Rapid magnetically directed assembly of pre-patterned capillary-scale microvessels

Abstract

Capillary scale vascularization is critical to the survival of engineered 3D tissues and remains an outstanding challenge for the field of tissue engineering. Current methods to generate micro-scale vasculature such as 3D printing, two photon hydrogel ablation, angiogenesis, and vasculogenic assembly face challenges in rapidly creating organized, highly vascularized tissues at capillary length-scales. Within metabolically demanding tissues, native capillary beds are highly organized and densely packed to achieve adequate delivery of nutrients and oxygen and efficient waste removal. Here, we adopt two existing techniques to fabricate lattices composed of sacrificial microfibers that can be efficiently and uniformly seeded with endothelial cells (ECs) by magnetizing both lattices and ECs. Ferromagnetic microparticles (FMPs) were incorporated into microfibers produced by solution electrowriting (SEW) and fiber electropulling (FEP). By loading ECs with superparamagnetic iron oxide nanoparticles (SPIONs), the cells could be seeded onto magnetized microfiber lattices. Following encapsulation in a hydrogel, the capillary templating lattice was selectively degraded by a bacterial lipase that does not impact mammalian cell viability or function. This work introduces a novel approach to rapidly producing organized capillary networks within metabolically demanding engineered tissue constructs which should have broad utility for the fields of tissue engineering and regenerative medicine.

Graphical Abstract

The rapid and scalable production of organized capillary beds that are critical to tissue survival and function remains an outstanding challenge for the field of tissue engineering and regenerative medicine. Here, we introduce a novel approach to rapidly generating organized capillary-scale microvessels by combining sacrificial microfiber scaffolds and magnetically assisted cell seeding.

1. Introduction

The presence and organization of microvasculature are critical to the proper function of most tissues and organs. Capillaries, the smallest length-scale of the microvasculature, most directly enable transport to and from tissue parenchyma due to their high surface area to volume ratio and sheer abundance. The organization and density of capillary beds vary widely across tissue systems, ranging from the random, sparse networks within adipose tissue to the dense and highly aligned arrays in the myocardium.^[1] In particular, the high transport demands of metabolically active tissues such as myocardium necessitates a high density of capillaries where each aligned cardiomyocyte bundle is juxtaposed with one or more co-aligned capillaries.^[2] The creation of 3D cardiac tissues at relevant scales for tissue replacement therapy has been in part hampered by the inability to integrate dense, organized capillary beds into these metabolically demanding tissues.^[3–10] While there have been many recent advances in tissue engineering larger-scale vessels across the entire vascular tree, the rapid fabrication and integration of organized microvessel networks at capillary scale (10–20 µm diameter) with parenchymal tissues has proved challenging.

There have been many recent advances across a diversity of approaches to engineering microvasculature. Bottom-up approaches include angiogenesis and vasculogenic assembly, while top-down approaches include 3D printing with sacrificial inks and two photon mediated photo-ablation of hydrogels to generate patent channels. Bottom-up approaches rely on cell-directed assembly which generally can achieve capillary-scale vessel diameters but are challenging to scale up to therapeutically relevant length-scales. Furthermore, approaches to drive angiogenesis into parenchymal cell-laden hydrogels can yield capillary-scale microvessels, however vessels take considerable time to form and are typically disorganized.^[8,11] Similarly, directing vasculogenic assembly can achieve appropriately sized perfusable 3D capillary networks, but result in disorganized capillary beds that take significant time to assemble and require support from other cells such as fibroblasts or mesenchymal stem cells for assembly.^[6,12–15] Alternatively, top-down approaches facilitate highly controlled vascular architectures and generally result in faster assembly of vessels but are unable to generate capillary-scale structures. Specifically, 3D bioprinting approaches (eg. using sacrificial inks to print patent fluidic channels) are rapid, highly tunable, and can generate organized microchannel architectures, but lack the resolution needed for generating capillary-scale networks.^[7,16–20] Other techniques such as photodegradation of hydrogels can generate capillary-scale channels, but photo-degradation may harm encapsulated cells and more importantly, capillary-scale channels (10–20 µm diameter) cannot be subsequently seeded with endothelial cells (ECs) via flow-mediated seeding as the diameter of an EC in suspension (>20 µm) exceeds the diameter of the channel.^[21] Notably, Arawaka, et al. were able to pattern channels in a photodegradable polyethylene glycol-based hydrogel spanning 10–200 μm in diameter but demonstrated EC seeding could only be achieved at channels sizes exceeding 45 µm diameter.^[21] Overall, the limited resolution of top-down approaches and the long temporal requirements and limited organization of cell-driven assembly processes motivates the need for new technologies that can template organized capillary-scale vasculature in 3D.

