Perfusable branching microvessel bed for vascularization of engineered tissues

Abstract

Vascularization is critical for the survival of engineered tissues in vitro and in vivo. In vivo, angiogenesis involves endothelial cell proliferation and sprouting followed by connection of extended cellular processes and subsequent lumen propagation through vacuole fusion. We mimicked this process in engineering an organized capillary network anchored by an artery and a vein. The network was generated by inducing directed capillary sprouting from vascular explants on micropatterned substrates containing thymosin β4-hydrogel. The capillary outgrowths connected between the parent explants by day 21, a process that was accelerated to 14 d by application of soluble VEGF and hepatocyte growth factor. Confocal microscopy and transmission electron microscopy indicated the presence of tubules with lumens formed by endothelial cells expressing CD31, VE-cadherin, and von Willebrand factor. Cardiac tissues engineered around the resulting vasculature exhibited improved functional properties, cell striations, and cell–cell junctions compared with tissues without prevascularization. This approach uniquely allows easy removal of the vasculature from the microfabricated substrate and easy seeding of the tissue specific cell types in the parenchymal space.

Tissue engineering may offer alternative treatment options for tissue and organ replacement, but challenges related to vascularization in vitro and in vivo remain. Organized vasculature is important for the integration of blood vessels with other tissue-specific cell types to create complex tissues (1), especially cardiac tissues that require small intercapillary distances for metabolically active cardiomyocytes (2). Moreover, organized vasculature allows easy isolation, manipulation, and implantation of the vascular structures (1). Strategies developed to engineer vascularized tissues generally have involved (i) coculture of tissue-specific cells with endothelial cells (ECs) or their precursors (3, 4); (ii) incorporation of preexisting endothelial networks or blood vessels (5); or (iii) engineering of proangiogenic scaffolds with peptides, growth factors, or geometric cues (6–11).

The incorporation of ECs can accelerate vascularization of the implanted biomaterial or engineered tissue and its functional anastomosis with the host vasculature (5, 12). In vivo, vascularization can be enhanced by cultivating cells around an arteriovenous loop (5) or by implanting materials into omentum (10). Alternatively, pore size alone (interconnected pores 30–40 μm in diameter) can promote angiogenesis in porous materials (6). Peptides and growth factors such as thymosin β4 (Tβ4) (13, 14), VEGF (7–9), and angiopoietin-1 (8, 9) lead to increased vessel density when incorporated into tissue-engineering biomaterials in soluble, encapsulated, or immobilized form. Recent studies have focused on improving the organization of ECs using microcontact printing and microgrooved substrates (15–17).

Although these methods suggest that improving vascularization enhances engineered-tissue function, limitations in the control over cell position and the ultimate organization of the vasculature lead to the formation of networks with random spatial distribution, with implications for network function (10, 18). Recent breakthrough studies enable microfabrication of microvascular structures with barrier function (19); however, when the geometry of tubules was controlled, there was no facile way to seed cells in the parenchymal space, because it was occupied by a synthetic or natural polymer (1, 19). Additionally, these approaches do not enable easy removal of the engineered microvascular bed from the microfabricated mold or microfluidic device used to house the vasculature.

Here, we devised a method to engineer a prototype vascular network consisting of two branching vessels that form a microvascular bed suitable for rapid vascularization of engineered tissues in vitro. Our approach integrates the use of topographical cues and sustained release of angiogenic factors (Fig. 1A). Rather than forcing isolated ECs into defined positions on a polymeric or hydrogel matrix, we used biomaterials to create a niche that directed sprouting and anastomosis of the vasculature from the two branching vessels in vitro. We demonstrated the utility of this vascular network for improving the function of engineered myocardium. Functional properties of myocardium are determined by cardiomyocytes, a cell type sensitive to oxygen limitations because of their high metabolic activity. The presence of a microvascular bed in an engineered cardiac tissue in vitro may provide a biomimetic milieu for functional cell assembly, leading to cardiomyocyte orientation and survival.

Fig. 1.

Substrate topography and controlled release of angiogenic peptide Tβ4 guide capillary outgrowths from arteries and veins. (A) Experimental setup. (i) PDMS substrates were fabricated using standard soft lithography methods and (ii) were coated with collagen-chitosan hydrogel with or without Tβ4. (iii) Arteries and veins were isolated and placed on the two ends of the substrate and (iv) were cultivated for 2 wk with HGF or VEGF supplementation or for 3 wk without any growth factors. (v) Cardiomyocytes were seeded onto the engineered vascular bed and cultured for additional 7 d to grow a beating, vascularized cardiac tissue. (B) Representative images of cell outgrowths from mouse artery and vein explants on nonpatterned substrates at day 14 of cultivation (n = 3 per group). The asterisks indicate the location of explants; arrows indicate locations of outgrowths. (Scale bars, 200 μm.) (C) Density of migrated cells from artery explants at various time points. (D) Density of migrated cells from vein explants at various time points. (E and F) Autocrine VEGF secretion was dependent on the Tβ4 dose and correlated with cell outgrowth density. (E) Total amount of VEGF measured with artery explants between days 12 and 14 (n = 3 per group; P = 0.0034, one-way ANOVA; P < 0.05 for control vs. both Encap100 and Encap1500). (F) Total amount of VEGF measured with vein explants between days 12 and 14 (n = 3 per group, P = 0.0036 for control vs. Encap1500; P = 0.0055 for Encap100 vs. Encap1500). (G–J) Tβ4 increased the branch length of capillary outgrowths from mouse artery and vein explants at day 14, with the most pronounced effect on substrates with grooves 50-μm wide. (G) Representative images of outgrowths (n = 10–12 per group). (Scale bars, 100 μm.) (H) Branch density on grooves with different widths (a branch is defined by the distance between two nodes or between a node and the end of a tube; P = 0.0002, two-way ANOVA; P < 0.01, Bonferroni posttest for control vs. Tβ4 gel groups on 100-μm grooves; P = 0.7078 within Tβ4 gel groups). (I) Average branch length on grooves with different widths (P < 0.0001, two-way ANOVA; P < 0.05 for control vs. Tβ4 gel on 25-μm grooves; P < 0.001 for control vs. Tβ4 gel on 50-μm and 100-μm grooves; P < 0.001 for 50-μm grooves vs. both 25-μm and 100-μm grooves in Tβ4 gel groups). (J) Average branch width on grooves with different widths (P = 0.3782). Micropatterned PDMS substrates had grooves 25, 50, and 100 μm in width and 65 μm in height. Experimental groups include hydrogel without Tβ4 (control) and hydrogel with 1,500 ng encapsulated Tβ4 (Tβ4 gel). The asterisks in graphs indicate statistically significant differences between groups (P < 0.05, two-way ANOVA with Bonferroni posttests for cell density, branch density, branch length, and branch width; one-way ANOVA with post hoc Tukey tests for VEGF expression).

