3D printed grafts with gradient structures for organized vascular regeneration

Synthetic vascular grafts suitable for small-diameter arteries (<6 mm) are in great need. However, there are still no commercially available small-diameter vascular grafts (SDVGs) in clinical practice due to thrombosis and stenosis after in vivo implantation. When designing SDVGs, many studies emphasized reendothelization but ignored the importance of reconstruction of the smooth muscle layer (SML). To facilitate rapid SML regeneration, a high-resolution 3D printing method was used to create a novel bilayer SDVG with structures and mechanical properties mimicking natural arteries. Bioinspired by the collagen alignment of SML, the inner layer of the grafts had larger pore sizes and high porosity to accelerate the infiltration of cells and their circumferential alignment, which could facilitate SML reconstruction for compliance restoration and spontaneous endothelialization. The outer layer was designed to induce fibroblast recruitment by low porosity and minor pore size and provide SDVG with sufficient mechanical strength. One month after implantation, the arteries regenerated by 3D-printed grafts exhibited better pulsatility than electrospun grafts, with a compliance (8.9%) approaching that of natural arteries (11.36%) and significantly higher than that of electrospun ones (1.9%). The 3D-printed vascular demonstrated a three-layer structure more closely resembling natural arteries while electrospun grafts showed incomplete endothelium and immature SML. Our study shows the importance of SML reconstruction during vascular graft regeneration and provides an effective strategy to reconstruct blood vessels through 3D-printed structures rapidly.

Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Cardiovascular disease is the leading cause of death in the world.The primary cause of cardiovascular disease is occlusion of the arteries, which causes a reduced blood flow to the target organs.Vascular surgeries, such as coronary artery bypass grafting surgery and arteriovenous shunts, require biologically responsive vascular grafts.Autografts, such as saphenous veins and internal mammary arteries, are the goldstandard grafts used to treat vascular occlusions.Usually, the first surgery is needed to harvest patients' own arteries or veins, which are then used as grafts to bypass clotted segments of arteries to restore blood flow.These surgeries can cause donor stie morbidity in patients.Meanwhile, patients who have no usable autologous grafts will lose opportunities for such vascular surgeries [1][2][3][4][5].Synthetic vascular grafts not only avoid the surgery to harvest autologous blood vessels, but also provide additional surgery opportunities to patients.Polyethylene terephthalate (Dacron)-and expanded polytetrafluoroethylene (ePTFE)-based synthetic grafts are quite successful in clinical settings to replace large arteries (>6 mm) [6][7][8].However, vascular grafts made of Dacron or ePTFE demonstrate disappointing clinical results when working in small-diameter (<6 mm) arteries, such as coronary arteries where the rate of blood flow is low, due to vascular thrombosis and stenosis [9][10][11].Thus, for patients who need vascular grafts to bypass vascular occlusions, synthetic grafts suitable for small-diameter arteries are in great need.Although significant progress has been made in manufacturing technologies, materials, and surface coating techniques for small-diameter vascular grafts (SDVGs), there are still no commercially available SDVGs in clinical practice.
Natural arteries have three distinct layers: the adventitia, the media, and the intima.The adventitia layer has a high percentage of collagen fibers, which increases the strength of the natural blood vessels and inhibits excessive dilation and rupture under physiological pressure [12,13].The media mainly comprises smooth muscle cells (SMCs), collagen, and elastin and can adapt to the contraction and relaxation of blood vessels under physiological pressures [14].The intima is mainly composed of endothelial cells (ECs) and plays an important role in preventing thrombosis.However, the regenerated ECs will gradually peel off and cause thrombosis if the microenvironment for ECs is not ideal.The media functions during contraction and dilation of blood vessels.In addition, the media can not only maintain healthy endothelium but also inhibit stenosis.For example, contact-dependent communication between SMCs and ECs can promote EC proliferation [15].In contrast, the injured SMCs can express matrix metalloproteinase and inflammation-associated genes [16].The injured SMCs will also over proliferate and cause stenosis [17].Electrospinning is currently a mainstream technology to fabricate SDVGs [18][19][20][21].However, due to small pore sizes caused by a deposition of micro-/nano-fibers, cellular infiltration and tissue regeneration within the vascular grafts is quite limited [22,23].With the time of implantation, the regenerated tissues within the grafts gradually regress and the vascular walls are calcified, finally resulting in the failure of the graft [24].However, till now, there are rare SDVGs with specific designs for rapid media regeneration.
The arrangement of natural extracellular matrix (ECM) collagen fibers in human tissues plays a significant role in maintaining tissue mechanics and functionality [25,26].In the three-layer structure of natural arteries, the different orientations of collagen fibers in each layer can regulate the mechanical behavior of the vessels and align the cells [27,28].The alignment of cells plays a crucial role in the maturation and regeneration of functional tissues [29].The collagen fibers in the smooth muscle layer (SML) are arranged circumferentially at an angle of 65 • [30], and SMCs also align circumferentially [31].Therefore, a SDVG that can guide the axial alignment of SMCs can promote blood vessel regeneration [32].
Many studies put emphasis on reendothelization of SDVGs and various strategies are used to promote endothelium regeneration for anti-thrombosis [33][34][35][36].However, the reconstruction of the SML plays a key role in vascular regeneration.Accelerating the SML reconstruction facilitates rapid restoration of compliance to adapt to the pulsation of blood vessels and induce spontaneous endothelialization, finally ensuring its long-term patency.Therefore, the rapid reconstruction of the SML is the first step for vascular regeneration.It is a good choice to increase the cell infiltration by increasing the pore size and porosity of scaffolds to rapid remodel SML, but the mechanical properties of scaffolds will decrease.Single-phase scaffolds are difficult to meet both mechanical requirement and high cell infiltration.However, in biological manufacturing, it is a great challenge to design and fabricate polymer multiphase scaffolds with adjustable pore size and porosity.
Melt electrowriting (MEW) is a high-precision additive manufacturing technology that can control the fiber diameter (800 nm-150 µm) and accurately deposit fibers to form various three-dimensional (3D) topologies structures, which is widely used in biomedicine and tissue regeneration engineering [37][38][39].Many researchers have used this technology to fabricate ultrafine fiber porous networks which can simulate natural ECM to regulate cell behavior, such as guiding cell directional growth and depositing collagen fibers [40,41].When the collection platform is changed to a cylindrical rod (diameter ⩾ 1 mm) [42,43], ultrafine fiber tubular networks (UFTNs) with controllable porosity and pore size can be manufactured.The UFTNs have been widely used in tissue engineering, such as renal tubules [44] human breast cancer [45], periosteum tissues [46].
Herein, we designed and manufactured a bilayer SDVG with a specific purpose for SML reconstruction (figure 1).The inner layer of the graft had the same fiber crossing angle (65 • ) as the collagen fiber arrangement in the media of the arteries.
Mechanically, it could contract and dilate as the media.In addition, it had large pore sizes (200 µm) and high porosity (41.61%), which allowed the rapid infiltration of vascular cells and guided their circumferential alignment.The outer layer, mimicking the adventitia of the arteries, had denser fiber intersections with smaller pore sizes (<50 µm) and a low porosity (12.52%).It prevented the over dilation of the vascular grafts and blood leakage.To demonstrate the superiority of our design, a rat aorta vascular graft interposition model was used.After one month of implantation in vivo, the bilayer graft was spontaneously and completely reendothelialized and promoted the SML maturation.It regenerated the three-layer structures of nature blood vessels and exhibited pulsation like that of the native arteries.The compliance of the regenerated blood vessels was 8.9%, which was close to that of the nature arteries (11.36%).

