Liquid-embedded (bio)printing of alginate-free, standalone, ultrafine, and ultrathin-walled cannular structures

Significance

In addressing the limitations associated with conventional filament-(bio)printing strategies for producing solid or hollow fibers, we describe the development of a technique that allows facile fabrication of standalone, alginate-free, solid or cannular structures with diameters and wall thicknesses both down to as small as few micrometers. It is anticipated that our method would serve as an enabling platform that may satisfy the needs for the different types of biomedical and other applications involving the use of fibrous structures.

Abstract

While there has been considerable success in the three-dimensional bioprinting of relatively large standalone filamentous tissues, the fabrication of solid fibers with ultrafine diameters or those cannular featuring ultrathin walls remains a particular challenge. Here, an enabling strategy for (bio)printing of solid and hollow fibers whose size ranges could be facilely adjusted across a broad spectrum, is reported, using an aqueous two-phase embedded (bio)printing approach combined with specially designed cross-linking and extrusion methods. The generation of standalone, alginate-free aqueous architectures using this aqueous two-phase strategy allowed freeform patterning of aqueous bioinks, such as those composed of gelatin methacryloyl, within the immiscible aqueous support bath of poly(ethylene oxide). Our (bio)printing strategy revealed the fabrication of standalone solid or cannular structures with diameters as small as approximately 3 or 40 μm, respectively, and wall thicknesses of hollow conduits down to as thin as <5 μm. With cellular functions also demonstrated, we anticipate the methodology to serve as a platform that may satisfy the needs for the different types of potential biomedical and other applications in the future, especially those pertaining to cannular tissues of ultrasmall diameters and ultrathin walls used toward regenerative medicine and tissue model engineering.

Fibrous structures are abundantly present in the human body, where they function broadly in different tissues or organs. These fibrous structures, such as those in neuronal and muscular tissues, vasculature, renal tubules, urethra, and the gastrointestinal tract, among others, whether solid or cannular, are distributed throughout the body with a wide range of diameters and wall thicknesses from centimeters down to the micrometer level.

To this end, the fabrication of solid and perfusable fibrous structures using functional biomaterials and cytocompatible biomanufacturing strategies is of great interest for applications in biomedicine, including but not limited to three-dimensional (3D) cell culture, tissue repair and regeneration, drug screening, and disease modeling (1–5). While solid or hollow fibers are geometrically simple, their fabrication nevertheless remains a consistent challenge in the field.

Previous studies have reported the fabrication of standalone solid or cannular fibrous structures using single or coaxial nozzles via inducing rapid physical cross-linking of sodium alginate-based hydrogel-precursors upon exposure to divalent cations such as Ca²⁺ or the so-called “wet-spinning” approach (6–9). When further integrated with 3D bioprinting as demonstrated in recent years, it allows adequate control over numerous parameters such as the diameters, wall thicknesses, and/or architectural complexities of the extruded filaments (10–12). However, the ability of wet-spinning and coaxial microfluidic bioprinting to adjust fiber diameters and wall thicknesses is oftentimes limited when it comes to the fabrication of tubes with ultrafine diameters/luminal sizes and those featuring ultrathin walls (9, 13, 14). Besides, the requirement of these technologies on the use of alginate as the sole component or one of the components of the bioink in most configurations also narrows their scope in terms of flexibility and material choices (10, 12).

On the other hand, all-aqueous strategies have commonly been used as a facile method to generate different types of soft material-based structures with cyto/biocompatible properties (15–21), which is further possible for use in biological systems in the presence of living cells or bioactive species. In terms of bioprinting, examples of adopting the all-aqueous strategies include the formulation of aqueous two-phase emulsion-enabled micropore-forming bioinks (22, 23). In addition, reports have preliminarily shown that one aqueous phase may be directly extruded into another thermodynamically incompatible aqueous phase to enable the drawing of 3D filamentous patterns (24, 25). However, the full capacities of all-aqueous bioprinting are far from being fully realized.

Herein, we present a robust method that takes on the challenge of fabricating ultrasmall luminal diameters and ultrathin walls of standalone cannular tissues through the merge of all-aqueous and coaxial microfluidic bioprinting technologies and in the absence of alginate. We demonstrate our strategy primarily based on gelatin methacryloyl (GelMA), a photocross-linkable gelatin-derivative, as the aqueous bioink, and poly(ethylene oxide) (PEO), a synthetic polyether, as the aqueous support bath. GelMA has been broadly used to produce cell-laden 3D tissue constructs due to its photoactivated on-demand cross-linking, cyto/biocompatibility, biodegradability, as well as intrinsic bioactivity, and tunable physical characteristics (26). Similarly, PEO (or more often its smaller-molecular weight (M_w) alternative, poly(ethylene glycol)) is prevalently used as hydrogel scaffolds for various biomedical applications, owing to its biocompatibility, inertness, and readily available chemical modification ability (27). GelMA and PEO have also been demonstrated to be immiscible at a wide range of concentrations where they form the two-phase aqueous system (28).