Our strategy involved controlled release of Tβ4 from the substrate coating to mimic the in vivo environment in which biomolecules are present to promote formation of vascular network. Tβ4 is an angiogenic and cardioprotective peptide that enhances cardiomyocyte survival by inducing coronary vascularization and up-regulation of Akt activity (14, 20). Tβ4 was shown previously to initiate angiogenesis and sustain vascular stability because of its ability to recruit endothelial and smooth muscle cells (13, 14). Compared with angiogenic growth factors, Tβ4 has advantages such as small size, high activity, minimal immunogenicity, and high solubility in water.

Soluble biomolecules are easily degraded and washed out, requiring multiple doses at increased levels to maintain bioactivity (21). We previously developed a collagen-chitosan hydrogel capable of releasing Tβ4 at nearly zero-order kinetics from day 3 to day 28, with the release behavior over 28 d well-fitted to Higuchi’s square root equation (22). Although collagen promotes cell attachment, survival, and proliferation, the interconnected pores in the collagen-chitosan hydrogel and the electrostatic interactions between negatively charged Tβ4 and positively charged chitosan contribute to the sustained release of Tβ4 (22). In contrast, Tβ4 was released quickly (within 3 d) from collagen-only hydrogel that was negatively charged and did not exhibit the interconnected porous structure of collagen-chitosan hydrogel (22). Here, we used Tβ4 to generate vascularized cardiac tissues because of its angiogenic and cardioprotective potential, and our hydrogel delivery system enabled localized and sustained activity. We applied the Tβ4-encapsulated hydrogel onto micropatterned polydimethylsiloxane (PDMS) substrates with 25-, 50-, or 100-μm–wide grooves (Fig. 1A and Fig. S1). Subsequently, mouse, rat, or human arterial and venous explants were placed at the two ends of the substrate to promote outgrowth of capillaries.

Results

To determine if the inclusion of Tβ4 improved cell recruitment and accelerated network formation, we exposed mouse arterial and venous explants to different amounts of Tβ4 encapsulated in collagen-chitosan hydrogel (0 ng for control, 100 ng for Encap100, and 1,500 ng for Encap1500). Increased cell outgrowth density was observed with a higher dose of Tβ4 (1,500 ng vs. 0 ng and 100 ng; Fig. 1 B–D). In the control group, higher outgrowth density was observed in venous explants than in arterial explants. However, there were no significant differences between the outgrowth densities in arterial vs. venous explants when Tβ4 was applied (compare Fig. 1 C and D).

Although Tβ4 alone can stimulate EC migration and differentiation (14, 23), Tβ4 also stimulates the synthesis and autocrine secretion of VEGF, as previously reported (24). Here, we observed enhanced VEGF secretion from explants cultivated on Tβ4 hydrogels compared with Tβ4-free gels (Fig. 1 E and F). Our basal culture medium contains serum, which may be the source of VEGF and was reported previously to stimulate quiescent ECs to proliferate (25). We measured the basal VEGF concentration in the culture medium as 0.06 ± 0.006 ng/mL by ELISA. However, this level was increased by autocrine secretion in the presence of cultivated explants (Fig. 1 E and F). Approximately 20% of the secreted VEGF was released to the medium, and 80% remained in the hydrogel (Fig. S2). In general, the basal level of VEGF secretion was higher with venous explants than with arterial explants cultivated without Tβ4 (compare Fig. 1 F and E), correlating with the greater cell density on control hydrogel in the presence of venous explants as compared with arterial explants (white bars in Fig. 1 C and D). The amount of VEGF measured in the culture medium for artery explants at day 14 was increased with the application of either 100 ng or 1,500 ng Tβ4. For vein explants, the amount of VEGF was increased with 1,500 ng Tβ4 gel as compared with either control or 100 ng Tβ4-gel. Thus, there was a correlation between VEGF secretion and cell outgrowth densities. A positive linear correlation was found when plotting the average VEGF concentration produced between days 12 and 14 for each group (Fig. 1 E and F) against the corresponding average cell density at day 14 (Fig. 1 C and D) (r = 0.85, P = 0.03; Fig. S3). The Encap1500 group (referred to as “Tβ4 gel”) was used for further studies because of the observed dose response of both VEGF secretion and cell outgrowth densities to Tβ4.

