused to ‘weld’ at fiber contact points [30], as confirmed by SEM images (Figure 1(b)). Compared to welded hybrid fibers, the pure silk nanofibers shown in Figure 1(a) were intact and did not show flash welding. Additionally, the fiber diameters showed almost the same size of 1,290 nm (SD = 370 nm) (Figure 1(b)) as the pure silk nanofibers (Figure 1(a)). The observed ‘weld’ at contact points may be due to the equilibrium water content, as was verified by thermogravimetric analysis (TGA) analysis shown later. Moreover, as the TMOS concentration increased to 15%, the fibers became belts and the juncture extended like a sheet which could not be identified as previously reported nanofiber mats (Figure 1(c)). So in this study we investigated the hybrid nanofibers with the TMOS concentration of 5%.

FTIR spectra of electrospun pure silk nanofibers and resultant silk/TMOS nanofibers were shown in Figure 2. Three important absorbance peaks: –Si–OH stretching at 950 cm⁻¹, –Si–O–Si– symmetric stretching at 790 cm⁻¹ and –Si–O– asymmetric stretching at 1,090 cm⁻¹ (each position was indicated by vertical dotted lines in Figure 2) were observed. The formation of siloxane group in the electrospun fibers was clearly evident from the –Si–O–Si– vibration band at 790 cm⁻¹, which is otherwise absent in the spectra of pure silk nanofibers. Furthermore, a slight increase in intensity at 1,090 cm⁻¹ and 950 cm⁻¹ indicating the Si–O and Si–OH stretching vibrations appeared in the hybrid nanofibers. The increase in intensity of the broad peak around 3,400 cm⁻¹ assigned to –OH groups, showed either the silanol group or the –OH group in the water or alcohol liberated due to the condensation reactions of TMOS. Indeed, the observed –OH peak occurred despite extensive drying of silicified fibers for 24 h at 25 °C in vacuum. Peaks corresponding to various groups in silk fibroin were also presented in the region of 1,200–2,000 cm⁻¹, indicating very little or no change in pure silk and silk/TMOS nanofibers.

TGA experiments enabled further quantification of water content in hybrid silk/TOMS nanofibers. We should note that other sources of moisture may exist and are not directly controlled in our experiments, such as trace moisture in TMOS, as well as moisture from the air during electrospinning. Figure 3 revealed that the silk/TMOS nanofibers had an inorganic content around 37 wt % at 600 °C which was lower than pure silk nanofiber. We reason that the silk/TMOS nanocomposite has higher equilibrium moisture content. As evident from Figure 3, a 10% weight loss of the silk nanofibers occurred at temperatures below 120 °C, roughly indicating the level of water absorption by the fibers. TGA characterization allowed us to estimate the water content in the silk/TMOS fibers. As a result, the present study revealed that the addition of TMOS resulted in higher water content in the resultant silk/TMOS fibers.

The hydrophilicity of electrospun nanofiber composites can be seen from Figure 4. Water contact angle showed a sharp decrease of electrospun silk nanofibers incorporated with TMOS than pure regenerated silk fibroin nanofiber from 116.2° to 84.8°. Although silk fibroin has many hydrophilic groups such as –OH and –COOH, the incorporation of TMOS results in higher water capacity in the resultant fibers due to the formation of spatial net structures formed via Si–O–Si– linkages as proved by TG result. Furthermore, the water contact value of silk/TMOS hybrid nanocomposites was closed to that of TCD (Tissue culture Dish) template (75.6°), which indicated that it may be more suitable for cell attachment than pure silk nanofibers because the optimum water contact angle of the surface for fibroblast adhesion is reported in the range between 55° and 75° [31]. The cell adhesion behavior will be discussed in the following text.

Lactate dehydrogenase (LDH) leakage assay results shown in Figure 5(a) suggest that cell culturing on silk/TMOS fibrous scaffolds cause LDH release near 8% while that on silk is near 7%. Both of them showed no significant difference in the LDH release, compared to TCD as a control. The results indicated that the incorporation of TMOS in the fibrous material did not affect the excellent biocompatibility of silk fibroin. From the live/dead fluorescence micrographs in Figure 5(b,c), the majority of cells incubated for 12 h on silk/TMOS and silk scaffolds were alive and parts of them revealed spindle shaped morphology. Cytotoxicity assays indicate that L929 cells on silk/TMOS fibrous scaffold have comparable viability on silk fibrous scaffold.

