Twenty-one passive cylindrical implants (Ø10 mm × 3 mm) were additively manufactured from commercially pure titanium powder (ASTM B348 Grade two spherical powder, Ø15-45 μm, Carpenter Additive, United States) using a powder bed fusion system (AM250, Renishaw, UK). Surface roughness was created on the bone-facing base of the implants by adding randomized sinusoidal-based contours to the part, as previously described (van Arkel et al., 2018), to yield roughness of Rz = 730 μm, Ra = 75 μm, Rq = 100 μm; this level of surface roughness is similar to that of contemporary cementless joint replacement implants.

Six of the passive implants were modified to deliver an active osseointegration stimulus. It has previously been shown that embedded ultrasound transducers are effective at diagnostically monitoring small changes at the implant/bone interface (Hall et al., 2021). Inspired by those findings, a system that delivers electrical power to an implant and stimulates tissue at the implant-bone interface via an embedded ultrasound transducer was designed. Thickness-mode piezoelectric transducers (PI Ceramic PRYY+0442; Ø10 mm × 1 mm) were embedded within a biocompatible polymer insert (Stratasys ABS-M30i) and bonded to the titanium discs using a thin layer of epoxy (Loctite Double-Bubble). Electrical contact was established to the positive electrode by soldering and to the negative electrode with a biocompatible Steel 316L grub screw. PDMS was used to encapsulate the circuitry and provide a 3-mm thick backing for the transducer.

Murine preosteoblasts from the MC3T3-E1 cell line (ATCC CRL-2593) were proliferated for three passages in cell culture medium: α-MEM (ThermoFisher 11095080) supplemented with 10% foetal bovine serum (ATCC 30-2020) and 1% penicillin-streptomycin (ThermoFisher 15070063). At 75% confluency on the final passage, the cells were trypsinised and seeded directly onto the rough surface of the additive-manufactured titanium discs (Figure 1; n = 8) at a density of 5,000 cells/cm². Specimens were cultured in twelve-well plates for 7, 14, 21, and 28 days (n = 2 per time point) in normoxic conditions at 37°C and 5% CO₂. Cell culture medium was replaced every 7 days.

Research was registered as part of the PI’s annual declaration in accordance with the host institution’s policy for research with animal tissue samples. Fresh trabecular bone cylinders (Ø10 mm × 5 mm; n = 24) were prepared from distal porcine femora (age: 18 weeks; weight: ∼75 kg); the donor animals had already been euthanised for unrelated research that did not affect bone metabolism or healing. Specimens were transected at the mid-femur and mid-tibia, exsected to the level of the stifle (‘knee’) joint capsule, and immersed in ethanol for 7 minutes before being transferred to a laminar flow hood. Cancellous bone cylinders were then extracted from the distal femoral condyles in a sterile manner (Figure 2): power tools, attachments, and clamps were autoclaved; the bioreactor was immersed in 70% ethanol after assembly and irradiated with ultraviolet (UV) light for 30 min on each side; and all other equipment was wiped with 70% ethanol and irradiated with UV.

Specimens were extracted using an oscillating saw blade and a diamond holesaw; ice-cold Dulbecco’s phosphate-buffered saline (DPBS; ThermoFisher 14190250) with 1% penicillin/streptomycin (ThermoFisher 15070063) was used to minimise thermal necrosis whilst cutting with low rotational speed. Specimens were washed in DPBS and vortexed for 30 s to reduce cutting debris. Within 5 hours of euthanisation, all bone cores were prepared and immersed in cell culture medium, which was composed of DMEM/F-12 with HEPES (ThermoFisher 11330057) plus 10% foetal bovine serum (ThermoFisher 10500064) and 1% penicillin/streptomycin (ThermoFisher 15070063). During culture, bone cores were incubated in normoxic conditions at 37°C and 5% CO₂.

The bone samples were cultured in a custom-built bioreactor (Figure 3) and divided between four conditions: A: the active osseointegrative implants (n = 6); B: passive implants treated with low-intensity pulsed ultrasonic delivered through the bone (n = 6); C: passive implant controls with no stimulation (n = 9); and D: bone-only controls (n = 3). For Condition B, the same transducers were used as in Condition A. The transducers were embedded into the base of the bioreactor using a thin layer of epoxy and a biocompatible polymer insert, which electrically isolated the transducers from the bone core.