In this work, we developed a novel method to rapidly create organized capillary-scale microvasculature that can be readily integrated with a variety of hydrogels containing parenchymal cells of choice. At the crux of our approach is a sacrificial capillary-templating polycaprolactone (PCL) microfiber lattice that can be fabricated using commonly utilized methods for generating polymeric microfibers, including solution electrowriting (SEW) and fiber electropulling (FEP).^[22–29] While traditional fused deposition modeling (FDM) 3D printing has a lower filament resolution limit of 100 μm, SEW and FEP can easily achieve filament sizes at or below the diameter of capillaries and are easily scalable to 3D arrays of filaments with highly controlled architecture. EC seeding onto highly porous microfiber lattices prior to encapsulation within a hydrogel-cell precursor solution can yield patent, EC-lined capillary-scale microvessels upon selective removal of the polymer microfiber. However, efficiently seeding highly porous polymer lattices is challenging due to the tendency for cells in suspension to rapidly settle before attaching to a 3D porous lattice. Recent work demonstrates magnetized cells and biomaterials can generate complex cell patterning into tubular structures around a cylindrical magnet,^[30] attach single cells onto magnetic posts,^[31] levitate cells at air-liquid interfaces,^[32] improve seeding of magnetic porous lattices,^[33] and magnetically align fibers within hydrogels to guide cell spreading and migration.^[34] Based on these recent advances, we hypothesized that magnetizing ECs and polymer microfibers would enable efficient EC seeding onto microfiber lattices which can be subsequently degraded via bacterial lipases that do not impact mammalian cell viability or function.^[30,35–37] Here, we introduce a novel method to rapidly generate scalable capillary networks with user-defined architecture integrated with 3D cell-laden hydrogel constructs with broad implications for engineering microvascularized tissues.

2. Results and Discussion

2.1. Fabrication of magnetized PCL lattices by FEP and SEW

To generate polymeric lattices for templating 3D capillary architectures, we adapted two previously established methods, SEW and FEP, for fabricating microfiber scaffolds of relevant length-scales to capillaries. FEP involves the translation of a collection surface and resulting mechanical forces to pull polymeric microfibers from solution onto a rotating mandrel system; electric fields are solely utilized to re-establish contact between the depositing microfiber and collection substrate should microfiber deposition be transiently disrupted.^[24–29] In contrast, SEW combines 3D printing with electrospinning at short spinneret to collector distances, such that electrostatic forces drive the formation and controlled deposition of polymeric microfibers onto an oppositely charged collection surface.^[22,23]

We sought to control the diameters of microfibers produced using either method by tuning the concentration and resulting viscosity of PCL solutions. Over the range of concentrations tested with FEP, 10 and 15 w/v% resulted in inconsistent microfiber formation and highly variable fiber diameter, while higher concentrations of 20 and 25 w/v% PCL yielded uniform fiber diameters and more consistent deposition (Figure 2A). Generally, fiber diameter increased with increasing PCL concentrations when solutions were processed by FEP (Figure 2A), but interestingly, the opposite relationship was noted when identical solutions were processed into microfibers using SEW (Figure 2A). Since FEP is mechanically driven, increasingly viscous solutions should correspond to larger fiber diameters due to the constant force applied to the solution at a given collection speed.^[38] In contrast, during SEW microfiber deposition, the force applied to the polymer solution and ensuing thinning of the drawn material is a function of charge density which should increase with polymer concentration.^[39]

Figure 2. Production and characterization of microfiber capillary-templating lattices produced by FEP or SEW.