Our method uses PEO solution as the support bath instead of the conventional solid gel-based baths, into which GelMA bioink could be directly extruded in an embedded manner with the patterns stabilized when the parameters of both aqueous phases are meticulously adjusted. As such, not only that alginate would not be a necessary component, but also standalone cannular structures could be readily generated by extruding solid filaments followed by surface-induced chemical cross-linking. Alternatively, adopting coaxial nozzles, hollow filaments could be directly produced as well, in this setup. Due to the unique properties of the all-aqueous system, unlike when using solid gel-based support baths that exert excessive friction forces on the nozzles, our strategy enables bioprinting of ultrafine solid filaments or hollow filaments with ultrathin wall thicknesses. As we will demonstrate, standalone solid or cannular structures with diameters ranging from 3 or 40 μm, respectively, to as large as one would wish, and wall thicknesses of cannular structures down to as thin as <5 μm, are conveniently obtained. Of note, this proposed strategy is compatible with cells, which may be embedded within the bioink during extrusion or post-seeded after bioprinting. The generation of ultrathin endothelialized or epithelialized hollow conduits is finally illustrated. We believe that our unique strategy for bioprinting ultrasmall and ultrathin-walled cannular tissues will find widespread future utilities in areas including but not limited to, tissue model engineering and regenerative engineering.

Results

Printability Behaviors of GelMA Bioinks in PEO Collecting Baths.

The setup for all-aqueous embedded bioprinting consisted of a custom-designed mechanical extrusion bioprinter and a collecting bath (Fig. 1). To evaluate the feasibility of the method, the printability behaviors of potential bioinks were assessed by taking GelMA as an exemplary bioink material. We have demonstrated previously that GelMA and PEO aqueous solutions can be utilized to induce the formation of an aqueous two-phase-separated system allowing facile bioprinting of micropore-forming hydrogels (22, 29, 30). Accordingly, we hypothesized that this same system, with further optimizations, may be able to enable direct all-aqueous embedded bioprinting to take place. Indeed, as shown in SI Appendix, Fig. S1A Harvard badge was readily printed using 8% (w/v unless otherwise noted) GelMA (low degree of methacryloyl-modification) containing 0.1% purple, fluorescent nanoparticles, into the aqueous support bath of 8% PEO (M_w: 300 kDa), to illustrate the concept. Using this system, the printed GelMA microstructures could be preserved within the bath for up to 7 d under the uncross-linked state in the absence of any external perturbation (SI Appendix, Fig. S2). Furthermore, to demonstrate the versatility of this all-aqueous bioprinting strategy, we printed several freeform 3D architectures, such as a planar circle, a vertical line, and a spring pattern (SI Appendix, Fig. S3). We also produced complex multimaterial structures by employing a dual-nozzle printhead (SI Appendix, Fig. S4) (31, 32).

Fig. 1.

Schematic illustration of the (bio)printing process for alginate-free, standalone, ultrafine, and ultrathin-walled cannular tissue-like structures via our two-phase aqueous embedded (bio)printing strategy.

We subsequently optimized the printing parameters. We initially chose PEO with an M_w of 300 kDa since it has been reported that the all-aqueous phase-separation in such a system is affected by the M_w of PEO; low-M_w PEO molecules do not effectively induce phase-separation (22, 33). Considering the factors that might affect the printing process, we assessed the influence of M_w of PEO (4, 100, 200, 300, and 400 kDa, 8%), concentration of PEO (M_w: 300 kDa; 4%, 6%, 8%, and 10%), degree of methacryloyl-substitution of GelMA (30%, 60%, and 90%), and concentration of GelMA (6%, 8%, 10%, and 12%). As revealed in SI Appendix, Fig. S5, with the increase of M_w of PEO, the settling velocity would decrease significantly; with increasing PEO concentration, longer suspension times were also observed (SI Appendix, Fig. S6). As shown in SI Appendix, Fig. S7, there were no significant differences between the different groups when using the three degrees of methacryloyl substitution of GelMA. In addition, with elevation of GelMA concentration, longer suspension times were observed (SI Appendix, Fig. S8).

To further increase the suspension time for achieving better printing stabilities, Ficoll, a biocompatibility polysaccharide widely used in density-gradient centrifugation (34), was introduced as an additive to the support bath. For enhanced visualization, GelMA was conjugated with N-hydroxysuccinimide-fluorescein (green). As revealed in SI Appendix, Fig. S9, we demonstrated that the addition of Ficoll into the PEO aqueous bath did not interfere with the phase-separated system between PEO and GelMA or direct writing of the GelMA bioink within the bath. Of note, due to the increase in the overall density of the PEO/Ficoll mixture aqueous bath when the Ficoll concentration was elevated from 1:12 to 1:6 (v/v), the suspension time of the printed GelMA pattern could be effectively prolonged. However, it was then noted that Ficoll and PEO themselves were also able to slowly phase-separate over time. Therefore, to further prolong the time of phase-separation of the printed constructs in the support bath, the ratio of Ficoll and PEO was increased accordingly. When increased to 1:4.5 (v/v), we found that the constructs could be suspended in the collecting bath for as long as 36 h before the they sedimented and attached to the interface formed by phase-separation between Ficoll and PEO, which at this ratio was only minimum during this time frame (SI Appendix, Fig. S10). Of note, for rest of the experiments, we mostly used pure PEO solution as the aqueous support bath and did not use the PEO/Ficoll mixture since we did not need extra-long suspension time in majority of these demonstrations, unless otherwise specified.

Size-Adjustability of Solid All-Aqueous-Printed Solid GelMA Filaments.

