Multiple channels with interconnected pores in a bioceramic scaffold promote bone tissue formation

Insufficient nutrition exchange and limited transportation of blood supply in a porous only scaffold often hinder bone formation, even though the porous scaffold is loaded with cells or growth factors. To overcome these issues, we developed a cell- and growth factor-free approach to induce bone formation in a critical-size bone defect by using an interconnected porous beta-tricalcium phosphate (β-TCP) scaffold with multiple channels. In vitro cell experimental results showed that multiple channels significantly promoted cell attachment and proliferation of human bone marrow mesenchymal stem cells, stimulated their alkaline phosphatase activity, and up-regulated the osteogenic gene expression. Multiple channels also considerably stimulated the expression of various mechanosensing markers of the cells, such as focal adhesion kinase, filamentous actin, and Yes-associated protein-1 at both static and dynamic culturing conditions. The in vivo bone defect implantation results demonstrated more bone formation inside multiple-channeled scaffolds compared to non-channeled scaffolds. Multiple channels prominently accelerated collagen type I, bone sialoprotein and osteocalcin protein expression. Fluorochrome images and angiogenic marker CD31 staining exhibited more mineral deposition and longer vasculature structures in multiple-channeled scaffolds, compared to non-channeled scaffolds. All the findings suggested that the creation of interconnected multiple channels in the porous β-TCP scaffold is a very promising approach to promote bone tissue regeneration.


Results
Cell attachment efficiency and proliferation on the scaffolds. Two geometrically different porous β-TCP scaffolds without channels and with five paralleled straight channels of 1 mm in diameter were fabricated in this study (Fig. 1a, b). The diameter of the scaffold is about 8 mm, and the height is around 5.5 mm. The pore size is around 550 μm. The interconnected porosity is approximately 80% 17 . From SEM images, we can see that the pores in the non-channeled scaffold are interconnected (Fig. 1c), and microCT images also showed the tortuosity of interconnected pores (Fig. 1d, e). After sintering, the porous scaffold was mainly composed of crystalline phased β-TCP according to the previous study 22 .
hBMSCs were rotationally seeded onto the scaffolds, and the number of attached cells per unit area were measured respectively after 4 and 16 h. It showed that the number of attached cells was increased along with time from 4 to 16 h, and multiple-channeled scaffold significantly increased cell attachment, compared with non-channeled scaffolds (Fig. 2a).
To investigate the effect of multiple channels geometry on cell proliferation, we used dsDNA concentrations of hBMSC to indirectly index cell numbers. The dsDNA concentrations of total hBMSCs grown on both types of scaffolds under two culture conditions (Fig. 2b) were measured. Under the static condition (Fig. 2c, i), hBMSCs proliferated on both non-channeled and channeled scaffolds from 3 to 7 days. There is no significant difference in the dsDNA concentrations were detected between the two groups within 7 days. However, the dsDNA concentration of hBMSCs on non-channeled scaffold decreased at 14 days compared to it at 7 days, which in contrast, slightly increased on the channeled scaffold (Fig. 2c, i). The channeled scaffold resulted in a considerable higher dsDNA concentration at 14 days. Nevertheless, this cell proliferation pattern was not applied to that in the dynamic culture condition. Under the dynamic condition (Fig. 2c, ii), hBMSCs on both types of scaffolds continuously proliferated during the experimental period. Remarkably, the cell proliferation was significantly stimulated by multiple-channeled scaffolds compared to non-channeled scaffolds (p < 0.05). From 7 to 14 days, this difference of cell proliferation between the two groups became even more pronounced (p < 0.01). From these results, it demonstrated that multiple channels can significantly stimulate cell proliferation within 14 days. More interestingly, the absolute values of the dsDNA concentration under dynamic condition (~ 1000 ng/mL at day 14) were all significantly higher than those under static condition (~ 100 ng/mL at 14), regardless of time and scaffold types, even though the initial cell number that we seeded on the scaffolds was consistently same. This interesting result implied that the dynamic culture itself can promote cell proliferation significantly. In addition to that, multiple channels can stimulate that even further. hBMSC differentiation. The level of alkaline phosphatase (ALP) was measured as one way to explore the effect of multiple channels on cell differentiation in vitro. Results showed that, regardless of the culturing conditions, the ALP concentration of hBMSCs on both types of porous scaffolds increased with time and peaked at 7 days, and then decreased afterwards (Fig. 2d). There is no significant difference between the two types of scaffolds at most time points except for 7 days, at which, the ALP activity was significantly stimulated by multiple channels. These results demonstrated that multiple channels promoted early-stage cell differentiation of hBMSCs.
