Pure titanium discs (>99.9%, 12 mm in diameter, 0.2 mm in thickness) and rods (3 mm in diameter) were purchased from Baoji Yuanda Metal Materials Co., Ltd. (China). Ammonium fluoride and ethylene glycol were obtained from Aladdin Biochemical Technology Co., Ltd. (China). Potassium hydroxide was acquired from Chron Chemicals Co., Ltd. (China).

The surface morphologies and distribution of element of the TNT and NTNF samples were analyzed by scanning electron microscopy (SEM, Hitachi S-4800, Japan) and using an energy-dispersive spectrometer (EDS, Oxford X-max80, United States). The surface morphology and roughness of TNT and NTNF were observed by atomic force microscopy (AFM, NX10, Park SYSTEMS, Korea). The chemical composition and phase of the specimens were measured by X-ray diffraction (XRD, PANalytical X’Pert PRO, Netherlands), at a test range of 10°–90°. The chemical compositions and states of the sample surfaces were studied by X-ray photoelectron spectrometry (XPS, Thermo ESCALAB 250, United States).

All values are presented as mean ± standard deviation (SD). Differences between the TNT and NTNF groups were evaluated using Student’s t-test with GraphPad Prism 8.0 software. Significance was determined at *p < 0.05 and **p < 0.01.

Dents and scratches can be seen on the surface of pure titanium in Figure 2 and Figure 3A. A TiO₂ nanotube array (TNT) with a pore size of approximately 70–80 nm and a thickness of approximately 7–8 μm was formed on the surface of pure titanium after electrochemical anodization at 60 V for 2.5 h (Figures 2Bi, 3B). After the nanotube array was shaken off by ultrasonic vibration, 5–6 polygonal honeycomb-like uniformly arranged nanobowls (TNB) were left on the surface of pure titanium (Figures 2Ci, 3C). Figure 2 shows SEM images of the surfaces of the Ti, TNT, and TNB substrates treated with 4M KOH solution at room temperature and pressure for 0 h, 1 h, and 2 h, respectively. After 1 h of alkali etching, the surface of Ti changed to uneven status from a relatively flat surface, and the corroded nano-holes can be seen in some areas (Figure 2Aii). TNT is corroded and collapsed into loose three-dimensional grid-like nanospheres that are easily exfoliated and attached to the titanium surface (Figure 2Bii). TNB is etched as mutually connected holes (Figure 2Cii). After 2 h alkali etching, the number of nanoholes corroded on the Ti surface increases, but they are not completely connected with each other (Figure 2Aiii). The nanospheres formed after TNT etching are further corroded and fused into large and loose three-dimensional grid-like nanospheres (Figure 2Biii). The surface of TNB is further etched into the nested nanofiber structure with connected holes (Figure 2Ciii). The same TNT and NTNF structures can also be seen on the titanium rods (Supplementary Figure S3).

A thin (1.5–10 nm) oxide layer is formed on the surface of the titanium when exposed to the air (Lee et al., 2019; Gianfreda et al., 2021). At the same time, there are slight dents and scratches on the surface (Wang et al., 2019), which leads to uneven and insufficient contact of -OH in the alkaline solution and pure titanium during alkaline etching (Figure 3A). Because etching efficiency is positively correlated with the concentration of -OH, a certain concentration of alkali solution can only partially etch the surface into discontinuous holes. TNT is a TiO₂ nanotube prepared by anodic oxidation (Sun et al., 2018). The inside and wall of the tube are fully contacted with -OH (Figure 3B). Although it can be sufficiently etched by alkali, it could cause the nanotubes to collapse and adhere to the titanium surface, and the adhesion is weak and easy to detach. On the other hand, the TNB has a regular and uniform nanostructure on the titanium surface and an increased specific surface area. The oxide layer is also thickened, so that the -OH in the alkaline solution can fully and uniformly contact the titanium surface (Figure 3C), which robustly increases chemical reaction efficiency (Hou et al., 2018; Hamans et al., 2020; Zhang et al., 2020) and can create nest-like nanofiber structures by soaking in alkaline solution for 2 h under normal temperature and pressure. It is also worth mentioning that the scratch test (Supplementary Figure S3) results show that the SEM images of the NTNF surface do not show the peeling layer after being scratched with the tip of the tweezers compared to that of TNT, indicating that the NTNF structure prepared by top-down etching under normal temperature and pressure has higher adhesion and stability.