Using our all-aqueous embedded printing strategy, we attempted to emphasize the facile and robust size-adjustability of the extruded GelMA filaments through modulations in various printing parameters such as nozzle diameter, ink flow rate, and printhead moving speed. As shown in Fig. 2A, solid GelMA microfibers with diameters conveniently controllable from 10 to 700 µm could be prepared by changing one or more of these mentioned parameters. To further enlarge the diameter range of the extruded solid fibers in the same printing session, we in addition adopted a multilayered coaxial nozzle (SI Appendix, Fig. S11A), using which the size of the GelMA filaments could be precisely tuned by switching off/on the layer combinations and by adjusting the flow rates in the different layers, an unconventional process. In this particular example, we illustrated that the diameters of the solid fibers were attainable across an intriguingly large scale from 20 (when only the central layer was used to deliver the GelMA ink at the flow rate of 0.5 mL h⁻¹) to 1,000 µm (all three layers were used to deliver the GelMA ink at the flow rate of 40 mL h⁻¹) using exactly the same nozzle (SI Appendix, Fig. S11B). In general, it was feasible to fabricate solid GelMA filaments with diameters as small as 10 µm using our all-aqueous embedded printing strategy adopting nozzles produced from commercial blunt metal needles.

Fig. 2.

Robust size-adjustability of the extruded GelMA filaments through modulations of various printing parameters. (A) Microscopic images of solid filaments printed with 14, 17, 21, 25, and 27G nozzles (printhead moving speed was 5 mm s⁻¹; GelMA extrusion rate was 2 mL h⁻¹), different extrusion rates of GelMA at 0.2, 0.5, 1, 3, and 10 mL h⁻¹ (printhead moving speed was 5 mm s⁻¹; nozzle was 27 G), and various velocities of the printhead movement at 0.1, 1, 5, 10, and 20 mm s⁻¹ (GelMA extrusion rate was 2 mL h⁻¹; nozzle was 27G). The right column shows corresponding quantitative analyses of the diameters. (B) Schematic illustration of the printing process with commercial ultrafine and self-drawn ultrafine-tipped glass capillaries, and the fluorescence images of printed solid filaments with different diameters. Printhead moving speeds were 5, 10, and 20 mm s⁻¹, respectively, and GelMA extrusion rate was 2 mL h⁻¹. (C) The comparison of two types of the printing systems, namely using the PEO/Ficoll (1:1, v/v) aqueous support bath system or gel support bath system (8% GelMA). Printhead moving speed was 5 mm s⁻¹. ***P < 0.001; one-way ANOVA (compared with the corresponding first groups); n = 40.

In addition to using these metal needles as nozzles, glass capillaries were also adopted to seek to print ultrafine filaments, which are potentially of significance for expanded applications in tissue biofabrication. As shown in Fig. 2B, both the commercial ultrafine and the self-drawn ultrafine-tipped capillaries were usable in our all-aqueous extrusion setup, where we were further able to refine the diameter range of our filaments to produce solid fibers from 60 μm down to as small as 3 μm. Oftentimes homogeneous hydrogels or granular hydrogels are utilized as a collecting bath for embedded extrusion printing (35–38). Nonetheless, these mechanically strong collecting baths would not typically let one print ultrafine filaments due to the inability of using the ultrathin glass capillaries that tend to break or deform. As we illustrate in Fig. 2 C, i and SI Appendix, Fig. S12, while the ultrathin glass capillary could be translocated smoothly in the PEO aqueous collecting bath due to its fluidity, the same capillary could not move at all when gel slurry was used as the bath, whose tip stayed in its original position even though the portion above the bath completely translocated already. Moreover, we showed that our custom-drawn, ultrafine-tipped glass capillary with a graded tip could readily translocate in both our PEO aqueous bath and the gel slurry bath from one side to the other (Fig. 2 C, ii and Top). Nevertheless, when such a nozzle moving speed became sufficiently fast (e.g., 1 mm s⁻¹ as shown in Fig. 2 C, ii and Bottom), significant amounts of bubbles along the path of the capillary were introduced into the gel bath system limiting the attainable printing speed, which remains as a general challenge for embedded bioprinting so far. Under the same nozzle moving speed of 1 mm s⁻¹, in contrast, no visible bubbles were present at all in our aqueous bath setup.

An additional advantage of using our all-aqueous strategy coupled with the mechanical extrusion method lies in that it would allow repeated revision and reconfiguration of the printed free-standing patterns (SI Appendix, Fig. S13A) (39). Of note, we used a coaxial nozzle to revise the printed structures. As such, the inner space was solely used to print, and the outer space was dedicated to retracting the printed constructs. This configuration was convenient, where the processes of extruding or retracting inks in inner or outer channels would not notably affect each other. To provide a proof-of-concept demonstration, we extruded a vertical line and then retracted the line operating in the reverse direction, as shown in SI Appendix, Fig. S13B. Similarly, we were able to print a spiral-shaped pattern and then retract it step-by-step (SI Appendix, Fig. S13C). This feature is convenient in scenarios where adding, retracting, and/or modifying printed structures are needed before the inks are solidified to stabilize the final shapes.

Besides, it was possible to dynamically morph the printed homogenous constructs into heterogeneous ones via printing in a gradient-density aqueous support bath (SI Appendix, Fig. S14A). As revealed in SI Appendix, Fig. S14B, the printed spiral pattern, which initially had a uniform thread thickness throughout the height, slowly became unequal in the different layers of the support bath featuring the gradient concentrations of PEO (hence different densities). Visualization of such an inhomogeneous sedimentation process was further aided using horizontally drawn guidelines shown in SI Appendix, Fig. S14C. We defined the vertical distance between the second and third spiral as unit 1 and serial number is defined as 0, and we neglected the distance between the first and second spirals because it was slightly influenced after printhead removal by the printhead. We then followed the changes in the distances between the two adjacent spirals with prolonged time and further measured the relative vertical distance according to unit 1 we defined. As the quantification table on the SI Appendix, Fig. S14 C, Right suggested, all distances became narrower over time, yet the closer to the top, the more sedimentation there was due to the lower densities of the PEO support bath segments that the spirals initially resided in. Therefore, with the use of dynamically configured gradient-concentration aqueous collecting baths, we may further achieve shape-reconfigurable four-dimensional printing ability in the future.