Osteogenic-related gene expression was performed and measured by real-time PCR to further investigate the effect of multiple channels on hBMSC differentiation. Results indicate that multiple channels up-regulated the expression of Runx2 gene by 1.6-fold under the static condition, and 2.1-fold under the dynamic condition after 7 days (Fig. 3a). However, after 14 days, no expression level variations were observed under both conditions. The expression level of alp was significantly promoted by multiple channels by 1.7-fold in the static condition, and 2.2-fold in the dynamic condition after 7 days (Fig. 3b). Both non-channeled and multiple channeled scaffolds significantly promoted alp expression under the dynamic condition compared to the static condition after 14 days. For bsp, the same trend as alp was observed at 7 days (Fig. 3c). At 14 days, under the static condition, no statistically significant difference was shown between the two groups (Fig. 3c). By contrast, under the dynamic condition, the multiple channels statistically enhanced bsp expression by 3.2-fold. The expression of oc was significantly promoted by multiple channels in the static condition at 7 days. At 14 days, oc expression remained almost the same level on both scaffolds in both culturing conditions, and there is no significant difference in the expression between the conditions and groups (Fig. 3d). In general, these results together indicated that multiple channels have the capacity to remarkably stimulate osteogenic-related genes expression of hBMSCs in vitro.
Expression of mechanosensing markers: FAK, F-actin, and YAP-1. To investigate whether the interconnected channel-pore structure with the flow of dynamic fluid played a role in initiating the mechanotransduction pathways of hBMSCs, which promoted cell differentiation, the expression of various mechanosensing markers, FAK, F-actin, and YAP-1 were observed. F-actin staining of hBMSCs under the static condition shows that the actin fiber was not fully and strongly polymerized yet after 7 days, regardless of the scaffold   (Fig. 4a). However, the actin skeleton started to show highly polymerized and widely stretched after being cultured for another 7 days. The fibers are thinner and weaker on non-channeled scaffold, compared with that on multiple-channeled scaffold, which is thicker and stronger. The dynamic culturing condition remarkably intensified F-actin. More F-actin filaments started to widely spread on multiple-channeled scaffolds after 7 days. Whereas, compared to that, less F-actin on non-channeled scaffold were observed. After 14 days, F-actin filaments were highly strengthened and intensively bundled with each other on both types of scaffolds. Furthermore, multiple-channeled scaffold supported on strengthening fibrous bundle, forming more compact fibers inside cells. Semi-quantitative evaluation demonstrates that multiple channels significantly promoted the polymerization of F-actin after 14 days under the static culture condition (Fig. 4b). Dynamic condition accelerated the progress, and multiple channels started to stimulate F-actin polymerization (Fig. 4c). This significant stimulation of the intensity of F-actin fiber by multiple channels was continued from 7 days till 14 days in the dynamic condition (Fig. 4d). For FAK, there are very few cells that expressed FAK on both types of scaffolds at 7 days, but there is relatively more FAK expression observed on the multiple-channeled scaffolds after 14 days under the static condition (Fig. 5a). However, interestingly, unlike static condition, more and stronger FAK signals were captured from the multiple-channeled scaffolds under the dynamic condition, whereas only a few cells positively expressed FAK on the non-channeled scaffolds after 7 days (Fig. 5b). After 14 days, there is no significant difference in the expression level of FAK on both types of scaffolds. Semi-quantification analysis shows that, under the static condition, though a slightly more FAK-positive cells exhibit at certain regions of multiple-channeled scaffolds, there is no statistically significant difference between two types of scaffolds on both two time points (Fig. 5c). Surprisingly, the combination of dynamic fluid and multiple channels together significantly enhanced the expression of FAK (p < 0.01) compared to non-channeled scaffold on 7 days (Fig. 5d). All these results imply that the multiple channels facilitate the dynamic fluid shear transmission inside scaffolds, which upregulates and activates the FAK expression.
YAP-1 was probed and there are only few cells positively expressing the protein on both types of scaffolds after 7 days under the static culture condition (Fig. 5e). They are weakly expressed and still mainly distributed in the cytoplasm area. However, after 14 days, multiple channels significantly stimulated YAP-1 expression in terms of the positively stained cell number, as well as the expressing strength and location inside the cells, which translocated from cytoplasm to nuclei, when compared with non-channeled scaffolds (Fig. 5e). Dynamic fluid accelerated this process and demonstrated the phenomenon earlier, in which condition, YAP-1 was strongly expressed in cell nuclei as well as cytoplasm area on multiple-channeled scaffold at 7 days, rather than on nonchanneled scaffold (Fig. 5f). This significant difference disappeared after 14 days. The expressions of YAP-1 in cell nuclei and cytoplasm on channeled scaffold did not show significant increment. Semi-quantification shows that multiple channels slightly promoted YAP-1 expression after 14 days in the static condition (Fig. 5g). Under www.nature.com/scientificreports/ the dynamic condition, the expression was significantly stimulated by multiple channeled scaffolds after 7 days (Fig. 5h). All these results demonstrate that the geometry of multiple channels can significantly activate YAP-1 expression and translocation. The dynamic culturing condition could accelerate this mechanism.
In vivo implantation and X-ray morphologies. Scaffold implantation was carried out on a rat calvarial defect model (Fig. 6a) and characterized through X-ray to evaluate the function of multiple channels on bone regeneration and mineralization in vivo. After 3 months, the implanted scaffolds with surrounding native tissues was scanned by X-ray (Fig. 6b, i). Results showed that both types of β-TCP scaffolds promoted bone formation. The images revealed that there was no healing in the non-implantation group, and has a defective area remained with no new bone filled in (Fig. 6b, ii). In the non-channeled group, there are still some remained defective areas which were surrounded by the implant, as well as some pore area within the implant without any new bone tissue fillings (Fig. 6b, iii, N(i)). In the multiple-channeled group, it is showed that all the created bone defects, the channels area, and the pore areas were filled with new formed bone ( Fig. 6b, iv, M(i)). Additionally, the radiopacity of regenerative bone is comparable to that of native calvarial bone in terms of the degree of mineralization (grey values of the image).

Histological analysis. Lateral section of the central part of the implant/defect samples was stained by H&E
staining. Results showed that there is a thin layer of 0.3 mm of soft tissue formed in the bone defect area, without any compact bone observed in the non-implant group. However, new bone tissue formed throughout the other two types of implants (Fig. 6c). After measuring the height of the new formed tissue in the defect zone, we found that the height of the bone defect (1.5 mm) was maintained by the implanted scaffolds after 12 weeks (Fig. 6d).
We further performed H&E staining on the transversal sections of the proximal, middle, and distal segments from the dura side to the top of the implanted scaffolds (Fig. 7a). We found that there is extensive bone growth in all the layers of both non-channel and multiple-channeled scaffolds, but the density of the formed bone is different. In the bottom layer, the density and dense area of the formed bone tissue are similar in the two groups ( Fig. 7b). However, with the location of the layers became further away from the dura layer, the formed bone tissue in the two groups became different. There are some small islands of fibrous bone tissue observed in the middle segments of the non-channeled scaffold, but larger areas of compact bone were formed inside multiple- www.nature.com/scientificreports/ channeled scaffolds (Fig. 7b). The semi-quantitative analysis shows that the ratio of compact bone to total tissue in the middle layer of the channeled scaffolds is much higher than that in the non-channeled scaffolds (Fig. 7c).
In the top layer, the density and dense area ratios in the two group became similar again. These results imply that new tissues grew easily from two ends (bottom layer and top layer, where tissues contacted the two ends of the scaffolds) into the scaffold. However, the new bone tissue may not easily reach to the middle part of the non-channeled scaffold, but the channels facilitated the ingrowth of new bone tissue to the middle part even the whole area of the channeled scaffold. These H&E staining results illustrated that multiple channels in porous β-TCP scaffold stimulated new bone formation and maturation. The formation of stimulated new bone tissue may be related to the effective vascularization in the channeled scaffold. To prove this point, we used immunohistochemical staining of angiogenic marker CD31 to observe the formation of blood vessels. Angiogenic marker CD31 staining showed that only small tubular-like blood vessels were observed on both non-channeled scaffold and no-implant group, but there was longer and tissue penetrative vasculature structure formed in the multiple-channeled scaffolds (Fig. 7d). We also observed that there are erythrocytes in the lumen of blood vessels (magnified image). Although there is no statistically significant difference in vasculature number between multiple-channeled scaffolds and non-channeled scaffolds (Fig. 7e), there is still significant difference in terms of the size of the formed blood vessels. This result implied that multiple channels have the potential of enhancing blood vessels formation, thus inducing more bone tissue formation in the channeled scaffolds.
Immunohistochemical staining. A series of immunohistochemical straining of bone matrix proteins including collagen type I, bone sialoprotein (BSP), and osteocalcin (OC) were carried out. The results showed that collagen type I was expressed in all three groups (Fig. 8a). The patterns of stained collagen type I in both non-channeled and multiple-channeled scaffolds are similar to that in native calvarial bone (Fig. 8a, ii, iii), www.nature.com/scientificreports/ compared to the no-implant group (Fig. 8a, i). There are few collagen fibers observed on the distal layer (from dura side) in the non-channeled scaffold group (Fig. 8a, ii), but remarkable collagen fibers were homogenously distributed around the whole multiple-channeled scaffolds (Fig. 8a, iii). Semi-quantification of the positively stained collagen type I from these images displays that the ratio of total collagen type I to the whole tissue area in the multiple-channeled scaffolds is significantly higher than that in the non-channeled scaffolds (Fig. 8d).
Besides that, this ratio in both two type of scaffolds is considerably higher than that in no implant group. For BSP, we found that it is highly and uniformly presented in the no-implant samples, but not in the non-channeled scaffolds (Fig. 8b, i). In the non-channeled group, BSP is only localized along the peripheral areas and at the center of mineralized bone tissue (Fig. 8b, ii). However, in the multiple-channeled scaffolds, it is not only localized at the same regions as non-channeled scaffolds, but also widely spread at the core regions of the channel's area (Fig. 8b, iii). Semi-quantitative analysis shows that the expression ratio of BSP in the multiple-channeled scaffolds is significantly higher than that in the non-channeled scaffolds (Fig. 8e). Surprisingly, positive BSP staining is statistically higher in the no-implant group compared with the non-channeled group as well, although no newly mineralized bone was observed inside this group. We further stained a late stage mineralized matrix protein, OC. We found that there is no positively stained OC signal found in the no-implant group (Fig. 8c, i), but OC was spread and confined to the area where compact bone formed, or mineralizing tissues are adjacent to the compact bone in both non-channeled and multiple-channeled scaffolds (Fig. 8c, ii, iii). Furthermore, multiple-channeled scaffolds express significant higher OC than non-channeled scaffolds (Fig. 8c, iii). Similarly, semi-quantitative analysis shows that the expression ratio of OC in the multiple-channeled scaffolds is significantly higher than that in the non-channeled scaffolds (Fig. 8f). Dynamic deposition of mineralized tissue in the non-channeled and multiple-channeled scaffolds is demonstrated through fluorescent images of the sections from fluorochrome labeled samples (Fig. 9). The results show that the amount of fluorescent alizarin red signal is much less than calcein green signal, and most of the alizarin red signal is overlapped and covered by calcein green signal. In the non-channeled scaffolds, the signals

Discussion
Using porous osteoconductive β-TCP bioceramic scaffolds for bone regeneration has been widely reported and studied 23,24 . However, challenges such as maintenance of sustainable nutrient diffusion that directly leads to an effective bone formation are still unsolved 25,26 . Insufficient vascularization and limited internal-external nutrient exchange delayed osteogenesis and bone regeneration [11][12][13][14][15] . Strategies on modifying the scaffolds with growth factors lost their superiority due to the challenges such as short half-life and unwanted physiological side effects 27,28 .
In our previous studies we've already proved that interconnected channels-pore geometry generated a relatively strong and homogeneously distributed fluid shear stress field inside the scaffold, which contributed to endothelial cell migration and promoted angiogenesis in vitro 17 . Further, in this study, we continued to utilize the discovered mechanotaxis property of multiple channels to induce bone mineralization and formation in vivo. We verified that the interconnected channels-porous architecture significantly promoted hBMSCs proliferation and differentiation in vitro through activating mechanotransduction pathway, and considerably stimulated osteogenesis, mineralization, and bone formation in vivo, compared to the non-channeled scaffolds. All these findings indicated that the architecture of multiple channels with interconnected pores play a significant role in cell proliferation, differentiation in vitro and bone tissue formation in vivo. These results also implied that without any additional biological components (cells, growth factors), the new architecture itself also activated an endogenous bone formation cascade. The channels facilitated the circulation of dynamic fluid and generated a homogenously higher fluid shear, which can be sensed by cells and in turn triggered the cells mechanotransduction signaling pathways that correlated with cell proliferation and tissue formation. This finding was consistent with the research of Kapur et al., in which they demonstrated that fluid shear stress triggered human osteogenic cell proliferation 29 . What's more, it is paramount that an uninterrupted differentiation process of multipotent mesenchymal stem cells needs to be initiated and sustained along with osteogenesis and bone formation 30,31 , which is a condition that can be effectively maintained by the interconnected channel-pore structure of the scaffold. Our results suggested that multiple channels play an indispensable role on hBMSCs differentiation by promoting alkaline phosphatase activity and up-regulating relative osteogenic gene expressions. The expression of the four genes investigated in this study follows a chronological order. According to that, interestingly, we found the dynamic fluid not only promoted the expression level of the genes, but also prolonged the expression time of each gene. This dynamic fluid condition synergizing with multiple channels makes the It's been also reported that fluid shear stress could activate Hippo/YAP pathway, the inducer of MSCs differentiation, through transferring the signal from FAK to YAP, which stimulated the production of Runx-2 and promoted osteogenesis 35 . In this study, it aroused our interest to see, without any biological components involved, whether mechanotransduction pathway of osteogenesis was activated by multiple channels, and acted as the underlying mechanism of enhancement of cell differentiation 36 . The immunofluorescent staining displays that multiple channels facilitated the diffusion of dynamic fluid, which activated the expression of FAK. FAK assisted the transmission of fluid shear stress signal and promoted the polymerization and formation of F-actin fibers, which upregulated as well as triggered the translocation of YAP-1 (Figs. 4, 5). The remarkable activation of proteins which are highly affiliated with osteogenic mechanotransduction pathway in multiple channeled scaffold, indicated that a relatively higher fluid shear force was initiated inside the scaffold, thus significantly promoting the expression of osteogenic genes and cell differentiation in vitro.
In vivo results further demonstrated that multiple-channeled scaffold promoted osteogenic collagen type I, BSP, and OC protein expression and sustainably enhanced the formation of dense bone. These results imply that during intramembranous bone formation, collagen fibers acted as a woven matrix for incorporating non-collagenous proteins such as BSP to initiate mineralization and modulate calcium phosphate mineral deposition [37][38][39] . Thus, a sustainable collagen production is essential for fully bone regeneration. In Roah's study, they found that the early osteogenic markers were not highly presented in heavily calcified bone 37 . In our study, we also found that more areas on multiple-channeled scaffold were filled with highly mineralized matrix, which has the appearance very similar to native calvarial bone. Collagen type I is homogenously distributed everywhere in the channeled scaffold, rather than that only distal layer from dura side is highly positively stained in the non-channeled scaffold. However, collagen is only observed on the above half of the non-channeled scaffold. BSP, one of the most abundant bone non-collagenous proteins, has collagen-binding sequence and negatively charged 40 . During the bone regeneration process, BSP binds to collagen fibers and then initiates the deposition of the calcium phosphate crystals to the site 41,42 . Therefore, BSP is essential on bridging upstream and downstream osteogenesis and bone mineralization. In our study we found that BSP protein expressed significantly higher in the multiple-channeled scaffolds than non-channeled scaffold (Fig. 8b). In addition to that, we observed that BSP distributed widely at the areas where un-matured fibrous tissue is located, such as the center of channel area, and no-implant samples. This observation is quite similar to that of collagen type I, where more staining was seen at the non-matured tissue areas. This finding could be explained by the study of Roah, in which collagen I and BSP were found to disappear when tissue became highly mineralized and calcified to bone matrix 37 . All these discoveries indirectly indicated the reinforcement of rapid bone formation by multiple channeled scaffolds, which might sustainably activate selfcirculative osteogenic mechanisms and promoted persistently bone formation. However, in the non-channeled scaffolds, the positive staining area of both collagen I and BSP is limited. This implies that bone mineralization www.nature.com/scientificreports/ and formation in the non-channeled scaffold are non-sustainable and delayed. Except for that, we found multiple channels significantly promoted the expression of late-stage bone mineralization marker osteocalcin, and considerably increased mineral deposition in the channeled scaffold through observing fluorochrome labeled samples. All these findings indicated that multiple channels prompted bone formation and maturation. The underlying mechanism may include but not limited to the surrounded cells and tissues in response to the mechanical shear force that was generated by the interconnected channels-pore architecture 17 . However, the detailed signaling pathways that correlate mechanosensation or mechanotransduction to osteogenesis and bone formation in vivo may need further studies to be verified. Even so, this new porous β-TCP scaffold with multiple channels provides a promising approach with cell-or growth factor-free component to induce bone tissue regeneration.

Conclusion
In this study, we demonstrated that the creation of multiple channels in a porous β-TCP scaffold promoted hBMSCs differentiation and osteogenesis in vitro, and sustainably induced bone formation in vivo. The interconnect channels-pore geometry in porous β-TCP scaffolds had significant functionality to stimulate new bone formation. The findings implied that the architecture of multiple channels-pores in the scaffold act as a promising stimulator to promote bone regeneration.

Materials and methods
Preparation of β-TCP scaffolds with multiple channels. Two geometries of interconnected porous β-TCP scaffolds were fabricated by using a template-casting method that was reported in the previous study 22 .
One is porous β-TCP scaffold, the other is porous β-TCP scaffold with five channels. The diameter of each channel is 1 mm. The five channels are homogenously distributed in the porous β-TCP scaffold. Briefly, β-TCP slurry was prepared by stirring mixture of β-TCP nano-powder (Nanocerox, Inc, Ann Arbor, Michigan), carboxymethyl cellulose powder, dispersant (Darvan C), and surfactant (Surfonals) with distilled water (Fisher Scientific). After filling the paraffin beads into the two types of customized modes, β-TCP slurry was casted into the molds and solidified in ethanol for two days, followed with a gradient ethanol dehydration. The completely dried β-TCP green-bodies were then sintered for 3 h at 1250 °C. The fully sintered scaffolds were then used in the in vitro and in vivo studies.
Characterizations. The pore morphologies of the porous scaffold were observed by scanning electron microscopy (SEM). The dried scaffold was sputter-coated with gold for observation under a benchtop SEM (JEOL, JCM-6000Plus). The tortuosity of pores was observed by a MicroCT scanner (Imtek Micro CAT II, Knoxville, TN) at a resolution of 80 μm.
In vitro cell attachment quantification. To measure whether hBMSCs' attachment efficiency was promoted by channels, six scaffolds of each type were placed in the wells of 24 ultra-low adherent well plates (Corning, NY). One scaffold per well was set. hBMSCs from Lonza were cultured with MSCGM medium (Lonza, Basel Switzerland) under a standard cell culture condition. Two milliliters of hBMSCs suspension with a concentration of 5 × 10 5 cells/mL were added into each well. To dynamically seed cells onto the scaffolds homogenously, the well-plates were then placed on a 3D platform rotator (Fisher scientific, 3D platform rotator, Hampton, NH) and rotated at 30 rpm in CO 2 incubator at 37 °C. At 4 and 16 h, 3 wells with each type of scaffolds were used to determine the cell attachment efficiency according to our published method 17 . Briefly, each scaffold was taken out and washed with 1 × PBS three times. All washed PBS, the leftover media, and the trypsinized content from each well were collected into a 15 mL centrifuge tube. The cell number was counted from each tube. Final cell attachment efficiency was determined by measuring the cell number per scaffold surface area and followed the equation: Na = (Ns − Nc)/As, where Na stands for the number of attached cells per unit area, Ns stands for the initially seeded cell number, and Nc stands for the collected cell number in the tube, and As stands for the surface area of each type of scaffolds. The surface area of each scaffold was calculated through the dimensions and volumes of the scaffold and its pores.

hBMSCs proliferation and osteogenic differentiation. Channels' function on hBMSCs proliferation
property was investigated in both static and dynamic culturing conditions. One hundred microliters of 10 5 hBM-SCs suspension was pre-seeded onto each scaffold and incubated for an hour at 37 °C, after which equal and enough MSCGM medium was added to each well to fully cover the scaffolds and kept culturing for another 24 h. The cells/scaffolds were kept culturing in 24-well plates with MSCGM medium for static condition measurement. A continuous 3-, 7-, and 14-days' culture with medium changed every 2 days was followed. For dynamic condition measurement, after 24 h of initial culture, all the cells/scaffolds were transferred into a dynamic circulating bioreactor system where fresh MSCGM medium was circulated through the cells/scaffolds at a rate of 10 μL/min for 3, 7 and 14 days as well, according to the setting of the dynamic culture system we used in previous study 17 . At the end of each time point, samples from both static and dynamic culturing conditions were collected and rinsed with 1 × PBS twice and stored at -80 °C. Fluorometric assay was utilized to measure the cell proliferation property quantitatively. Briefly, the cell lysate was collected by immersing all the stored cells/scaffolds samples with 0.2% Triton X-100 in 1 × TE buffer solution and followed by three freeze-thaw cycles. For each cycle, cells/scaffolds were frozen at − 80 °C for 20 min, and then thaw to 37 °C for another 20 min. After that, the content of dsDNA from each scaffold was measured by using a Quant-iT™ PicoGreen™ dsDNA Assay Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's instruction. All the samples were read through Spectra Max Gemini EM plate reader with 480/520 nm excitation/emission wavelengths 43  www.nature.com/scientificreports/ Alkaline phosphatase (ALP) activity was measured as one way to investigate the effect of channels on hBMSCs osteogenic differentiation. The cells/scaffolds were cultured in both static and dynamic systems as described above as well. However, after 24 h pre-seeding in MSCGM medium, the culturing media was changed to osteogenic differentiation media, which contained 10% FBS, 10 mM β-glycerophosphate, 10 nM dexamethasone, 50 mg/mL ascorbic acid, and 1% PSA, other than MSCBM. The cells were continuously cultured for 3, 7 and 14 days, and samples were rinsed with 1 × PBS, collected and stored in − 80 °C at each time points for further tests.
To determine the cell ALP activity quantitatively, the total protein was isolated through three times of 80 °C to 37 °C freeze-thaw cycles as well by immersing all the stored cells/scaffolds osteogenic differentiation samples within 0.2% Triton X-100 solutions. The total protein was measured through Pierce™ BCA Protein Assay Kit (Thermofisher Scientific), and read by Spectra Max 190 plate reader at the wavelength of 562 nm. The total ALP was determined through Quantitative Alkaline Phosphatase ES Characterization Kit (EDM Millipore, CA), and read by Spectra Max 190 plate reader at the wavelength of 405 nm. Final cell ALP activity was determined through the following equation: C Aa = Ca/Cp, in which C Aa stands for ALP activity, C a stands for the total alkaline phosphatase amount and C p stands for the amount of total protein of each sample 44 .

Real time PCR.
Osteogenic differentiation related genes were run by real time PCR. Human BMSCs were cultured with osteogenic differentiation media in the same static and dynamic conditions as mentioned above. At the end of 7 and 14 days, total RNA of cell/scaffolds samples were extracted by using RNeasy Mini Kit (QIA-GEN, Hilden, Germany). The concentration of RNA was measured through NanoDrop™ One/OneC Microvolume UV-Vis Spectrophotometer (Thermofisher Scientific). mRNA of all samples was then transcribed into cDNA as templates in real-time PCR by iScriptTM cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA). Quantitative real-time PCR was performed afterwards by using iQ™ SYBR ® Green Supermix Kit (Bio-Rad Laboratories) on an AriaMx Real-Time PCR System (Agilent Technologies, Santa Clara, CA), and followed the instructions according to the manufacturers. Specific primers (expressed by italic, lower cases) including runtrelated transcription factor 2 (Runx2), alkaline phosphatase (alp), bone sialoprotein (bsp), osteocalcin (oc), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) with the sequences shown in (Table 1) were purchased from Invitrogen. The 2 − C T method was used to relatively quantify genes expression, where C T = C T,Multiple − C T,Non Target gene − C T,Multiple − C T,Non GAPDH , in which expressed target genes were normalized to the expression level of a housekeeping gene GAPDH 45,46 . Immunofluorescent staining. The samples were immuno-probed with FAK and YAP-1 antibodies by following the instruction of manufacturers to investigate whether the addition of multiple channels inside scaffolds initiates the markers along with mechanotransduction pathway of hBMSCs. Briefly, the immunofluorescent staining of FAK and YAP-1, both non-channeled and multiple-channeled scaffolds were incubated with primary anti-FAK (1:100) and anti-YAP-1 (1:50) antibodies (Abcam, Cambridge, UK) separately at 4 °C overnight, followed with a subsequently fluorescent conjugated secondary antibody Alexa Fluor 594 (Invitrogen) labeling for 2 h at room temperature. F-actin fiber of cells on the scaffolds was stained by a fluorescent Phalloidin kit (Cytoskeleton, Denver, CO). Briefly, Actin-stain 555 phalloidin solution was used for the incubation of two types of scaffolds at room temperature for 30 min by following the instruction of manufacturer. After that, all the samples were co-stained with 4′6-diamidino-2-phenylindole (DAPI). The stained scaffolds were then placed on a glass cover slide on an inverted fluorescent microscope. Stained cells on the scaffolds were observed and imaged through a Nikon TE-2000 fluorescent microscope. Three random regions of each type of scaffold from the bottom to an inner area of the scaffold that the lens reached were pictured. The number of positively stained cells by FAK, YAP-1, and F-actin, and the total cells of each image per type of scaffolds were semi-quantified by FIJI ImageJ (NIH). The ratio of positively stained cells to total cells were calculated.
In vivo implantation and fluorochrome labeling. Animals  In vivo implantation. To evaluate the function of the channel-pore architecture on bone regeneration, scaffolds were implanted in rat critical-sized calvarial bone defects. The animals were randomly divided to three groups, including non-channeled, multiple channeled scaffolds, and no-implant as a control. Six rats were used for each group (three female and three male rats). A circular 8 mm defect in diameter was created on each rat's calvarial bone using a dental drill, and a scaffold with 8 mm in diameter and 1.5 mm in height was implanted. All the scaffold samples were shaped with sandpapers, washed with 70% ethanal and 1 × PBS for 3 times each, and autoclaved before utilization in the implantation surgery. Fluorochrome labeling was performed postoperatively. Alizarin red and Calcein green were injected subcutaneously at week 6 and 10 weeks separately after the surgery. Both fluorochromes were purchased from Sigma-Aldrich (Munich, Germany), and solutions was prepared as shown in (Table 2). Before injections, both fluorochrome solutions were adjusted to pH 7.4 and sterilized through a 0.22 µm filter. The dosage was calculated in accordance with the body weight (BW) of each rat.
Tissue harvest, fixation and X-ray characterization. All the animals were euthanized with 5% CO 2 after 3 months of surgery. The entire defect with implanted scaffolds, and adjacent native bone tissue were harvested, rinsed with 1 × PBS, and fixed with 10% buffered formalin solution (Thermofisher Scientific) for 48 h. All the samples were characterized by a SkyScan microCT machine (Bruker, Billerica, MA) associated with the digital image analysis software (CT Analyser Version 1.18.4.0, SkyScan, Bruker microCT) to evaluate the regeneration and osseointegration of new bone on different implants.
Histological and histomorphometric analysis. All harvested samples were processed for histological analysis by decalcifying with a rapid decalcifier solution (RDO Rapid Decalcifier, IL) for 7 days until the samples were softened. The samples were then embedded with paraffin and sectioned laterally and transversely. Hematoxylin and eosin (H&E) staining was carried out to evaluate newly formed bone density and height. For newly formed bone density analysis, each implant was analyzed from proximal, middle, and distal three transversal segments, and 0.5 mm height for each segment (bottom, middle, top). Three different regions of each segmental section were calculated, and three samples of each type of implant was considered. For all the other semi-quantitative evaluations, only central lateral segmental sections were considered for each implant.
Immunohistochemical characterization. The central lateral sections of each group sample were probed with collagen type I, bone sialoprotein (BSP), osteocalcin (OC), and angiogenic marker CD31. Briefly, the samples were retrieved with 1 × citrate buffer (Sigma-Aldrich, St. Louis, MO) for 30 min at 95-100 °C. Followed which, all sets of groups were incubated with primary anti-BSP (1:500), anti-collagen type I (1:250), anti-OC (1:100), and anti-CD31 (1:50) antibodies (Abcam, Cambridge, UK) separately in a humidity chamber at 4 °C for overnight, and were subsequently marked with anti-biotinylated binding IgG secondary antibodies for at room temperature for 30 min. All immunohistochemical images were taken from a Nikon TE-2000 microscope, and three regions of each section were captured for semi-quantitative analysis using FIJI ImageJ (NIH). For BSP, collagen type I, and OC staining, percentage of positively stained tissue area to the whole tissue area were utilized on osteogenesis analysis. For CD 31 staining, the amount of blood vessels that formed on tissues were calculated for angiogenesis analysis 47,48 . All the other chemicals which were used in the characterizations were purchased from Vector Laboratories (CA, US). Statistical analysis. GraphPad Prism 6 (GraphPad, San Diego, CA) was used for conducting statistical analysis and plotting graphs. ANOVA with Tukey multiple comparison test was applied, and the results between groups are statistically significant when p < 0.05. All the experiments were performed in triplicates using one hBMSC donor.