We next aimed to determine if we can guide and control the sprouting from mouse arterial and venous explants through the integrated use of topographical cues (25-, 50-, and 100-μm–wide grooves) and the controlled release of Tβ4. Previously, exogenous, soluble Tβ4 was found to increase sprouting from coronary artery rings in an angiogenesis assay (23), but the directionality of the outgrowths was not controlled. We evaluated three parameters to describe the sprouting process: branch density, branch length, and branch width. Increased branch density correlates with a higher level of sprouting often found during angiogenesis. Increased branch length is necessary for rapidly forming connections between the parent explants. Ideally, the branch width should be similar to that of the capillaries in the native vasculature.

Both the branch density and length within the outgrowths from mouse vascular explants (Fig. 1 H and I) were increased with the addition of encapsulated Tβ4 in hydrogel coating. Branch density for the Tβ4 gel groups was not affected by groove width (Fig. 1H). However, branches were significantly longer on 50-μm grooves than on other groove widths (Fig. 1I). The lower branch lengths on substrates with smaller groove size likely were related to the hindrance of tube growth by spatial limitations for the sprouts growing inside the grooves. At larger groove widths, capillary outgrowths tended to branch more because of the wide channels available for expansion of the vascular structure, thus lowering branch length (Fig. 1 G and I). There was no significant difference in tube widths in the control and Tβ4 gel groups (Fig. 1J). Tubes grown on substrates with 25-, 50-, and 100-μm grooves showed comparable widths of ∼10–12 μm, typical of capillaries in the native mouse myocardium (∼10–20 μm) (26). Thus, we used 50-μm–wide grooves in further studies. Arterial and venous explants did not differ significantly in branch density, length, or width, and the results are shown as lumped rather than separately for arteries or veins.

Although both cell outgrowth density and autocrine VEGF concentration were higher with vein explants (Fig. 1 D and F), there were no significant differences in branch density, length, and width in artery and vein explants (Fig. 1 H–J). VEGF receptor 2 (VEGFR2) is considered to be the major mediator of physiological effects such as proliferation, migration, survival, and tube formation caused by VEGF₁₆₅. Binding of the VEGF to VEGFR2 causes receptor dimerization and autophosphorylation, followed by the activation of a number of downstream pathways to mediate different processes. For example, activation of the ERK pathway regulates proliferation (27); activation of the p38 MAPK, PI3K, and Rac pathways regulates migration (28–30); and activation of Src regulates both migration and vascular permeability (29). Tube formation by VEGFR2–VEGF₁₆₅ signaling was reported to be mediated by MAPK (31, 32), p38 (31), and PI3K (30, 32, 33). Therefore, because of the complexity and interconnectivity of the intercellular signaling pathways, increased cell proliferation and migration may not translate proportionally to increased tube formation.

Our results clearly indicate that the enabling factor in tube formation and guidance was the presence of topographical cues, whereas the presence of angiogenic factor Tβ4 simply accelerated the formation of the tubes and enhanced their length (Fig. 1). Introduction of topographical cues was essential for the formation of branches with open lumens, because Tβ4 gel alone led to cell outgrowth without the formation of luminal structures (Fig. 1B). Topographical cues themselves can enable tube formation by contact guidance, changing the local mechanical properties of the environment, or locally increasing the concentration of autocrine growth factors. We aimed to understand further the factors by which topographical cues affect tube formation by considering appropriate controls: (i) flat PDMS substrates with control or Tβ4 hydrogel, and (ii) uncoated PDMS substrates with 50-μm grooves and Tβ4 placed in the culture medium (Fig. 2A). Flat controls demonstrated cell migration from the artery or vein explants, with more cell outgrowths on Tβ4 gel-coated substrates than on control gel-coated substrates. However, the cells did not form any vascular tube structures on flat substrates, thus indicating the importance of topography for tube formation and organization. Although controls with uncoated grooved substrates led to tube formation by migrated cells, these vascular structures were not well organized. This result shows the importance of the localization of the Tβ4 molecules to the groove/ridge space, because Tβ4 was present in the culture medium but was not localized within the hydrogel coating in this case (Fig. 2A).

Fig. 2.

Topographical cues promote tube formation by providing contact guidance and locally increasing the concentration of autocrine growth factors but not by affecting the local mechanical properties. (A) Contact guidance. YFP mouse artery and vein explants were cultivated on flat PDMS substrates or uncoated substrates. Flat substrates were coated with either control gel (Top) or Tβ4 gel (Middle). Explants were cultivated for 21 d. The vein on uncoated substrate (Bottom Right, Inset) shows unorganized tube formation. Asterisks indicate the locations of artery or vein explants, and yellow arrows indicate the presence of tube formation by outgrown cells. (Scale bars, 50 μm.) (B) Local mechanical stiffness. Local Young’s modulus for hydrogel-coated PDMS substrates without grooves and with 25-, 50-, and 100-μm–wide grooves, as measured by atomic force microscopy in a 50 × 50 μm area spanning a part of the gel-coated groove and a ridge for each substrate. Uncoated flat PDMS substrate also was measured for comparison. The local stiffness of uncoated flat PDMS substrate was significantly higher than that of all coated substrates (P < 0.0001, one-way ANOVA), but no differences were found among coated substrates. (C) Local concentration of autocrine growth factors. A simplified mathematical model of the VEGF concentration gradient in grooved and smooth samples, showing the effect of grooves on the local increase in the concentration of autocrine growth factors. (i and ii) Concentration profiles generated for different substrates. (i) Cross-sections showing a single groove (small rectangular region at the bottom) and the height of the culture medium on top of the substrate for each substrate. (ii) Cropped 100 × 200 μm close-up view showing the cross-section of a single groove for each substrate. The double-ended arrow indicates the width of the channel (25, 50, or 100 μm). The vertical dotted line indicates symmetry. The centrally positioned cell is shown as a black semicircle. The bottom of the image represents the top surface of the substrate for the flat substrate case. (iii) Horizontal concentration profile of VEGF along the bottom of the channel for grooved substrates and at the surface of the flat substrate, shown as relative VEGF concentrations with all values normalized to the value for flat substrate at the point (0, 0). In this steady-state model, VEGF was assumed to be secreted at a zero rate from a cell centrally positioned at the bottom of the groove or on a flat substrate.