The adhesion ratio of L929 cells on pure silk, silk/TMOS nanofibers and TCD controls are shown in Figure 6(a). The cell adhesion ratio of silk/TMOS nanofibers was significantly higher than pure silk nanofibers and TCD controls in all the culture times, it reached as high as 95% after 90 min' cultivation while that on pure silk was near 85%. Although an increase in adhesion ratio on both pure silk nanofibers and TCD controls after 30 to 90 min of cell culture were observed, adhesion ratio of silk/TMOS nanofibers showed excellent attachment behavior to L929 cells which could be attributed to the melioration of hydrophilicity. The incorporation of TMOS on silk fibroin nanofibers had enhanced the adhesion behavior of L929 cells as expected.

Immunofluorescence microscopy of L929 cells grown on pure silk, silk/TMOS fibrous scaffold and TCD after 6 h cultivation are shown in Figure 6(b–d). Blue fluorescence of cell nuclei revealed round, well-spaced, and regularly distributed nuclei across the surface of both fibrous scaffolds. Compared to the L929 cells (round shaped morphology) on pure silk, the cytoskeletal organization (green fluorescence) of most cells on silk/TMOS scaffold showed obvious spindle-shaped morphology, while both round and spindle-shaped L929 cells have been investigated on TCD as a control. Moreover, only L929 cells on silk/TMOS showed vinculin signals (red fluorescence) at the extremities of cellular extensions. Consistent with the adhesion ratio in Figure 6(a), these results reveal a better adhesion and stretching behavior for L929 cells on silk/TMOS nanofibrous scaffolds than on a pure silk scaffold.

Accordingly, intensive researches have been carried out in order to manipulate cellular behavior by modifying the relative properties of materials. M. Kuzuya et al. induced durable hydrophilicity on the hydrophobic of polystyrene surface and further modified it by Arg-Gly-Asp (RGD) sequence which can be recognized by the receptor protein on the cellular membrane to enhance the adhesion and proliferation of PC12 cell [32]. M. Lensen et al. induced stable cell adhesion by manipulating the surface topography to the hydrogel poly (ethylene glycol) although fibroblast is intrinsically non-adhesive to the smooth surface [33]. M. Hamdan et al. found that the positive surface of the titanium cylinder resulted in favorable NCTC clone 929 fibroblast cell adhesion [34]. The results in our study suggested that the cell adhesion ratio and spreading on silk/TMOS have been enhanced compared to the pure silk. This can be explained by the change of fibrous surface properties in terms of hydrophilicity and surface morphology change. First of all, water contact angle showed that silk/TMOS had better hydrophilicity than neat silk because of the formation of spatial net structures formed via Si–O–Si– linkages. Studies about the wettability, initial protein adsorption, and the cell adhesion showed that one of the fibronectin states had more active conformation (via secondary structure rearrangements) being on a hydrophilic surface [35,36]. This will consequently lead to more spreading of fibroblasts and ultimately sufficient cell adhesion and spreading. It has been reported that the optimum wettability of the surface for fibroblast adhesion is in the range between 55° and 75° [31]. The TCD used in this study as control showed a water contact of 75.6° (data not shown) while the incorporation of TMOS has changed to the relatively hydrophobic silk surface from 116.2° to 84.8°. Secondly, SEM images in Figure 1(b,c) showed the interesting adjacent fibers in silk/TMOS hybrid electrospun nanofibers caused by ‘weld’ morphology at contact points. It has been known that the substrate's topography has a great influence on the behavior of cells at interface. It showed that contact guidance happened to cells of different types on different materials with different sizes and shapes of patterns [37–39]. Probably, this kind of ‘weld’ morphology in silk/TMOS nanofibrous mats influence the surface microstructure of the fiber that might has positive effect to the L929 cell adhesion, although more intensive study is necessary for the conclusion. Nevertheless, considering the complexity of cell surface interaction, which involves protein absorption and specific binding, the function groups that existed in TMOS and net charges presented on the silk/TMOS hybrid scaffolds might also influence the protein adsorption and therefore cell adhesion to some degree [40,41].

Mineralization of nHA was successfully deposited on pure silk fibroin nanofibers after 3 cycles [42] of Ca–P treatment as depicted in Figure 7. As shown in Figure 7(b,c), the diameter of obtained silk fibroin nanofibers was around 242 ± 34 nm. It was observed that nHA was homogenously formed on pure silk nanofiber substrates. As evidenced in the high resolution FE-SEM micrograph (Figure 7(d)), nHA particles formed on silk fibroin nanofibrous scaffolds were nanocrystalline in structure and the deposition occurred predominately on the surfaces of the nanofibrous scaffolds. The size of nHA particles was approximately 30–35 nm in diameter, which was proved by XRD analysis (see below).