Material
The raw material for MEW was PCL pellets (CAPA6800, Perstorp Ltd, Sweden) with a molecular weight of 80 000 g•mol −1 and a melting point of 60 • C.

Design and fabrication of the ultrafine fiber tubular network
The UFTN was designed using the EFL_PotatoE software (Suzhou Intelligent Manufacturing Research Institute, Suzhou, China).The software provides two basic unit patterns (rectangular and diamond), and it allows for the setting of geometric parameters for the porous diamond tubular network, including the length of the tubular network (start and end points), fiber crossing angle (0 • -180 • ), and the number of fiber crossing nodes on the circumference.The software then generates the corresponding G-code.After determining the diameter of the tubular network, the network's porosity was controlled by adjusting the number of fiber crossing nodes on the circumference.
A custom-built MEW printer (EFL-BP6601, Suzhou Intelligent Manufacturing Research Institute, Suzhou, China) was used to print the UFTNs.PCL pellets were put into syringe with a 150 µm nozzle.When printing non-woven and regular network pattern, the nozzle and a syringe were preheated to 120 • C and 90 • C, respectively.The printing environment was maintained at room temperature 20 • C. The printing parameters are in table 1. Electrospun vascular grafts were fabricated with 14% (w/v in 2,2,2-trifluoroethanol) PCL solution at a flow rate of 30 µl per min under a voltage of 12 kV.A 2 mm thick mandrel was used to collect fibers at a speed of 500 rpm.The fabricated grafts had an inner diameter of 2 mm and wall thickness of 200 µm.The electrospun vascular grafts were used as a control in this study.

SEM scanning of UFTNs
The UFTNs were sputter coating with gold utilizing a sputter coater (Ion Sputter E-1045, Hitachi, Tokyo, Japan) for 2 mins.Then the UFTNs were imaged by electron microscopy at a 3 kV acceleration voltage (SEM, Hitachi SU-8010, Japan).Image J software was applied to estimate the fiber diameter (n = 50), pore diameter (n = 50), and porosity (n = 5).

Assembling vascular grafts using UFTNs
Based on the diameter of the abdominal aorta, the diameter of the rotating axis was meticulously controlled, ensuring precise regulation of the inner diameter of the outer layer and the outer diameter of the inner layer in the MEW printing process.With a consistent number of stacked layers, the inner layer maintains a nearly uniform thickness, promoting a stable fit between the outer and inner layers.The substantial friction between the layers offers an inherent anti-slip function.