We also measured local Young’s modulus on hydrogel-coated grooved and flat substrates by atomic force microscopy nanoindentation (Fig. 2B) and found no significant differences in the stiffness of the groove area and the space between grooves. Tβ4 encapsulation in the hydrogel is not expected to change the mechanical properties appreciably, based on our previous study (22). Only plain collagen-chitosan hydrogel was tested here, because we previously found no significant differences in our rheological measurements with or without Tβ4 encapsulation (22). All hydrogel-coated substrates had similar local mechanical properties, irrespective of the groove dimensions. The stiffness of uncoated flat PDMS substrate was measured as 5.85 ± 0.16 MPa. This value is comparable to the local Young’s modulus (6.6 MPa) previously obtained for the same PDMS material using atomic force microscopy but is much higher than the bulk Young’s modulus of PDMS (2.5 MPa) as measured by tensile testing (34). Previous studies comparing local mechanical properties measured by atomic force microscopy and bulk mechanical properties as determined by a tensile tester found that local measurements usually give higher values (34). Because the local stiffness of the hydrogel was found to be unaffected by the grooves, tube formation by cells was not the result of the differences in local mechanical properties. However, cells degraded the hydrogel to migrate and form tubes, as shown by the steady degradation of the hydrogel over 21 d with the cultivation of artery and vein explants on top of the gel (Fig. S4).

In addition to providing physical guidance for the sprouting capillaries, the grooves also may influence the local concentration of autocrine growth factors released from the endothelial cells actively participating in angiogenesis (Fig. 2C). A simplified mathematical model was derived to show that the presence of grooves and the groove width affected the local concentration of VEGF₁₆₅ assumed to be secreted as a point source from one single cell. The model values are not absolute, because several simplifying assumptions were used. However, the relative local VEGF concentrations, as normalized to the value for flat substrate at point (0, 0), show that the grooves can increase the local concentration of the autocrine growth factors. Thus, one way that the grooves contribute to the differences observed in capillary growth is through the effect of physical barriers, as provided by the PDMS groove walls, on the profiles of the autocrine growth factor concentration. This process likely affects the optimal groove width for the tube formation: Although the narrower grooves concentrate the autocrine growth factors more, grooves that are too narrow impose physical constraints on the migrating cells; thus the optimal groove width would be in the intermediate range (e.g., 50 μm), as observed here (Fig. 1I).

When an artery and a vein were placed at the opposite ends of a PDMS stamp with Tβ4 hydrogel, the outgrowths from each vessel followed the topographical cues, approaching each other over time (Fig. 3A, days 7 and 14). By day 21, the capillary outgrowths from the artery explants connected with those from the vein explants, forming an arteriovenous loop with capillaries aligned in the direction of the microgrooves (Fig. 3A). The average branch length (Fig. 3C) was similar to the capillary length of ∼600 μm in the native rat myocardium (35). The achieved branch density was much lower than the capillary density of >4,000 capillaries/mm² in the native mouse heart (36). Multiple engineered vascular structures could be stacked and cultivated for an additional period to increase capillary density. Additionally, hypoxia could be used to increase sprouting from the formed vascular bed. Another option for increasing capillary density is simply to increase groove density on the PDMS substrates by decreasing the spacing between grooves.

Fig. 3.

Engineering of a connected microvascular bed using topographical cues and hydrogels with encapsulated Tβ4. (A) Fluorescence microscopy images showing the time course of capillary outgrowths extending between a YFP⁺ mouse artery (on the right) and a vein (on the left) during in vitro cultivation on substrates with 50-μm grooves and hydrogel coating containing 1,500 ng encapsulated Tβ4. The connection was achieved at day 21. Arrows indicate locations of artery and vein explants. (Scale bars, 200 μm.) The artery and vein explants were identified by the presence of bright fluorescence in the YFP mouse explants. The artery was placed on the right and the vein was placed on the left of the PDMS substrate. The formed vascular structure was identified by the presence of fluorescent tube-like structures extending between the artery and the vein. The progress of the formation of the vascular structures was monitored by fluorescence microscopy at the same location of the sample at different time points. (B–G) Soluble angiogenic growth factors enhance the outgrowths. (B) Branch density at different time points when the explants were cultivated in culture medium with no growth factors or supplemented with VEGF or HGF. (C) Average branch length at different time points with no growth factors or supplemented with VEGF or HGF (P < 0.0001, two-way ANOVA; P < 0.05 for Tβ4 gel vs. Tβ4 gel + VEGF on day 14; P < 0.001 for Tβ4 gel vs. Tβ4 gel + HGF on day 14). An asterisk indicates a statistically significant difference between groups. (D–G) Fluorescence microscopy images of YFP⁺ mouse artery and vein outgrowths with soluble growth factor supplementation. The connection was achieved at day 14 using VEGF supplementation (E) and HGF supplementation (G). (Scale bars, 100 μm.) These figures are composites of multiple images to illustrate the entire capillary bed structure. n = 6 per group.