Surface Chemical Cross-Linking-Induced Wall Thickness-Controllable Hollow Filaments.

With our ability to directly print solid aqueous GelMA filaments within the aqueous PEO collecting bath as illustrated above, we anticipated that a surface cross-linking method might enable the formation of hollow fibers that feature tunable wall thicknesses across a wide range. Specifically, microbial transglutaminase (MTG) was chosen for this purpose. MTG, derived from multiple bacterial strains such as Streptomyces mobaraensis, is an enzyme of the class of transferases well-known to chemically cross-link most proteins by catalyzing the formation of an isopeptide bond between an amine group and an acyl group (40–42).

As schematized in Fig. 3A, our surface cross-linking strategy involved a series of optimized steps: i) a solid aqueous GelMA filament is extruded into the aqueous PEO collecting bath also dissolved with 2% MTG at 20 °C; ii) the setup is transferred to 37 °C post-printing to facilitate outside–inward chemical cross-linking of GelMA for a desired period of time; iii) afterward, the container is briefly relocated into −20 °C to rapidly decrease the ambient temperature and again minimize the cross-linking reaction, which also would temporarily solidify (not freeze) the interior, uncross-linked portions of the GelMA filament due to the temperature-sensitive property of GelMA; iv) at last, the GelMA filament containing the chemically cross-linked solid wall as well as the physically solidified interior is stably removed from the PEO bath, following which v) a warm medium is then added to wash away the uncross-linked GelMA, obtaining a hollow fiber.

Fig. 3.

Surface chemical cross-linking-induced hollow filament-formation. (A) Schematic illustration of the printing procedure of cannular fibers via the surface-cross-linking strategy. (B) Fluorescence microscopy images of cannular fibers printed under the same parameters but with the different cross-linking times, as well as the corresponding quantitative analyses of outer diameters and wall thicknesses. GelMA extrusion rate was 2 mL h⁻¹, printhead moving speed was 5 mm s⁻¹, and nozzle was 27 G. ***P < 0.001; one-way ANOVA (compared with the corresponding preceding groups); n = 40.

By controlling the MTG cross-linking time, we were able to precisely adjust the wall thickness of the resulting hollow fiber (Fig. 3B and SI Appendix, Fig. S15). At 1 min of MTG cross-linking, the wall thickness could be attained at as thin as 8.23 ± 1.15 μm; as the chemical cross-linking time was prolonged, the wall thickness gradually increased (58.15 ± 5.46 μm at 2 min and 136.45 ± 8.75 μm at 5 min). By 10 min however, it could be observed that the entire filament would already be entirely cross-linked, becoming a solid hydrogel fiber without any hollow interior. In addition to the chemical cross-linking time, we further evaluated the influences of the temperature of the collecting bath and the concentration of MTG. As revealed in SI Appendix, Fig. S16A, in general, at the same MTG concentration of 2% and chemical cross-linking time of 5 min, the lower temperatures of the PEO bath would favor shape maintenance of the printed filaments. At the same temperature of 20 °C and chemical cross-linking time of 5 min, the higher MTG concentration could better stabilize the printed pattern (SI Appendix, Fig. S16B).

To this end, we have shown that this surface chemical cross-linking strategy allowed convenient tunability of wall thicknesses of hollow fibers simply obtainable from all-aqueous-printed solid GelMA filaments, without needing the slightly more complex coaxial nozzles as we will demonstrate in the next section. While the flexibility may not as high as that when using coaxial nozzles, this surface chemical cross-linking-induced hollow fiber-formation method would still enable the generation of ultrathin-walled standalone cannular filaments with wall thicknesses down to the micrometer-range, which is otherwise hard to obtain using conventional extrusion-based embedded bioprinting strategies.

Coaxially Printed Standalone Cannular Constructs Featuring Broad Size Ranges.

Coaxial wet-spinning has been extensively utilized to produce standalone hollow filaments at high throughput (1, 10). When equipping coaxial nozzles onto a motorized stage, the coaxial microfluidic bioprinting method further enables good control over the 3D morphologies of perfusable filamentous patterns, as reported by both others and ourselves (43–46). Nevertheless, two key limitations have faced coaxial extrusion. First, it usually needs to be conducted in the presence of alginate, a biomaterial that can be physically cross-linked through divalent cations such as Ca²⁺, to ensure rapid shape fidelity during the extrusion process. As such, either alginate has to be used alone as the bioink, or it requires to be a component in the hybrid bioink formulations. Unless modified or subsequently removed, alginate oftentimes does not facilitate cellular spreading and functions (47). Although we recently proved that using a temperature gradient by extruding a gelatin-compounded bioink into an ice bath, it was possible not to include alginate (48), it does not necessarily address the second challenge. In general, the diameters and the wall thicknesses of the coaxially extruded hollow fibers remain at relatively large scales. According to our own experiences and the numerous literatures, the cannular fibers produced with coaxial extrusion cannot readily go below approximately 100 µm in inner diameters with the wall thickness in the range of tens of micrometers the smallest. Adopting a post-printing shrinking method, we suggested the potential of reducing the inner diameters to the sub-50-µm range (49), despite that the walls did not become too much smaller (in the same sub-50-µm range), and further reducing these dimensions was shown to be difficult due to the inability to generate hollow filaments that start with smaller sizes.