XRD results clearly demonstrated the presence of nHA in the mineralized silk/nHA nanofibrous scaffolds (Figure 8(b): nHA residues and Figure 8(c): mineralized silk/nHA nanofibers). The broad halo peak at 20.6° in Figure 8(c) was attributed to the silk II form of β-sheet crystalline structure [43,44]. All the peaks in Figure 8(b,c) were consistent with the peaks associated with pure nHA (Figure 8(a)), suggesting that rapid mineralization approach used in our study was effective in producing nHA phases on the silk nanofibrous substrates. Unfortunately, both the Energy Disperse X-ray Analysis (EDX) and XRD analyses indicate the poor crystallinity of the nHA formed on silk nanofibers. It can be explained that the hydroxyl or amide group, which existed in the silk fibroin macro chains, captured calcium or phosphate ions in the solution by chelation. The supply of calcium or phosphate ions to the apatite nuclei was retarded, and the apatite crystals grew under the condition that calcium or phosphate ions were not sufficiently supplied. Therefore, the crystal growth of apatite was inhibited along a particular axis and resulted in random orientations of crystals in the mineralized fibroin [45].

Figure 9 shows EDX (Energy Disperse X-ray Analysis) patterns of mineralized silk/nHA nanofibers. It was found that the Ca/P atomic ratio in the nHA prepared in this work was lower than 1.67 of theoretical atomic ratio in the pure HA, which might be indicative of the presence of Ca-deficient apatites. In addition, the full width at half-maximum (FWHM) of the strongest characteristic peak (211) was used to estimate the average crystallite size by applying the following Debye–Scherrer equation, D = κλ/βcosθ, where X-ray wavelength λ was 1.5402, k the shape factor which was often assigned a value of 0.89 if the shape was unknown, D the average diameter of the crystals in angstroms, θ the Bragg angle in degrees, and β the full width at half-maximum of the strongest characteristic peak in radians. It was found that the average crystallite size of pure nHA was 30 nm while both nHA residues and nHA nanoparticles deposited on the silk nanofibers were ca. 32 nm, which also agreed well with FE-SEM result, as shown in Figure 7(d).

FT-IR spectrum was carried out to confirm the functional groups in the mineralized silk/nHA nanofibrous scaffolds. Our observations demonstrated that both phosphate and carbonate groups were presented in mineralized silk fibroin nanofibers (Figure 10(c)), suggesting the successful mineralization by a Calcium–Phosphate (Ca–P) alternate soaking method. The O–P–O bending mode was characterized at approximately 559 cm⁻¹ and 603 cm⁻¹. The bands centered at 1,010 cm⁻¹ and 959 cm⁻¹ corresponded to the P–O stretching vibration modes, which were caused by the PO₄³⁻ group in HA [46,47]. For carbonate groups, the peak was observed at around 873 cm⁻¹ which was ascribable to C-O stretching vibration mode on CO₃²⁻ group [48]. Furthermore, the C=O stretch vibration of amide I (1,625 cm⁻¹) and N–H bending modes of amide II (1,529 cm⁻¹), which were characteristic bands for silk fibroin, were also observed in the IR spectra as seen in Figure 9(b,c). However, when the spectra of mineralized silk fibroin/nHA composites was compared with that of pure silk fibroin nanofibers, the intensity of the two amide peaks of fibroin/nHA composites decreased severely, because of the formation of the bonding between Ca ions and C=O bonds. It has been reported that in the initial stage of collagen mineralization, collagen prefers to chelate calcium ions in solution and these chelate ions subsequently form nucleation sites of calcium phosphate crystals [49]. The oxygen atoms of carboxyl groups and carbonyl groups on the surface of fibroin molecules were incorporated with calcium ions and then served as the nucleation sites for apatite formation and, consequently, the nHA crystals were precipitated on the surfaces of silk fibroin nanofibers.

From our TG results, there was no significant weight loss of pure nHA when the temperature was raised from 50 °C to 600 °C (Figure 11(a)). A weight loss of about 7% around 90–105 °C was detected in pure silk fibroin nanofibers, which was ascribed to the evaporation of water. As the temperature is increased further, the weight residue started to decrease sharply at 295–305 °C (Figure 11(c)) due to the thermal degradation of silk protein [50]. At 600 °C, the weight residue was about 28.5%. The similar water content (about 5%) in the mineralized silk fibroin/nHA composite scaffolds was also observed around 90–105 °C (Figure 11(b)). Moreover, similar thermal degradation behavior with a dramatic decrease in weight residue was observed at around 300 °C. About 48.4% residue was remained at 600 °C for the mineralized silk/nHA composite scaffolds. It was therefore found that the amount of nHA in mineralized silk nanofibers was ca. 20%.

