Recombinant DTβ4-inspired porous 3D vascular graft enhanced antithrombogenicity and recruited circulating CD93+/CD34+ cells for endothelialization.

Matching material degradation with host remodeling, including endothelialization and muscular remodeling, is important to vascular regeneration. We fabricated 3D PGS-PCL vascular grafts, which presented tunable polymer components, porosity, mechanical strength, and degrading rate. Furthermore, highly porous structures enabled 3D patterning of conjugated heparin-binding peptide, dimeric thymosin β4 (DTβ4), which played key roles in antiplatelets, fibrinogenesis inhibition, and recruiting circulating progenitor cells, thereafter contributed to high patency rate, and unprecedentedly acquired carotid arterial regeneration in rabbit model. Through single-cell RNA sequencing analysis and cell tracing studies, a subset of endothelial progenitor cells, myeloid-derived CD93+/CD34+ cells, was identified as the main contributor to final endothelium regeneration. To conclude, DTβ4-inspired porous 3DVGs present adjustable physical properties, superior anticoagulating, and re-endothelializing potentials, which leads to the regeneration of small-caliber artery, thus offering a promising tool for vessel replacement in clinical applications.

INTRODUCTION

Developing nonthrombogenic lumen remains a key challenge in artificial blood vessels, including vascular prosthesis and tissue engineering grafts (–). During the past decades, efforts have been constantly made to fabricate vascular grafts favorable for antithrombogenicity and host remodeling. The highly porous, fast-degrading elastomer may recruit a sufficient number of vascular cells to assemble neoartery (). Functionalizing bioresorbable grafts through conjugating growth factors such as vascular endothelial growth factor (VEGF) () or seeding bone marrow cells were expanding the success in regenerating small-caliber blood vessels in animals (). In addition to the application of tissue engineering vascular grafts for hemodialysis access, encouraging results in Fontan operation were also reported in clinical trials (). Despite the desire to transform bioresorbable grafts into natural vessels, mismatches between mechanical, degrading properties of grafts and host remodeling remain thorny issues in clinical translation (). We have fabricated bilayered grafts by poly(glycerol sebacate) (PGS) and polycaprolactone (PCL) previously (). Adjusting sheath thickness and strength may amend limits of porous PGS scaffolds in degradation and mechanical strength, while they may compromise muscular remodeling and recruitment of perivascular cells (). Recently, three-dimensional (3D) printing of PGS from You’s group opened a new avenue to use this elastomer and relevant thermosets (, ), which enabled facile customization of 3D constructs from PGS including tubular scaffolds (, ). Compared with salt leaching approach, 3D printing technologies may rationalize the properties of porous PGS through changing squeezed polymer components, porosity, and clinically available sizes, which may produce appropriate vascular grafts for different animal models and therefore hold great clinically translational merit.

Enhancing the antithrombogenicity of biodegradable scaffolds remains a technical challenge, mainly owing to uncontrollable surface topography, as well as rapidly dissolved anticoagulant blood-contacting interface. Mechanistically, endothelialization of vascular grafts could be contributed by endothelial cells (ECs) from four sources: in vitro EC seeding, transanastomotic migration, transmural infiltration, and adherence of circulating endothelial progenitor cells (EPCs) (–). In situ endothelialization of vascular graft shares common features with reparative angiogenesis in ischemic tissues, which encompasses chemotactic effects, recruitment, and adhesion of EPCs in neovascularization sites (). Although specific markers remain defined, early EPCs express CD34, CD133, and VEGF receptor 2 (, ). Coating bioactive substances and covalently immobilizing cytokines on grafts may accelerate the endothelialization of polymeric grafts (); however, bioactive molecule–inspired lumen not only favored EC adhesion but also adhered to other blood cells such as platelets and white cells, which may induce thrombogenesis. Therefore, in situ endothelialization requires conjugating peptides that can selectively adhere to vascular cells ().

Thymosin β4 (Tβ4), a G-actin–sequestering peptide regulating cell motility, migration, and differentiation, is highly expressed in the cardiovascular niche during fetal development and after injuries (). An adequate dosage of Tβ4 may inhibit platelet aggregation by an as yet unknown mechanism (). As a water-soluble peptide conserving 43 amino acids, Tβ4 is rapidly metabolized and unstable in vivo (). We have produced a recombinant human Tβ4 dimer that contains two complete Tβ4 molecules (), which may double the molecular groups upon limited binding sites. We proposed that dimeric Tβ4 (DTβ4) may augment interaction with EPCs when grafted on the luminal surface of vascular grafts as compared with Tβ4. These thoughts, together with the expected extending of metabolizing period and enhanced bioactivity, led us to investigate whether DTβ4 might act as the appropriate EPC homing signal and its effect on the endothelialization of the vascular grafts.

In this study, using sucrose particles as molding and foaming agents, 3D additive composed of PCL and PGS prepolymer presented adequate strength and degrading rate. The rational design of PCL/PGS ratio and inherent pores significantly increased the amount of heparin-binding DTβ4, which promoted patency rate through accelerating endothelialization of vascular grafts for carotid arteries in rabbits. Through “single-cell RNA sequencing” (scRNA-seq), the influx and differentiation of CD34+ EPCs in 3D vascular grafts (3DVGs) were proven to be durable and efficient, and the CD93+ subset was further identified as the main contributor of newly formed endothelium, which directs the carotid artery regeneration in rabbits.

RESULTS

3D printing enabled tunability of mechanical strength, microstructure, and amount of covalently conjugated heparin

3DVGs contained porous tube and nanofibrous sheath (Fig. 1A). To be used in rabbits, 3DVGs presented an inner diameter of 1.62 ± 0.12 mm and an outer diameter of 2.43 ± 0.16 mm. Polymer/sucrose (w/w) was rationalized as 1/2 (fig. S1, C and F) to facilitate smooth squeezing. Since the printed PCL melted during heat curing, increasing PCL ratio decreased the surface porosity of the columns (100/0, 70.14 ± 3.26%; 85/15, 65.25 ± 5.64%; 70/30, 29.05 ± 4.83%), which also compromised heparin conjugation (100/0, 15.44 ± 0.76 mg/cm3; 85/15, 14.09 ± 1.07 mg/cm3; 70/30, 6.08 ± 1.44 mg/cm3), and decreased DTβ4 content (100/0, 261.58 ± 11.38 ng/cm3; 85/15, 264.58 ± 14.24 ng/cm3; 70/30, 147.95 ± 19.29 ng/cm3) (Fig. 1, C, O, and P). Through heparin binding, the 85/15 3DVGs could conjugate Tβ4 (262.74 ± 25.09 ng/cm3) and VEGF (334.83 ± 23.73 ng/cm3), respectively (fig. S1K). Scanning electron microscopy (SEM) revealed that fusing columns constituted wavy luminal and exterior surfaces, with peak and valley thicknesses (396.25 ± 7.43 μm and 245.13 ± 11.18 μm), and ultrathin PCL sheath (33.62 ± 12.84 μm) (Fig. 1J). More PGS content improved the fusion thickness between columns (100/0, 261.64 ± 12.03 μm; 85/15, 240.36 ± 13.81 μm; 70/30, 138.93 ± 16.25 μm) (Fig. 1C and fig. S1G). Micro–computed tomography (micro-CT) identified high porosity and interconnectivity among micropores in 3DVGs (fig. S1I and table S2). Different from rapid DTβ4 release in soaked grafts, heparin-conjugated grafts presented controlled DTβ4 release, and 100/0 3DVGs acquired the highest accumulating amount (fig. S1H). Energy-dispersive spectroscopy (EDS) on 3DVGs visualized even distribution of the sulfur element in heparin and the nitrogen element in DTβ4 (fig. S1J). Using 85/15 3DVGs as representative samples, heparin conjugation and hep-DTβ4 binding significantly enhanced the hydrophilicity of grafts, as shown by decreased water contact angle (Fig. 1K). On the contrary, heparin immobilization on grafts did not change the axial elastic moduli (fig. S1, D and E). Mechanical properties significantly varied with PCL addition. Axially, the ultimate tensile strength (UTS) of 100/0, 85/15, and 70/30 3DVGs was 101.93 ± 16.35 kPa, 122.86 ± 9.92 kPa, and 247.24 ± 12.71 kPa, respectively (Fig. 1D), and corresponding elastic moduli were 0.29 ± 0.04 MPa, 0.45 ± 0.05 MPa, and 0.83 ± 0.04 MPa (Fig. 1H). The axial stress-strain curve suggested that maximum strain in the 70/30 group (34.89 ± 6.27%) was poorer than 100/0 (58.43 ± 5.76%) and 85/15 group (42.89 ± 7.14%) (Fig. 1D). Radially, the UTS of 100/0, 85/15, and 70/30 3DVGs was 247.62 ± 12.47 kPa, 351.74 ± 10.96 kPa, and 668.45 ± 9.77 kPa, respectively (Fig. 1, E and H). The compressive moduli were 100/0, 0.17 ± 0.04 MPa; 85/15, 0.35 ± 0.05 MPa; and 70/30, 0.71 ± 0.05 MPa (Fig. 1, F and I). The 3DVGs also showed elasticity after compression (fig. S1B), maintained round lumen upon benching and folding, and therefore met the requirement of neck movement (Fig. 1, L and M). As described in our previous study (), the potential of further shape morphing (the straight tube was morphed into the “S” shape) also met the variant anatomic requirements in humans (Fig. 1N).