To this end, we here for the first time report the all-aqueous coaxial printing strategy, using which we illustrate the feasibility of freely adjusting both the diameter and wall thickness of cannular filaments across extremely wide ranges previously almost absolutely unattainable. Instead of the single-layered nozzle used to extrude solid aqueous GelMA filaments into the PEO bath, we switched to our custom-assembled dual-layered coaxial nozzles, where the GelMA ink was delivered from the sheath while PEO was codelivered from the core layer, into the aqueous PEO support bath. As such, since PEO is immiscible with GelMA at the concentrations we chose, the core PEO would function as additional structural support to maintain the shape fidelity of the aqueous GelMA shell within the PEO bath, finally leading to the formation of hollow GelMA hydrogel fibers after photocross-linking, including those that are ultrafine and ultrathin-walled (Fig. 4A).

Fig. 4.

Coaxial printing of GelMA-based cannular fibers. (A) Schematic illustration of the printing procedure for generating cannular fibers and the experimentally results, using the core-sheath coaxial nozzle used as the printhead. The GelMA ink is delivered through the sheath flow and the PEO is codelivered through the core flow besides the use as the aqueous support bath. (B–D) Photographs showing the adjustment of diameters and wall thicknesses with different parameters including the extrusion rates of GelMA at 0.2, 0.5, 1, 3, and 10 mL h⁻¹, printhead moving speed was 5 mm s⁻¹, PEO extrusion rate was 2 mL h⁻¹, and nozzle was 18/27G; the printhead movement speed at 1, 5, 10, and 20 mm h⁻¹, GelMA extrusion rate was 2 mL h⁻¹, PEO extrusion rate was 2 mL h⁻¹, and nozzle was 18/27G; as well as nozzle combinations (18-21/27G; GelMA extrusion rate was 2 mL h⁻¹, PEO extrusion rate was 2 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹). (E) Corresponding quantitative analyses of outer diameters and wall thicknesses. (F) Coaxial 3D printing of GelMA-based constructs including kidney and spiral structure composed with cannular fibers. *P < 0.05, **P < 0.01, ***P < 0.001; one-way ANOVA (compared with the corresponding preceding groups); n = 40.

As experimentally illustrated in Fig. 4A as well, we could successfully obtain a hollow fiber using our proposed all-aqueous coaxial printing method. Of critical importance, we demonstrated extremely wide tunability of the printed hollow fibers. As revealed in Fig. 4B, not only the diameter of the hollow fiber was able to be conveniently adjusted from 420.71 to 129.50 μm via simply changing the GelMA flow rate from 0.2 mL h⁻¹ to 10 mL h⁻¹ but the thickness of the wall was decreased accordingly from to 134.51 μm to an intriguingly thin value of merely 4.00 μm. Besides, the printhead’s moving velocity and a series of nozzle combinations were examined to understand their effects on the produced hollow filaments. As shown in Fig. 4 C and D, tailor-made hollow fibers were attainable, such as a combination of 69.01 μm in diameter and 15.88 μm in wall thickness, which is not something possible at all with the conventional coaxial wet-spinning or bioprinting setups. In Fig. 4E and SI Appendix, Fig. S17, we measured and listed all the quantified outer diameters, inner diameter, and wall thicknesses of the obtained hollow filaments corresponding to Fig. 4 B–D.

The engineered vasculature should allow efficient oxygen-exchange. To illustrate such a key function, hollow filaments with different wall thicknesses were fabricated. We analyzed the oxygen diffusion across the conduits by using our customized perfusion device, as schematically shown in SI Appendix, Fig. S18A. The two ends of each conduit were fixed onto two blunt needles and externally connected to a syringe pump from one of the ends. We injected suspensions of oxygen-carrying live human red blood cells (40% w/v) at a constant flow of 6 mL h⁻¹ and compared the diffusion profiles of the oxygen across the conduits with different wall thicknesses for up to 10 min. As shown in SI Appendix, Fig. S18B, there were apparent differences in diffusion rates of oxygen for the various groups. Namely, the oxygen diffusion was significantly faster with the wall thickness of the conduit becoming thinner. The results illustrated that the printed conduits featuring ultrathin walls could accelerate the diffusion of oxygen and likely other biomolecules, as well in better-mimicking the substance transport of capillary or small vessels than those with thicker walls fabricated by the conventional coaxial approaches. Moreover, the mechanical properties impact the utilities and biological activities of a hydrogel construct. To evaluate the mechanical properties of the printed hollow GelMA conduits, we fabricated the conduits with different wall thicknesses and carried out tensile testing (SI Appendix, Fig. S19). The different wall thicknesses gave slightly varying tensile moduli, although no significant differences were noted across the three groups, in consistency with our previous observations (50). Generally, during the printing process of tubes with thinner walls, the shear forces were higher than those for tubes with thicker walls (44), which could result in enhanced molecular chain orientation, and hence the enhancement in tensile moduli further balanced out by their thinner geometry.

A step forward, to illustrate the versatile adaptability of our unique printing strategy, we took on the challenge of printing cannular structures with designable size changes in different ways. As schematized in SI Appendix, Fig. S20A, fluorescence images indicated the varying diameters and wall thicknesses of the hollow conduits along the microchannel direction (SI Appendix, Fig. S20 B–D). In the first example, the inner and outer diameters of the conduit were gradually reduced simultaneously (SI Appendix, Fig. S20B). The second demonstration maintained the outer diameter of the conduit unchanged, but the inner diameter was increased by reducing the wall thickness correspondingly (SI Appendix, Fig. S20C). Finally, we illustrated the possibility of keeping the inner diameter of the conduit roughly consistent while decreasing the outer diameter along the length (SI Appendix, Fig. S20D). This set of examples suggested the broad adjustability of our all-aqueous coaxial bioprinting method.