Figure 4A shows the XRD patterns for the TNT and NTNF samples. The majority phase for both samples can be indexed to pure Ti (PDF05-0682). Since the top surface of the TNT sample is covered by amorphous TiO₂ nanotubes, the intensity of the XRD peaks of TNT is lower. In addition, amorphous TiO₂ nanotubes do not show any additional peaks. When TNB is immersed in KOH aqueous solution, the oxide layer on the surface is partially dissolved due to the corrosion of hydroxyl (-OH). At the same time, titanium reacts with alkaline solution through hydration reaction, and it further attacks on the hydroxyl group of hydrated titanium dioxide (TiO₂.nH₂O) to produce negatively charged hydrates ([HTiO₃¯]. nH₂O) (Shin et al., 2016). Therefore, the formed titanite (red asterisk) can be observed on the XRD pattern of NTNF (Raveendra et al., 2018). Supplementary Figure S4 shows the full-scale XPS patterns for the TNT and NTNF samples. All the major peaks corresponding to Ti_3p , Ti_3s , Ti_2p , Ti_2s , C_1s , O_1s , and F_1s are labeled. Generally, except for the F_1s peak, the XPS pattern changes little after KOH application. The F₁ peak for the TNT sample indicates that F ions are introduced during the electrochemical process in NH₄F solution. After applying the KOH aqueous solution, F ions are removed. Figure 4B shows the XPS patterns around the Ti_2p peaks for the TNT and NTNF samples. As is shown, both samples show two combining energy peaks for Ti_2p3/2 and Ti_2p1/2. For the TNT sample, the Ti_2p3/2 and Ti_2p1/2 peak positions are 458.6 and 464.4 eV, respectively, which is a signature of Ti⁴⁺ ions (Yang et al., 2005). After applying KOH, the peak positions of Ti_2p for NTNF sample shift to higher energy positions, namely, from 458.6 eV Ti_2p3/2 and 464.4 eV Ti_2p1/2 for the TNT sample to 459.28 eV Ti_2p3/2 and 465.08 eV Ti_2p1/2 for the NTNF sample. This finding demonstrates that oxidation state of Ti on the surface of NTNF sample is not +4. Figure 4C shows the XPS patterns around O_1s peaks for the TNT and NTNF samples. The combining energy ∼529.88 eV for the TNT sample is a typical value for lattice-O, while ∼531 eV for the NTNF sample is a typical value for absorbed O₂ (Yang et al., 2006). Generally speaking, the Ti-O layer on the TNT sample surface was removed by KOH dealing. As shown in Figure 4D, the XPS pattern around K_2p peak for NTNF sample indicates the existence of K ions on the surface of NTNF sample. Generally, the existence of K ions and the shift of Ti_2p peaks from XPS for NTNF sample confirm that the new peaks from XRD pattern should be titanite.

EDS analyzed the chemical composition of the TNT and NTNF samples (Supplementary Figure S5; Supplementary Table S1). Figure 5A and Supplementary Figure S6 show the EDS mapping images of the TNT and NTNF [Ti (yellow), O (green), C (red), and F (blue)]. Apparently, the oxygen ratio for the NTNF sample is smaller than that of the TNT sample. The surface roughness of the implants influences cellular adhesion, dispersal, and proliferation in vitro and in vivo (Zhang et al., 2018; Bernhardt et al., 2021). The surface roughness of the materials refers to the unevenness of the surface with small peaks and valleys, which is correlated with surface smoothness, a critical factor for cell adhesion and proliferation (Zhou et al., 2021). Ra, the arithmetic average roughness, is a commonly used parameter of roughness. Rq is the root-mean-square roughness, which is the root-mean-square value of the deviation between the contour and the average line. For the TNT group, the Ra value was 26.82 ± 1.32 nm and the Rq value was 34.68 ± 0.70 nm, and for the NTNF group, the Ra value was 43.80 ± 2.78 nm and the Rq value was 59.29 ± 4.80 nm, which was significantly greater than that of the TNT group (Figure 5B, p < 0.01). The AFM topography and phase images of NTNF and TNT are shown in Figure 5C. Cracks between nanotube arrays could be seen on the surface of TNT. The surface of the NTNF group had a rougher appearance than that of the TNT group, which is consistent with the SEM results.

Next, the adhesion and morphology of rBMSCs on the surface of TNT and NTNF were evaluated by SEM after 1 day of cell culture. As shown in Figure 6A, the SEM images showed that a large number of slender filamentous pseudopodia protruded from the rBMSCs on the NTNF surface, and these pseudopodia were marked by white arrows. Intriguingly, it was observed that the ends of the pseudopodia all protruded into the grid structure of the nest-like nanofiber structure (white dotted line). However, the filamentous pseudopodia of the rBMSCs on the TNT surfaces were obviously fewer and shorter, and the ends of the pseudopodia were spread flat in the vicinity of the nozzle of TiO₂ nanotubes (blue dotted line). The pseudopod of cell is related to cytoskeleton remodeling and cell adhesion, which is critical for cell spreading and cell migration (Jia et al., 2005). The length of filamentous pseudopodia of the rBMSCs on the NTNF surface was longer than that of the TNT surface (p < 0.05, Figure 6B). To further evaluate the morphology of rBMSCs on the TNT and NTNF surfaces, the cytoskeleton was stained with phalloidin for immunofluorescence detection by CLSM. As shown in Figure 6C, the number of rBMSCs on the surface of the NTNF samples increased slightly and was arranged more tightly, with extensively stretched morphologies, compared with that on the TNT samples. Moreover, the pseudopodia of the rBMSCs on the surface of the NTNF samples were interlaced with each other. These results indicated that NTNF effectively promoted the early adhesion and dispersed morphology of rBMSCs compared with TNT, which may be attributed to the nest-like nanofiber structure of NTNF.