Fig. 1.

Fabrication of 3DVGs and optimization.

(A) Schematic of the fabrication of 3DVGs and DTβ4 immobilization design on heparinized 3DVGs. (B) Luminal surface of the 3D printed column examined with SEM. (C) Toluidine blue (TB) staining for the sagittal surface of the 3D printed column. (D) Axial stress-strain curves of the 3DVGs. (E) Radial stress-strain curves of the 3DVGs. (F) Radial compressive stress-strain curves of the 3DVGs. (G) Surface porosity of the 3DVGs. (H) Quantitative comparison of axial elastic moduli of the 3DVGs (n = 5 independent samples). (I) Quantitative comparison of radial compressive moduli of the 3DVGs (n = 5 independent samples). (J) SEM examination showed microstructures of 3DVGs. (K) The hydrophilicity of grafts was detected by water contact angle (n = 3 independent samples). (L and M) Simulated carotid bending test in vitro. White arrow reveals the kink of the mold casting graft. (N) Straight tube printed was bent into the “S” shape and experienced heat curing for shape morphing. (O and P) Quantitative comparison of heparin and DTβ4 amount conjugated onto the 3DVGs. (Q) In vivo evaluation of 3DVGs with different ratios of PCL (n = 5 in each group). EDA, ethylenediamine; HEP, heparin; NHS, N-hydroxysuccinimide; EDC, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (black marks, 100:0 group; red marks, 85:15 group; blue marks, 70:30 group; PGS:PCL, w/w). Plotted data represent means ± SD. For (G to I), (K), (O), and (P), significance was determined by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc analysis. ns, P > 0.05; *P < 0.05; and **P < 0.01.

In vivo, short-term observation of DTβ4-conjugating 3DVGs interpositionally implanted into the common carotid artery (CCA) of rabbits showed that the 100/0 3DVGs frequently burst (5 of 5) in 2 weeks postoperatively, whereas 70/30 3DVGs occluded (5 of 5). 85/15 3DVGs presented a high patency rate (4 of 5) and were free of rupture (Fig. 1Q) and therefore were chosen for further studies.

The antithrombogenic properties of DTβ4-3DVGs both in vivo and in vitro

To evaluate the acute antithrombogenicity of grafts, 3DVGs implanted in rabbit CCAs were observed for 2 hours (Fig. 2A, a). DTβ4-3DVGs showed a high patency rate (5 of 5), as revealed by a hollow lumen and cleaning luminal surfaces (Fig. 2A, e). In contrast, phosphate-buffered saline (PBS)–3DVGs, Tβ4-3DVGs, and VEGF-3DVGs were all embolized (Fig. 2A, b, d, and f). Lumens of Hep-3DVGs were partially blocked by mural thrombosis, which eventually led to occlusion in 24 hours (Fig. 2A, c). To explore the mechanism of antithrombogenesis of DTβ4-3DVGs, different groups of 3D grafts (3DGs) were incubated with recalcified whole blood and isolated platelets. Following incubation with recalcified whole blood, a trace amount of thrombus was spotted on the surface of DTβ4-3DVGs, while much more clots formed on the surface of other groups (Fig. 2A, g to l). This finding was confirmed by pre- and post-mass quantification (PBS-3DGs, 63.26 ± 5.44 mg; Hep-3DGs, 32.68 ± 3.72 mg; Tβ4-3DGs, 41.02 ± 3.23 mg; DTβ4-3DGs, 12.56 ± 1.42 mg; VEGF-3DGs, 39.72 ± 3.66 mg) (Fig. 2B). Following the incubation of platelet-rich plasma (PRP) at 37°C for 1 hour, lactate dehydrogenase (LDH) assay in DTβ4-3DGs decreased significantly (PBS-3DGs, 1.33 ± 0.06; Hep-3DGs, 0.99 ± 0.05; Tβ4-3DGs, 1.22 ± 0.12; DTβ4-3DGs, 0.98 ± 0.11; VEGF-3DGs, 1.19 ± 0.04). SEM showed that activated platelets with extending filopodia were flooding in 3DGs, Tβ4-3DGs, and VEGF-3DGs (Fig. 2D, f, h, and j), while significantly fewer platelets adhered in Hep-3DGs (Fig. 2D, g). DTβ4-3DGs presented the cleanest appearance, with no platelet detected in pores (Fig. 2D, i). After 30 min of incubation with whole blood, a significant number of clots formed in 3DGs, Tβ4-3DGs, and VEGF-3DGs appeared as fibrous-like thrombus embedding erythrocytes, while Hep-3DGs and DTβ4-3DGs were almost free of clotting (Fig. 2D, p to t). The effects of Tβ4 and DTβ4 on platelet aggregation were further quantitatively analyzed by flow cytometry (Fig. 2H). It was shown that activated platelets proportionally decreased from 24.62 ± 0.97% to 10.18 ± 0.93% (PBS versus DTβ4) (Fig. 2, I and J), suggesting a superior effect of DTβ4 on inhibiting platelets activation. The blood coagulation test confirmed that DTβ4 mainly affected activated partial thromboplastin time (APTT) and inhibited fibrin formation (Fig. 2, E and F). Mixing fresh blood with DTβ4 revealed that, in contrast to Tβ4, the anticoagulant effect of DTβ4 was significantly enhanced at low concentration (0.1 μg/ml), and the anticoagulant effect reached a peak point at 1.0 μg/ml (Fig. 2G). Together, both in vivo and in vitro results suggested that DTβ4 was stronger in inactivating platelets and fibrinogenesis inhibition than Tβ4 when they were both loaded on grafts with low dosage. On the basis of markedly different results among control groups and DTβ4-3DVGs, the biological activity of the grafts could be attributed to the consequence of the heparin-binding DTβ4.

Fig. 2.

Anticoagulation evaluation for 3DVGs in vivo and in vitro.

[(A), a to f] Cross-sectional images of different types of grafts exposed to carotid circulation for 2 hours (n = 5 in each group). [(A), g to l] Gross view of grafts through incubation with recalcified whole blood for 30 min (n = 5 in each group). (B and C) Quantitative analysis of pre- and post-mass weight and LDH assay (n = 5 independent samples). (D) SEM examination for the morphology of 3DGs with platelet and recalcified whole blood adhesion (n = 5 independent samples). (E to G) Blood coagulation test of Tβ4 and DTβ4 (n = 5 independent samples). (H and I) Rates of activated platelets by flow cytometry. (J) Quantitative analysis of platelet aggregation (n = 5 independent samples). Plotted data represent means ± SD. For (B), (C), (E), (F), (G), and (J), significance was determined by one-way ANOVA followed by Tukey’s post hoc analysis. ns, P > 0.05; and **P < 0.01.