In both treatment conditions (A and B), bone cores were stimulated daily for 30 min. The ultrasound stimulus replicated clinical low-intensity pulsed ultrasound systems (frequency 2 MHz, power density 30 mW/cm²) (Harrison and Alt, 2021). Control specimens were acoustically isolated from treated specimens. Treatment began after 24 h of acclimatisation to the ex vivo environment and lasted 21 days, at which point the culture was terminated. Fresh media (8 mL) was supplied to the specimens on the days 0, eight and 15, and retrieved on days 7, 14 and 21. The media on days 7–8 and days 14–15 contained fluorescent dyes for mineral labelling and was replaced after 24 h.

A two-colour assay (ThermoFisher L3224; Calcein-AM/Ethidium homodimer) was used to stain live/dead cells to determine cell viability, proliferation and distribution on the titanium implant surfaces: (i) after 7, 14, and 28 days of in vitro culture (n = 1 per timepoint) and (ii) after 21 days of ex vivo culture (n = 1 per condition). Specimens were incubated with the stains for 2 hours and washed three times with phosphate-buffered saline to remove the excess of residual dye. Confocal images (Leica Microsystems TCS SP5) were acquired from all specimens; widefield microscopy images (Zeiss Axio Observer) were acquired from ex vivo discs to determine ongrowth at the macroscopic level.

Fluorophores (Calcein, Sigma-Aldrich C0875; Alizarin Red S, Sigma-Aldrich A5533) were administered on the seventh and 14th days of stimulation through the media (50 μg/mL) for 24 h. Calcein (excitation: 470 nm; emission: 509 nm) and Alizarin Red S (excitation: 532 nm; emission: 620 nm) bind to the calcium ions in growing calcium phosphate/carbonate crystals, forming fluorescent complexes that label newly mineralised bone at each timepoint (van Gaalen et al., 2010). Post-incubation, all titanium discs were imaged using widefield microscopy (Zeiss Axio Observer). Mean fluorescent intensity on the disc for each stain was recorded and background fluorescence subtracted.

Titanium discs (n = 2 per in vitro timepoint; n = 3 per ex vivo condition) were fixed in 2.5% glutaraldehyde with 0.1 M phosphate buffer (pH 7.4) overnight, dehydrated through a series of ethanol solutions (20%-100%; 7 min per solution), and air-dried for analysis. The discs were then sputter-coated with chromium (thickness: 15 nm) and imaged using a scanning electron microscope (TESCAN MIRA) for signs of cellular ongrowth.

Supernatants from each timepoint in the ex vivo culture were retrieved and analysed for calcium concentration using a colorimetric assay (Sigma-Aldrich MAK022; n = 3 for each condition). The chromogenic reagent (o-Cresolphthalein complexone) forms a violet-coloured complex (absorbance: 575 nm) with Calcium ions (Ca²⁺) in alkaline solution. Absorbance readings were taken in triplicate for each sample, fresh media, and blank media without any reagent, as per manufacturer’s instructions.

The retrieved media from the ex vivo culture was analysed for alkaline phosphatase activity (ALP) using a fluorometric assay (Sigma-Aldrich MAK411; excitation: 360 nm; emission: 440 nm). ALP is a major regulator of osteoblast differentiation and bone extracellular matrix mineralisation, with a vital role in the supply of inorganic phosphate ions for the formation of hydroxyapatite, the primary mineral component of bone (Blair et al., 2017).

Cell culture media (ThermoFisher 11330057) was supplemented with phenol red–a non-toxic pH indicator that is colour-sensitive in the range 6.8–8.2 (yellow to fuchsia). The colour of the media was monitored during the culture to ensure near-physiological pH (∼7.4) was maintained. Secondary measures of pH were acquired directly from the media upon retrieval using indicator strips (VWR 315082P).

Bone specimens were fixed in 10% neutral buffered formalin overnight at room temperature, washed in 0.1 M phosphate-buffered saline (pH 7.4), and decalcified over 21 days in an acid solution (8% EDTA; 0.8% NaOH) with daily changes. After decalcification, specimens were washed with tap water, dehydrated through a graded series of ascending alcohol solutions and toluene, and embedded in paraffin wax. Sections perpendicular to the bone-implant interface were cut from the centre of each specimen using a rotary microtome (Leica Biosystems RM2235) to a thickness of 7 µm and mounted on microscope slides. Slides were deparaffinised, rehydrated, and stained with Haematoxylin and Eosin (H&E; BioGnost HE-RTU-100T), Masson-Goldner’s trichrome (BioGnost MGT-100T), Movat’s Pentachrome (BioGnost MOV-100T), and Picrosirius Red (Abcam ab150681), according to manufacturer protocols.