We next aimed to determine if we could accelerate the sprouting rate further to achieve connection between the artery and vein in a shorter time frame. For this purpose, we introduced the angiogenic growth factors VEGF and hepatocyte growth factor (HGF) to the culture medium at 100 ng/mL (37) or 20 ng/mL (38), respectively. The doses were chosen based on concentrations previously used for angiogenesis assays (37, 38). VEGF promoted the infiltration of ECs into collagen gels to form capillary-like structures (39) and induced sprouting of rat aortic rings (40). In addition, the increased sprouting in our system was correlated with autocrine VEGF secretion (Fig. 1 B–F). HGF stimulates migration and proliferation of vascular ECs and accelerates their organization into capillary-like tubes in vitro (38). It also induces the formation of blood vessels in vivo (38). Previous studies demonstrated that HGF-induced vascularization in vivo involved the induction of VEGF expression in ECs (41).

HGF or VEGF supplementation during explant cultivation increased the branch length significantly at day 14 (Fig. 3C). As a result, the vessel outgrowth rates were accelerated with HGF or VEGF, and connections between an artery and a vein could be attained after 14 d of culture (Fig. 3 B–G). There was no difference between VEGF and HGF in accelerating anastomosis, branch density, or branch length in this system (Fig. 3 B and C ).

Our strategy was not species dependent. Similar results were shown using rat femoral artery and vein explants and mouse cardiac tissue explants (SI Results and Figs. S5 and S6), suggesting that it is not necessary to isolate specific arteries and veins. However, there are advantages in using vascular explants rather than cardiac or other specific-tissue explants. First, arteries and veins are more readily available for isolation. Second, the anchorage of the resulting capillary structure by parent vascular explants could allow implantation and perfusion of the structure. The maximal branch density obtained using a single mouse or rat vascular explant was similar, ∼10/mm² (Fig. 1 and Fig. S5). Cardiac explants, in contrast, gave a density of ∼30/mm² (Fig. S6), likely because of the presence of a large number of blood vessels in cardiac explants that could all sprout concurrently.

Confocal microscopy indicated that the mouse capillary outgrowths had characteristics of developed vascular structures with open lumens (Fig. 4 A–F and Movie S1). Transmission electron microscopy demonstrated the presence of lumens formed by cells in the peripheral position (Fig. 4H), as is typical of capillaries, which are made of a single layer of ECs. Cells forming the capillary outgrowths were positive for common endothelial markers, including CD31, VE-cadherin, and von Willebrand factor (Fig. 4 G, I, and J). Importantly, the formation of lumens by von Willebrand factor-positive ECs was clearly evident (Fig. 4G) by confocal microscopy z-stacks. Smooth muscle actin- and NG2-positive cells associated with these capillary outgrowths appeared in the peripheral position, typical of pericytes (Fig. S7). Smooth muscle actin-positive cells also were found between the vessels and the basement of the grooves (Fig. S7A).

Fig. 4.

Capillary outgrowths from mouse arteries and veins contained open lumens formed by endothelial cells and remained intact upon removal from the substrate. (A–F) Confocal microscopy showing lumens of YFP⁺ capillary outgrowths (n = 3). (A) Full view z-stacked x–y image. (Scale bar, 50 μm.) (B) Cross-sectional lumens in the y–z plane at the location indicated by the dotted line labeled “yz” in the z-stacked image. (Scale bar, 20 μm.) (C and D) Cross-sectional lumens in the x–z plane at the locations indicated by the dotted lines labeled “xz-1” (C) and “xz-2” (D) in the z-stacked image. (Scale bars, 50 μm.) (E and F) Longitudinal lumens at the locations indicated by orange boxes labeled “xy-a” (E) and “xy-b” (F) in the z-stacked image. (Scale bars, 10 μm.) (G) Confocal microscopy images of von Willebrand factor staining (red) (n = 3). (i) Full view z-stacked x–y image. (Scale bar, 50 μm.) (ii) Cross-sectional lumens in the y–z plane at the location indicated by the dotted line labeled “yz” in the z-stacked image. (Scale bar, 20 μm.) (iii) Cross-sectional lumens in the x–z plane at the location indicated by the dotted line labeled “xz” in the z-stacked image. (Scale bar, 50 μm.) (iv) Longitudinal lumen at the location indicated by the orange box labeled “xy” in the z-stacked image. (Scale bar, 20 μm.) (H) Lumens formed by endothelial cells as shown by transmission electron microscopy (n = 3). (Scale bar, 2 μm.) (I and J) Representative immunostaining for (I) CD31 and (J) VE-cadherin (n = 3). Red indicates positive staining; blue indicates counterstaining of the nuclei with DAPI; arrows show localization of protein; dotted lines outline a capillary outgrowth. (Scale bars, 20 μm.) (K and L) The capillary network maintained its structural integrity after removal from the PDMS substrate. Fluorescence microscopy images showing capillary outgrowths before (K) and after (L) removal (n = 3). (Scale bars,100 μm.)

Because the microvascular bed originates from artery and vein explants, we investigated the phenotype of the newly generated sprouts. Ephrin-B2 and EphB4 selectively mark arterial and venous ECs (42). As expected, outgrowths on the artery and vein sides of the engineered capillary structure were stained positively for Ephrin-B2 and EphB4 respectively at day 21 (Fig. S8). Interestingly, there also were Ephrin-B2⁺ cells within capillaries on the vein side and EphB4⁺ cells on the artery side, suggesting an integration of arterial and venous outgrowths. It has been demonstrated that Ephrin-B2 and EphB4 can interact physically in growing blood vessels (43), as is consistent with the lack of preferential localization of these markers to either the venous or arterial side observed here. Ephrin-B2 is known to be up-regulated during pathological and physiological angiogenesis (42, 44). It also regulates the internalization and signaling activity of VEGFR2 (45), the receptor of VEGF₁₆₅ that was up-regulated using Tβ4 gels here.