(A) Transmitted light images and fiber diameter quantification of FEP (left) and SEW (right) fibers as a function of polymer solution concentration. (B) Transmitted light images of FEP (left) and SEW (right) SPION- and FMP-loaded fibers. (C) Fiber diameter quantification of FEP and SEW SPION- and FMP-loaded fibers formed from 25 w/v% PCL with histograms displaying the fiber diameter frequency of fibers loaded with 40 mg mL⁻¹ FMPs. (D) SEM images of FEP and SEW single and multi-layer PCL (25 w/v%) microfiber lattices without FMPs. (E) SEM images of FEP and SEW PCL microfibers (25 w/v%) with 40 mg mL⁻¹ FMPs. n ≥ 14 lattices. All data are mean ± SD. *P < 0.05.

To magnetize lattices, magnetic particles were doped into the PCL solution prior to fiber fabrication. Superparamagnetic iron oxide nanoparticles (SPIONs, ~8 nm in diameter) and ferromagnetic microparticles (FMPs, ~5 ~m in diameter) were readily incorporated into microfibers with limited impact on microfiber deposition and resulting morphology (Figure 2B). Above 40 mg mL⁻¹, magnetic particle concentration proved too high to consistently form fibers due to frequent interruptions in microfiber deposition, similar to previous reports.^[40] When doping PCL solutions with FMPs, we found that resulting fiber diameters decreased when processed into microfibers by FEP but not by SEW. FMP-doped fibers produced from a 25 w/v% PCL solution were smaller when generated by FEP (11.87 ~m +/− 7.67 ~m) as compared to SEW (19.07 ~m +/− 4.71 ~m) (Fig. 2C). This decrease in fiber diameter specifically with FEP is likely due to the interruption of polymer extrusion caused by the presence of the particles. Distinct from SPIONs, the innate magnetic poles within each FMP can be reoriented by briefly placing the particles in a high strength magnet. SPIONs, in contrast, are only transiently magnetized when in the presence of a magnetic field. Due to the potential to permanently reorient the magnetic poles of FMPs, the highest concentration of FMPs (40 mg mL⁻¹) was used in all subsequent studies. Both printing methods could fabricate lattices composed of multi-layered microfibers to generate user-defined capillary architectures (Figure 2D). Due to the 3D printing nature of SEW (as compared to FEP which requires manual reorientation of the collection substrate), this method could be utilized in the future to create scalable, customized multi-layer lattices of desired geometries for specific tissues of interest. Further, SEM images show that FMP–doped microfibers do not have significantly different morphology from non-loaded microfibers (Figure 2E).

Using either method to fabricating capillary-templating lattices, the spacing between microfibers can be controlled. Utilizing FEP, fiber diameter can be controlled via the linear rotation speed of the collecting mandrel. For 25 w/v% PCL, the slowest rotation speed enabling constant fiber deposition (8 cm s⁻¹) yielded deposited fibers at ~25 ~m diameter. Increasing mandrel rotation speed decreased average fiber diameter to ~15 ~m; however, rotation speeds higher than 66 cm s⁻¹ prevented consistent fiber deposition due to fiber breaking from mismatched polymer extrusion and deposition rates (Supp. Figure 1A–B). For SEW, the travel path of the 3D printer nozzle could be encoded to space fibers at defined distances apart (Supp. Figure 1C). A commensurate increase in fiber spacing with increasing coded printing space was observed, however the actual spacing between fibers was consistently ~75 ~m greater than the intended distance, likely due to variations in the electrostatic field (Supp. Figure 1D). These inconsistencies in electric field likely arise from the deposition of nearby deposited fibers which contribute static charge buildup and repel subsequently deposited fibers.