To further showcase the patterning ability using our motorized stage of the bioprinter, volumetrically sophisticated 3D structures such as a grid (SI Appendix, Fig. S21), a kidney-like construct (Fig. 4 F, i), and a cylindrical construct (Fig. 4 F, ii) all composed of cannular fibers were produced. In tissue engineering, hydrogel patterns with embedded vascular structures are essential for tissue repair in vivo or building their models in vitro toward various utilities. In addition to printing volumetrically sophisticated 3D constructs with cannular structures shown in Fig. 4F, we further printed hollow filaments with designed spatial patterns including straight or spiral conduits within the cellularized hydrogel matrices to obtain vascularized tumor tissues, as indicated in SI Appendix, Figs. S22 and S23. Such an ability to control the overall patterns of the conduits suggested potential impactful future applications in cannular tissue engineering.

To investigate whether this all-aqueous coaxial extrusion approach was generally applicable to other biocompatible hydrogel-forming polymers as well, we synthesized hyaluronic acid methacrylate (HAMA) and prepared nonmodified gelatin bioinks for further assessments. As SI Appendix, Fig. S24 illustrates, cannular filaments based on HAMA with a series of diameters and wall thicknesses were readily generated; the wall thickness could be obtained as thin as 15.13 ± 1.92 µm in this particular nozzle combination. Similarly, as revealed in SI Appendix, Fig. S25, cannular filaments based on gelatin with a series of diameters in the range of 109.82 ± 4.86 µm to 306.82 ± 8.68 µm (outer diameters) and wall thicknesses from 19.94 ± 2.20 µm to 107.85 ± 3.75 µm were printed. These results confirmed the versatility of our all-aqueous coaxial extrusion strategy in the direct printing of hollow fibers possessing excitingly wide ranges of both diameters and wall thicknesses using various ink formulations, all alginate-free. Further of note, although the same fabrication conditions were used, the size parameters of these hollow fibers were ink-dependent, which could be primarily ascribed to their differences in viscosities and possibly surface tensions in interacting with the aqueous PEO core phase and support bath.

Cellular Behaviors.

Since the ultimate goals are aimed at engineering functional cannular tissues, we subsequently set to evaluate the cytocompatibility of our all-aqueous embedded printing method. We first bioprinted GelMA-based solid fibers encapsulating NIH/3T3 fibroblasts (SI Appendix, Fig. S26). The cells were cultivated for up to 7 d, where the cell viabilities and metabolic activities were characterized by Live/Dead assay and PrestoBlue test, respectively. It was found that the cell viabilities remained as high as 90% during the entire course of culture, and the cells showed steadily increasing cluster sizes and metabolic activities. Like fibroblasts, C2C12 myoblasts indicated high viability values and metabolic activities across the 7-d culture period, as shown in SI Appendix, Fig. S27.

Cytocompatibility of our all-aqueous coaxially printed GelMA cannular fibers was also assessed by seeding endothelial cells into their lumens. To provide stable medium-perfusion for long-term cultures, we first designed a microfluidic chip device to connect the hollow filaments (SI Appendix, Fig. S28). Each printed vascular conduit was perfused with endothelial growth medium-2 at 200 μL min⁻¹ and kept in the incubator for roughly 6 h before seeding with human umbilical vein endothelial cells (HUVECs). Next, 200 μL of the HUVEC suspension at a cell density of 1.5 × 10⁷ cells mL⁻¹ was injected into each channel through one of the open ends. The device was then placed back into the incubator under static conditions for another 6 h. Perfusion of the fresh medium was then initiated, and nonadherent cells were gently flushed out of each channel lumen. Adherent cells would then start to grow during the following culture period until they reached full confluency and circumscribe the lumens, which could take anywhere between 3 to 7 d depending on the lumen size.

As illustrated by the fluorescence images in Fig. 5A and SI Appendix, Fig. S29, using the method described above, the cells well-attached at day 1 after seeding into the printed thin-walled conduit (outer diameter: 647.71 µm; wall thickness: 47.25 µm; SI Appendix, Fig. S30). From day 4 to day 7, the cells gradually spread to form the spindle shape. Cell viability and metabolic activity assays revealed that at day 1, 96% of the HUVECs attaching on the channel lumens were viable, and after 7 d of culture, the cell viability elevated to as high as 98% (Fig. 5A and Fig. 5 B, i). Besides, HUVECs retained strong metabolic activities with increasing cellular spreading areas over the 7-d culture period (Fig. 5 B, ii and iii). To further verify the spreading and morphologies of the HUVECs attached to the inner surfaces of the GelMA vascular conduits, F-actin staining was performed. As revealed in Fig. 5C and SI Appendix, Fig. S31, the cells were well-distributed and tightly attached already at day 1 of culture; cellular growth and reorganization were apparent, eventually forming a very tight monolayer covering the entire lumen surface after 7 d. Confocal images and reconstructions further validated the 3D nature of these endothelialized vascular conduits (Fig. 5D and SI Appendix, Fig. S32). Of note, we additionally demonstrated that the endothelial cells exhibited noticeable intercellular junctions over the culture period by immunostaining for ZO-1. While no ZO-1 expression was observed at day 1 of culture, increasing amount of its staining became apparent as the HUVECs gradually increased confluency from day 4 to day 7 (Fig. 5E and SI Appendix, Fig. S33). These results indicated that our all-aqueous coaxial bioprinting strategy was cytocompatible and would likely enable vascular engineering with excellent endothelial cell growth.