The capillary structure was removed easily without damage to the morphology by peeling the hydrogel coating off the PDMS substrate (Fig. 4 K and L). More importantly, the engineered vascular structures were perfusable with fluorescently labeled dextran (Movie S2).

Using the same strategy, we engineered an oriented capillary bed using human umbilical arteries and veins (Fig. 5 A and B). The capillary outgrowths were composed of CD31⁺ ECs (Fig. 5C). Connections between the parent arterial and venous explants were made by day 21 (Fig. 5D) and the outgrowths contained open lumens (Fig. 5E).

Fig. 5.

Oriented and connected capillary outgrowths were formed from human umbilical artery and vein explants. Day 21 microvascular bed on micropatterned PDMS substrates with a coating of collagen-chitosan hydrogel containing 1,500 ng encapsulated Tβ4 (n = 5 per group). (A) Brightfield images of outgrowths on PDMS substrates with groove widths of 25, 50, or 100 μm. (Scale bars, 50 μm.) (B) Viable staining of outgrowths on PDMS substrates with groove widths of 25, 50, or 100 μm. Green represents carboxyfluorescein diacetate (CFDA) staining of live cells. (Scale bars, 50 μm.) (C) Fluorescence microscopy image showing positive CD31 staining of endothelial cells that make up the capillary outgrowths on PDMS substrate with a groove width of 50 μm. Red indicates positive CD31 staining; blue indicates counterstaining of nuclei with DAPI. Arrows indicate CD31-positive endothelial cells that make up the outgrowths. (Scale bar, 100 μm.) (D) Brightfield image showing connection of capillary outgrowths between the artery (on the right) and vein (on the left) explants on PDMS substrate with a 50-μm groove width. Connected outgrowths are shown between pairs of dotted lines. Arrows indicate the parent explants. (Scale bar, 100 μm.) The figure is a composite of multiple images to illustrate the entire capillary bed structure. (E) Confocal microscopy image of CFDA-stained sample showing lumens of the capillary outgrowths: (i) z-stacked x-y image; (ii) cross-sectional lumens in the y–z plane at the location indicated by the dotted line in the z-stacked image. (Scale bars, 20 μm.)

To determine whether cardiac tissues with improved function could be engineered using the established capillary structure, cardiomyocytes were cultivated for 7 d on the PDMS substrate containing mouse vascular explants and the associated capillary outgrowths created after 14 d of culture with HGF supplementation (Fig. 6). Culture medium without HGF was used to cultivate the cardiomyocytes. This type of coculture precisely defines the position of ECs and cardiomyocytes, with beating cardiomyocytes (Movie S3) found in parenchymal spaces around the capillaries, as in the native heart (Fig. 6A). Cardiac tissues grown on vascular structures had a significantly lower excitation threshold than those grown on control and Tβ4 gel-coated substrates without vascular structures, indicating improved functional properties (Fig. 6B). The maximum capture rates for the cardiac tissues were similar with or without vascular structures (Fig. 6C) and were comparable to values from our previous studies (46). Troponin T staining demonstrated that cardiomyocytes grown on vascular structures contained better-organized sarcomeres than those grown on hydrogel coating only (Fig. 6 D and E). Increased Connexin-43 staining in cardiomyocytes cultivated on vascular structures (Fig. 6F) indicated better cell–cell junctions.

Fig. 6.

Engineering of vascularized cardiac tissue. Neonatal rat cardiomyocytes were seeded around day 14 microvascular bed and were cultivated for additional 7 d (n = 6 per group). HGF was added to the culture medium during cultivation of capillary outgrowths. (A) A movie frame of the beating cardiac tissue from the group cultivated on Tβ4 gel with capillaries, showing the position of the cardiomyocytes relative to the capillary outgrowths that run parallel between the parent explants. The asterisk indicates the location of the vascular explant. YFP⁺ capillary outgrowths are shown in green and are indicated by arrows. Cardiomyocytes are shown as brightfield with an added dark background to enhance contrast. (Scale bar, 200 μm.) (B and C) The functionality of engineered cardiac tissues was evaluated by measuring (B) the excitation threshold (P = 0.0036, one-way ANOVA; P < 0.01 post hoc Tukey test for control vs. Tβ4 gel with capillaries; P < 0.05 for Tβ4 gel vs. Tβ4 gel with capillaries) and (C) maximum capture rate (P = 0.2226, one-way ANOVA). An asterisk indicates a statistically significant difference between groups. (D) Troponin T immunostaining shows striations in cardiomyocytes grown on a microvascular bed. The Tβ4 hydrogel alone exhibits poorly developed cells. Troponin T staining is shown in red. Hoechst dye staining of the cell nuclei is shown in blue. (Scale bars, 20 μm.) (E) High-magnification confocal microscopy images showing Troponin T immunostaining (red). (Scale bars, 20 μm) (F) Images of Connexin-43 staining show more cell–cell junctions in the cardiac tissues grown on the microvascular bed. Arrows indicate positive punctate Connexin-43 staining. Connexin-43 staining is shown in red. Hoechst dye staining of the cell nuclei is shown in blue. (Scale bars, 50 μm.) Experimental groups include seeding of cardiomyocytes on PDMS substrates with a coating of Tβ4-free hydrogel (control gel), with a coating of Tβ4-encapsulated hydrogel (Tβ4 gel), and with day 14 capillary outgrowths on a coating of Tβ4-encapsulated hydrogel (Tβ4 gel with capillaries).