Mechanical test
A universal tensile machine (UTM2203, China) equipped with a 20N weighing sensor was used to measure the mechanical properties of the axial and radial directions of the graft.For the axial mechanical performance test of the graft, a 20 mm tubular graft was clamped onto the universal tensile machine's jaws, ensuring that each test sample's tensile length was 10 mm.Then, the sample was stretched at a 10 mm min −1 speed until it fractured.The axial modulus of elasticity was obtained between 1% and 3% strains.For the radial mechanical performance test of the graft, a 10 mm long graft was clamped with two 'U'-shaped fixtures, and the samples were stretched at a speed of 10 mm min −1 until they fractured.The burst pressure of the graft was evaluated using Laplace's law [47], where the graft underwent failure when the thin-walled tube was subjected to a pressure Pi.The Laplace equation was used to calculate the burst pressure, where D is the outer diameter of the graft, and L is the length of the graft.

Suture retention force test
The 10 mm graft was sutured by a 6-0 suture thread with the needle (round 3/8 arc, 1 × 5, nylon thread, Ningbo Medical Suture Needle Co., Ltd, China) after wetting with physiological saline.The graft's thin wall at a 2-3 mm distance from the edge of the graft was pierced by the needle and tied.The other end of the graft and the suture thread were fixed to the universal tensile machine, and the samples were kept vertically.The upper jaw of the machine was then stretched at 50 mm min −1 speed until the suture thread was pulled out.

Vascular graft implantation in vivo
All animal experiments were approved by the Chinese Institutional Animal Care and Use Committee at Sir Run Run Shaw Hospital, School of Medicine, Zhejiang University.Male Sprague Dawley rats (age: 8-10 weeks; weight: 350 grams-400) were provided by Laboratory Animal Center in Zhejiang Province (Hangzhou, China).Ten rats were used for aortic implantation of vascular grafts fabricated with electrospinning (the Espin groups) and another 10 rats were used for aortic implantation of vascular grafts fabricated with 3D printing (the 3D groups).Briefly, the abdominal aorta was exposed through a midline incision on the abdomens of rats following anesthesia via isoflurane inhalation.After blocking the blood flow of the aortas at distal and proximal ends with microvascular clamps, the aortas between the two clamps with a length of around 10 mm were cut off.The electrosup or 3Dprinted grafts were anastomosed to the aorta end-to-end with 8-0 nylon monofilament sutures and the microvascular clamps were removed to recover the blood flows after anastomosis.The rats were sacrificed for sample harvest 14 or 30 d post the surgeries.

Ultrasonic evaluation of the implanted vascular grafts
The blood flow and inner diameter of the implanted grafts were checked with a small animal ultrasound imaging system (VisualSonics, Vevo 3100, FUJIFILM) on day 14 and 30 after implantation.To detect the blood flow of the grafts, the aortas were exposed through an incision on the abdomen and ultrasound probes worked by directly touching the grafts after anesthesia of the rats.B mode, color mode, and PW mode images in cross sections were obtained.The images of B mode were used to measure the diameters of the grafts; the images of color mode were used to evaluate graft patency; and the images of PW mode were used to calculate the mean velocity of the blood flow.

Chemical staining
Samples were rinsed with phosphate buffered saline (PBS), fixed in 4% paraformaldehyde at 4 • C overnight, and soaked in 30% sucrose solution at 4 • C for 24 h.The samples were then embedded vertically into the freezing media (Optimal Cutting Temperature Compound, Sakura Finetek, USA), frozen at −80 • C, and cryo-sectioned.Hematoxylin and eosin (H&E), Masson's trichrome, and elastic Verhoeff-Van Gieson were used to stain tissue slides.H&E staining images of three different samples (n = 3) in one group were used to measure luminal areas and wall thicknesses.

Biochemical assays
The elastin and collagen contents of the explanted samples and the native abdominal aortas were measured with the FastinTM Elastin Assay kit (Biocolor, UK) and Total Collagen Assay kit (Abcam, USA), respectively.Samples were weighed first, and then minced finely before digestion for elastin or collagen extraction.After quantification, the elastin and collagen contents were normalized to the wet weight (µg mg −1 ) of each sample.Three different samples (n = 3) in one group were measured.

Assessment test of compliance in regenerating blood vessels
A custom-designed device was used to measure circumferential mechanical properties of the vascular grafts that were harvested on day 30 after implantation.The explanted vascular grafts were fixed onto the device through two 19 G needles.A syringe pump was connected to one end of the device, and a pressure monitor was connected to the other end of the device.PBS was infused at a rate of 2 ml hr −1 and the pressure within the grafts was recorded.A digital camera was used to record the outer diameters of the explanted vascular grafts.Stress-stretch curve and compliance of the explanted vascular grafts were calculated, respectively.The following equation: C = (D high − D low )/D low (P high − P low ) was used to calculate the compliance, which was expressed as a percentage of diameter change per 100 mmHg (%/100 mmHg).

Macrophage polarization on 3D-printed scaffolds in vitro
Rats were injected intraperitoneally with 1 ml of 6% starch dissolved in PBS to induce sterile peritoneal inflammation.3 d later, the peritoneum of the rats was washed with 10 mL PBS using a 21-G needle and the PBS was collected and centrifuged to pellet cells.Red blood cells were lysed before cell seeding.The macrophages were then seeded on 3D-printed or electrospun PCL scaffolds and cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum for 3 d.The supernatant was collected to measure concentrations of IL-10, TNFα, and IL-1β with enzyme-linked immunosorbent assay kits (Proteintech, Wuhan, China) and the macrophages were observed after labeling with carbocyanine dyes DiO (Invitrogen, USA).