Fig. 5.

Cellular behaviors on coaxially printed cannular tissues. (A) Bright-field and fluorescence microscopy images of vascular conduits seeded with HUVECs in the interior lumen surfaces after 1, 4, and 7 d of culture. (B) Quantification results of cell viabilities, metabolic activities, and cellular spreading sizes. (C) Fluorescence images of the HUVECs grown on the inner surface of the vascular conduit after 7 d of culture stained for F-actin (green) and nuclei (blue) with Alexa Fluor 488-phalloidin and DAPI. The images in the Upper panel and the rightmost corner of the Lower panel were focused on the top lining of the cells, where the left two images in the lower panel were the cross-sectional views of the vascular conduits. (D) Confocal fluorescence images of the HUVEC-seeded vascular conduit on day 7 of culture, stained for F-actin (green) and nuclei (blue). (E) Fluorescence micrographs of immune stained vascular conduits exhibiting expression of ZO-1 (red) counterstained for nuclei (blue). ***P < 0.001; one-way ANOVA (compared with the corresponding preceding groups); n = 3.

The perfusability and permeability are critical functional characteristics of the vascular conduits. These properties were also assessed using our customized perfusion device (SI Appendix, Fig. S28). We injected solutions of fluorescein isothiocyanate (FITC)-coupled dextran (M_w: 5 kDa) at a constant flow of 80 μL min⁻¹ and compared the diffusion of the molecule across the tubular wall in the control (cell-free) and the endothelialized groups. After injection, fluorescence images were captured every 4 min until 30 min using a wide-field fluorescence microscope. Interestingly, SI Appendix, Fig. S34A clearly revealed the delayed diffusion of FITC-dextran into the surrounding chamber in the endothelialized conduit than the control sample that did not have any endothelial cells. The cross-sectional intensity profiles quantitively illustrated such differences (SI Appendix, Fig. S34B).

We subsequently expanded our demonstration in a sprouting assay in further illustrating the unique biological relevancy of our GelMA-only conduits. As discussed, sodium alginate has remained as a widely used material for the coaxial fabrication of standalone solid or cannular fibrous structures due to its fast physical cross-link mechanism. Nevertheless, alginate cannot easily interact with the cells. Therefore, alginate would mostly function like a solid wall that constrains cell movement and remodeling, as well-reported by others and ourselves (12). To investigate whether our GelMA-only vascular conduits were generally applicable to endothelial cell sprouting and multiscale vascularization, the vessels endothelialized with HUVECs were embedded additionally into the bulk GelMA matrix containing vascular endothelial growth factor (50 ng mL⁻¹) and phorbol myristate acetate (50 ng mL⁻¹). As illustrated in SI Appendix, Fig. S35 A and B, from day 1 to day 14, the cells gradually sprouted from the lumen through the GelMA wall to the outside matrix. The cell sprouting was significant on day 21, demonstrating that the GelMA-only conduits could be readily remodeled by the cells, unlike alginate-based tubular constructs that do not necessarily support cellular remodeling.

Beyond vascular cells, epithelial cells could be populated into the all-aqueous coaxially printed cannular constructs as well. Human renal proximal tubular epithelial cells were taken as an example and seeded in the lumens using the same method as with HUVECs. Fluorescence images indicated that the cells spread well and exhibited a confluent monolayer on the lumen surface of the conduit after 7 d of culture (SI Appendix, Fig. S36). SI Appendix, Fig. S37 shows the corresponding quantified outer diameter, inner diameter, and wall thickness of the obtained hollow filaments.

Engineering Standalone Vascular Conduits across A Wide Size Range of All-Aqueous Coaxially Printed Cannular Filaments.

The cannular tissues in the human body feature wide size ranges including their diameters and their wall thicknesses. For example, the blood vessels are highly hierarchical consisting of larger vessels to smaller vessels, as well as microvessels such as capillaries in the range of several centimeters down to as small as sub-10 µm (51). These vessels play key roles operating in union to efficiently maintain the homeostasis throughout our biological system. Hence, engineering vessels across a wide size range is an essential step toward the successful reproduction of vascular structures and functions for translational applications. Given the fact that conventional coaxial microfluidic bioprinting typically would not allow the formation of vascular conduits featuring ultrafine diameters and/or ultrathin walls, we finally sought to preliminarily demonstrate the flexibility of our unique strategy to achieve such a broad size range of the engineered endothelialized conduits.

In a series of illustrations, we successfully fabricated vascular channels populated with confluent monolayers of HUVECs on their inner luminal surfaces with diameters from 2,000 to 40 μm and corresponding wall thicknesses from 500 to 5 μm (Fig. 6). As shown in SI Appendix, Fig. S38, we measured and listed all the quantified outer diameters, inner diameters, and wall thicknesses of the illustrated hollow filaments corresponding to Fig. 6. Particularly, one instance used a cannular filament as small as 40 µm with only 5 μm of wall thickness, which allowed endothelialization with only a few HUVECs laterally (Fig. 6G). This configuration again was not quite a possibility when conventional coaxial microfluidic bioprinting is utilized. This size range could potentially satisfy the need for different types of applications due to the sizes being close to or covering the vessel size range in the human body, especially for those that feature very small diameters or/and wall thicknesses. It is worthy of noting that however, while further smaller sizes of the channels were printable, seeding any cells into the lumens would become impractical and hence were not demonstrated for cell culture here.