Decalcified sections were deparaffinised in Citroclear (Genta Medical CIT050; 2 × 15 min) and taken through secondary rinses in 99% industrial methylated spirit (Solmedia; IMS005; 2 × 30 min). Endogenous peroxidases were blocked by incubating sections in 1% hydrogen peroxide in absolute methanol for 10 min, and sections were then rinsed in cold-running tap water for at least 10 min. Antigen retrieval was carried out by incubating sections in citrate buffer (pH 6) overnight: sections were put into the buffer whilst still cold and were allowed to warm gradually. The following day, sections were rinsed in PBS and a PAP pen ring was drawn around the sample. Sections were blocked by incubating in 5% normal goat serum (Sigma N502L) in PBS for 1 h at room temperature; excess serum was tapped off; and sections were rinsed in PBS (2 × 2 min). Endogenous biotin, biotin receptors, and avidin binding sites were blocked with an Avidin/Biotin Blocking Kit (2BScientific SP-2001), which was used in accordance with the manufacturer’s instructions, and sections were again rinsed in PBS (2 × 2 min). Positive sections were incubated overnight with polyclonal rabbit anti-osteopontin primary antibody (Abcam ab231736; 1.25 μg/mL) in 5% goat serum; isotypic controls were incubated in Rabbit IgG at the same concentrations; and PBS controls were maintained in PBS overnight at 4°C. All sections were rinsed the following day in wash buffer solution (pH 7.4; 3 × 5 min), which was composed of 1% Triton X-100 (Millipore 648643) in PBS. Sections were then incubated in biotinylated goat anti-rabbit secondary antibody (Jackson ImmunoResearch 111–066-046) at 1:250 PBS dilution for 1 h at room temperature, followed by rinses in the wash buffer (3 × 5 min). The VectaStain ABC kit (2BScientific PK-4000) was then applied to all sections according to the manufacturer’s instructions, and sections were again rinsed in wash buffer (3 × 5 min). Sections were incubated in DAB substrate (2BScientific SK-4100; 2-10 min per sample) with microscopic control: the reaction was stopped with cold-running tap water when the target antigen was crisp brown and the background was clear. Nuclei were counterstained with Harris Haematoxylin for 20 s and blued in running tap water. Finally, sections were dehydrated, cleared, and mounted.

Slides were analysed quantitatively for osteopontin (OPN) expression: all OPN-positive and OPN-negative osteocytes within 500 µm of the implant-bone interface were counted. Osteopontin expression was reported as a percentage of total osteocytes. Osteopontin is an adhesive extracellular matrix protein with a prominent role in early osseointegration: it is secreted onto the damaged trabecular and implant surfaces (McKee et al., 2011) and becomes mineralised as ‘cement lines’ whilst devoid of collagen fibres. De novo bone formation proceeds from these cement lines on both the damaged trabeculae (distance osteogenesis) and the implant itself (contact osteogenesis).

Cell-seeded specimens were cultured in vitro for 28 days; high viability (>95%; Figure 4) was maintained in all samples extracted on days 7, 14, and 28. Few dead cells were observed. Additionally, uniform cell proliferation and cell distribution was observed on all timepoints during the culture.

Ex vivo specimens (n = 24) were successfully explanted from the medial and lateral porcine femoral condyles; 22 specimens were cultured for 22 days and used for analyses; two cultures (1 C and one D specimen) were terminated on the 10th day due to contamination and were not included in the analyses. Ongrown tissue of high viability was observed on the titanium discs of stimulated specimens (Figure 5A, B), whilst ongrown tissue of less expanse and less viability was observed on the titanium discs of the passive controls (Figure 5B, C). All active osseointegrative implant specimens (A) and all-but-one distant ultrasound specimens (B) had a weak degree of adhesion–enough for the bone to adhere to the implant under its own weight - at the implant-bone interface when the culture was terminated; no adhesion was observed in any of the passive controls (C). The media of active osseointegrative implant specimens was lighter upon retrieval (peach-coloured; pH ∼7.1) than distant ultrasound specimens and passive controls (red-coloured; pH ∼7.4); pH values were confirmed using indicator strips.

MC3T3-E1 murine preosteoblasts had proliferated over the entire rough titanium surface within 7 days (Figure 4A); the preosteoblasts had flattened on the surface with filopodia and there was early evidence of organic matrix deposition (Figure 4B). By 14 days, cells had continued to proliferate (Figure 4C) and organic structures spanning up to 200 µm between semi-sintered particles were observed (Figure 4D). High density cell clusters were observed after 28 days of in vitro culture (Fig, 4e), indicating osteoblast maturation, which further resulted in more expansive new extracellular matrix formation than previous timepoints (Figure 4F).

Specimens from the active osseointegrative implant (WFM; Figure 6A) exhibited mineralisation across almost the entire surface on the seventh day of culture (Calcein Green) with continued mineralisation by the 14th day (Alizarin Red S). Mineralisation at both time points was more than ten times higher than the passive controls (WFM; Figure 6C) as measured by mean fluorescent intensity (Figure 7; p < 0.05). By the end of the 3-week culture, there was near-total coverage of the titanium implant surface with extracellular matrix in active osseointegrative implant specimens (SEM; Figure 6D); continuous tissue formations had spanned distances up to 2 mm; and irregular woven bone structures had bridged between and enveloped semi-sintered particles on the implant surface (BSE; Fig, 6g). A large mature bone fragment was found integrating with one active osseointegrative implant specimen (BSE; Figure 8A); continuous bone formation emanating from the fragment had begun to integrate with the semi-sintered particles on the titanium implant surface (BSE; Figure 8B).

A lesser amount of day-seven mineralisation was observed in specimens that had been subject to distant ultrasonic stimulation and the passive controls (WFM; Figures 6B, C), dissipating to near-zero mineralisation by the 14th day. Extracellular matrix had also been deposited onto the titanium by the endpoint analysis in distant ultrasound specimens (SEM; Figure 6E); tissue formations were less expansive and appeared less mature compared to active osseointegrative implant specimens; but similar woven structures were observed (BSE; Figure 6H). Some cells and organic deposits were observed on the titanium implant surface of the static control specimens (SEM; Figure 6F), but the surface was devoid of any mature, expansive organic structures (BSE; Figure 6G).

ALP was expressed by all specimens throughout the 21-day culture period (Figure 9), though levels of ALP secretions decreased steeply from an initial peak in the first 7 days. In this initial period, ALP levels of the active osseointegrative implant specimens were 52% lower than those of the passive control specimens (p = 0.010); Distant ultrasound treatment had no effect on ALP secretions (p = 0.808); and ALP secretions without any titanium implant were increased by 55% (p = 0.048). There was no difference in ALP secretions beyond the initial 7-day period.

Calcium levels in the active osseointegrative implant specimens were significantly elevated versus the passive control at every timepoint (Figure 10; all p < 0.01). Calcium levels in the active osseointegrative implant specimens did decrease throughout culture, but remained higher than all other samples, which showed constant calcium levels, without any differences detected between them.

Histological analysis of the specimens treated with the active osseointegrative implant revealed bone fragments–cleaved from the bulk trabecular structure during the initial extraction–amidst newly-formed non-collagenous fibrous tissue (Figure 11). There was evidence of new bone matrix deposition (Figure 11; 12A) in direct apposition to the implant surface, growing from both the bulk trabecular structure and the resection debris fragments. Furthermore, bone remodelling (Figure 12: formation and resorption) was seen in the bulk tissue of the active osseointegrative implant specimens. Less osteogenic activity was observed at the surfaces open to the media and in contact with the polymer base of the bioreactor (Supplementary Figures S2-S3); similar resection debris and non-collagenous fibrous tissue formations were seen at these surfaces.

There were also similar bone fragments and non-collagenous tissue formations near the interface in the distant ultrasound specimens and the passive controls, but less osteoid seams were observed in apposition to the titanium implant surface; (Figures 13A, B). Remodelling was evident in the bulk tissue, the surfaces open to the media, and the surfaces in contact with the polymer base of the bioreactor in the distant ultrasound and passive control specimens (Figures 13C, D). There was no indication of mass resorption of cancellous bone in any of the specimens.

Additional histology images are available as Supplementary Material.

Evidence of osteopontin expression was observed in all specimens (Figure 14). At the bone-implant interface, the number of osteopontin-expressing osteocytes was 47 percentage points (pp) higher in the active osseointegrative implant compared to the passive implant controls (2.5 times greater, p < 0.005; Figures 14A, G; Figure 15). Similarly, the distant ultrasound led to a 32 pp higher osteopontin expression compared to the passive implant controls (2 times greater, p < 0.05; Figures 14E,G; Figure 15). Comparing the two stimulated cases, the active osseointegrative implant led to a 15 pp higher osteopontin expression versus the distant ultrasound (1.2 times greater, p < 0.05; Figures 14A, E; Figure 15). Osteopontin was also expressed by osteocytes in the bulk of tissue of all specimens, with further expression by osteoblasts and osteoclasts in remodelling zones (Figure 14B, F, H).