DTβ4-3DVGs maintained a high patency rate and presented matching degradation time and remodeling in rabbit carotid artery

In vivo remodeling of DTβ4-3DVGs was further evaluated through 12-week implantation in the CCA model. 3DVGs merged with the native vessel and remodeled by the host tissue progressively (Fig. 3A). Blood flow in 3DVGs was monitored by Doppler ultrasound at 2, 4, and 12 weeks. At 2 weeks, the overall patency rate for DTβ4-3DVGs was 80% (16 of 20 rabbits), and patency grafts were maintained in the remaining 16 rabbits before sacrificing at each time point (Fig. 3, B and F). Flow rate and diameter of neoarteries measured by ultrasound showed no significant difference with native CCA over 12 weeks of observation, suggesting no dilation or thrombosis of t DTβ4-3DVGs (Fig. 3, B, D, and E). CT angiography further revealed a high patency rate of DTβ4-3DVGs at 12 weeks postoperatively (Fig. 3C).

Fig. 3.

Gross morphology and ultrasound monitoring of 3DVGs at each time point.

(A) Gross morphology of grafts at 2, 4, and 12 weeks after implantation. The yellow dotted line shows the anastomosis. (B) Representative Doppler ultrasound images of the grafts at 2, 4, and 12 weeks after implantation. (C) Carotid artery CT angiography at 12 weeks after implantation. The black dotted line shows the graft. (D) Peak velocity of blood flow of 3DVGs at each time point after implantation (n = 5 independent samples). (E) Average inner diameter of 3DVGs at each time point after implantation (n = 5 independent samples). (F) Number of patent 3DVGs at each time point. Plotted data represent means ± SD. For (D and E), significance was determined by one-way ANOVA followed by Tukey’s post hoc analysis. ns, P > 0.05.

Because of rapid degradation, 100/0 DTβ4-3DVGs ruptured before acquiring efficient host remodeling (fig. S2, A and B). Increasing PCL in printing additives prolonged the degradation time. For 85/15 and 70/30 3DVGs, approximately 30 and 50% gross weight was lost following 5-hour alkaline treatment (Fig. 4E). In vivo evaluation presented a consistent trend. At 2 weeks, hematoxylin and eosin (H&E) staining revealed that PGS degradation was delayed in the 85/15 group as compared with 100/0 ruptured samples (100/0 versus 85/15, 7.61 ± 1.64% versus 25.19 ± 4.93%; P < 0.01) [Fig. 4, A (a, e, and i) and B, and fig. S2, B and D], while totally invisible at 4 and 12 weeks [Fig. 4, A (b, f, j, c, g, and k) and B]. In H&E staining slides, the PGS residual appeared as red anucleated porous structures on the luminal side (Fig. 4A, e). Under polarized light, the PGS residual showed dark red without birefringence structure, while the PCL appeared as bright fragments and compact fibers, which was frequently observed at 2 weeks (Fig. 4A, i). In contrast, PCL fragments in 3D printing part degraded completely, while fibers turned messy and loose at 4 weeks (Fig. 4A, j). Only approximately 3% of original PCL sheath remained visible in neoarteries at 12 weeks [Fig. 4, A (k) and C]. Together, the matching degrading rate of 3DVGs in rabbit carotid was rationalized through 15% PCL mixing.

Fig. 4.

Examining material degradation and host remodeling for 3DVGs.

(A) H&E staining and polarized H&E for the transversal surface at the midportion of DTβ4-conjugated grafts after being implanted for 2, 4, and 12 weeks, with native vessel as control. (B and C) Quantitative assessment of polymer residuals through 2, 4, and 12 weeks (n = 5 independent samples). (D) Wall thickness of neoarteries at each time point (n = 5 independent samples). (E) Quantitative assessment of polymer degradation in vitro (n = 3 independent samples). (F) Masson and EVG staining of cross sections and longitudinal sections for neoarteries, with native vessel as control. (G to I) Quantitative analysis of the collagen, elastin, and desmosine content in neoarteries at each time point (n = 5 independent samples). Plotted data represent means ± SD. For (B to D) and (G to I), significance was determined by one-way ANOVA followed by Tukey’s post hoc analysis. ns, P > 0.05 and **P < 0.01.

Remodeling 3DVGs presented similar trilayer vessel walls to native vessels, maintained even wall thickness, and turned to be more and more compact through 2 to 12 weeks (Fig. 4D). Along with regularly aligned smooth muscle cells (SMCs), organized structure consisted of densely layered collagen in an interconnecting lamellar network presented at 12 weeks (Fig. 4F and fig. S3A). Similar to native vessels, extracellular matrix (ECM) in neoarteries was positively stained with Col-I and Col-III (fig. S3A) and acquired a similar amount of collagen content with native vessels (Fig. 4G). Different from diffusive distribution of elastin in immunostained slides, Elastic Van Gieson (EVG) staining showed less cross-linked elastic fibers with black coloring than native vessel. Elastin quantification suggested that elastin content in 3DVGs was similar to that of the CCA (Fig. 4H), while desmosine, a unique amino acid in elastic fibers, was weakly generated in neoarteries compared with native CCA even through 12 weeks of remodeling in vivo (2 weeks, 0.25 ± 0.11 ng/mg, versus 4 weeks, 0.71 ± 0.13 ng/mg, versus 12 weeks, 1.93 ± 0.61 ng/mg, versus 12 weeks: 7.13 ± 0.98 ng/mg; n = 5) (Fig. 4, G to I). These differences suggested limited transformation of tropoelastin into load-bearing cross-linked elastin in neoarteries, as was also proven by mechanical test. Although the UTS of neoarteries at 12 weeks was similar to that of CCA, the burst pressure and compliance of neoarteries were relatively lower than native CCA (fig. S3, B to D).

DTβ4-3DVGs recruited CD34+ cells and enabled rapid endothelialization

Underlying the high patency rate of DTβ4-3DVGs was rapid endothelialization. SEM examination revealed that fast cellularization occurred on the luminal surface at 12 weeks (fig. S7, A to C). Immunostaining revealed that CD31+ cells covering the suturing site, middle, and quarter sites were all progressively increased through 2 to 12 weeks. The cells covering the suturing site were distinct in morphology and alignment from those on mid and quarter sites at 2 weeks, whereas they tended to be similar through 4 and 12 weeks (Fig. 5, B, D, and F). Furthermore, immunofluorescence staining on cross-sectional slides revealed the tempospatial distribution of recruited cells in grafts (Fig. 5, A, C, and E). Represented as images from midgrafts, a significant amount of CD34+ cells was recruited in 3DVGs already at 2 weeks (36.66 ± 5.68 per field), while only minority of them was CD31+ (2.33 ± 0.57 per field). Unexpectedly, cell ratio for CD34+ and CD31+ was totally reversed at 12 weeks, with more continuous EC lining (4 weeks, 53.78 ± 7.49%; 12 weeks, 91.69 ± 4.21%) on luminal surface and diminishing CD34+ cells (4 weeks, 17.66 ± 4.72 per field; 12 weeks, 1.33 ± 1.52 per field) inside grafts suggesting the formation of neo-endothelium on 3DVGs (Fig. 5G). En face immunostaining–detected mature ECs were migrating from host artery and only covered suturing gap at 2 weeks; very few CD31+ cells were detected at quarter and middle sites (Fig. 5, B and H). At 4 weeks, the migrating distance of ECs on suturing sites remained limited, as presented by discontinuities of EC lining between suturing and quarter sites. However, CD31+ cells increased significantly at the middle and quarter of the neoarteries (Fig. 5, D and H). The increasing ECs progressively merged through 12 weeks and almost covered the whole luminal surface (Fig. 5, F and H). To visualize the endothelialization process of neoarteries, longitudinal sections at each time point were also acquired for CD31 and CD34 immunofluorescence staining. The CD31+ cells predominantly covered the suture site, while they only sporadically appeared on the midportion of grafts at 2 weeks. These cells were consistently increased and covered most of the luminal side after 12 weeks. On the contrary, a large amount of CD34+ cells recruited into 3DVGs, whereas they decreased gradually through 12 weeks (fig. S4, A to E). Together, the multipoint, discontinuous regeneration of endothelium suggested the predominant contribution of circulating EPCs for endothelialization of 3DVGs.

Fig. 5.

The differentiation of CD34+ cells into CD31+ cells during host remodeling.