The cell viability of rBMSCs on the surface of the TNT and NTNF samples was evaluated by LIVE/DEAD staining on days 1, 4, and 7. As shown in Figure 7A, the green-stained live rBMSCs exhibited a normal morphology and adhered to the surface of all samples, in which few red-stained dead rBMSCs were found. As the cultivation time increased, the number of rBMSCs on the surface of the TNT and NTNF increased. To evaluate the cell viability and proliferation rate with a more quantitative approach, a CCK-8 assay was performed on days 1, 3, and 5 (Figure 7B). The number of rBMSCs on the surface of NTNF gradually increased during the culturing period, and obvious cytotoxicity was not observed compared with TNT samples. Although no prominent difference in proliferation rate was observed between the TNT and NTNF samples on days 1 and 3, a higher cell proliferation of rBMSCs on the surface of NTNF was observed on day 5, compared with TNT (p < 0.05). These results showed that rBMSCs exhibited satisfactory cell viability and proliferative performance on the surface of NTNF.

Encouraged by the satisfactory biocompatibility of NTNF, we next explored whether the nest-like nanofiber structure of NTNF promotes the osteogenetic process of rBMSCs. As an early marker of osteogenic differentiation, alkaline phosphatase (ALP) is mainly distributed in the cell membrane as a transporter to promote osteoblast maturation and calcification (Gonzalez Ocampo et al., 2019). The quantitative detection of ALP can reflect the differentiation level of osteoblasts. As shown in Figure 7C, the rBMSCs exhibited more distinct ALP staining after 7 days of culture on the NTNF samples than that on the TNT samples. However, no prominent difference was observed between the TNT and NTNF samples on day 4. To further explore the ALP activity quantitatively, the ALP activity was measured after culturing rBMSCs on the TNT and NTNF samples for 4 and 7 days. The quantitative analysis demonstrated that there was no prominent difference in the ALP activity between the TNT and NTNF groups on day 4. However, the activity level of NTNF was significantly higher than that of TNT on day 7 (Figure 7D, p < 0.05). Taken together, these data illustrated that compared with TNT, NTNF exhibited enhanced osteogenic capacity by upregulating the ALP activity.

To further investigate the osteogenesis-inducing capability of NTNF, RT-qPCR was used to quantify the expression of osteoblast-related genes, including ALP, COL1, RUNX2, and BMP2. As described in Figure 7E, the NTNF samples significantly enhanced the expression of COL1, ALP, and BMP2 at day 7 (p < 0.05 for COL1 and ALP, p < 0.01 for BMP2). Similar results could also be detected at day 14, when the expression of COL1, ALP, BMP2, and RUNX2 were significantly upregulated in the NTNF group compared with the TNT group (p < 0.05 for RUNX2, p < 0.01 for COL1, ALP, and BMP2). Overall, these results suggested that NTNF promoted highly efficient osteogenic differentiation at the transcriptional level.

After evaluating the in vitro biocompatibility and osteogenic performance, the in vivo application of NTNF by establishing a beagle alveolar implant model was explored (Figure 8A). The reason for choosing large animals such as beagle is that its jaw is similar to that of humans, which can better recapitulate the osseointegration of implants in vivo (Hakanen et al., 2018). Micro-CT and fluorescent labeling analyses were utilized to detect osteointegration after 3 months of implantation (Figure 8B; Supplementary Figures S7, S8). As shown in Figure 8C, the percent bone volume (BV/TV), trabecular thickness (Tb.Th), and bone mineral density (BMD) of the NTNF group were higher than those of the TNT group (p < 0.05), while the bone surface/volume ratio (BS/BV) of the NTNF group was lower than that of the TNT group (p < 0.05).

The new bone tissue was qualitatively assessed employing double-fluorescence labels at 9 and 11 weeks after titanium bar implantation. As shown in Figure 9A, green and red represent calcein (CA) and alizarin red (ARS) injected into the beagle at 9 and 11 weeks, respectively, and the distance between them represents the formation of new bone. The quantification shows that the dividing distance between CA and ARS was significantly higher in the NTNF group than that in the TNT group (Figure 9B, p < 0.01). These data indicated that the newly formed bone area around the bone implants in the NTNF group was higher than that in the TNT group, which is consistent with the micro-CT observations. Consistent with the in vitro study, the in vivo data illustrated that the nest-like nanofiber structure provided a favorable microenvironment for osteoblast adhesion, dispersal, and differentiation, thus contributing to satisfactory bone implant osseointegration in the beagle model.