Discussion

Here we focused on proving that the engineered vascular network is suitable for cultivation of parenchymal cells, looking at the effects of the network alone. The phenotype of cardiomyocytes in this system could be enhanced further in future studies by applying electrical (46, 47) or mechanical (48) stimulation. EC cords previously led to synchronized contraction of cardiomyocytes and increased Connexin-43 expression because of the improved survival and spreading of cardiomyocytes (49). In future studies, the vascular structures could be connected to a perfusion system to engineer more complex and metabolically active cardiac tissues. This strategy can provide a native-like supply of oxygen and nutrients while preventing the cells from experiencing hydrodynamic shear. Importantly, anchoring of the parent artery and vein explants provides a means for surgically connecting the host vasculature to the engineered vascular structure, thus potentially allowing the engineered tissue to be integrated fully with the native tissue.

A number of notable approaches have been used previously to create branching microchannels via microfabrication in polymers such as polystyrene (50) and silk fibroin (51). Endothelialized microvessels also have been created in hydrogels such as collagen (19, 52, 53) and alginate (54). Embedding EC-laden alginate microfibers into smooth muscle cell-laden agar-based matrix has been used for defined vascular cell coculture (55). These approaches rely on providing defined areas for the attachment of ECs. In polymeric devices, the parenchymal space often is occupied by polymer, making cell seeding in the area around the microvasculature difficult. Although hydrogels are more amenable to coculture, most are often fragile and difficult to handle, especially when fabricated as microfibers or long, hollow cylinders. In these approaches, the engineered vascular bed cannot be removed from the hydrogel or the microfluidic device and manipulated. In contrast, we did not force cells to attach in microfabrication-defined attachment regions. Instead, our approach was inspired by nature: Topographical cues, hydrogel, and angiogenic factors were used to create a supportive niche for the controlled and directed sprouting of arterial and venous explants.

Vascularization in vivo involves EC proliferation and angiogenic sprouting in which tip cells acquire motile behavior under the influence of VEGF growth factors (56). Anastomosis is accomplished via the connection of extended cellular processes followed by lumen propagation through intercellular and intracellular vacuole fusion (57, 58). Here, we used the angiogenic factor Tβ4 to enhance outgrowth cell density and increase autocrine VEGF secretion, and topographical cues were critical for the formation and propagation of extended cellular processes. In this approach, ECs are not forced to attach and proliferate on a prefabricated structure. Instead they migrate and proliferate under the direction of externally added and autocrine cytokines and assemble into luminal structures directed by topographical cues. Thus, the approach described here enables the engineering of a prototype vascular network consisting of two branching vessels suitable for the cultivation of different parenchymal cell types.

In summary, we have shown that Tβ4 in collagen-chitosan hydrogels in conjunction with the application of topographical cues guided endothelial outgrowths from an artery and a vein, enabling their organization into microvasculature. The resulting vascular bed was used to cultivate engineered vascularized cardiac tissues with improved function. The engineered microvascular bed was perfusable and removable from the PDMS substrate. Its functionality and ease of manipulation and the presence of the flanking artery and vein overcome key limitations of currently available approaches for in vitro vascularization. Although we have demonstrated seeding and cultivation of cardiomyocytes as a parenchymal cell type here, the vascular bed is generic and could be used for the vascularization of other tissue types.

Materials and Methods

Preparation of Hydrogel.

The hydrogel solutions were prepared by mixing 10× PBS, Tβ4 (stock solution of 1 mg/mL in distilled water) (HOR-275; ProSpec), chitosan in distilled water (75–90% deacetylation; PROTASAN, catalog no. G113; NovaMatrix), collagen I (4.08 mg/mL rat tail collagen I in acetic acid) (catalog no. 354236; BD Biosciences), and 1 N NaOH. The volume of 10× PBS in the hydrogel solution was 1/10th of the final volume, and the volume of NaOH was 0.025 times the volume of collagen I. The volumes of Tβ4, chitosan, and collagen I were calculated based on the final desired concentrations of 10 μg/mL (Encap100 group) or 150 μg/mL (Encap1500 group), 1.25 mg/mL and 2.5 mg/mL, respectively. Distilled water was added instead of Tβ4 in the control group.

Fabrication of Micropatterned PDMS Substrates.

PDMS substrates were fabricated as previously described (59), using standard soft lithography. The resulting substrates consisted of lanes 25, 50, or 100 μm in width and 65 μm in height. The substrates were cut into 5 × 5 mm squares and were fixed to the bottom of individual wells. Hydrogel solution, 10 μL, was pipetted evenly onto each PDMS substrate to coat the surface, followed by incubation at 37 °C for 1 h. To show the influence of topography on the tube formation and organization, flat PDMS substrates coated with control or Tβ4 hydrogel were added as controls.

Isolation and Cultivation of Explants.