To selectively degrade PCL microfibers after endothelialization and encapsulation in a hydrogel of choice along with desired parenchymal cells (Figure 1A), we used a bacterial lipase that efficiently and selectively cleaves ester bonds within the PCL backbone. Lipase from Pseudomonas sp. has previously been shown to rapidly degrade electrospun PCL nanofibers,^[41] but the ~10 to 100-fold larger diameter of microfibers used in our studies warranted an investigation into degradation rate as a function of lipase concentration. Fluorophore-doped PCL microfibers were treated with varying concentrations of Pseudomonas sp. lipase and imaged over 24 hours (Figure 3A, Supp. Figure 2).^[41] Exposure to bacterial lipase degraded PCL microfibers as a function of lipase concentration, where the highest concentration (1 U mL⁻¹) completely degraded microfibers within 12 hours (Figure 3C; Supp Video 1–2). Importantly, exposure of ECs to even the highest concentration of lipase tested did not negatively impact EC viability (Figure 3B, 3D). Based on these studies, 1 U mL⁻¹ lipase for 24 hours was determined to be sufficient for PCL microfiber degradation and was used in all subsequent cell studies.

Figure 1. Versatile approach to generating engineered capillary beds via magnetic lattice fabrication and EC seeding.

(A) Schematic overview of fabrication scheme for generating capillary-scale endothelialized microchannels within engineering parenchymal tissues composed of one or more cell types and a hydrogel of choice. (B) Schematics and images of the SEW and FEP methods utilized in this work for fabricating capillary templating PCL microfiber lattices.

Figure 3. Bacterial lipase-mediated degradation of PCL fibers does not negatively impact endothelial cell viability.

(A) Fluorescent time course images and (C) normalized fluorescent intensity of PCL microfibers upon lipase-mediated PCL degradation within collagen gels. (B) Fluorescent time course images and (D) quantification of EC viability in the presence of lipase as determined by propidium iodide and Hoechst staining. n = 6 fields of view. All data are mean ± SD. *P < 0.05.

2.2. Endocytic loading of endothelial cells with magnetic particles

The highly porous nature of lattices composed of microfibers presents a challenge for EC seeding, as ECs in suspension rapidly settle prior to lattice attachment resulting in non-uniform seeding. We sought to enhance EC attachment to magnetized microfiber lattices by loading ECs with magnetic particles. ECs were fed SPIONs or FMPs at various concentrations diluted in cell culture media and cultured overnight to allow for particle uptake via endocytosis. To assess the degree of cell magnetization, EC suspensions were plated onto tissue culture plastic (TCP) well-plates positioned over a 3D printed plate holder containing N52 neodymium magnets (Figure 4A). The percentage of cells adhering above magnets was interpreted as a measure of cell magnetization efficiency, where equal distribution of cells on magnet and non-magnet positions results in a magnetization efficiency of 50% (Figure 4B–C). All particle concentrations tested increased the percent of cells seeded over magnets (Figure 4D). To examine whether endocytosis of magnetic particles or magnetized seeding induced cell death, cell viability was analyzed by propidium iodide staining (Figure 4C). Notably, only 1000 ~g mL⁻¹ SPIONs resulted in a significantly lower cell viability than control cells that were not fed magnetic particles (Figure 4E). Despite magnetic particle loading with either SPIONs or FMPs, ECs retained their ability to localize VE-Cadherin to adherens junctions, the cell-cell adhesion critical to endothelial barrier function (Figure 4F).

Figure 4. EC loading with either SPIONs or FMPs enables cell magnetization with limited impact on cell viability and barrier function.

(A) Schematic of 3D printed magnetization assay plate containing four neodymium cube magnets fixed at set locations. (B) Image of hematoxylin-stained 6-well plate plated with variably magnetized ECs positioned over the 3D printed magnetization assay plate. (C) Fluorescent images of propidium iodide and Hoechst-stained pre-magnetized ECs seeded above and away from neodymium magnets and respective quantifications of (D) cell magnetization efficiency and (E) cell viability. (F) Fluorescent images of EC monolayers loaded with magnetic particles. n = 4 fields of view. All data are mean ± SD. *P < 0.05.