(A, C, and E) Immunofluorescence staining of midpoint cross sections from neoarteries (left: CD34, red; right: CD31, red). (B, D, and F) Immunofluorescence staining of en face from different regions at different time point samples (CD31, red). (G) Quantitative analysis of CD34- or CD31-positive cells per field (n = 5 independent samples). (H) Comparison of the endothelialization of different sites in neoarteries (n = 3 independent samples). The yellow dotted line marked the anastomosis. Plotted data represent means ± SD.

scRNA-seq analysis identified two genotypes of ECs

We performed correlation analysis on the vascular cell sequencing results at 2 and 12 weeks. Hosted cells could be genetically classified into macrophages (clusters 0, 1, 2, 3, 9, 10, 11, and 12), ECs (cluster 13), SMC (cluster 15), T&B cells (cluster 6), fibroblasts (clusters 4 and 8), myofibroblasts (clusters 7 and 14), neutrophils (cluster 5), plasma cell (cluster 17), and mast cells (cluster 16) via uniform manifold approximation and projection (UMAP) analysis (Fig. 6, A and B). In scRNA-seq profiles, ECs were defined as the genetic expression of vWF and PECAM1. Macrophages expressed the marker gene CSF1R. SMCs and myofibroblasts were labeled with ACTA2 and MYL9. Fibroblasts expressed classic DCN and COL1A1 markers (fig. S5). Clustering heatmap analysis was performed via differential expression gene analysis of each cluster (fig. S6A). Enrichment pathway analysis performed according to the highly expressed genes revealed that ECs were enriched in phosphatidylinositol 3-kinase (PI3K)–AKT and Wnt pathway (fig. S6B), while SMCs were enriched in vascular smooth muscle contraction and PI3K-AKT pathway (fig. S6C). Gene Ontology (GO) analysis indicated that ECs might be involve in angiogenesis (fig. S6D), while SMCs might be involved in vasculogenesis (fig. S6E), both suggesting that ECs and SMCs actively participated in the regeneration of the carotid artery.

Fig. 6.

Single-cell RNA-seq analysis of carotid artery remodeling.

(A) UMAP analysis atlas of carotid artery remodeling at 2 and 12 weeks after implantation. (B) Specific cluster identification map of the cell type. (C) Arrangement of dominant cells in neoarteries (ECs, SMCs, macrophages, fibroblasts, and myofibroblasts) along the differentiation trajectory. (D) Specific cluster identification map of the EC type. (E) Analysis of the differentiation trajectory of endothelial 1 and endothelial 2. (F) Comparison of the expression changes of EC marker genes PECAM1, vWF, CDH5, and CD34 at 2 and 12 weeks after implantation.

Pseudo-time analysis on the key cell clusters in vascular remodeling was performed on neoarteries at 2 and 12 weeks, including ECs, SMCs, myofibroblasts, and fibroblasts. It was shown that cell differentiation presented three kinds of cell fates. After refining each cell cluster specifically, we found that EC and SMC clusters reached the later stages of overall differentiation (Fig. 6C). Pseudo-time analysis identified two different ECs, endothelial 1 and endothelial 2 (Fig. 6, D and E). Further EC-related gene analysis revealed that the expression of vWF, CDH5, and CD34 up-regulated in endothelial 2 at 12 weeks as compared with 2 weeks, whereas CD34 down-regulated in endothelial 1 significantly (Fig. 6F).

To further study the characteristics of ECs involved in vascular remodeling, we grouped all ECs into five types: a, b, c, d, and e (Fig. 7, A and B). UMAP showed that cluster b ECs deviated significantly from other clusters (Fig. 7B). The differential genes of all taxa were accordingly presented in the heatmap (Fig. 7C). CD93 was highly expressed in a, c, d, and e, whereas it was negatively expressed in b (Fig. 7D). Cluster b specifically expresses MMRN1, a gene of mature endothelium (Fig. 7C). However, a, c, d, and e expressed CD34, which indicated that a, c, d, and e were immature ECs and b was mature ECs (Fig. 7D). Immunostaining on neoarteries showed that CD93+ cells emerged at 2 weeks, while only few cells were coimmunostained with CD31 (Fig. 7F). Through 12 weeks, it appeared that more CD93+ cells in newly formed endothelium were coimmunostained with CD31, while abundant CD93+/CD31− cells were scattering in adventitial tissues and failed to trace the ingrowing microcapillaries, which suggested their less relationships with ECs in mature microcapillaries (Fig. 7G). En face immunofluorescence staining of neoarteries showed that CD93 and CD31 were coexpressed on the luminal surface, which confirmed that CD93+ cells were involved in the process of endothelialization (Fig. 7, H to J).

Fig. 7.

Identification of ECs in remodeling 3DVGs grafted in carotid.

(A and B) UMAP plot of EC showing five distinct subtypes colored by cluster. (C) Heatmap showing the differentially expressed genes in the five EC clusters. (D and E) Violin plots for expression of marker genes (CD34 and CD93) in the five EC clusters. (F and G) Representative immunofluorescent costaining of CD93 and CD31 in neoarteries at 2 and 12 weeks postoperatively. (H and I) Representative images showed immunofluorescent en face staining (CD93 and CD31) for neoarteries at 2 and 12 weeks postoperatively. CD93, green; CD31, red; 4′,6-diamidino-2-phenylindole (DAPI), blue. (J) Quantitative analysis of CD93+ cells from immunofluorescent en face staining. Plotted data represent means ± SD. For (J), significance was determined by Student’s t test. ns, P > 0.05 and *P < 0.05.

As a classic marker gene of myeloid cells (Gene Card) (–), CD93 was also found to promote angiogenesis (). The CD93 may be derived from the proliferated arterial endothelium at both ends of the surgical anastomosis.

CD93+ cells regenerate endothelium during the remodeling of DTβ4-3DVGs

The UMAP profile from neoarteries at 2 and 12 weeks identified two types of ECs, the majority of them were CD93, and the minority were CD93 (Fig. 8A). To understand how the context-dependent communications among different types of cells enabled remodeling processes, we used CellPhoneDB to infer cell-cell interactions from combined expressions of multisubunit ligand-receptor complexes. CellPhoneDB analysis of ECs, macrophages, SMCs, and fibroblasts showed that CD93 were extroverted cells and frequently communicated with fibroblasts and SMCs, while CD93 were autistic and seldom communicated with other cells (Fig. 8B and fig. S8, A to D). To examine the role of immune cells in vascular remodeling, we performed macrophage receptor-ligand analysis based on the CellChat platform. CellChat analysis identified macrophages and fibroblasts as dominant communication hubs, with ligands being secreted by macrophages to fibroblasts (Fig. 8C). The results also showed that the recruited macrophages communicated frequently with fibroblasts and seldom with ECs (fig. S8I). Signal transductions between ECs, SMCs, and fibroblasts included transportations of collagen (such as COL1A1) and fibroblast growth factor family proteins (fig. S8, B to D). Both CD93 and CD93 expressed the vWF (Fig. 8F), while only CD93 expressed endothelial gene MMRN1 (Fig. 8D). It was unusual to find that CD14, a marker gene of bone marrow monocytes (BMMNCs), was expressed in ECs (Fig. 8E), which suggested that BMMNCs might differentiate into ECs. The expression of marker gene CD34 was showed by the violin plots (Fig. 8G). It therefore confirmed the determining roles of BMMNCs on the endothelialization of DTβ4-3DVGs (Fig. 8, H and I, and fig. S9, A and B). At the same time, we found a large number of macrophages in the remodeling 3DVGs. Immunofluorescence staining revealed that most of them were M2 macrophages, which was consistent with our previous study () and confirmed promoting roles of DTβ4-3DVGs in M2 polarization (fig. S9, C to F). Furthermore, cell communication results also suggested that communications between macrophages and fibroblasts were active during smooth muscle differentiation. These findings suggested that immunomodulation played a pivotal role in muscular remodeling of vessel wall, while it was less related to endothelialization.

Fig. 8.

Single-cell RNA-seq analysis of remodeling 3DVGs at 2 weeks after implantation.