Statistical analysis
All data was presented as mean ± standard deviation.Each experiment was repeated 3 times at least.The statistical significance was calculated with one-way ANOVA.The Turkey's post-hoc test or the two-tailed Student-t test was used to check the significance between two different groups.P < 0.05 was statistically significant in this study.

Fabrication of biomimetic bi-layered vascular grafts
To facilitate rapid SML regeneration, we designed and fabricated a biomimetic bilayer small-diameter vascular graft (BSDVG) (figure 2(b)).The inner layer and outer layers were printed separately by MEW rotational printing technology (figure 2(a)) and then assembled into the vascular grafts with a lumen diameter of 1.5 mm by way of pipe casing.As shown in figure 2(c), there were two scaffolds in the inner layer and one scaffold in the outer layer, exhibiting gradient pore sizes and porosity from the inside to the outside.The inner scaffold (IS) had a regular pattern and relatively large pore sizes, which could enhance cell infiltration and direct cell growth.On the other hand, the outer scaffold (OS) mimicked the adventitia of the natural arteries to improve the strength of BSDVG with relatively small pore sizes and numerous fiber intersections.
The pore size and porosity of tissue engineering scaffolds are critical factors in determining induced tissue regeneration.Large pore size or high porosity in scaffolds facilitates effective nutrient supply and removal of metabolic waste.The pore sizes of scaffold ranging from tens to hundreds of micrometers are optimal for cell infiltration and growth [48][49][50].However, ultra-larger pore size or higher porosity may reduce intercellular signal transmission and compromise the mechanical performance of the scaffold.Furthermore, the SMCs in natural blood vessels are aligned along the circumferential direction, requiring the IS to guide cell growth along the circumferential direction.Studies have shown that when the fiber diameter is similar to the cell size, it can induce cell-directional growth along the fibers [40,51].Hence, we designed and printed three biomimetic ISs with cross-fiber angles of 65 • and pore sizes (distances between adjacent nodes on the circumference) of 200 µm, 300 µm, and 500 µm (IS200, IS300, IS500), with an initial printing diameter of 25 µm (figure 2(d)).Although MEW can deposit fibers in the range of micrometers or submicrometers, forming porous 3D scaffolds with adjustable pore sizes and porosity, the minimum distance between fibers is limited by some factors, such as residual charges in the deposited fibers [52,53].This limitation generally results in functional MEW scaffolds with pore sizes above 100 µm [37].When pore size was reduced to 100 µm, severe fiber deposition misalignment occurred due to the influence of residual charges in the deposited fibers (figure S1(a)).However, MEW resolution is enough for meeting the pore size and porosity of the designed scaffold.As shown in figure 2(h), as the pore size decreased, the fiber diameter increased continuously, with the fiber diameter of IS200 ((35 ± 2.8) µm) increasing by 1/3 compared to the original diameter.Due to the repeated heating of the deposited fibers by the high-temperature nozzle heating block (90 • C) in the printing process and the slower heat dissipation with smaller pore size, the crystallization and solidification of the fibers was slowed down (figure S2(b)), resulting in fusion and collapse of the fibers between layers (figure 2(d)).In addition, with an increase in the number of layers (12 layers), 'spherical' structures appeared at the nodes (figure S1(a)-(i)), which would reduce the strength of the scaffold [54].Therefore, the presence of molten protrusions at the nodes and the high local temperature field limits the number of layers that can be stacked (around 10 layers).The SDVGs require high mechanical properties, but increasing the number of printed layers cannot increase the strength of the scaffold or improve the cell adhesion area.Therefore, two IS200 scaffolds were assembled to form the inner layers of the BSDVG (with a thickness of approximately 200 µm) to improve their mechanical strengths.The large pore size of the inner layer of SDVG can accelerate cell infiltration and not impede cell proliferation, maturation, and production of ECM.However, if the pore size is large, the surface area of the scaffold will decrease relatively, and the adhesion points provided to the cells will also reduce.In the designed scaffolds, the IS200 ensures a large pore size (10 times greater than the cell size), and its fiber surface area is also the largest.Two IS200 scaffolds could be assembled to form the BSDVG inner layer (with a thickness of approximately 200 µm) to improve strength and increase fiber surface area.Therefore, the IS200 is the best choice as the IS of BSDVG.
The BSDVG needs to have both sufficient strength and the ability to prevent blood leakage after implantation.Therefore, to ensure that BSDVG can withstand abnormal physiological vascular mechanics without bursting when implanted in the body for a long period, the pore size of the outer layer of BSDVG needs to be smaller than that of the ISs, and the cross points of the fibers should be denser.The jet-whipping effect in the MEW process could be utilized to create non-woven fabrics characterized by dense fiber intersections and small pore size.We designed a printing path with a fiber crossing angle of 65 degrees and a distance of 50 µm between adjacent nodes on the circumference to print the OS of the BSDVG by filling the gaps between the fiber paths with the jet-whipping effect.As shown in figure 2(e), the morphology of OS was characterized.The results showed that due to the higher local temperature, there were fusion and spherical bead phenomena between the fibers.As shown in figures 2(i) and S1(c), the fiber diameter and pore size of OS were measured, with an average fiber diameter of (16.79 ± 6.3) µm and an average pore size of (25.84 ± 6.7) µm, which is almost smaller than 50 µm.On the other hand, as shown in figures S1(d) and (e), the electrospinning scaffold (ES) had a smaller average diameter ((0.62 ± 0.28) µm) and pore size ((0.55 ± 0.12) µm) compared to OS, and the pore size was smaller than the cell size.Compared to the morphology of the OS, the ES was densely packed with no fiber fusion.As shown in figure 2(j), comparing the porosity of the OS and ES, the ES had a high porosity of 79.42%, while the OS had a porosity of 12.52%.Although ES has a high porosity, its pore size is less than 1 µm (about one tenth of the cell size), resulting in poor cell infiltration.The OS with a large fiber diameter and minor porosity can promote BSDVG with high strength.The IS200 had a higher porosity than the OS (41.61%), nearly four times.The results indicate that the designed BSDVG was a multi-phase vascular graft with a gradient pore size and porosity from the inner to the outer layers.

Mechanical properties of biomimetic bi-layered vascular grafts
The mechanical strength of BSDVG is essential to ensure successful vascular regeneration.Before implantation, it is necessary to evaluate the mechanical properties of the manufactured BSDVG to meet transplantation requirements.The BSDVG could recover its original shape after repeated axial compression (figure 3(a)).Although it exhibited automatic resilience to folding back to its original shape (movie 1), further quantitative assessment of the mechanical strength of the graft was conducted through the uniaxial tensile test (figure 3(b)).As shown in figure 3(c), the stress-strain curves of the ISs with different pore sizes were used to calculate the axial tensile modulus, strength, and elongation at break.As shown in figures 3(d) and (e), the tensile modulus and strength continuously decrease with the increase in pore size of the ISs.The fracture elongation of the ISs increases with the increases in pore size (figure 3(f)).Among them, the axial tensile modulus and strength of the IS200 were the largest, reaching (1.14 ± 0.096) MPa but slightly lower than the healthy human artery ((1.44 ± 0.87) MPa) [55].The tensile strength of IS200 was (367.24 ± 32) kPa, also slightly lower than the maximum stress of healthy human arteries (0.51 MPa-3.20 MPa) [55], and the fracture elongation was (0.71 ± 0.18), exceeding the range of fracture elongation of healthy human arteries (0.54 ± 0.25).The suture retention force is also a key factor for the successful implantation of artificial blood vessels.The suture retention force of the IS200 was only (0.59 ± 0.02) N, slightly lower than that of natural human arteries ((1.4 ± 0.01) N) [56] and that of mouse carotid arteries ((0.64 ± 0.04) N) (figure 3(h)).Furthermore, after implantation of BSDVG in vivo, they will endure long-term pulsation of normal blood flow or pressure beyond the normal blood pressure range.The bursting pressure of BSDVG is also a critical factor for successful implantation.The bursting strength of the BSDVG was evaluated through circumferential ringlet tensile experiments.As shown in figure 3(i), with the increased pore size, the bursting pressure of the ISs decreased.The IS200 had the highest bursting pressure of (2345.2 ± 134.14) mmHg, far less than the typical carotid artery bursting pressure of 5000 mmHg [57].The results indicated that the mechanical strength of the IS200 was insufficient.As shown in figure S1(f), MEW can create programmable porous tubular scaffolds.The axial elastic modulus and strength of OS were (2.3 ± 0.22) MPa and (0.311 ± 0.062) MPa, respectively.Adding an OS could compensate for the insufficient mechanical strength of the IS200 (movie 2).
To fully demonstrate the gradient structures for improved vascular regeneration, we set up two experimental groups: 3D-printed BSDVG (IS200/OS) with the gradient pore sizes and electrospun vascular graft as the experimental control group.As shown in figure 3(g), the tensile modulus and strength of the IS200/OS were (5.6 ± 1.2) MPa and (0.76 ± 0.064) MPa, respectively, slightly higher than natural arterial vessels.In comparison, the electrospun vascular graft had a tensile modulus and strength of (0.915 ± 0.147) MPa and (1.1 ± 0.242) MPa, respectively.The experimental results indicated that 3D-printed and electrospun vascular grafts met the implantation strength requirements.The 3D-printed graft had a bursting pressure of (5001 ± 266) mmHg, lower than the electrospun vascular graft ((8625 ± 456) mmHg) but still sufficient to withstand physiological blood vessel pressures.

Fast remodeling of 3D-printed grafts after in vivo implantation
To test if the 3D-printed BSDVG (IS200/OS) could promote vascular regeneration in vivo, we implanted the 3Dprinted into the abdominal aortas to evaluate their performances in vivo with electrospun vascular grafts as controls.The 3D-printed vascular grafts were well perfused with blood and showed red color after implantation (figure 4(a) and movie 3).Interestingly, no leakage happened to the 3D-printed vascular grafts despite their highly porous structures.This is because bi-layer graft has gradient porosity and pore size.When implanted in vivo, the gradient porosity traps fibrinogen in the blood with micro-thrombosis within pore structures of the graft, forming a composite graft with a strong sealing effect on the blood.In contrast, the electrospun vascular grafts were yellowish with no blood perfusion.The 3D-printed vascular grafts had regular and large pores (around 200 µm), while the electrospun vascular grafts had irregular and small pores ((0.55 ± 0.12) µm).As such, blood could easily perfuse the 3D-printed vascular grafts but hardly penetrate the walls of the electrospun vascular grafts.14 d after in vivo implantation, the 3D-printed vascular grafts were transparent and like native vascular tissues.However, the electrospun vascular grafts showed limited remodeling as compared to those upon implantation.30 d post-implantation, the 3D-printed vascular grafts were even more like the native arteries, while the electrospun vascular grafts still showed a high percentage of materials (movie 4).These results indicate a fast remodeling of 3D-printed vascular grafts after in vivo implantation.
To evaluate performances of vascular grafts in vivo, ultrasound imaging was used to detect the blood flow in the grafts.As shown in B mode images of ultrasound in figure 4(b), both types of grafts showed increased diameters with time.The 3Dprinted vascular grafts had larger diameters than the electrospun ones on both day 14 and day 30 (figure 4(d)).C mode images of ultrasound showed that both types of grafts were patent from day 14 to day 30.PW mode images showed a higher mean velocity of the blood flow in the electrospun vascular grafts than in the 3D-printed vascular grafts on day 14 while an opposite mean velocity of the blood flow on day 30 (figure 4(e)).With the time of implantation, both types of grafts showed movements of vascular walls, with more obvious movements observed in the 3D-printed vascular grafts on day 30 (arrows in M mode images of ultrasound in figure 4(b)).Due to larger fiber diameters and more porous structures, the 3D-printed vascular grafts had weaker mechanical properties than the electrospun ones.Therefore, an increase in graft diameters under blood pressure could be detected in the 3Dprinted grafts.Further enhancement of mechanical properties of the 3D-printed grafts are requisite for clinical translation of the grafts.To enhance the mechanical properties of the vascular grafts, introduction of nanosystem into 3D-printed vascular grafts is an option, since nanosystem can improve load transfer efficiency in the grafts [58][59][60][61][62]. Another option is to enhance the mechanical properties of 3D printed vascular grafts by electrospinning PCL sheaths in the outermost of the vascular grafts.
Tissue regeneration within the vascular grafts was evaluated with histology (figure 4(c)).H&E staining showed quite limited tissue regeneration within electrospun fibers in the electrospun vascular grafts on day 14 and day 30 post-implantation.On the contrary, numerous tissues were regenerated surrounding large 3D printed fibers.In addition, the 3D-printed vascular grafts had larger luminal areas than the electrospun grafts on day 14 and day 30 (figure 4(f)).The vascular walls of the 3D-printed vascular grafts were also thicker than the electrospun grafts (figure 4(g)).The extremely porous structures of the 3D-printed vascular grafts encourage fast cell migration, whereas limited spaces between electrospun fibers inhibit cell infiltration.Due to more tissue regeneration, the 3D-printed vascular grafts were more transparent than the electrospun grafts from macroscopic views (figure 4(a)).Similarly, due to higher cellular components, the 3D-printed vascular grafts were softer than the electrospun grafts.Thus, the 3D-printed vascular grafts showed more obvious vascular wall movement under blood pressure as revealed by PW images in figure 4(b).

Native artery-like structures of 3D-printed grafts after in vivo implantation
Native arteries had three distinct layers: the endothelium, the media, and the adventitia.To investigate whether the remodeled vascular grafts had structures similar to the native arteries, we stained samples with eNOS, an EC marker; MYH11, a mature SMC marker; and PDGFRα, a fibroblast cell maker.As shown in figure 5(a), the native arteries had an intact endothelium, which could be positively stained by anti-eNOS antibodies.Both the electrospun and 3D-printed grafts had no endothelium coverage on day 14 post-implantation.On day 30, the 3D-printed grafts had about 93% endothelial coverage, while the electrospun grafts had only 61% endothelial coverage (figure 5(d)).Meanwhile, the mature SMCs in the media of the native arteries expressed both MYH11 and αSMA (figure 5(b)).The electrospun grafts only express αSMA on day 30 (figure 5(b)), indicating the immaturity of the SMCs in the regenerated grafts.The 3Dprinted grafts, on the other hand, had a significantly higher percentage of MYH11 + areas on day 30, revealing more mature SMCs (figure 5(e)).Finally, the native arteries were positively stained with PDGFRα in the adventitial areas (figure 5(c)), which were missing in the electrospun grafts on both day 14 and day 30.However, the 3D-printed grafts had PDGFRα positive areas in the adventitia areas on day 30, which was significantly larger than the electrospun grafts.All these data indicate that the 3D-printed vascular grafts had mature endothelium, media, and adventitia, which were very like the native arteries.The 3D-printed vascular grafts were bilayer.The inner layer consisted of two 3D-printed scaffolds with large pores and the outer layer consisted of one 3D-printed scaffold with smaller pores.It is possible that the large pores in the inner layers of the 3D-printed vascular grafts allow maturing of the infiltered SMCs to form the media, while the smaller pores in the outer layers of the grafts favor the growth of fibroblasts to form the adventitia.The ECs then attach onto the infiltrated SMCs to form the endothelium.Therefore, three-layer structures could be observed after in vivo implantation of the 3Dprinted grafts.The design of the 3D-pritned vascular grafts in this study greatly induces the regeneration of vascular grafts in an ordered manner.On the contrary, because the electrospun grafts had only one pore size, the distribution of SMCs and fibroblasts presents in chaos, showing no preferences.