Cardiac tissues, thoracic arteries, and inferior vena cava were isolated from 5-wk-old YFP transgenic mice [129-Tg(CAGEYFP) 7AC5Nagy/J; Jackson Laboratory], and femoral arteries and veins were isolated from Sprague–Dawley rats according to a protocol approved by the University of Toronto Committee on Animal Care. Human umbilical arteries and veins were kind gifts from John Davies at Tissue Regeneration Therapeutics (Toronto, ON, Canada). For studies with vascular explants, one piece of artery and one piece of vein were placed on the two sides of the PDMS substrate, ∼0.5–1 mm apart. For studies with cardiac explants, the top part of the heart containing the dense vasculature was used. As a control to determine the importance of the ECM interactions, the explants were cultivated on PDMS substrates with 50-μm grooves without hydrogel coating. In this case, the YFP mouse vascular explants were pinned down at the two ends of the PDMS substrates, and soluble Tβ4 was added to the culture medium. The amount of Tβ4 added to the medium was based on our previous study (22), in which the average rate of Tβ4 release from Tβ4-encapsulated hydrogel was found to be ∼37.5 ng/mL/d over 28 d. The mouse explants were cultivated in culture medium consisting of DMEM with 1% (vol/vol) penicillin/streptomycin, 1% (vol/vol) Hepes, and 15% (vol/vol) FBS. In some cases, culture medium was supplemented with 20 ng/mL HGF or 100 ng/mL VEGF. The rat explants were cultivated in culture medium consisting of DMEM with 1% (vol/vol) penicillin/streptomycin, 1% (vol/vol) Hepes, and 10% (vol/vol) FBS. The human explants were cultivated in Clonetics Endothelial Cell Growth Medium-2 (EGM2) with BulletKit (catalog CC-3162; Lonza), suitable for the culture of human endothelial cells.

Image Analysis.

At different time points (days 7, 9, 11, 14, 21), the samples were imaged under optical or fluorescence microscope (Olympus IX2-UCB), and the images were analyzed using ImageJ (SI Materials and Methods).

Immunohistochemical Analysis.

The staining was carried out as described previously (22). Briefly, the samples were fixed in 10% formalin and stained with specific antibodies against CD31 (1:50; Abcam), VE-cadherin (1:50; Abcam), von Willebrand factor (1:100; Abcam), SMA (1:100; Abcam), NG2 Chondroitin Sulfate Proteoglycan (1:200; Millipore), EphB4 (1:50; Hycult Biotech), Ephrin-B2 (1:100; GenScript), Troponin T (1:100; Fisher), and connexin-43 (1:100; Abcam) followed by rhodamine- or Texas red-labeled secondary antibodies (1:100; rhodamine goat anti-mouse, Jackson ImmunoResearch; rhodamine goat anti-rat, Jackson ImmunoResearch; or Texas red goat anti-rabbit, Abcam). Nuclei were stained with Hoechst dye or DAPI (1:100).

Dextran Perfusion.

Rhodamine-labeled dextran (Invitrogen) in distilled water, 200 μg/mL, was injected into the artery explant using a 30-G 1/2-in needle. Fluorescence microscopy was used to capture the perfusion of dextran through the capillary outgrowths.

Transmission Electron Microscopy.

After 21-d cultivation, samples were fixed in glutaraldehyde for 1 h and washed in PBS. Osmium tetroxide was added to the samples for 1 h. The samples then were serially dehydrated in 70, 90, and 100% ethanol. After dehydration, the samples were imbedded in resin, sectioned, mounted on glass slides, and stained.

VEGF ELISA.

Culture medium was collected at 12 and 14 d of explant cultivation on hydrogel. VEGF ELISA was performed on the culture medium samples using the Murine VEGF ELISA Development Kit (Peprotech) to determine the VEGF secretion between days 12 and 14. The time points were chosen based on the plateau in cell outgrowth and the initiation of the majority of tube formation.

Seeding Cardiomyocytes onto Capillary Structures.

After mouse arterial and venous explants were cultured for 14 d with HGF supplementation to form connected capillary outgrowths, 100,000 neonatal rat cardiomyocytes were seeded on the capillary structure for each sample (SI Materials and Methods). The cardiomyocytes were cultivated for 7 d in 1 mL culture medium consisting of DMEM with 1% penicillin/streptomycin, 1% Hepes, and 15% FBS. At the end of the 7-d cultivation, the engineered cardiac tissues were removed from the PDMS substrates, and their functional properties were measured (SI Materials and Methods) as previously described (46).

Atomic Force Microscopy.

PDMS substrates without grooves or with 25-, 50-, and 100-μm–wide grooves were fabricated, coated with plain collagen-chitosan hydrogel solution, and incubated at 37 °C for 1 h for gelation. Culture medium was added to the dish to immerse the PDMS substrates fully in liquid. Local mechanical stiffness of the hydrogel coating was measured for a selected 50 × 50 μm area, spanning a part or all of one groove/ridge region, on each sample using an atomic force microscope (Bioscope Catalyst; Bruker) under force volume mode. The pyramidal tip used (MSCT-D; Bruker) had a semivertical side angle of 17.5°, a typical height of 5–8 μm, and a nominal cantilever spring constant of 0.03 N/m. The measured local mechanical stiffness values were averaged for each substrate type, because no pattern was found in stiffness with regards to the location of measurement on the substrates.

Mathematical Modeling of Topographical Effects on Local Concentration Profiles of Autocrine VEGF.

A simplified mathematical model was formulated to examine the potential differences in local autocrine growth factor concentrations caused by the presence of PDMS grooves. A representative 2D steady-state model of VEGF₁₆₅ diffusing out of a single cell was solved using finite element model software, COMSOL Multiphysics 3.5. The solver used was DIRECT(UMFPACK). All initial concentrations were equal to 0. Because the purpose of this model was simply to investigate the effect of geometry on local VEGF concentration, several complexities (e.g., kinetic receptor and ECM binding, basement membrane formation, proteolysis, internalization, multiple VEGF splice isoforms) were ignored. The details for the mathematical modeling are given in SI Materials and Methods.

Statistical Analysis.

Statistical analysis was performed with Graphpad Prism 5.0 software. One-way ANOVA with post hoc Tukey tests and two-way ANOVA with Bonferroni posttests were used to compare the differences among groups, with P < 0.05 considered statistically significant. The data are presented as mean ± SE.