Fig. 6.

Wide size tunability of the coaxially printed cannular tissues. (A–G) Schematic illustration and fluorescence image of a series of printed vascular channels with broadly adjustable diameters and wall thicknesses. Fluorescence images of the HUVECs cultured on the inner surfaces of the vascular channels after 7 d of culture stained for F-actin (green) and nuclei (blue). The printing conditions were as follows: (A) GelMA extrusion rate was 20 mL h⁻¹, PEO extrusion rate was 0.5 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹; (B) GelMA extrusion rate was 20 mL h⁻¹, PEO extrusion rate was 4 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹; (C) GelMA extrusion rate was 2 mL h⁻¹, PEO extrusion rate was 4 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹; (D) GelMA extrusion rate was 0.2 mL h⁻¹, PEO extrusion rate was 4 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹; (E) GelMA extrusion rate was 2 mL h⁻¹, PEO extrusion rate was 2 mL h⁻¹, and printhead moving speed was 5 mm s⁻¹; (F) GelMA extrusion rate was 2.0 mL h⁻¹, PEO extrusion rate was 2 mL h⁻¹, and printhead moving speed was 10 mm s⁻¹; (G) GelMA extrusion rate was 1.0 mL h⁻¹, PEO extrusion rate was 1.0 mL h⁻¹, and printhead moving speed was 20 mm s⁻¹. The same coaxial nozzle (18/27G) was used in all these printing processes.

Discussion

The human cardiovascular system consists of a sophisticated hierarchical network of blood and lymphatic vessels that conduct fluids to and from tissues and organs. The large blood vessels branch into progressively smaller vessels, even capillaries, to control local blood pressures and volumetric flows to the tissues and cells within each organ system (52). While there has been considerable success in the biofabrication of large- and small-diameter vessels, the generation of functional ultrathin microvasculature at the microscale remains a particular challenge given the limitations of the current technologies (53–55). The same applies to the engineering of many other cannular tissues that are widely present in the human system beyond the vasculature, such as the tubules within the kidney (56, 57). Advances in 3D bioprinting have led to the bioengineering of perfusable cannular structures via the use of techniques such as coaxial bioprinting. However, the ability of creating ultrathin hollow conduits for emulating these tissues remains insufficient.

In this work, we presented a facile method that took on the challenge of fabricating ultrasmall luminal diameters and ultrathin walls of standalone cannular tissues through the merge of all-aqueous and coaxial microfluidic bioprinting strategies, and in the absence of alginate, which is a necessary component of the bioink used in conventional coaxial bioprinting. As we demonstrated, standalone solid or cannular structures with diameters ranging from approximately 3 or 40 μm to as large as one would wish, and wall thicknesses of hollow conduits down to as thin as <5 μm were readily obtained, which may satisfy the needs for the different types of potential applications, especially those relating to cannular tissues of ultrasmall diameters and thin walls. In addition, a surface-cross-linking method was adopted to also generate hollow filaments with ultrathin walls as small as 8 μm, by acting on printed solid filaments using the single-layered nozzle. Further expansion of our technology set included dynamic morphing of the printed 3D patterns using a gradient-density aqueous support bath, the ability to vary the diameters of the coaxially printed hollow filaments, and printing with non-GelMA inks. Cellular experiments illustrated that the bioink was cytocompatible, allowing either direct encapsulation of the cells during bioprinting or post-seeding the printed conduits with cells in their lumens. Preliminary functional assays were conducted, which suggested that the endothelialized conduits served as barriers to efficiently decay the diffusion of biomolecules across the walls, as well as the amenability of the GelMA-alone conduits to endothelial sprouting otherwise not possible with alginate-based conduits.

Of note, our technology does not come with no limitation, mostly caused by the ultrasmall size parameters of the generated conduits. For example, the consistency of quality of the products may be easily affected by the bioprinting conditions, where even tiny deviations of parameters or slight differences in the setup could lead to variations in the sizes (diameters or/and wall thicknesses) of the filaments, thus necessitating tightly controlled fabrication procedures. Moreover, while it was possible to produce conduits as small as few 10 of micrometers in diameter and few micrometers in wall thickness, as these sizes become smaller, handling of the conduits would also become increasingly more difficult. Such fragility of conduits with small sizes, along with the limited lumen area, would naturally lead to hardness of seeding cells into them. Nevertheless, our study presented an enabling platform technology for production of tubular tissues, in particular those that feature ultrasmall diameters and ultrathin walls. With continued improvements, it may find versatile applications in cannular tissue engineering, disease modeling, and pharmaceutical screening.

Materials and Methods

Materials and methods are expanded in the SI Appendix, which include Synthesis of GelMA, Synthesis of FITC-Conjugated GelMA, Preparation of GelMA Bioinks, Synthesis of HAMA, Preparation of HAMA Bioinks, Preparation of Gelatin Bioinks, Extrusion (Bio)Printing of GelMA Constructs Using the Two-Phase Aqueous System, Extrusion (Bio)Printing of GelMA Hollow Fibers by the Surface-Cross-linking Strategy, Extrusion (Bio)Printing of Hollow Fibers by the Coaxial Strategy, Oxygen Diffusion Assay, Mechanical Property Evaluations, Cell Culture, Cell Viability Assay, Cell Seeding in Printed Hollow Conduits, Immunostaining, Fabrication of Vascularized Tissues, Endothelial Cell Sprouting Assay, and Statistical Analyses.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All data supporting the findings of this study are available in the article and the SI Appendix.