We performed additional studies to test whether loading ECs with magnetic particles alters long-term EC function. ECs were passaged up to three times after magnetic particle loading, and at the first and third passage were assessed for proliferative capacity and ability to robustly form adherens junctions (Figure 4A). Maximal concentration of 400 ~g mL⁻¹ SPIONs were tested in these studies due to noted decreases in cell viability at 1000 ~g mL⁻¹ (Figure 4E). For both particle types across all tested concentrations, proliferation was unaffected as quantified by EdU assay (Figure 4B). Additionally, ECs were able to form comparable adherens junctions across all conditions, suggesting retained capacity to form robust cell-cell adhesions required for endothelial barrier function (Figure 4B). Endocytosis of SPIONs appeared more efficient than FMPs, as shown in representative SEM images (Figure 4C). Notably, both SPIONs and FMPs accumulated around the nucleus. Due to the difference in particle sizes between SPIONs (~8 nm in diameter) and FMPs (~5 ~m in diameter), we posit that the smaller SPIONs, which should not form aggregates prior to endocytosis, are more easily endocytosed than FMPs. Based on these results, 400 ~g mL⁻¹ SPIONs was utilized in subsequent studies given their more homogeneous distribution and undetectable impact on EC viability compared to FMPs.

Figure 5. Cell magnetization with an intermediate particle loading concentration does not affect long-term cell proliferative ability or capability of forming VE-cadherin enriched adherens junctions.

(A) Fluorescent images of magnetized ECs passaged once or thrice, assayed for proliferation by EdU incorporation and VE-cadherin immunolocalization. (B) Quantifications of EdU+ cells and VE-cadherin expression. (C) SEM images of SPIONS, FMPs, naive ECs, and of ECs magnetized by endocytic loading with either SPIONs or FMPs. n = 6 fields of view. All data are mean ± SD. *P < 0.05.

2.3. Magnetic lattice seeding of magnetized ECs

FMPs innately contain small internal magnetic domains with randomly aligned magnetic dipoles. This random alignment of the magnetic domains within a particle cancels out the resulting net magnetic field, effectively resulting in an “unmagnetized” state. However, the magnetic domains in FMPs can be rewritten by applying a strong external magnetic field to saturation, thereby aligning each domain’s magnetic pole to strengthen the net particle magnetic field. When the strong external field is removed, the now polarized particles preserve the co-aligned pole and magnetic field (high remanence) (Figure 5C). In contrast, superparamagnetic particles such as SPIONs only temporarily magnetize under a stronger magnetic field and will return to their non-magnetic state upon removal of the external magnetic field (low remanence). For this reason, a high concentration 40 mg mL⁻¹ FMPs solution was used for lattice magnetization with the thought that magnetically polarized FMPs (either prior to or after lattice fabrication) would enhance microfiber magnetization and attract SPION-carrying ECs during seeding.

Figure 6. Lattices fabricated with pre-polarized FMPs result in the highest degree of EC spreading along PCL microfibers.

(A) Transmitted light timelapse images of ECs magnetically attracting to a FMP-doped PCL fiber (random magnetic polarization) during lattice seeding; white arrowheads indicate locations of FMP aggregates within the PCL fiber; red arrowheads indicate magnetized cells. (B) Schematic of magnetic polarization orientations and (C) the resulting remanence as measured in a vibrating-sample magnetometer and reported in milli-electromagnetic units (mEMU). (D) Fluorescent images of EC seeded on PCL lattices 24 hours after seeding; note that PCL fibers are autofluorescent. Quantification of (E) EC density along PCL fibers and (F) resulting diameter of endothelialized structures. (G) Fluorescent images of EC attachment 24 hours after seeding as a function of seeding density. (H) Corresponding quantification of EC density along PCL fiber length (G). n ≥ 6 fields of view. All data are mean ± SD. *P < 0.05.

