**4. Discussion**

In this study, we used extrusion based DIW printing technology to fabricate PCL scaffolds. DIW allowed us to 3D print scaffolds directly from PCL pellets, which were melted within and extruded from a steel syringe attached to the print head. PCL was selected as a model polymer as it is a "Generally Recognized As Safe" (GRAS) polymer by the U.S. Food and Drug Administration and widely used to 3D print tissue engineering sca ffolds for both in vitro and in vivo studies [27,35,75]. We used human adult mesenchymal stem cells (hMSCs) as the main cell line due to their ability to proliferate, migrate, and di fferentiate into a wide range of tissue specific phenotypes including bone, cartilage, and muscle. Stem cells are known to feel and respond to their microenvironment (matrix sti ffness, topography, and bioactivity) by regulating their behavior [56,60]. Here, we focused on the topography, or sca ffold architecture. To investigate the e ffects of 3D sca ffold architecture on stem cell osteogenesis, we constructed sca ffolds using struts in sinusoidal waveforms, systematically varying the amplitude and the wavelength (Figures 1 and 2, Table 1). The sinusoidal waveform design created highly curved strut surfaces forming 3D sca ffolds with wavy patterns. Our motivation to create wavy sca ffolds was based on previous studies, which clearly showed the importance of substrate curvature on stem cell osteogenesis [69,70].

The minimum wavelength and amplitude achievable for a strut size around 500 μm were 2 mm and 0.5 mm (A0.5W2). While keeping the amplitude constant at 0.5 mm, the wavelength was increased to 3 mm (A0.5W3) and 4 mm (A0.5W4). For the 4 mm wavelength, the amplitude was increased to 0.75 mm (A0.75W4) and 1 mm (A1W4). These geometrical constraints allowed us to create sca ffolds with an average strut-to-strut distance of approximately 350 μm (Figure 2, Table 1). Note that the effect of pore size in bone sca ffolds has been well studied [15,24,76–81], and a minimum pore size of ~150 μm is usually required for cell migration and tissue ingrowth [3,82–84]. We then investigated the effect of sca ffold design on mechanical properties of the sca ffolds (Figure 3). The compressive modulus (*E*) values were determined by the design, i.e., strut-to-strut contacts between layers, and the overall sca ffold porosity. *E* values were significantly the highest for orthogonal sca ffolds (12.5 MPa) mainly because these sca ffolds inherently displayed more strut-to-strut contacts, considering that this design had 16 struts per layer, whereas all of the wavy designs had 15 struts per layer. This design also had one of the lowest porosities with ~56%. When wavy sca ffolds were compared, A0.5W4 showed the significantly highest porosity (~62%) corresponding to the significantly lowest *E* value of 9.5 MPa followed by A0.75W4 (58%, 10.7 MPa), A1W4 (57% 11.3MPa), and A0.5W3 (56%, 11.5 MPa). A0.5W2 (56%, 10.5 MPa) was an exception and did not follow the trend. This was due to reduced strut-to-strut contacts due to the design (Figure 2). Although the overall sca ffold modulus determines the mechanical support level that a sca ffold can provide when implanted, it did not a ffect the stem cell behavior in our study. This is because the stem cells feel the mechanics of the individual struts (which was uniform for all sca ffold groups) that they reside on when seeded on to the sca ffolds [74].

First, the growth study was conducted to determine the attachment and proliferation of the hMSCs cultured on our sca ffolds. The metabolic activities of the cells were not significantly di fferent from each other at each culture day, but increased significantly with culture day, reaching a maximum at Day 7 (Figure 4A). The same trend was observed when the DNA was quantified (Figure 4B). Note that this trend was not true for the A1W4 and A0.75W4 sample groups, for which the metabolic activity reached a maximum at Day 4 and did not change significantly at Day 7. Yet, the DNA count did not show this unexpected trend for these two sample groups, which represented the cell proliferation more accurately. F-actin staining at Day 7 confirmed that cells attached onto the struts and formed confluent layers at Day 7, taking the shape of the struts. Cells on wavy sca ffolds were highly elongated, especially on the curved edges with well-defined F-actin filaments aligned with the sca ffold curvature as compared to much bulkier cells on orthogonal sca ffolds (Figure 5). In addition, stem cells on wavy sca ffolds showed mature vinculin (focal adhesion marker) patches as compared to di ffused vinculin staining of cells on orthogonal sca ffolds at Day 7 (Figure 8). Focal adhesion is a vital step in osteogenesis [85] in which vinculin directs the interaction between talin and actin to direct the focal adhesion process [86]. We investigated if these significant changes in stem cell morphology, F-actin expression, and focal adhesion on wavy sca ffolds as compare to orthogonal sca ffolds correlated with stem cell osteogenesis on wavy sca ffolds. It was also noted that the curvature had a direct e ffect on cell proliferation, and studies have shown that curvature induced contractility enhances proliferation and

cell growth [87–89]. In our study, we did not observe a significant difference in proliferation between sample groups. This was not contradictory to the literature as each of our wavy scaffolds displayed both concave and convex curvature, and the overall cellular behavior was collective rather than distinct for each type of curvature.

The differentiation studies were conducted after the cells reached a confluent state at Day 7, as shown by the growth studies (Figures 4 and 5). At Day 7, the growth media was replaced with osteogenic induction media, and cells were cultured for 14 additional days in induction media, a total of 21 days in culture. To assess the osteogenic differentiation of hMSCs, we quantified calcium deposition and ALP activity and performed immunostaining for osteocalcin. The AR assay was used to probe the deposition of calcium. Optical microscope images revealed that wavy scaffolds showed more stained regions than the orthogonal group. When quantified, all the wavy scaffolds showed higher calcium deposition than the orthogonal group, and in particular, two groups (A0.5W2 and A0.75W4) showed significantly higher calcium deposition (Figure 6). These results indicated that the overall contribution of the curvature on these two scaffolds on cellular contractility induced calcium deposition was the highest. ALP is a well known biological marker for stem cell osteogenesis [90]. ALP activity increased significantly for all of the scaffold groups from Day 14 to Day 21 (Figure 7). All the wavy groups showed higher ALP activity than the orthogonal group. However, the differences between wavy groups and the orthogonal group were not as significant as the results from the AR assay. This could be because the ALP expressed at earlier stages of the osteogenesis process. At Day 14, A0.5W3 showed significantly higher ALP activity (*p* < 0.05) when compared to the orthogonal group. At Day 21, both A0.5W2 and A0.5W3 were substantially higher than the orthogonal group (*p* < 0.15). To supplement our quantitative differentiation assays, we performed OC immunostaining (Figure 9) as a marker for osteogenesis. Qualitatively, we observed increasing OC staining with culture day, and wavy scaffolds showed more OC staining, in particular in the curved regions of the scaffolds. The enhanced osteogenesis behavior on wavy scaffolds could be explained as the effect of the curvature, which led to a highly aligned and stretched cellular morphology (Figure 5) with mature focal adhesions (Figure 8), leading to highly contractile cells promoting osteogenesis. We strongly believe that our results clearly showed the importance of scaffold architecture on hMSC osteogenesis and would help to develop novel scaffold architectures for bone tissue regeneration.