In Figure 12, immunofluorescence of actin filaments demonstrates the cytoskeletal organization (green). Since the high surface area to volume of nanofibers which is used to mimic the extracellular matrix environment of cells, the MC3T3-E1 cells in Figure 12(I) is investigated to spread in spindle or polygonal morphology after 24 h cultivation. Moreover, intensive vinculin signals can be found along the stretching cellular axis. The MC3T3-E1 cell's adhesion activity in Figure 12(I) suggested that the mineralization of nHA on silk fibrous mats did not outweigh the benefit of silk nanofibrous scaffold. 3D network culturing morphology of MC3T3-E1 in Figure 12(II) was determined by laser depth-of-focus scanning about 20μm of the silk-nHA scaffold. Together with the cross-section image of Figure 12(III) where a portion of cells are penetrated into fabricated channels, silk-nHA fibrous mats were proved to be suitable for supporting MC3T3-E1's 3D cultivation.

As evidenced in the FE-SEM micrographs, osteoblasts were successfully seeded on both pure and mineralized silk nanofibers where the cells were partly adhered to nHA regions in the silk/nHA nanofibers (Figure 13(b)). The deposition of nHA did not affect the MC3T3-E1 attachment compared to those grown on pure silk nanofibers after 1 day cultivation (Figures 13(a,b)) [51,52]. Likewise, cell spreading in a spindle-like shape was also observed on HA-based composites after 2 days of cell culture due to the physical contacts between cells which is maintained via the formation of filopodia or lamellipodia [53]. As seen in Figure 13(d), the cells were strongly anchored on the silk/nHA nanofibrous scaffolds, with preferential attachment of the pseudopodia to nHA regions. In addition, a greater cell spreading on silk/nHA nanofibers was observed after 2 days of cell culture (Figure 13(d)), compared to that after 1 day (Figure 13(b)). When the culture period is prolonged in our study, full cell coverage was found on the nanofibers, and eventually osteoblasts covered most of the nanofiber surfaces after 7 days of cell culture with extended lamellipodia, creating a cell multi-layers on the fibers (Figure 13(f)).

Figure 14 shows cell proliferation on pure and mineralized silk nanofibers after 3 days of cultivation. It was observed that when compared with the pure silk nanofibers, the cell numbers were smaller for mineralized silk/nHA nanofiber scaffold and TCD until 7 days cultivation. This is different from what was observed in other studies where osteoblast proliferation was improved on nanophase HA materials [54,55]. Probably, the difference was due to the size effect of hydroxyapatite nanoparticles on proliferation as well as the density or bulk distribution. Moreover, previous studies showed that curved nHA surface at a nanometer level could decrease osteoblast-like cells on early period of proliferation [56]. The previously reported results [57–59] also coincided with those observed in our study: surface topography had a crucial influence on cell behavior, which was accompanied by attenuated proliferation rates on rougher surfaces. Nevertheless, after 14 days of cultivation, cell number on mineralized silk is of no significant differences between pure silk and TCD controls (p ≥ 0.05), suggesting that the addition of nHA had no negative effect on later period of cell proliferation.

One of the properties of nHA is its bioactive nature which promotes osteoblastic differentiation in vitro [60–62]. ALP-hydrolyzed phosphate esters play an essential role in the initiation of the cell differentiation process. Thus, ALP activity, as a marker of osteoblastic activity and a standard to evaluate the differentiation of MC3TC-E1 cells, were measured and shown in Figure 15. There was a slight reduction in ALP activity on the pure silk and mineralized silk/nHA nanofibers than TCD after 5 days of cell culture, while a significant increase in ALP activity on both pure silk and mineralized silk/nHA nanofibers after 7 to 14 days of cell culture, compared to TCD counterparts. Results of ALP activity of pure silk and mineralized silk/nHA nanofibers were comparable to an early stage after 5 days of cell culture, but after 7 days of cell culture ALP activity was meliorated in mineralized silk/nHA than pure silk substrates. The incorporation of nHA on silk fibroin nanofibers had enhanced the differentiation activity of MC3T3-E1 from day 7 to 14. After 14 days of cell culture, ALP activity on mineralized silk/nHA nanofibers was nearly 1.6 times higher than that of pure silk nanofibers. One noteworthy observation was that ALP activity in pure silk and mineralized silk/nHA nanofibers was superior to that of TCD as a control from 7 to 14 days of cell culture.