(A) UMAP plot of the neoarteries at 2 weeks. (B) Heatmap of cell-cell communication analysis based on CellPhoneDB. (C) Strength of ligand-receptor interactions between macrophages and other cell population pairs based on CellChat analysis. Edge width is proportional to the number of ligand-receptor pairs. Circle sizes are proportional to the number of cells per cluster. (D to G) Violin plots for expression marker genes (MMRN1, CD14, vWF, and CD34) in each cell cluster. (H and I) Immunofluorescence staining of myeloid cell marker CD14 in neoarteries at 2 and 12 weeks.

To further verify the role of DTβ4, BMMNCs were directly cocultured with 3DGs immobilized with heparin, hep-Tβ4, hep-DTβ4, and hep-VEGF, separately (Fig. 9, A and B).

Fig. 9.

Verification of the relationships between DTβ4 and CD93+ cells in vitro and in vivo.

(A) Schematic illustration of BMMNCs cocultured with grafts in vitro. (B) SEM and immunofluorescent staining of CD93+ (red) cells and nuclei (blue) after 12 hours of coculturing with grafts immobilized with heparin, hep-Tβ4, hep-DTβ4, and hep-VEGF, separately. (C and D) Count of BMMNCs infiltrating into the grafts after 12 hours of coculturing (n = 5 independent samples). (E) Schematic illustration of GFP+ bone marrow cells transplantation (BMT). (F) Representative immunofluorescence images of the cross sections showed BMT-GFP+ cells (green) infiltrated into the graft at 1 and 2 weeks after implantation. (G) Representative immunofluorescence images of en face staining showed BMT-GFP+ cell (green) recruitment onto the luminal surface of 3DVGs at 1 and 2 weeks after implantation. (H) Number of GFP+ cells infiltrating into the 3DVGs at each time point (n = 3 independent samples). (I) Percentage of GFP+ cells infiltrating into the grafts at different time points (n = 3 independent samples). Plotted data represent means ± SD. For (C and D), significance was determined by one-way ANOVA followed by Tukey’s post hoc analysis. For (H and I), significance was determined by Student’s t test. *P < 0.05 and **P < 0.01.

SEM revealed that hep-DTβ4 had high affinity to BMMNCs (PBS-3DGs, 75.32 ± 12.63; Hep-3DVGs, 76.69 ± 13.46; Tβ4-3DVGs, 123.18 ± 18.05; DTβ4-3DVGs, 156.37 ± 16.07; VEGF-3DVGs, 132.64 ± 14.92) (Fig. 9C). Immunostaining confirmed that CD93+ cells more frequently adhered to hep-DTβ4–conjugated grafts (PBS-3DGs, 1.33 ± 1.63; Hep-3DVGs, 1.33 ± 1.96; Tβ4-3DVGs, 3.33 ± 1.32; DTβ4-3DVGs, 6.66 ± 1.07; VEGF-3DVGs, 2.33 ± 1.92) (Fig. 9D). Using Sprague-Dawley (SD) rat as in vivo cell tracing model, green fluorescent protein–positive (GFP+) bone marrow cells, including CD93+ cells, were detected to participate in the host remodeling of grafts (Fig. 9, E to I). GFP+ cells had infiltrated into the 3DVGs in both groups since 1 week after implantation, and DTβ4-3DVGs presented higher affinity to circulating bone marrow–derived GFP+ cells as compared with PBS-treated grafts (en face 1 week, 5.33 ± 1.53% versus 0.33 ± 0.58%; 2 weeks, 6.33 ± 1.53% versus 0.33 ± 0.58%) (Fig. 9, H and I). Immunostaining of CD31 on GFP+ cells confirmed that recruited bone marrow progenitors might differentiate into ECs in 3DVGs at 2 weeks (Fig. 9, F and G).

Promoting effects of DTβ4 on migration and proliferation of BMMNCs were also verified in vitro (fig. S10, A and B). Two-week EGM-2 incubation enabled high expression of both CD93 and CD11b in BMMNCs, and CD93 was also coexpressed with the endothelial nitric oxides (eNOS) (fig. S10C) and therefore confirmed its potential of EC differentiation. We cultivated human umbilical cord vein ECs (hUVECs) and BMMNCs in the normal culture system and EGM-2 system, respectively, which further proved that BMMNCs expressed CD93 (fig. S10G) and failed to express MMRN1 (fig. S10G). The expression of vWF in BMMNCs increased significantly after feeding EGM-2 (fig. S10G), and CD34 expression was also significantly increased (fig. S10I). Together, CD93+ cluster, derived from BMMNCs, actively participated in endothelialization on DTβ4-3DVGs.

DISCUSSION

Different from the rodent model, transanastomotic endothelialization is very limited in humans. Designing vascular grafts capable of recruiting progenitor cells transmurally would be more clinically translational. Talacua et al. () showed that circulatory progenitor cells might constitute endothelium and smooth muscles during host remodeling of acellular grafts, while the amount of circulatory EPCs may vary greatly among different animal models. Several heparin-binding peptides, such as SDF-1a and Arg-Glu-Asp-Val (REDV), have also been conjugated on grafts to recruit CD34+ EPCs for luminal endothelialization (). Smith Jr. et al. () developed an acellular vascular graft that was based on small intestinal submucosa with immobilized heparin and VEGF, showing that circulating monocytes may amend rarity of EPCs, transdifferentiate into functional ECs, and enhanced endothelialization. Different from smooth luminal surface in electrospun grafts or decellularized ECM, 3DVGs were highly porous and rough in luminal side and therefore provides much more cell migrating channels. We believed that dual roles of antithrombogenicity and homing CD93+ EPC subset in DTβ4-3DVGs are unique. Antithrombogenicity of DTβ4 not only prevents acute thrombogenesis but also inhibits the shielding of pores by clotted fibrin, thereafter offers opening channels for consistent infiltration of circulating CD93+ cells. As the result, despite the rarity of CD93+ subset in blood, there was a wide distribution of these cells in degrading graft at 2 weeks and acquired rapid endothelialization in rabbit carotid artery. The above findings also suggested that the diversity of progenitors for ECs in circulation and binding DTβ4 may enhance the efficiency of capturing CD93+ EPCs, a highly proliferative, migratory subset (). Further studies remain to be performed to reveal which kinds of progenitors are more competitive in human beings.

Although macrophages were proven to play key roles in vascular repair and angiogenesis, we failed to detect their strong direct connection with endothelialization of 3DVGs. CellChat analysis revealed all cell communications among recruited cells, which stressed the involvement of macrophages in fibroblasts and smooth muscle differentiation (Fig. 8C). This finding confirmed our previous claims about promoting roles of M2 macrophage polarization on muscular remodeling of vascular grafts (), while suggesting less influence of perivascular macrophages on recruiting EPCs at the luminal side. M1/M2 immunostaining also revealed macrophages predominantly infiltrated in adventitial part of grafts (fig. S9C), suggesting that their modulatory effects may be predominant on perivascular progenitor cells.

Circulating EPCs are the most probable cells for rapid endothelialization, owing to the direct contact between luminal surface and bloodstream, as well as abundant provision of bone marrow. However, selectively recruiting vascular progenitor cells while excluding the platelet aggregation and thrombosis has long been the challenge in designing vascular grafts targeting circulating EPCs. In this study, we proved that DTβ4-3DVGs played dual roles, including antiplatelets and recruiting bone marrow EPCs, and therefore realized rapid endothelialization and guaranteed the high patency rate in the rabbit carotid arteries. The vital contribution of marrow-derived progenitors was confirmed from three aspects: (i) Despite slow transanastomotic EC migration in rabbits, host remodeling at 2 weeks presented sufficient CD34+ cells homing in graft, especially at the quarter and mid sites. (ii) At 2 weeks, when nanofibrous sheath separating blood flow and perivascular tissues remains compact and intact, CD34+ cells already occupied the luminal part of the grafts, suggesting that cells may come from the blood stream. (iii) GFP+ cells injected were traced, covering the luminal surface of the neovessel. To the best of our knowledge, affirming the decisive contribution of DTβ4 in selective recruitment of bone marrow–derived EPCs and endothelialization of fast degrading polymer is new to vascular substitutes. This report is the first step that identifies the peptide selectively adhering EPCs in circulation for vascular regeneration.

3D printing approach allowed tunable ratio of PGS and PCL, as well as variant amount of sugar particles in polymers, which results in adequate degrading rate and tunable porosity (). Our previous study revealed that PGS almost degraded in 7 days in rat abdominal aorta (), while the burst of grafts frequently occurred in rabbits (fig. S2A). The regenerative potential varied significantly among animal species; tunable mechanical strength and degrading time are therefore prerequisites to clinical translation. 3DVGs extended the degrading period of PGS/PCL tubes to about 2 weeks and thereafter matched the cell recruitment in rabbits, thus warranting the success of endothelialization. In addition, the highly porous structure of 3DVGs resulted in a thicker wall compared with native vessels and would offer more space for cell homing and peptide conjugating. PCL sheath enforced the strength of porous PGS/PCL and prevented blood leakage. As the slowly degrading part of graft, it has been previously revealed that such a nanofibrous interface guided muscular remodeling through regulating “Macrophage-Sca-1+ cells cross-talk” and therefore added much mechanical strength for grafts when porous PGS degraded in vivo (, ). Together, 3DVGs enabled timely degradation to maintain the mechanical strength matching vascular cell infiltration.

Microporous structure not only provides space for cell infiltration but also offers abundant heparin-binding sites for peptides. Although smooth lumens with or without topographical modification have shown superiority in counteracting platelet aggregation, they are poor in recruiting vascular cells (). The porous surface provides more PGS area for blood contacting, thus offering more opportunity to adhere stem cells. 3D printing technique acquired desired porosity, pore size, and PGS ratio by adjusting the sucrose particles-PGS/PCL mixing ratio. As shown in Fig. 1G, the complementarities of PGS and PCL were sufficiently balanced through 3D printing, and acquired enhanced peptides mediated antithrombogenicity. Note that the luminal surface of 3DVGs appeared as a wavy structure. As compared with smooth lumen in most vascular grafts, the wavy structure provided more area for DTβ4 conjugation, thereafter facilitating the antithrombogenicity of DTβ4-conjugated grafts. Moreover, the wavy luminal surface offered more area for blood contacting and slowed blood flow contacting the graft wall, which was supposed to be favorable for EPC infiltration.

The in vitro experiment emulated the placement of the PGS/PCL constructs in the circulation to answer two questions: (i) Does DTβ4 take a stronger effect in anticoagulation as compared with heparin or Tβ4? (ii) Can DTβ4-grafted surface selectively adhere EPCs? We unexpectedly found that DTβ4 was stronger in antiplatelets and inhibiting fibrinogen than Tβ4 when administered at a low dosage. The same trend was also presented in the proliferation, migration, and EC differentiation of bone marrow–derived CD34+ cells. This property favored peptides grafting on material surfaces. Limited binding sites and areas always restricted the amount of peptide loading. Although dimers will double the dosage as compared with monomers, the loading peptides remain at a low level. Although the extended metabolism period may contribute to this superiority, the detailed mechanism may also be involved in the molecular superposition effect of dimers, and evidence requires further investigation. One limit in this study was that we failed to use human BMMNCs to evaluate the ability of DTβ4-3DGs to capture CD93+ cells, mainly owing to the preciousness of fresh human bone marrow and ethical limitation. We will supplement related results in our further studies on large animals.

scRNA-seq allows large-scale profiling of cell properties in complex tissues, which has identified cellular heterogeneity in skin repair, carcinogenesis, and adult mammalian lungs (–). Here, two main genetic CD34 subsets were identified in vascular grafts. CD34/CD93 cells dominate, whereas CD34 cells only constituted 10% of the CD34+ cells in 2-week grafts. Considering the degrading of DTβ4-3DVGs and compromised peptide activity at this time, CD34+/CD93− cells may bear more anticoagulant tasks through adenosine diphosphate–induced platelet integrin receptor αIIbβ3 activation and P-selectin membrane expression (–). CD93 is a transmembrane receptor that is up-regulated in tumor vessels in many cancers and is closely related to the proliferation and migration of ECs. Unexpectedly, analysis of 2-week samples revealed that most regenerated immature ECs were gene CD93, pseudotime, and RNA velocity analyses predict CD93+ EPC-to-EC differentiation trajectories. It was also shown that abundant CD93+ cells were scattering in adventitial tissues at 12 weeks; we suspected that these cells were transmurally migrated from blood stream. Most of these cells were CD93+/CD31− and failed to trace the ingrowing microcapillaries (CD31+), suggesting its fewer relationships with ECs in mature microcapillaries. As the evidence, tracing GFP+ cells also revealed the appearance of GFP+/CD93+ inside 3DVGs (Fig. 9). Most of CD93+ cells transmurally migrated in 3DVGs failed to contact blood stream, were shielded from shear stress stimulus, a key factor to drive EC differentiation, and therefore constituted CD93+/CD31− clusters. These findings affirmed the pivotal roles of CD93+/CD34+ cells in the endothelialization of vascular grafts. Note that CD34+ cells promoted tissue healing and reparative angiogenesis, mainly through secreting growth factors and extracellular vesicles (–). Seeding EPCs on vascular grafts may have a direct vasculogenic ability, as well as a strong paracrine effect to recruit host ECs (). Together, CD34+/CD93− cells may promote the migration, proliferation of CD34+/ CD93+ cells, and their differentiation through paracrine effects.

MATERIALS AND METHODS

Fabrication of 3DVGs conjugated with peptides

To improve the mechanical properties of carotid grafts, the grafts were fabricated from blended polymer PGS/PCL via the combination of 3D printing () and electrospinning technology. PGS was synthesized in-house as previously reported (). Blended polymer PGS and PCL (Mn 80,000 g mol −1; Sigma-Aldrich, MO, USA) were dissolved in N,N-dimethylformamide (20%, w/v) at mass ratios of 100:0, 85:15, 70:30, respectively. The polymer solution was then mixed with sucrose particles (30 to 38 μm) at 1:1, 1:2, and 1:3 weight ratios (polymers:sucrose particles). The solution was carried out with magnetic stirring for 12 hours at 70°C for solvent evaporation to obtain the printing ink. The mixture ink was printed into tubular grafts and further cured in a vacuum oven at 150°C (−0.1 MPa) for 24 hours. Electrospinning technology was used to fabricate PCL sheath on the PGS/PCL-sucrose core as previously described (). Briefly, PCL was dissolved in trifluoroethanol at 14% (w/v) concentration. The PCL solution was fed for 3 min at a rate of 1.5 ml/hour through a 22-gauge stainless steel needle using a syringe pump, and a high voltage (26 kV) was applied. After electrospinning, the PGS/PCL grafts were immersed in distilled water for 24 hours to remove the sucrose porogen completely. Direct immobilization of peptides on the scaffold would lead to the denature of peptides (). Therefore, we used heparin-binding strategy to immobilize peptides onto the 3DVGs. 3DVGs were then immersed in 0.1 M ethylenediamine solution with gentle shaking for 60 min and rinsed with distilled water three times. The 3DVGs were then transferred to MES solution (pH 5.6) containing 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (0.1 M), N-hydroxysuccinimide (0.1 M), and heparin (10 mg/ml). The heparinization proceeded for 12 hours at 4°C. 3DVGs were sufficiently rinsed three times with distilled water to obtain heparinized 3DVGs. Subsequently, the 3DVGs were immersed in the solution of PBS, VEGF165 (1 mg/ml) (MedChemExpress, Shanghai, China, HY-P7311), Tβ4 (1 mg/ml; 4963.49 Da; China Peptides Ltd., Shanghai), or DTβ4 (1 mg/ml) for 2 hours in the ice bath. Last, 3DVGs were lyophilized in a vacuum, sterilized by ethylene oxide, and preserved at −20°C until use.

Characterization of the 3DVGs

Micromorphology of 3DVGs was observed via SEM (s-4800, Hitachi, Japan). The specimens were fixed in 2.5% glutaraldehyde, dehydrated with an alcohol gradient, soaked in hexamethyldisilane, and air-dried to prepare the specimens. Then, the dehydrated specimens were cut into segments of different sizes and sputtered with platinum to examine the inner, outer, longitudinal, and cross sections of the grafts. The elements on the 3DVGs were measured by SEM equipped with EDS.

micro-CT analysis

Specimens were cut into 6-mm-long segments and scanned by a Siemens micro-CT system (Siemens, Germany). Raw data were reconstructed into 2D tomograms. Average pore size and wall thickness were performed by Inveon Research Workplace. Pore interconnectivity was calculated by inverting the solid and pore spaces, eliminating any disconnected closed pores, and determining the percentage of connected pore space to total pore volume (). To assess porosity, pore space and whole graft volume were identified. The ratio of pore volume to the whole volume was acquired as “porosity.”

Mechanical characterization

Tensile strength and compressive strength of the grafts were measured using the force test system (BOSE, ElectroForce 3200 Series II, USA). To measure tensile and compressive strength, 3DVGs were cut into 5-mm-long segments (n = 5). The uniaxial tensile force was applied to segments at a rate of 5 mm/min with an initial force of 0.1 N until the break. Each 3DVGs (n = 5) was compressed to a strain of 50% for measuring compressive force. Burst pressure and compliance of neoarteries were measured by a perfusion system (MFCS-EZ Microfluidic Flow Control System, Fluigent, France).

Quantification of heparin and peptides on grafts

The heparin-conjugated 3DVGs were qualitatively observed by toluidine blue (TB) staining. Briefly, 0.05% TB solution was prepared in 0.01 M HCl with 0.2% (w/v) NaCl, and heparin-bonded grafts were incubated in the TB solution for 12 hours; grafts without heparin bond were incubated as control. The samples were incubated in the solution of 100% ethanol and 0.1 M NaOH (4:1, v/v) with gentle shaking to dissolve TB completely. The resultant solution’s absorbance was obtained at a wavelength of 530 nm via a spectrophotometer (NanoDrop One, Thermo Fisher Scientific, USA). Quantitative analysis of each sample was calculated from the heparin standard curve. The quantification of peptides was measured using an enzyme-linked immunosorbent assay (ELISA) kit (BIOMATIK, catalog no. EKU07666) according to the kit instruction. The quantification of VEGF was measured using an ELISA kit (BIOMATIK, catalog no. EKA51932) according to the kit instruction.

Quantification of PGS/PCL residuals in grafts

Polymer residuals were also calculated according to the approaches reported previously (). Polarized H&E images (three per sample) from different samples (n = 5) in each group at the designated time points were used to quantify the area of polymer residual. The PCL residual visualized with polarized light presented as specific white birefringence, while there was no birefringence on PGS. The PGS residual visualized with bright field appeared as a slightly stained porous structure. The areas of polymer residuals were measured using ImageJ software (National Institutes of Health).

Surface contact angles

The 3D printing ink was squeezed into a circular polytetrafluoroethylene (PTFE) mold to fabricate porous 3DGs to test hydrophilicity. Water contact angles were measured by a drop shape analysis system (Krüss, EASY DROP K100, Germany) followed by image processing of a drop of distilled water at room temperature via DSA 1.90.0.2 software.

Platelet and recalcified whole blood adhesion and flow cytometry analysis

The anticoagulant ability of 3DVGs in vitro was determined by platelet and recalcified whole blood adhesion assay. Fresh rabbit whole blood was maintained in the anticoagulant tubes containing 3.8% sodium citrate. After the addition of calcium chloride (CaCl2) solution (100 mM), 3DGs were incubated with recalcified whole blood for 30 min. The 3DGs with clotted blood were rinsed with PBS and weighed. The PRP was obtained by centrifuging the rabbit whole blood at 200g for 15 min. 3DGs were incubated with PRP at 37°C for 1 hour. After gently rinsing the 3DGs with PBS, the amount of platelets adhered to 3DGs was determined by an LDH kit (BC0685, Solarbio, Beijing, China) according to the instructions. Then, the morphology of 3DGs with platelet and recalcified whole blood adhesion was examined by SEM.

Fresh rat whole blood was drawn into EDTA anticoagulation tubes. We added DTβ4 (10 μg/ml), mixed well, took 10 μl, and diluted to 1 ml with PBS. We also added 2 μl of CD62P (BioLegend, B311813) and CD41 (BioLegend, B282877) antibodies, incubated for 2 hours in the dark, and then detected using a FACSCalibur flow cytometer (Becton Dickinson, Franklin Lakes, USA).

Animal experiments

All the experiments involving rabbits and rats were purchased from the animal center of Fourth Military Medical University, Xi’an, China. All animal procedures were approved by the Animal Experiments Ethical Committee, Fourth Military Medical University and complied with animal welfare regulations. Male New Zealand white rabbits (6 months old; n = 35) weighing 3.5 to 4.5 kg underwent left carotid artery interposition grafting with 15-mm-long 3DVGs (5 in 100% PGS group, 5 in 85% PGS group, 5 in 70% PGS group, and 20 in DTβ4-3DVGs group). Male SD rats (8 weeks old) weighing 300 to 350 g underwent abdominal aorta interposition grafting with 8-mm-long 3DVGs (n = 10 in PBS group; n = 10 in DTβ4 group). For acute anticoagulation evaluation test in vivo, 15-mm-long 3DVGs (five in bare 3DVGs group, five in Hep-3DVGs group, five in Tβ4-3DVGs group, five in DTβ4-3DVGs group, and five in VEGF-3DVGs group) were used for carotid artery interposition and explanted after 2 hours. No systematic anticoagulation treatment was administrated postoperatively, and all animals were maintained on a 12:12-hour light cycle. The surgeries were performed by operators blinded to the drug-loading grafts.

Bone marrow transplantation

GFP transgenic rats (2 to 4 weeks old, 70 to 100 g) were purchased from Cyagen Biosciences Inc. (Guangdong, China). After 1-week acclimatization to the new environment, GFP transgenic rats were used as donors for the extraction of GFP+ bone marrow cells (BMCs). The recipient SD rats were lethally irradiated with two apart doses of 4.5 grays (Gy) each (9.0 Gy in total) using a Cobalt-60 gamma source. GFP+ BMCs (1 × 107) were injected via tail vein into recipient rats before implantation of 3DVGs into the abdominal aorta. The recipient SD rats were sacrificed to analyze cell recruitment in remodeling 3DVGs.

Biochemical evaluation

Tissues from neoarteries and native carotid were acquired for collagen and elastin quantification (n = 5 independent samples). To measure the collagen content, the total collagen of each explant was quantified by the Sircol Collagen Assay Kit. Collagen concentration in pooled supernatants was measured following the kit instruction, and insoluble collagen per wet weight of each sample was calculated from the standard collagen curve. The total elastin of each explant was extracted and measured using a Fastin Elastin assay (F2000; Biocolor, Carrickfergus, UK) according to the instructions, then pooled, and normalized to sample wet weight to yield the mass of insoluble elastin per tissue wet weight (μg/mg). The desmosine content of neoarteries was evaluated by the desmosine ELISA kit (MBS751798; Belgium) as described previously ().

Histology evaluation

The neoarteries were embedded in OCT (Sakura Finetek, USA) and processed for cryosections with a thickness of 7 μm. The sections were immersed in hematoxylin for 10 min followed by 70% ethanol and eosin for 5 min, then dehydrated in alcohol, and treated with xylene for discoloration to complete H&E staining. Masson staining (G1340, Solarbio, Beijing, China) and EVG (ab150667, Abcam, MA, USA) followed the instructions provided by the reagent supplier. All histological images were observed with an upright fluorescence microscope (DM6000B, Leica, Germany) in bright field or under polarized light.

Cell culture

hUVECs were obtained from the American Type Culture Collection Cell Bank (Shanghai, China). BMMNCs were derived from the bone marrow of 2-week-old SD rats. After the rats were sacrificed, femurs and tibias were taken out and the bone marrow was flushed out and cultured overnight. Transfer the supernatant to a new petri dish, which was BMMNC. The cells were cultured in Dulbecco’s modified Eagle’s medium/high glucose (HyClone, Logan, UT, USA) containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA). When the BMMNC density reached 70%, the EGM-2 medium (CC-3162, Lonza, Switzerland) containing recombinant protein VEGF165 (20 ng/ml; HY-P7311, MedChemExpress, Shanghai, China) and macrophage colony-stimulating factor (1 ng/ml; HY-P7247, MedChemExpress, Shanghai, China) was replaced to induce differentiation. All cells were passaged when the density reached 70%, and the culture medium was replaced every 2 days. The experiment was performed when the cells were in two to four generations.

Immunofluorescence staining

Explant samples were fixed using glacial acetone, permeabilized using 0.5% Triton X-100, and then blocked in 5% bovine serum albumin (BSA) for 30 min. Thereafter, samples were sequentially incubated with primary antibody [elastin, 1:200 (Novus, NB100-2076); CD93, 1:100 (Bioss, bs-10232R); CD14, 1:100 (Bioss, bs-1192R); vWF, 1:1000 (Abcam, ab6994); Ki67, 1:200 (Abcam, ab15580); CD31, 1:200 (Novus, NB600-562); CD11b, 1:200 (Novus, NB600-1327); CD34, 1:200 (GeneTex, GTX28158); α-SMA, 1:200 (Abcam, ab242395); Collagen I, 1:200 (GeneTex, GTX26308); Collagen III, 1:200 (GeneTex, GTX26310); CD31, 1:200 (Abcam, ab24590); eNOS, 1:200 (Abcam, ab76198)] overnight and fluorescent secondary antibody (1:1000; Abbkine) for 1 hour at 37°C. Last, nuclei were stained with 4′,6-diamidino-2-phenylindole. After each step of incubation, samples were rinsed with PBS five times, 5 min each time. Images were captured using a laser scanning confocal microscope (FV 1000, Olympus, Japan) and a stereo fluorescence microscope (Leica, MF205, Germany). Different cell populations were determined on the basis of cell counts from each immunostained images on five different parts (the center and 12, 3, 6, and 9 o’clock position, respectively). The data were collected from at least three different 3DVGs in each group.

Single-cell RNA library preparation and sequencing

The neoarteries were harvested at 2 and 12 weeks, rinsed with PBS three times. scRNA-seq analysis was performed by experimental personnel in the laboratory of NovelBio Bio-Pharm Technology Co. Ltd. The tissues were surgically removed and kept in MACS Tissue Storage Solution (Miltenyi Biotec) until processing. The tissue samples were processed as described below. Briefly, samples were first washed with PBS, minced into small pieces (approximately 1 mm3) on ice, and enzymatically digested with collagenase I (2 mg/ml; Worthington) and Dispase II (1 mg/ml; Worthington) for 50 min at 37°C, with agitation. After digestion, samples were sieved through a 70-μm cell strainer and centrifuged at 300g for 5 min. After the supernatant was removed, the pelleted cells were suspended in red blood cell lysis buffer (Miltenyi Biotec) to lyse red blood cells. After washing with PBS containing 0.04% BSA, the cell pellets were resuspended in PBS containing 0.04% BSA and refiltered through a 35-μm cell strainer. Dissociated single cells were then stained for viability assessment using Calcein AM (Thermo Fisher Scientific) and Draq7 (BD Biosciences).

BD Rhapsody system was used to capture the transcriptomic information of the single cells. Single-cell capture was achieved by the random distribution of a single-cell suspension across >200,000 microwells through a limited dilution approach. Beads with oligonucleotide barcodes were added to saturation so that a bead was paired with a cell in a microwell. The cells were lysed in the microwell to hybridize mRNA molecules to barcoded capture oligos on the beads. Beads were collected into a single tube for reverse transcription and Exo I digestion. Upon cDNA synthesis, each cDNA molecule was tagged on the 5′ end (that is, the 3′ end of an mRNA transcript) with a unique molecular identifier (UMI) and cell barcode indicating its cell of origin. Whole transcriptome libraries were prepared using the BD Rhapsody single-cell whole-transcriptome amplification (WTA) workflow, including random priming and extension (RPE), RPE amplification polymerase chain reaction (PCR), and WTA index PCR. The libraries were quantified using a High Sensitivity DNA chip (Agilent) on a Bioanalyzer 2200 and the Qubit High Sensitivity DNA assay (Thermo Fisher Scientific). Sequencing was performed by Illumina sequencer (Illumina, San Diego, CA) on a 150–base pair paired-end run.

Bioinformatic analysis

scRNA-seq data analysis was performed by NovelBio Bio-Pharm Technology Co. Ltd. with the NovelBrain Cloud Analysis Platform. We applied fastp () with default parameter filtering the adaptor sequence and removed the low-quality reads to achieve the clean data. UMI-tools () were applied for single-cell transcriptome analysis to identify the cell barcode whitelist. The UMI-based clean data were mapped to rabbit (OryCun2.0 Ensembl) using STAR () mapping with customized parameters from the UMI-tools standard pipeline to obtain the UMIs counts of each sample. Cells contained more than 200 expressed genes; mitochondria UMI rate below 20% passed the cell quality filtering, and mitochondria genes were removed in the expression table. Seurat package (version: 3.1.4; https://satijalab.org/seurat/) was used for cell normalization and regression on the basis of the expression table according to the UMI counts of each sample and percent of mitochondria rate to obtain the scaled data. Principal components analysis (PCA) was constructed on the basis of the scaled data with top 2000 high variable genes, and top 10 principals were used for t-distributed stochastic neighbor embedding (tSNE) construction and UMAP construction.

Using the graph-based cluster method (resolution, 0.8), we acquired the unsupervised cell cluster result based on the PCA top 10 principal and we calculated the marker genes by FindAllMarkers function with Wilcox rank sum test algorithm under the following criteria: (i) lnFC > 0.25, (ii) P < 0.05, and (iii) min.pct > 0.1. To identify the cell type detailed, the clusters of the same cell type were selected for re-tSNE analysis, graph-based clustering, and marker analysis.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was used to find out the significant pathway of the differential genes according to the KEGG database. We turned to Fisher’s exact test to select the significant pathway, and the threshold of significance was defined by P value and false discovery rate (FDR).

GO analysis

GO analysis was performed to facilitate elucidating the biological implications of unique genes in the significant or representative profiles of the differentially expressed gene in the experiment. We downloaded the GO annotations from NCBI (www.ncbi.nlm.nih.gov/), UniProt (www.uniprot.org/), and the GO (www.geneontology.org/). Fisher’s exact test was applied to identify the significant GO categories, and FDR was used to correct the P values.

Pseudo-time analysis

We applied the single-cell trajectory analysis using Monocle2 [Monocle (cole-trapnell-lab.github.io)] using DDR-Tree and default parameter. Before Monocle analysis, we select marker genes of the Seurat clustering result and raw expression counts of the cell passed filtering. On the basis of the pseudo-time analysis, branch expression analysis modeling was applied for branch fate determining gene analysis.

Cell-cell communication analysis

To systematically analyze cell-cell communication, intercellular interactions based on ligands and receptors were inferred using the CellPhoneDB () and the recently developed CellChat platform (). Briefly, cell-cell interaction heatmaps showing the number of interactions were generated using pheatmap R package, and ligand-receptor interactions were visualized using ggplot2 R. The differential edge list was passed through CircosDiff (a wrapper around the R package “circlize”) and netVisual_chord_gene in CellChat to filter receptor ligand edges and generate Circos plots. Significant mean and cell communication significance (P < 0.05) were calculated on the basis of the interaction and the normalized cell matrix achieved by Seurat normalization.

Quantitative real-time PCR

Total RNA was extracted from cells using TRIzol reagent (Takara Bio, Otsu, Japan) according to the manufacturer’s instructions. After measuring the concentration, the RNA was stored at −80°C until use. About 500 ng was reverse-transcribed into cDNA using PrimeScript RT Enzyme Mix (Takara Bio) and used for quantitative real-time PCR, which was performed on ABI StepOne Plus system (Applied Biosystems, Foster City, CA, USA) using SYBR Premix Ex Taq (Vazyme Biotech, Nanjing, China). The relative expression levels of the genes of interest were calculated with the 2−ΔΔCt method, with β-actin used as an internal reference gene. Primer sequences are shown in table S1.

Statistical analysis

All the replicate experiments were biological replicates that were repeated at least three times. All analyses involved the use of SPSS23 (IBM). All data were represented as means ± SD. Comparison of two groups was determined by Student’s t test, and that of multiple groups was determined by one-way analysis of variance (ANOVA) with Tukey’s post hoc test. P < 0.05 was considered statistically significant.