Positive remodeling of ECM in 3D-printed grafts
The native arteries had a clear and well-organized distribution of ECM, with elastin in the inner layers and collagen in the outer layers (figure 6(a)).The elastin functions during elastic contraction and dilation of blood vessels, while the collagen can prevent over-stretch of blood vessels.This specific distribution of ECM makes blood vessels soft under physiological pressures (60 mmHg-120 mmHg) but can resist over-dilation under extremely high blood pressures (over 200 mmHg).Both types of grafts expressed elastin and collagen increasingly with the time of implantation, as shown in figure 6(a).Interestingly, the 3D-printed vascular grafts had collagen distribution only in the outer layers but not in the inner layers, while the electrospun grafts had a transmural collagen distribution.The elastin content of the 3D-printed grafts was close to that of the native arteries and significantly higher than that of the electrospun grafts (figure 6(b)), whereas the eletrospun grafts had a higher content of collagen than the 3D-printed grafts (figure 6(c)).Although the 3D-printed grafts promoted vascular regeneration, the elastin expressed in the electrospun and 3D-printed vascular grafts still exhibited a transmural and unordered distribution.This result indicates that the regenerated SMCs could not produce sufficient elastin and well organize the produced elastin.In addition to physical signals, such as mechanical properties and pore sizes, the maturation of SMCs needs biochemical stimulation as well.In our future studies, we will try to surface modified 3Dprinted grafts with functional biomolecules to further improve the maturation of SMCs recruited in the grafts.On the other hand, collagen in the outer layer of the 3D-printed vascular grafts after 30 d was much thinner than that of the native arteries.The current study observed the regeneration of the grafts in vivo only for 30 d.It is possible that the collagen expression can get closer to that of the native arteries with the time of implantation time in vivo.Performance of the 3D-printed vascular after 1 yr implantation in vivo will be investigated in our future studies.
Finally, a house-made device was used to test the mechanical properties of the grafts, as shown in figure 6(d).The stress-strain curves of the 3D-printed vascular grafts were like that of the native arteries, while the pressure of the eletrospun grafts increased exponentially with the increase of the stretch (figure 6(e)).The compliance of the 3D-printed vascular grafts was also significantly higher than that of the eletrospun vascular grafts (figure 6(f)).These data indicate that the 3D-printed grafts were softer that the eletrospun grafts in mechanical properties.The eletrospun vascular grafts had a high collagen content while low elastin content.Meanwhile, the collagen distribution was transmural.The adverse ECM remodeling greatly constrains the dilation of the graft, making the whole graft stiff.In contrast, by controlling the fiber diameter and porosity, the 3D-printed vascular grafts had a positive ECM remodeling, with collagen and elastin content close to that of the native arteries.More importantly, the 3D-printed vascular grafts showed well-organized collagen distribution, which was like that of the native arteries.Therefore, the 3D-printed grafts could dilate as the native arteries with the increase of the pressure, showing a soft mechanical property.

M2 polarization of macrophages in 3D-printed grafts
To explore the reason for the improved vascular regeneration of the 3D-printed grafts, we stained the tissue sections with αSMA and proliferative markers cyclin D1.Since αSMA did not co-localize with cyclin D1 (figure S2(a)) and the 3D-printed vascular grafts had fewer cyclin D1 + cells than the electrospun grafts on day 30 (figure S2(b)), the facilitated vascular regeneration of the 3D-printed vascular grafts was not because of SMC proliferation.We then stained tissue sections with CD68 and CD206 to verify the phenotypes of the infiltrated macrophages.Both types of grafts had macrophage infiltration (figures 7(a) and (b)) on day 14 and the number of CD68 + cells was comparable between the two groups (figure 7(c)).However, the electrospun vascular grafts were positively stained for CD68 but not CD206, while the 3D-printed vascular grafts were positively stained for both CD68 and CD206.The number of CD206 positive cells in the 3D-printed grafts was much higher than those in the electrospun grafts (figure 7(d)).The ratio of CD206 positive cells to CD68 positive cells was also higher in the 3D-printed grafts as compared to the electrospun grafts (figure 7(e)).These results indicate that the infiltrated macrophages in the electrospun grafts were M1, whereas the macrophages in the 3D-printed grafts were M2.On day 30, there were still quite a lot of CD68 positive cells but no CD206 positive cells for the electrospun grafts, while the 3D-printed grafts had fewer CD68 positive or CD206 positive cells (figures 7(a) and (b)).The quantification data also confirmed these results, with fewer CD68 positive cells, more CD206 positive cells, and a higher ratio of CD206 positive cells to CD68 positive cells in the 3D-printed grafts as compared to the electrospun grafts (figures 7(d) and (e)).In addition, the macrophages seeded on 3D-printed scaffolds secreted high levels of IL-10, while those seeded on electrospun scaffolds secreted high levels of TNFα and IL-1β (figures S3(a) and (b)).
Macrophages with M2 phenotypes can secrete woundhealing cytokines, such as IL-10, to promote vascular regeneration, while macrophages with M1 phenotypes secrete pro-inflammatory factors, such as TNFα and IL-1β, to affect vascular regeneration negatively [63].In vivo studies showed that the 3D-printed vascular grafts had more M2 macrophages than eletrospun vascular grafts did.Through in vitro cell experiments, we also showed that the macrophages seeded on 3D-printed scaffolds secreted high levels of IL-10, while those seeded on electrospun scaffolds secreted high levels of TNFα and IL-1β.These results indicate that 3D-printed vascular grafts with thicker fibers tended to polarize macrophages into M2 phenotypes, while eletrospun vascular grafts with thinner fibers polarize macrophages into M1 phenotypes.The secreted IL-10 can modulate inflammation responses and promote vascular regeneration in vivo.Previous studies show that macrophages cultured on thicker-fiber (5-6 µm) scaffolds can polarize into M2 phenotypes, while those cultured on thinner-fiber (0.1 µm-6 µm) scaffolds can polarize into M1 phenotypes [23].Our study also shows that thick-fibers ((3.75 ± 0.65) µm in axil direction and (1.13 ± 0.24) µm in circumferential direction) supports macrophage M2 polarization as compared to thin fibers (1.01 ± 0.18) µm in axil direction and (0.37 ± 0.03 µm in circumferential direction) [63].For electrospun grafts, the larger fiber diameters are, the larger pores there are between fibers.When there is sufficient space for macrophage spreading, macrophages prefer to polarize into M2 phenotypes [23].Therefore, the thicker fibers may induce M2 polarization of macrophages more easily than thinner fibers.However, due to technique limitation of electrospinning, the pore sizes between fibers are usually less than 1 µm, which is much smaller than the size of macrophages (10 µm).On the other hand, the 3D printing technique can control pore sizes precisely.The pore sizes of the 3D-printed grafts were 200 µm, which is greatly beyond the limitations of electrospinning.The larger pores of the 3D printed grafts may polarize macrophage into M2 phenotypes more easily than the smaller pores of the electrospun grafts, which greatly improved vascular graft regeneration in vivo.