The magnetic field produced by lattice-embedded FMPs was sufficiently strong to attract SPION-loaded ECs and promote rapid, selective attachment during seeding with magnetized cell suspensions (Figure 5A and Supp. Video 3). To further enhance the efficiency of seeding, two additional approaches to polarizing lattice-embedded FMPs by re-writing magnetic microdomains were tested: (1) pre-polarization of FMPs prior to mixing into PCL solutions and lattice fabrication, polarization after lattice fabrication with the magnetic field oriented (2) orthogonal to the direction of the fibers, or (3) parallel to the direction of the fibers (Figure 5B). The resulting remanence (magnetization at zero field) of the microfibers in a vibrating-sample magnetometer varied by magnetic microdomain (Figure 5C). Control, non-magnetized scaffolds exhibited very low remanence since particles were not polarized. ‘Random’ scaffolds contained prepolarized FMPs that lacked alignment after microfiber printing. Due to the random alignment of FMPs with respect to each other, the net magnetization at zero field is low. However, orthogonal and parallel samples exhibit a non-zero remanence since they were polarized after microfiber fabrication in uniform directions. Parallel alignment exhibits the highest remanence of the sample set and provides more homogeneous magnetization and field lines around the diameter of the fiber. Orthogonal polarization results in anisotropic fields, which would be less favorable for homogeneous cell seeding and coverage of the fibers. Without pre-polarizing FMP microdomains, SPION-loaded ECs failed to attach to the lattice due to weak FMP-generated magnetic fields and low remanence, in agreement with Figure 5C (Figure 5D). In contrast, all three pre-polarization techniques seeded significantly more efficiently than nonpolarized, low-remanence controls (Figure 5D). Parallel FMP polarization seeded at the highest efficiency, while pre-polarized FMPs with putatively random orientation with PCL microfibers proved less efficient, as measured by the density of attached cells (Figure 5E). Upon closer analysis, the morphology of cells attached to orthogonal or parallel polarized lattice appeared more rounded and less spread, perhaps due to an excessive density of attached cells that subsequently prevented cell spreading (Figure 5D). The density of attached ECs correlated with the resulting diameter of cell-seeding PCL microfibers (Figure 5F). Further, the concentration of cells in suspension correlated with the degree of attachment, as expected. A seeding density of 2,000,000 cells mL⁻¹ resulted in complete cellular coverage of PCL microfibers while allowing for EC spreading along microfibers (Figure 5G–H). Given the efficient seeding, resulting capillary-scale diameters, and uniform distribution of ECs on PCL microfibers containing pre-polarized FMPs, 2,000,000 cell mL⁻¹ seeding density and parallel FMP pre-polarization was employed in all subsequent experiments.

2.4. Engineered capillary characterization and parenchymal cell support

Fluorescent microsphere perfusion demonstrated fluid flow through the templated microchannels (Figure 6A), so we next examined the morphology and function of endothelialized microvessels. To assess microvessel assembly, EC spreading and morphology were investigated after 2 days of culture on PCL lattices encapsulated within a type I collagen hydrogel. Collagen was cast around the lattices followed by lipase degradation of PCL. ECs maintained their spread morphology reflecting the original PCL lattice design (Figure 6B). Interestingly, occasional EC extension and anastomosis between adjacent, parallel microvessels was observed when microvessels were positioned less than ~100 ~m apart (Figure 6B). Microvessels expressed VE-cadherin at cell-cell junctions and demonstrated proper apical-basal polarity and lumenization, as visualized by the apical/lumenal marker podocalyxin (Figure 6C). Lumenization was further validated by fluorescent bead perfusion of acellular 4 mg mL⁻¹ collagen hydrogels and GFP HUVEC and GFP MVEC lined vessels in 10 mg mL⁻¹ fibrin hydrogels (Figure 6C–D, Supp. