Combination of polyetherketoneketone scaffold and human mesenchymal stem cells from temporomandibular joint synovial fluid enhances bone regeneration

Therapies using human mesenchymal stem cells (MSCs) combined with three-dimensional (3D) printed scaffolds are a promising strategy for bone grafting. But the harvest of MSCs still remains invasive for patients. Human synovial fluid MSCs (hSF-MSCs), which can be obtained by a minimally invasive needle-aspiration procedure, have been used for cartilage repair. However, little is known of hSF-MSCs in bone regeneration. Polyetherketoneketone (PEKK) is an attractive bone scaffold due to its mechanical properties comparable to bone. In this study, 3D-printed PEKK scaffolds were fabricated using laser sintering technique. hSF-MSCs were characterized and cultured on PEKK to evaluate their cell attachment, proliferation, and osteogenic potential. Rabbit calvarial critical-sized bone defects were created to test the bone regenerative effect of PEKK with hSF-MSCs. In vitro results showed that hSF-MSCs attached, proliferated, and were osteogenic on PEKK. In vivo results indicated that PEKK seeded with hSF-MSCs regenerated twice the amount of newly formed bone when compared to PEKK seeded with osteogenically-induced hSF-MSCs or PEKK scaffolds alone. These results suggested that there was no need to induce hSF-MSCs into osteoblasts prior to their transplantations in vivo. In conclusion, the combined use of PEKK and hSF-MSCs was effective in regenerating critical-sized bone defects.


Osteogenic characterisitic of hSF-MSCs on PEKK in vitro. The alkaline phosphatase (ALP)
activity of hSF-MSCs seeded on PEKK scaffolds was higher than that on TCP, in osteogenic medium at day 4 and day 7 (p < 0.05) (Fig. 3a). At day 14 and 21, no statistically significant difference was found between the osteogenically-induced hSF-MSCs seeded on PEKK scaffolds (i.e. PEKK + OS group) versus osteogenically-induced hSF-MSCs cultured on plastic (TCP + OS group). Results of quantitative reverse transcription polymerase chain reaction (qRT-PCR) (Fig. 3b) demonstrated that osteogenically-induced hSF-MSCs on PEKK scaffolds (PEKK + OS) showed a significantly higher upregulation for OPN (9.6 folds), OCN (6.8 folds), COLІA1 (1.8 folds), and RUNX2 (2.2 folds) than osteogenically-induced hSF-MSCs cultured on plastic (TCP + OS) at day 21 of culture. Control groups (PEKK + SF and TCP + SF) were detected with negligible expression of the above-mentioned genes, and the differences between these two groups in ALP activity and gene expression were statistically not significant.
In vivo evaluation of hSF-MSCs seeded on PEKK scaffolds. Using micro-computed tomography (micro-CT) and 3D image reconstruction, the regenerated bone within the critical-sized defects could be measured. The PEKK material/implant was radiolucent and could be discriminated from the newly-formed bone. Micro-CT analysis revealed that bone regeneration occurred in all four groups at 4 and 12 weeks post-surgery, but at different rates (Fig. 4g,h). Detailed and quantitative bone growth measurements were shown in Fig. 5. Each original critical-sized defect margin was reconstructed using a hypothetical red-dotted circle of 8 mm in SCiENTiFiC RepoRTs | (2019) 9:472 | DOI: 10.1038/s41598-018-36778-2 diameter (Fig. 5a). Bone regeneration started at the margin of the surgically-created osseous defect and gradually moved toward the center of the defect. Fig. 5b showed cross-sectional images at three different depths of the defects (8 mm diameter by 1.5 mm height). These images demonstrated that new bone regenerated faster at the top (superficial) of the defect than at the bottom of the defects. Defects implanted with hSF-MSCs seeded on PEKK scaffolds (PEKK + SF) had the highest volume of regenerated bone at 12 weeks (Fig. 5a,b). These observations were supported by measurements of bone volume normalized to the total volume of the defect (BV/TV) indicating that, at 12 weeks, hSF-MSCs seeded on PEKK (group PEKK + SF) had ~20% bone volume versus 9-10% for the other 3 groups ( Fig. 5c; p < 0.05). In terms of trabecular structure, the cross-sectional images of the upper, middle, and lower portions/depths of the defects showed that bone grew in an organized pattern around the connecting pores of the scaffolds for the PEKK + SF group (Fig. 5b). The distribution of the regenerated bone in the PEKK + SF group was relatively uniform across the scaffold depth. No significant differences in trabecular thickness (Tb.Th) were found among the four groups (Fig. 5d). However, trabecular numbers (Tb.N) in the PEKK + SF group was higher than the other groups (0.80 ± 0.05 mm −1 , p < 0.05) (Fig. 5e). This meant that PEKK scaffolds with hSF-MSCs regenerated more bone tissue in vivo than scaffolds without seeded cells. Additionally, hSF-MSCs provided more bone regeneration than osteogenically-induced hSF-MSCs (PEKK + OS group). Histological staining of the explanted calvarial specimens was performed to examine the mineralization status of the newly formed bone in the critical-sized defects (Fig. 6). Four weeks and 12 weeks post-surgery, von Kossa staining identified very little mineralization in defects without scaffolds (Control group) or with cell-free scaffolds (PEKK group). Conversely, a considerable amount of mineralized bone tissue was found between the microchannels of PEKK seeded with hSF-MSCs (PEKK + SF group). Bone marrow-like tissue was observed within the new bone. Larger amounts of newly formed mineralized bone were observed at 12 weeks when compared to 4 weeks. There was relatively less new bone formed in the PEKK + OS group, and the newly-formed bone was located at the periphery of the defect.
The mechanical push-out test was used to measure the functional strength of the regenerated bone. It was observed that PEKK + SF (87.6 ± 5.4 N, p < 0.05) and PEKK + OS (89.9 ± 4.0 N, p < 0.05) groups had significantly greater push-out strengths than the PEKK group (Fig. 7). No statistical difference was found between PEKK + SF and PEKK + OS. The positive control group was the native intact calvarial bone (114.9 ± 4.4 N), while the lowest strength was observed in PEKK group without cells (68.7 ± 0.7 N).

Discussion
Cells used in this study were isolated from human synovial fluid of TMJ from TMD patients. In our pilot studies, hSF-MSCs of healthy volunteers were obtained and isolated, but these were not used in the current study. Those MSCs from healthy donors shared similar flow cytometric characteristics to MSCs from TMD patients. However, the main difference was that hSF-MSCs could only be isolated from ~16.7% of healthy donors TMJ synovial fluid 45 . Also, the number of colony forming units (CFUs) of hSF-MSCs from healthy donors at plating (passage 0) was much lower, and this resulted in an insufficient cell number for our subsequent experiments. Another reason for using hSF-MSCs from TMD patients was the availability of the synovial fluid following the arthrocentesis treatment; while synovial fluid collection would not have been indicated for healthy patients (i.e. without TMD). Furthermore, our long-term goal is to use autologous hSF-MSCs to reconstruct the mandibular condyle of patients suffering from severe TMJ osteoarthritis, and thus this was another reason for selecting hSF-MSCs from TMD patients for this study. hSF-MSCs displayed typical MSCs characteristics based on their cell surface markers (CD73, CD90, CD105, CD45, etc), multipotency, and self-renewal capacity (Fig. 1). In our previous study about hSF-MSCs 34,46-48 , Flow cytometric experiments were done from the third to sixth passages of hSF-MSCs. At passage 6 th , more than 96% of hSF-MSCs were positive for MSC markers and less than 2% of hSF-MSCs were positive for (CD45, CD34, CD11b, CD19, and HLA-DR). These flow cytometric results were not statistically different when compared to those cells at passages 3 rd to 5 th . CD44 is a receptor for hyaluronic acid, which is a major component of articular surfaces 49 . Previous reports focused on the chondrogenesis function of hSF-MSCs and their potential applications in cartilage repair 24,26,35 . hSF-MSCs were also reported to have comparable biological characteristics with BMSCs and possessed a greater osteogenic potential than other MSCs obtained by non-invasive procedures, such as the dental pulp or exfoliated deciduous teeth stem cells 23,26 . However, their contributions in regenerating bone defects have barely been investigated in the literature. By testing the osteogenic ability of hSF-MSCs on 3D-printed scaffolds both in vitro and in vivo, our study proposed an additional source of non-invasively harvested human MSCs for bone regeneration as well as an optimal source of autologous cells to repair bone defects of TMD patients.
The first finding of this study was that the surface of 3D-printed PEKK scaffolds was biocompatible for cell attachment and growth. PEKK belongs to the polyaryletherketones family (PAEKs) which are materials with mechanical properties that can coexist with human bone 37 . PEKK is considered a promising material to replace metals and ceramics currently used in orthopedics 36,38,50 and in dentistry 42,43,[51][52][53] . However, there are only a few studies 36,39 reporting the effect of PEKK on stem cells. This study tested a 3D-printed PEKK scaffold with 750-1000 µm interconnecting channels. The laser sintering process resulted in a rough and porous surface, when observed at the SEM level (Fig. 2a,d). This surface topology might have favored cell attachment and their   osteogenic differentiation. Although hSF-MSCs proliferated slower on PEKK than on tissue culture plastic (TCP; 2D culture), cells on PEKK showed multiple filopodia and lamellipodia (Fig. 2b,e,f). Both structures are composed of actin fibers and transmembrane adhesion complexes. Filopodia act as sensors to explore the extracellular environment whereas lamellipodia can induce cell migration by supporting traction forces from the cell actomyosin network 54,55 . These images (Fig. 2b,e,f) suggested that the surface of our 3D-printed PEKK had adhesive interactions with the seeded cells and might modulate cellular signalings toward osteogenesis.
The second finding of this study was that the PEKK scaffold by itself could not induce osteogenesis in hSF-MSCs but required the osteogenic cultured media. As expected, hSF-MSCs cultured on either PEKK or TCP, and in non-osteogenic media did not show an osteogenic differentiation. However, when hSF-MSCs were cultured in osteogenic media, this cultured condition enhanced the osteogenic differentiation of hSF-MSCs as the result shown in groups PEKK + OS and TCP + OS (Fig. 3). ALP activity levels were higher in hSF-MSCs grown on PEKK than on TCP, as early as day 4 and day 7 (Fig. 3a). Also, an upregulation of osteogenesis-related genes was detected at day 21 for cells grown on PEKK (Fig. 3b). ALP is a transient early marker of osteogenic differentiation for MSCs 56 , indicating their ability to differentiate into osteoblasts 57 . An initial rise in ALP activity, with a peak at day 14, was observed in PEKK + OS and TCP + OS groups, which was in line with several previous studies [57][58][59] . This observation was confirmed with the downregulation of the ALP gene at day 21. Unlike ALP, other osteogenic genes such as OPN, OCN, COLІA1, and RUNX2, all exhibited a significant upregulation when cultured in osteogenic media for 21 days. These markers were commonly used in many other studies to demonstrate the osteogenic potential of MSCs, and whether the MSCs were at their early or late stage of osteogenic differentiation. To obtain an overall ability of hSF-MSCs in maintaining their osteogenic phenotypes when cultured on PEKK scaffolds, the mRNA levels of these markers were measured at day 21, which was a widely accepted timepoint in the literatures. RUNX2 encodes a key transcription factor with an essential role in the commitment of MSCs towards osteoblasts 60 . However, its expression pattern during osteogenesis varies. Huang X et al. 61 reported that the pattern of RUNX2 expression fluctuated during osteogenic differentiation of MSCs. The expression of RUNX2 was found to increase initially and then to decrease in human dental pulp MSCs, while it remained consistently high in human bone marrow MSCs during their osteogenic/odontogenic differentiation 61 . It was also reported that the mRNA level of RUNX2 correlate positively with the level of OCN during osteogenesis 62 . Therefore, we measured both the osteogenic late markers and their upstream regulator (RUNX2) to allow us to detect their correlations. The mRNA level of RUNX2 was significantly higher in the PEKK + OS group. Not surprisingly, the expression of OCN on PEKK was 5.8 folds higher than the TCP + OS group, suggesting an enhanced mineralization of the extracellular matrix (ECM) and more active osteoblasts. ECM provides a template for mineralization, which mainly consists of collagens (especially collagen type І) 63 . A higher expression of COLIA1 indicated a more active formation of ECM on PEKK relative to TCP. The highest upregulated gene that we measured during hSF-MSCs culture was OPN. This indicated a proliferation of pre-osteoblasts in vitro as OPN is secreted prior to matrix mineralization 64 . The increased gene expression levels of OPN, OCN, COLІA1, and RUNX2 implied that hSF-MSCs cultured on PEKK 3D-scaffolds were induced into an accelerated osteogenic differentiation in vitro. Such benefits of PEKK (i.e. increased mRNA levels toward osteogenesis) along with its biocompatibility and biomechanical properties confirmed its use as an adequate scaffold for bone tissue engineering.
The third finding of this study was that (naïve/non-differentiated) hSF-MSCs were more effective in regenerating bone when compared to osteogenically-induced hSF-MSCs. Our in vivo results demonstrated that the volume of bone formed (~20%) by hSF-MSCs was twice higher than the bone volume (~10%) of osteogenically-induced hSF-MSCs. This was a surprise to us because some researchers had previously reported that it was better to osteogenically-induced MSCs prior to their in vivo transplantation to increase bone regeneration 17,65 . We hypothesized that the undifferentiated/naïve hSF-MSCs possessed a greater cell plasticity and responded to the rich cocktail of healing stimulus in the critical-sized defects 16,17,66 . Nevertheless, the decision to transplant naïve/non-differentiated versus osteogenically-induced MSCs to increase bone formation remained controversial 16,17,65,67 . This discrepancy might be due to studies using MSCs from different tissues, different scaffold materials, and different animal models. Results from this study indicated that there was no need to osteogenically-induced human MSCs from TMJ synovial fluid when cultured on PEKK prior to their in-vivo implantation into the rabbit cranial critical-sized defect model.
Last but not the least, our in-vivo results supported the use of PEKK as a scaffold for hSF-MSCs. Our micro-CT and histological results (Figs 5 and 6) revealed a distinct increase in bone regeneration when either hSF-MSCs (PEKK + SF) or osteogenically-induced hSF-MSCs (PEKK + OS) were seeded on PEKK scaffolds. According to our histological analysis, non-differentiated hSF-MSCs (from treatment group PEKK + SF) formed larger and more matured osteoid, while osteogenically-induced hSF-MSCs (from treatment group PEKK + OS) produced fewer and smaller mineralized tissue, which was mainly located to the periphery/border of the critical-sized defect (Fig. 6). However, newly-formed peripheral bone still contributed to the integration of PEKK into the critical-sized defect because results from the Push-Out Strength test of both PEKK + SF and PEKK + OS groups were higher than those from the PEKK group (that had no seeded hSF-MSCs) (Fig. 7). Histologically, we did not detect an increased number of inflammatory cells when PEKK was transplanted with hSF-MSCs. This result suggested no (observable) immunological rejection of the xenogeneic (human) hSF-MSCs in the rabbit model. Our observations were in agreement with earlier studies supporting the immunomodulatory properties of human MSCs 68 . Still, the rabbit calvarial site might be more immune-privileged 69 and we could not exclude this possibility. On the contrary, Adamzyk et al. 36 found no added effect of transplanted ovine BMSCs on bone regeneration in sheep calvarial defects. Adamzyk and colleagues reported, in their flow cytometric data, a lower percentage of MSCs from ovine bone marrow. Also, the sheep model mounted an immunological response to transplanted allogeneic cells 36 . These two factors might explain a decreased potential of bone regeneration observed by these authors. Although the bone defect was not completely healed within the 12-week duration of this study, the regenerated bone in the group that was implanted with PEKK seeded with hSF-MSCs demonstrated a functional recovery. Additional studies with a longer follow-up period are required to investigate whether complete closure of the bone defect would be established with our proposed grafting material combined with hSF-MSCs. Also, to further confirm the benefits of PEKK scaffolds, additional studies comparing PEKK to other typical bone scaffolds will be needed.

Conclusions
In conclusion, hSF-MSCs from the TMJ could be easily harvested with minimal invasiveness to the patients. hSF-MSCs possessed high multipotency and could be an alternate cell source to BMSCs for bone regeneration, as well as an optimal autologous cell source for TMJ repair. The combination of hSF-MSCs and PEKK (a biocompatible bone-mimetic polymer) resulted in an increased osteogenic ability, both in vitro and in vivo. The combined implantation of PEKK and hSF-MSCs is a promising therapy to regenerate large bone defects.
Methods PEKK Scaffold fabrication. PEKK scaffolds were designed using SOLIDWORKS ® (version 2015, Vélizy-Villacoublay, France). The porosity was set at approximately 60% with a strut size of 1.15 mm, resulting in a printed pore size of 750-1000 μm. With an EOSINT P 800 printer (EOS GmbH, Germany), the PEKK scaffolds were printed using a laser sintering technique (Oxford Performance Materials, CT, USA) and then cut into cylinders (8 mm diameter and 1.5 mm thickness).
Isolation and culture of hSF-MSCs. hSF-MSCs were isolated from the synovial fluid aspirated from the TMJ cavity of TMD patients undergoing arthrocentesis 34 . The synovial fluid sample was centrifuged at 300 g for 5 min and cells were resuspended and cultured in standard culture media, which is complete alpha-minimum essential medium (α-MEM; Gibco BRL, NY, USA) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin/streptomycin (Gibco), at 37 °C in 5% humidified CO 2 . The cultured media was replaced every 2-3 days until 80% cell confluence. Cells were then detached with 0.25% trypsin EDTA (Gibco), resuspended in standard culture media, and replated at a density of 5000 cells/cm 2 for expansion. Cells used for experiments in this study were from the 3 rd to 5 th passages.
Cell Proliferation on scaffolds. The PEKK scaffolds were sterilized by high-pressure steam (Yamato SM 300, Japan). The scaffolds were seeded with hSF-MSCs at a cell concentration of 5.0 × 10 4 cells/mL. hSF-MSCs were also seeded, using the same concentration and volume, on tissue culture plastic (TCP) as controls. Cell proliferation was evaluated at 1, 3, 5, and 7 days by the Alamar Blue assay (Invitrogen, Carlsbad, CA). Cell/scaffold complex were incubated in 0.5 mL of complete culture medium with 10% (v/v) Alamar blue solution for 3 h. The fluorescence intensity of the culture media at 570/585 nm was measured in triplicates using a microplate reader (SpectraMax ® M2, USA).

SEM.
After culturing hSF-MSCs on PEKK for 7 days, the cell/scaffold complexes were collected, washed three times with PBS, and fixed in 2.5% (v/v) glutaraldehyde for at least 24 h at 4 °C. Then, the samples were rinsed three times, followed by serial dehydration in 30%, 50%, 70%, 90%, and 100% ethanol. Hexamethyldisilazane was used before the samples were air-dried and sputtered with gold (Leica EM ACE600, Wetzlar, Germany). The surface morphology was analyzed by FEG Quanta Scanning Electron Microscope (Inspect F50, FEI Company, OR, USA) at 5 kV to image. ALP activity. Cell suspension, at a density of 2 × 10 5 cells/mL, was evenly added onto PEKK/TCP dropwise. Four groups were tested in vitro: (1) PEKK + OS (PEKK seeded with hSF-MSCs in osteogenic media), (2) PEKK + SF (PEKK seeded with hSF-MSCs in standard culture media), (3) TCP + OS (TCP seeded with hSF-MSCs in osteogenic media), and (4) TCP + SF (TCP seeded with hSF-MSCs in standard culture media). After 3 days of culture, the standard culture media in the OS groups was changed into an osteogenic media and renewed twice a week while the SF groups were renewed with standard medium at the same frequency. The cellular ALP activity was detected after 1, 4, 7, 14, and 21 days 70 For anesthesia, acepromazine (0.75 mg/kg), xylazine (5 mg/kg), and ketamine (20-35 mg/kg) were injected intramuscularly followed by isoflurane (1.5-3%) delivered in 100% oxygen (flow rate 0.8-1.5 L/min for induction and 0.4-0.8 L/min for maintenance) through endotracheal intubation inhalation 72 . Ancef (12 mg/kg) was given intramuscularly prior to incision. A 4-cm long skin incision over the linea media was made, and the tissue was dissected to adequately expose the calvaria. Using an electric hand piece with a trephine bur, four circular bicortical defects (8 mm in diameter) were created symmetrically to the midline of the calvaria in the parietal bones under constant saline irrigation (Fig. 4a,b). After the transplantation procedure (Fig. 4c), the layers were closed separately with resorbable 4-0 sutures. Baseline computed tomography images (at day 0) of the rabbits were taken immediately after the surgery (Fig. 4d). Buprenorphine (0.05 mg/kg) was administered for 48 h to reduce post-operative pain. Rabbits were sacrificed either at the 4th (Fig. 4e) or 12th week (Fig. 4f) postoperatively, and their calvaria were removed for further analyses.
Micro-CT analysis. The explanted specimens were fixed with 10% buffered formalin and then stored in 70% ethanol. Micro-CT was performed to quantify the volume of bone formation within the defects using the Skyscan 1072 scanner (Bruker-MicroCT, Kontich, Belgium) at 55 kV, 181 µA, and a spatial resolution of 15.96 μm pixel through an Aluminum filter 72 . The micro-CT data were reconstructed by NRecon software (Bruker-MicroCT), realigned to the primary axes of the cylindrical defect space by DataViewer (Bruker-MicroCT), and the 3D image stack was analyzed by CTAn (v1.13; Materialise, Leuven, Belgium). A cylindrical shape of 8 mm in diameter and 1.5 mm in height was selected at the center of the bone tunnel/defect as the volume of interest. The same threshold was used among all groups. Bone volume and trabecular pattern were analyzed afterwards, followed by 3D visualization using CTVol (v13.0; Materialise).
Histology. Samples were dehydrated through a series of ethanol concentrations and then embedded in methyl methacrylate. Thin undecalcified tissue slices (7 µm) were cut perpendicularly to the bone surface using a rotary microtome (Leica RM2255, Germany). These slices were stained with von Kossa and counterstained with van Gieson to distinguish the mineralized non-mineralized tissues.
Push-out test. At week 12 post-operatively, the explanted specimens were stored in cold PBS and tested by a push-out machine (TestResources 313, Shakopee, USA). Parietal bone samples without surgical defects were used as positive controls. The specimens were placed on a customized fixed rigid platform with a 9 mm hole and a 6 mm flat indenter centered in the defect area. The surface of the defect was aligned parallell to the surface of the indenter. The force and displacement were recorded when the indenter moved into the repair area at a speed of 5 mm/min until the specimen failed/broke. XY Software (Version 4.00.10, Shakopee) was used to analyze the data and identify the peak compressive force for each set of data.
Statistical analysis. All data were analyzed using SPSS 19.0 software. Statistical analysis was conducted using one-way analysis of variance followed by the Bonferroni test for multiple comparisons among the experimental groups. These data were presented as mean ± standard error of the mean (SE). Value of p < 0.05 were considered significant.

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding authors.