Conclusion
In summary, bioinspired by the collagen alignment of SML, the inner layer of our grafts had larger pore sizes and high porosity and the outer layer had low porosity and minor pore size.One month after implantation, the regenerated arteries had good pulsation and three-layer structures similar to natural arteries, and their compliance (8.9%) was close to that of natural arteries (11.36%).These results indicate that the SML was successfully reconstructed with native functions and provides an effective strategy to reconstruct blood vessels through 3Dprinted structures rapidly.

Figure 1 .
Figure 1.Schematics of vascular regeneration induced by 3D printed grafts with gradient structures.

Figure 2 .
Figure 2. Morphology characterization of bi-layered small diameter vascular graft.(a) Schematic of MEW rotational printing.(b) Images of bi-layered vascular graft.(c) SEM images of vascular grafts' transverse cross-section, inner layer scaffold and outer layer scaffold.(d) SEM images of inner layer scaffolds with different porosity and the partial enlarged detail of IS200 and IS300.(e) SEM images of the outer layer scaffold and (f) electrospun scaffold.(g) The partial enlarged detail of electrospun scaffold.(h) The inner layer scaffolds' fiber diameter.(i) The outer layer scaffolds' fiber diameter.(j) The porosity of the inner layer scaffold with different pore sizes, electrospinning scaffold and outer layer scaffold.

Figure 3 .
Figure 3. Mechanical characterization of bi-layered vascular graft.(a) Picture shows that the retention of its original shape of bi-layered vascular graft after subjected to external force in the radial direction.(b) Representative photographs of testing axial and radial mechanics.(c) Stress-strain curves of ISs with pore sizes.(d) Axial mechanics testing of ISs with pore sizes for tensile modulus, (e) tensile strength and (f) elongation at break.(g) Axial mechanics testing of OS, ES and BSDVG.(h) Suture retention force of different vascular grafts and natural arteries.(i) Burst strength of different vascular grafts.

Figure 6 .
Figure 6.Positive remodeling of ECM in the 3D-printed grafts.(a) EVG and MTC staining of electrospun and 3D-printed grafts 14 and 30 d post-surgeries.* indicates the elastin positive areas.In both 3D-printed and electrospun vascular grafts, the distribution of the elastin was transmural.# indicates the collagen positive areas.The collagen was mainly in the outer layer of the 3D-printed vascular grafts, while the distribution of the collagen in the electrospun vascular grafts was transmural.(b) Elastin and (c) collagen quantification of electrospun and 3D-printed vascular grafts 30 d post-surgeries.(d) Customized device for mechanical test of the explanted vascular grafts.(e) Circumferential pressure-stretch curves of electrospun and 3D-printed vascular grafts 30 d post-surgeries.(f) Compliance of electrospun and 3D-printed vascular grafts 30 d post-surgeries.* * indicates that there was a significant difference between two groups (p < 0.01).

Figure 7 .
Figure 7. M2 polarization of macrophages in the 3D-printed grafts.(a) Immunofluorescence staining of electrospun and (b) 3D-printed grafts 14 and 30 d post-surgeries with CD68 and CD206.(c) Quantification of CD68 positive cells, (d) CD206 positive cells, and ratio of CD206 positive cells to (e) CD68 positive cellsin electrospun and 3D-printed grafts 14 and 30 d post-surgeries.* * indicates that there was a significant difference between two groups (p < 0.01).