Enhancing mesenchymal stem cell response using uniaxially stretched poly(ε-caprolactone) film micropatterns for vascular tissue engineering application

Department of Mechanical Engineering, Engineering Drive 1, Singapore 117 576, Fax: +65 6779 1459; Tel: +65 6516 5233 School of Chemical and Biomedical Engine Singapore Department of Reproductive Medicine, KK Buikit Timah Road, Singapore 229899, Si Fax: +65 6394 1618; Tel: +65 6394 1051 Department of Electrical and Compute Singapore, Singapore Department of Obstetrics and Gynaecology, University of Singapore, Singapore Cancer and Stem Cell Biology, Duke-NUS G † Electronic supplementary informa 10.1039/c4tb00522h Cite this: J. Mater. Chem. B, 2014, 2, 5898


Introduction
Cardiovascular diseases are the leading cause of death globally ($38% and $50% of death in the U.S. and Europe, respectively). 1,2 Tissue engineering (TE) could offer a novel therapy to improve or regenerate diseased vessels using a combination of biomaterials, cells, and engineering techniques. 3 However, current vascular TE gras remain limited by mismatched compliance, thrombogenicity and/or inappropriate burst strength. 2,4 A suitable scaffold for engineered, autologous and commercially available vascular gras has yet to be developed.
In healthy blood vessels, the circumferentially aligned architecture of tunica media comprised of smooth muscle cells (SMCs) and elastin/collagen-enriched extracellular matrices (ECM) play important roles in maintaining the endothelium phenotype, mechanical strength and regulation of blood pressure. 3,5 Geometric cues serving as articial cell micro-environments could facilitate vascular TE scaffolds to replicate the complex architecture of tunica media. [6][7][8] It has been known that geometric cues could trigger cell response via contact guidance. 9,10 Human SMCs cultured on micropatterned arrays organized into alignment as in native vessels 7,11,12 and could obtain differentiation towards a contractile status. 6,8 Despite these great advances in basic research, the incorporation of geometric cues into threedimensional (3D) tubular vascular TE scaffolds remains a challenging issue. Geometric features presently have oen been fabricated on stiff and non-biodegradable substrates (e.g. Silicon 13 and polydimethylsiloxane 8 ) that are not exible and have little value for clinical translation. Although fabrication of biodegradable materials (e.g. poly(3-caprolactone-r-L-lactide-r-glycolide) 6 and poly(L-lactide-co-3-caprolactone) 14 ) could be realized by using electrospinning, so-lithography and direct laser writing, these techniques oen involve tedious procedures, unstable yield in a large area, and/or chemical solvent usage and residue. 15 Such limitations related to the fabrication of geometric cues hindered the generation of commerically available, non-cytotoxic, and cost-effective vascular gras.
Traditionally, isolated mature SMCs were utilized for tissueengineered vascular gras (TEVGs), 16 but were found to dedifferentiate, overproliferate, and decrease the secretion of collagen/elastin in vitro. [17][18][19] Recently, mesenchymal stem cells (MSCs) have shown great potential for vascular TE, 17 and TEVGs constructed from MSCs can replicate an intact tunica media with histological and molecular phenotypes similar to that of the native vessels. 18 Compared to the fully differentiated SMCs, MSCs were more easily obtained from the umbilical cord, bone marrow and adipose tissue, 20 and they were more expandable in vitro up to millions of folds in culture, 17 antithrombogenic, 21 and immunocompatible due to the lack of expression of major histocompatibility complex II antigens. 22 However, biochemical factors, such as transforming growth factor (TGF-b1), plateletderived growth factor and ascorbic acid, were oen required to differentiate MSCs for obtaining SMCs contractile phenotypes, which were considered to be costly and difficult in controlling the optimal concentration for effective stimulation. [17][18][19]23 Although geometric cues have been used to differentiate SMCs, they have not been reported on regulating MSCs towards SMClike contractile phenotypes.
Herein, we describe the development and application of exible poly(3-caprolactone) (PCL) lm micropatterns via a reproducible, simple, and solvent-free method of uniaxial thermal stretching. Vascular TE scaffolds based on the PCL lm micropatterns could have sufficient theoretical burst strength and potential to promote stem cell response for tunica media regeneration.

Fabrication of geometric cues on PCL lms
Geometric cues were fabricated on different PCL lms. Briey, solvent-cast PCL lm (SC-PCL) was generated through casting PCL solution (2 wt% in dichloromethane) onto a glass mold and evaporating the solvent in a fume hood. 24 Cast-stretch PCL lm (CS-PCL) was obtained using an in-house developed machine via cast extrusion. 25 Heat-press PCL lm (HP-PCL) was generated through pressing PCL solid mass at 80 C and 300 MPa. 24 The PCL lms were then cut into rectangular shapes (5 Â 3 cm 2 ) and subjected to uniaxial stretching at a temperature of 54 C and a draw ratio of 4 ( Fig. 1a) for generating microstructured topographies as previously described. 15 For convenience, the stretched PCL lms were named as UX SC -PCL, UX CS -PCL and UX HP -PCL. There were no obvious changes in M w and M n for PCL lms before and aer uniaixal stretching. For cell studies, PCL lms were surface-modied using alkaline hydrolysis (3 mol L À1 NaOH solution) as previously described, at room temperature for 24 hours to obtain a water contact angle of $52 for facilitating cell adhesion. 15,26

Morphological characterization
The microstructured topographies of PCL lms were characterized using eld-emission scanning electron microscopy (FE-SEM; S-4300, Hitachi, Japan). Samples were sputter-coated with gold at 10 mA for 30 s, and images were taken under different magnications at a voltage of 15 kV. The PCL lm morphologies were further conrmed using atomic force microscopy (AFM; MFP-3D, Asylum Research, USA) via a tapping scan over an area of 30 Â 30 mm 2 . Inter-ridge distance was measured from SEM images by drawing lines perpendicular to the stretch direction and presented as a length of the line divided by the crossed ridge number. Ridge depth was analyzed from AFM images with topographical proles obtained along lines perpendicular to the ridges using the built-in function of Igor Pro6 (USA). Three samples were used for each kind of PCL lm, and results were averaged as the mean.

Measurement of mechanical properties
The mechanical properties of PCL lms were measured using a tensile testing machine (Model 3345, Instron, USA). 26 Samples were prepared into strips of 5 Â 30 mm 2 with a major edge along the stretching direction. The thickness of each sample was measured using a digimatic micrometer (APB-1D, Mitutoyo Corporation, Japan) at ve random positions. Testing was performed using a load cell of 100 N at a pulling rate of 10 mm min À1 . An offset-strain of 0.005 was used for the stretched lm samples to determine their yield stress and strain, and for unstretched lm samples, a low yield point was used to determine the yield stress and strain at the linear-elastic region. Theoretical burst strength was calculated based on Laplace's law (burst strength ¼ materials yield stress Â thickness/radius). 27 For each kind of PCL lm, three sample copies were tested.

In vitro cellular isolation and culture
Human tissue collection for research purposes was approved by the Domain Specic Review Board of National Healthcare Group, in compliance with international guidelines regarding the use of fetal tissues for research. 28 In all cases, patients gave separate written consent for the use of their collected tissues.
MSCs were isolated from human fetal bone marrow. Through ushing the marrow cells out of femurs and passing a cell strainer (40 mm), single-cell suspension was obtained. Cells were then seeded in a tissue culture ask (75 ml) at 10 6 per cm 2 and cultured in D10 medium (DMEM + 10% FBS + 1% PS) for three days. Non-adherent cells were removed by changing the medium. The adherent cells (MSCs) were further cultured and recovered from the primary culture aer one week. As previously described, the obtained cells were checked for standard stem cell markers through immunochemistry analysis using ow cytometry (FACS, Beckman, USA; Table S1 †). 28,29

Characterization of MSCs organization and morphologies
The organization and morphologies of MSCs were characterized through cytoplasmic staining. 15 MSCs seeded on substrate samples (unstretched and uniaxially stretched PCL lms) were cultured for a pre-determined period to maintain a sub-conuence towards single-cell analysis. Aer washing with PBS, cells were incubated with FDA (5 mg L À1 ) at room temperature for 10 min, and aer washing thrice, they were immediately visualized using confocal laser scanning microscopy (CLSM, FV1000, Olympus, Japan). Images were analyzed using the builtin function of NIH ImageJ soware (USA). All cells in contact with other cells and image edges were manually removed from the data sets. The cell angle was determined using the orientation of the major elliptic axis of the individual cell. A preferential cell orientation has been determined and set as 0 . All cell angles were then normalized to the preferential cell orientation. Cell number within each degree from À90 to +89 was calculated and normalized such that the total sum was unity. A perfectly isotropic sample would be expected to have an even distribution of cell angles in each degree (a value ¼ 1/180 Â 100%). Cells with angles falling into AE10 were considered to be aligned. 15 The isotropic sample would have $11.7% of cells in AE10 (total 21 , a value ¼ 21 Â 1/180 Â 100%). Values above 11.7% indicated increased alignment. Cell elongation was described as a long-to-short elliptic axis index (a ratio of the major and minor elliptic axes). Three samples were used for each PCL lm group, and CLSM images were taken at four random regions of each sample for data analysis.

Characterization of MSCs cytoskeleton and nucleus organization and morphologies
The organization and morphologies of MSCs subcellular organelles were characterized through F-actin and DNA staining. Aer culturing for a pre-determined period, MSCs were xed (3.7% PFA in PBS), permeabilized (0.1% Triton-X 100 in PBS) and blocked (2% BSA in PBS). The cells were sequentially incubated with TRITC-conjugated phalloidin (1 : 200 in PBS) and DAPI (1 : 1000 in PBS) for F-actin and DNA labeling, respectively. CLSM images were analyzed using the built-in function of NIH ImageJ soware. Cell nuclei with angles falling into AE10 were considered to be aligned. A perfectly isotropic sample would be expected to have an alignment efficiency of $11.7%. Nucleus elongation was described using a cell nucleus shape index (CNSI, circularity ¼ 4 Â p Â area/perimeter 2 ), with a CNSI of 1 representing a circle. Three samples were used for each PCL lm group, and CLSM images were taken at four random regions of each sample for data analysis.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)
The expression of gene targets (Table S2 †) in MSCs was investigated using real-time qRT-PCR. ACTA2, CNN1 and MYH11 were the targeted gene sequences of SM-a-actin, calponin, and SM-MHC, respectively, and related to human vascular SMCs early-, middle-and late-term differentiation, respectively. 8,19,30,31 MSCs seeded on TCP and PCL lm samples were cultured for ve days in a D10 medium. Total RNA was extracted using RNeasy Mini Kit, and cDNA was synthesized using RT 2 First Strand Kit. qRT-PCR was performed on an ABI Prism 7000 realtime detection system (AB Applied Biosystem, USA) for 45 cycles. Primers specic to the targeted genes were obtained from PrimerBank (http://pga.mgh.harvard.edu/primerbank/). The relative expression levels of each gene targeted were determined by comparing the quantied cDNA transcript levels to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH, used as the internal standard). 15 The results were nally normalized to the relative expression levels of each gene of the TCP group. Three experimental samples were used for each testing group, and detection was performed in triplicates of each sample for data analysis.

Immunocytochemistry assays
The protein expression of SMCs contractile markers in MSCs was analyzed through qualitative and quantitative immunocytochemistry assays. MSCs seeded on PCL lm samples were cultured for ve days under identical conditions in a D10 medium.
Intracellular qualitative analysis. Aer culturing, MSCs were xed, permeabilized, and blocked. MSCs were then incubated with the primary monoclonal antibodies specic-targeted to SMCs contractile markers or the related IgG isotypes as negative control (Neg Ctrl) at room temperature for 60 min (Table S3 †). Aer washing thrice with PBS, cells were further incubated with the uorescence-labeled second antibody (goat anti-mouse IgG-FITC) at room temperature for another 60 min. Aer washing thrice with PBS, cells were nally incubated with DAPI (1 : 1000 in PBS) for 5 min for nucleus visualization. Cell images of each SMCs contractile marker were captured by CLSM using identical parameters for all groups.
Intracellular quantitative analysis. Aer culturing, MSCs were treated with trypsin-EDTA (0.25%) and collected through centrifugation (400 rcf, 5 min). Cells aer xation, permeabilization and blocking were immunolabeled with the primary monoclonal antibodies or the related IgG isotypes (Neg Ctrl), followed by the incubation of the uorescence-labeled second antibody. Human perivascular SMCs isolated from the arteries of umbilical cord (UCPVCs) were used as a positive control as previously described. 19 The intracellular uorescence of 10 000 cells of each sample was detected using a FACS equipped with an argon laser emission of 488 nm. A gate was set to ensure that a positive cell number in Neg Ctrl was less than 0.5%. Data were analyzed using the built-in function of Summit V4.3.02 soware (Beckman, USA). Three experimental samples were used for each testing group, and detection was performed in triplicates of each sample for data analysis.

Statistical analysis
Data analysis was performed on Prism 5 Soware. Results were reported as mean AE SD. A value of p < 0.05 was considered to be statistically signicant.

Results and discussion
3.1. Fabrication of 3D self-assembling micropatterns on exible PCL lms The development of a suitable scaffold is crucial for vascular TE to achieve engineered, autologous and commercially available gras. Herein, we present the development of a readily engineered, exible and light-weight lm with micropatterned geometries for functionalized vascular TE scaffold construction. Fig. 1a illustrates the designed set-up for uniaxial thermal stretching of PCL lms. As shown in Fig. 1b, the uniaxially stretched PCL lms were exible enough to allow bending and folding, and readily recover to their original states. SEM images show that the uniaxially stretched PCL lms obtained micropatterned topographies comprised of highly orientated ridges and grooves along the stretching direction, although different morphologies were observed for the unstretched PCL lms ( Fig. 2a and b). AFM examination conrmed the microstructured surfaces on uniaxially stretched PCL lms (Fig. 2b). Comparatively, the AFM image of UX CS -PCL reveals additional brous structures in nanoscale. Such laments also existed on the ridges of UX SC -PCL and UX HP -PCL based on the observations at a high magnication of SEM (e.g. a magnication of x25 000, data not shown). These observations suggested that uniaxial stretching as a newly developed method for surface micropatterning could be compatible with current lm fabrication techniques such as solvent cast and heat press. Quantitative analyses of the SEM and AFM images showed that among the stretched PCL lms, UX SC -PCL had the widest inter-ridge distance (vs. UX HP -PCL and UX CS -PCL: 9.0-10.2Â, p < 0.001; Fig. 2c) and the deepest ridge depth (vs. UX HP -PCL and UX CS -PCL: 2.9-17.5Â, p < 0.001; Fig. 2d). UX HP -PCL as compared to UX CS -PCL revealed a considerable increase in ridge depth (5.1Â increase, p < 0.001), although they shared an approximate inter-ridge distance of $5 mm (p > 0.05). These results suggested that the pre-fabrication of PCL lms could have an effect on the formation of uniaxial stretching-induced micropatterns. It should be noted that uniaxial stretching under different conditions could also result in ridges/grooves with different sizes. 15 With a decrease in stretching temperature (54 / 21 C) and draw ratio (4 / 2), the ridge length reduced and inter-ridge distance increased. Such discontinuous ridges with smaller ridge lengths and wider grooves could facilitate cell-cell contacts to interfere with cell-substrate interaction, and lead to the delayed and reduced cell alignment. [32][33][34] Therefore, in this study, uniaxial stretching of the PCL lms was performed at an elevated temperature of 54 C and a draw ratio of 4. Higher temperatures and draw ratios were not chosen, because the PCL melting point is $60 C, 15 and under these conditions, PCL lms were easily broken. Previously, we proposed that the surface micro-ridges of uniaxially stretched PCL lm were generated from recrystallized PCL crystals. 15 During the fabrication processes, PCL molecular chains were orientated because of stretching and formed into ordered, multi-layered lamellas to self-assemble into new crystals following the direction of the stretching. 35 Because the crystals were more rigid than the amorphous structures, they would have less deformation under uniaxial stretching, whereas the amorphous regions became thinner and formed into grooves to reveal the crystals from the lm surface as ridges. 15 In this study, the observed different sizes of the ridges/grooves on three uniaxially stretched PCL lms were in agreement with the previously proposed mechanism, as lm-processing history has been known to affect polymeric crystallization. 36 For potential application for vascular TE, we then investigated the mechanical properties of PCL lms. As shown in Fig. 3a, tensile stress-strain curves showed elastic deformation up to the proportional limit for UX SC -PCL, UX CS -PCL and UX HP -PCL, followed by plastic deformation characterizing with considerable strain increase but non-corresponding rise in the loading stress. Quantitatively, UX HP -PCL as compared to UX SC -PCL and UX CS -PCL exhibited higher yield stress (1.3-1.5Â, p < 0.05; Table 1), increased Young's modulus (3.1-3.6Â, p < 0.001), and higher fracture stress (1.8-3.4Â, p < 0.001) and strain (1.3-1.5Â, p < 0.001). The yield strain of UX HP -PCL was comparable with that of UX CS -PCL (p > 0.05) and larger than that of UX SC -PCL (0.9Â increase, p < 0.001). Comparing UX HP -PCL and HP-PCL, it was found that uniaxial stretching transformed the lm's stress-strain curve, and the latter experienced a reduced stress aer achieving the proportional limit (Fig. 3b). Uniaxial stretching conferred UX HP -PCL with a signicant increase in yield stress (2.5Â increase, p < 0.001; Table 1), Young's modulus (2.1Â increase, p < 0.001) and ultimate stress (2.6Â increase, p < 0.001) without sacricing the lm's elasticity (p > 0.05). Moreover, the lm thickness of UX HP -PCL was reduced (0.2Â of HP-PCL, p < 0.001), resulting in a low weightto-surface ratio of $5 mg cm À2 .
Previously, micropatterned ridges/grooves have shown robust regulation on vascular cell behavior. 6,8 However, their application for functionalizing vascular TE scaffolds remained limited in research, 12,37 as current fabrication techniques, such as so-lithography and direct laser writing, were oen applied on stiff and/or thick substrates, and involved tedious procedures and usage of various solvents. 14,15,38 Comparatively, our study realized micropatterns on a exible and light-weight PCL lm via uniaxial stretching. This development imparted the microstructures with exible lm properties to facilitate rolling into vascular TE scaffolds and also enabled solvent-free and reproducible topographical geometries with simple procedures. Previously, although electrospinning was considered as an established strategy that could realize the specic circumferential alignment of native arteries, it has not yet generated a vascular gra with sufficient burst strength from biological materials alone. 37,39 Interestingly, the present method of uniaxial stretching, in addition to surface structuring, has also enhanced lm substrates with improved mechanical performances. Based on Laplace's law, the theoretical burst strength of UX HP -PCL-based vascular TE scaffolds at a diameter of 2.3-4.6 mm could be >5000 mm Hg, which was higher than those of scaffolds based on UX SC -PCL and UX CS -PCL (1.5-3.7Â increase, p < 0.001; Table 1), and is sufficient to withstand the normal inside pulsatile pressure of the human mammary artery. 27

Flexible PCL lm micropatterns to align MSCs organization
We further investigated the biological functions of uniaxially stretched PCL lms. The in vitro isolated and expanded MSCs grew as plastic-adherent cells and revealed a typical stem cell immunophenotype (Table S1 †), which was negative (<1.5%) for the haemopoietic and endothelial markers: CD19, CD45 and CD34, and positive (>95%) for the mesenchymal markers: CD73 (SH-3, SH-4) and CD105 (SH-2) and cell adhesion molecules: CD90. Culturing MSCs on PCL lms, CLSM images of live-cell cytoplasmic staining showed the ordered cell organization on UX SC -PCL and UX HP -PCL in a direction along the ridges (Fig. 4a). Cellular orientation and morphological analysis showed that MSCs cultured on UX SC -PCL and UX HP -PCL achieved a signicant increase in alignment (2.7Â increase, p < 0.001; Fig. 4b) and elongation (0.5Â increase, p < 0.001) efficiencies as compared to those of the un-stretched PCL lm groups, suggesting that the cells obtained a capability of self-ordering into the aligned organization as observed in the tunica media. 3 In contrast, MSCs grew on UX CS -PCL and the un-stretched PCL lms in a random fashion. It should be noted that MSCs grown on CS-PCL exhibited more localized cell alignment than those of the cells on SC-PCL and HP-PCL. This could be attributed to the higher number of cells obtained on SC-PCL, as MSCs with in vitro culture approaching conuence have been known to self-organize into orientation within a small area. 15 However, this localized alignment was inconsistent in the cell orientation (Fig. 4a) and has not resulted in CS-PCL with better cell alignment efficiency as compared to others (Fig. 4b). It should also be noted that compared to CS-PCL, UX CS -PCL did not impart MSCs with an increase in alignment efficiency (p > 0.05), although more elongated cell morphologies were observed. Among the three uniaxially stretched PCL lms, UX HP -PCL elicited ordered MSCs growth with the highest alignment (1.3-3.9Â, p < 0.001) and elongation (1.1-1.4Â, p < 0.01) efficiencies.  During the cell-substrate interaction, ridge depth is crucial for determining to what degree the cells can recognize given lateral dimensions. 40 In agreement with this, we observed that MSCs achieved effective alignment on UX SC -PCL and UX HP -PCL but failed to align on UX CS -PCL. Comparatively, although UX CS -PCL had an inter-ridge distance similar to that of UX HP -PCL, it revealed a signicantly lower ridge depth than that of UX HP -PCL and UX SC -PCL, indicating that a threshold of ridge depth (e.g. $116 nm of UX CS -PCL) might exist for the aligned MSCs on PCL lms. In addition to ridge depth, inter-ridge distance could also affect MSC alignment on PCL lms. Although UX SC -PCL possessed deeper ridges/grooves than UX HP -PCL, it led to a lower efficiency of cell alignment, suggesting that a wider interridge distance (e.g. >50 mm of UX SC -PCL) could hinder substrates in eliciting cellular alignment. 33,34 These observations would facilitate the understanding on how uniaxially stretched PCL lms were able to modulate MSC alignment, and in this study, UX HP -PCL was thus selected as a candidate for further investigation towards the construction of functionalized vascular TE scaffolds.

Flexible PCL lm micropatterns to upregulate MSCs differentiation
Additional in-depth inuences of UX HP -PCL have been linked to MSCs myogenic differentiation towards vascular TE applications. Fig. 5a shows qRT-PCR analysis of SMCs contractile gene markers' expressions in MSCs, which revealed upregulated levels of ACTA2 (2.7Â increase, p < 0.001), CNN1 (3.8Â increase, p < 0.001) and MYH11 (2.4Â increase, p ¼ 0.15) for the cells cultured on UX HP -PCL for ve days compared with those of the HP-PCL groups. These observations demonstrated the altered genetic expression of MSCs on UX HP -PCL. It has been known that, in healthy arteries, SMCs possess the highly aligned organization with elongated shapes, whereas during in vitro culture transform into synthetic phenotypes with good spread and random organization. 8,30 Such transition was similar to the observed changes of cell morphology and genetic expression for MSCs cultured on UX HP -PCL and HP-PCL.
With protein detection, CLSM images of immunocytochemistry-labeled MSCs revealed positive expressions of SMCs contractile markers: SM-a-actin, calponin and SM-MHC on both UX HP -PCL and HP-PCL lms (Fig. 5b-d). In contrast, no cellular uorescence was observed for the corresponding negative control with no indication that nonspecic immune conjunction occurred. In the same manner, FACS quantitative analysis conrmed the positive expression of SMCs contractile markers for MSCs on UX HP -PCL and HP-PCL (positive cell number of each marker: <0.5% for Neg Ctrl vs. >30% for PCL lm groups, p < 0.001; Fig. 6a-c). When comparing UX HP -PCL and HP-PCL groups, CLSM images show that the staining of SM-a-actin, calponin and SM-MHC was higher in the UX HP -PCL groups with higher levels of SM-a-actin expressed than late-term myogenic differentiation protein markers such as SM-MHC (Fig. 5b-d).
Our results through qRT-PCR and the immunocytochemistry analyses using CLSM and FACS demonstrated that UX HP -PCL upregulated the expression of SM-a-actin, calponin and SM-MHC in MSCs at both the gene and protein levels. These markers associated with SMCs, and they are expressed in early, middle and late SM differentiation. 8 Moreover, CLSM images of MSCs on UX HP -PCL revealed the aligned organization of SM contraction-related laments in a controlled direction along the ridges (Fig. 5b-d), whereas randomly organized laments were observed for the cells on HP-PCL. Such induced lament alignment on UX HP -PCL might also facilitate the contractile SMC-like phenotype of the MSCs, as a recent study on cardiomyocytes indicated that contraction behavior relied on both the positive expression and ordered organization of cellular stress laments. 32 Thus, UX HP -PCL could be used to promote MSCs as potential cell sources for vascular TE applications. Despite the enhanced response of MSCs on UX HP -PCL, it should be noted that these MSCs, as compared to Pos Ctrl, still exhibited a lower expression of SMCs markers at the positive cell number (vs. UCPVCs: 0.9Â for SM-a-actin, p < 0.001; 0.8Â for calponin, p < 0.01; and 0.9Â for SM-MHC, p ¼ 0.54; Fig. 6ac), although their average cell expression intensity for SM-aactin and SM-MHC was comparable (p > 0.05; Fig. 6d and f).
To understand the reasons behind the observed MSCs differentiation, we investigated the alterations of subcellular organelles on UX HP -PCL. Through cytoskeletal F-actin and nucleus DNA labeling, Fig. 7a shows the CLSM images of MSCs on UX HP -PCL with typically spindle-like morphologies, reduced spreading and ordered cytoskeleton bers, which have been known to associate with SMCs in differentiated status. 8,30,31 In contrast, MSCs spread well with random cytoskeletal laments on HP-PCL. Correspondingly, MSCs nuclei on UX HP -PCL deformed into a more elongated morphology along the ridges with increased alignment efficiency (2.5Â higher, p < 0.001; Fig. 7a and b) and reduced CNSI (0.3Â lower, p < 0.001) as compared to those of the HP-PCL groups, respectively.
Geometric structures to trigger cell differentiation have been known via mechanotransduction pathways, 32,41 during which cells recognized the substrate geometrices with altered focal adhesion distribution, migration and extension to align into elongated shapes. Such elongation of cell shape during direct mechanotransduction could result in aligned cytoskeletonal bers triggering nucleus deformation, leading to structural adjustment of the DNA and nucleus pores, changed accessibility of transcriptional factors, and nal alterations in gene expression. [41][42][43] In this study, MSCs cultured on UX HP -PCL demonstrated altered cell morphology, cytoskeletonal reorientation and nucleus elongation, suggesting that the upregulation of SM-a-actin, calponin and SM-MHC in MSCs could be, in part, attributed to the ridges/grooves of UX HP -PCL via direct mechanotransduction. Such a differentiation effect of geometric cues on MSCs has not been reported, which agreed with the previous observations on SMCs, including marker expressions of SM-a-actin, 6 calponin, 6 smoothelin 6 and SM-MHC. 6,8 In addition to topographical geometries, UX HP -PCL, compared to HP-PCL, had a higher Young's Modulus, which might also have accounted for the elevated SMCs marker expression for MSCs cultured on UX HP -PCL as recent evidence showed that the stiffer substrate could result in MSCs with the up-regulated expression of SM-a-actin and calponin-1. 31 On the other hand, the source of stem cells might inuence the selectivity of gene expression on geometric cues. Bone marrow-divided stem cells (e.g. human fetal bone MSCs) with cultures in non-differentiating medium have been reported to differentiate towards myocardial lineages on anisotropic bers 44,45 and ridges/grooves, 15,34,43 whereas human embryonic stem cells on both patterned nanoscale ridges/grooves 46 and aligned electrospun bers 47 tended to enhance the differentiation towards neurons. However, these studies lacked investigation on the SMCs contractile markers, although MSCs have shown promise for vascular TE application. 17,18 Herein, we demonstrated that the ridges/grooves of UX HP -PCL could induce improved gene and protein expressions of SM-a-actin, calponin and SM-MHC in MSCs. Thus, this conrms the ability of geometric cues in modulating MSCs differentiation along myogenic lineage. 15,34,[43][44][45]

Construction of functionalized 3D tubular scaffolds with double-surface micropatterns
We nally applied UX HP -PCL to construct 3D tubular vascular TE scaffold ($2.3 mm in diameter; Fig. 8a and b) through the rolling of the lm around a rod, and then attaching it through heat welding. SEM images of the inner and outer scaffold surfaces revealed similar circumferential anisotropy of ridges/ grooves in a direction perpendicular to the tube (Fig. 8b). Seeding MSCs on both surfaces of the opened UX HP -PCL-based scaffold, an engineered hybrid vascular wall construction comprising of layered MSCs/PCL/MSCs was achieved with anisotropic organization along the ridges (Fig. 8c). Previously, different strategies, such as mechanical stimuli 17 and biochemical factors, 18 have been studied for functionalized generation of vascular gras. However, in this study, we developed exible lm micropatterns of UX HP -PCL. Vascular TE scaffolds with incorporation of these geometric cues could regenerate the ordered architecture of tunica media with aligned MSC layers and regulate MSCs towards contractile SMCs-like phenotypes in a more controllable and cost effective way, as well as avoiding the introduction of side effects such as atherosclerosis from biochemical cues. 23,46 Beneting from uniaxial thermal stretching, UX HP -PCL achieved improved mechanical performance without sacricing the lm's elasticity, which ensured the safety considerations for engineered vascular TE scaffolds and provided them the unique tubular patency to facilitate operative storability and simplicity of  surgical handling when the scaffolds were in contact with biological liquids (Fig. S1 †). 2,4 Furthermore, our previous study demonstrated that the micropatterns of UX HP -PCL had enhanced stability against erosive environments; 26 thus, they could provide long-term cell regulation for the engineered vascular TE scaffolds even aer implantation in vivo. In short, the UX HP -PCL-based vascular TE scaffold possessed unique mechano-and bio-functions. Moreover, with the incorporation of MSCs, it has the potential to be used for generating engineered, autologous, and commercially available vascular gras with the desired tunica media architecture of native vessels.

Conclusions
Herein, we presented the development of exible lm micropatterns for the functionalized construction of vascular TE scaffolds. PCL lms aer uniaxial stretching resulted in topographical micropatterns comprised of ridges/grooves, and mechanical enhancement in terms of yield stress, Young's Modulus and fracture stress without sacricing the lm's elasticity. Culturing on such PCL lm micropatterns, MSCs selfaligned along the ridges into more elongated morphology as compared to that of the unstretched lm groups. MSCs obtained a contractile SMC-like phenotype with upregulation and ordered organization of SM-a-actin, calponin, and SM-MHC laments. The PCL lm micropatterns could be rolled into a small-diameter 3D tubular scaffold with circumferential anisotropy of ridges/grooves and for the incorporation of MSCs, which facilitated a hybrid sandwich-like wall construction with ordered cell architecture, similar to that of the tunica media. These results provide insight into how geometric cues are able to regulate stem cells with desired functions and have signicant implications for designing functionalized vascular TE scaffolds with appropriate topographical geometries for guiding tunica media regeneration with microscale control of cell alignment and genetic expression.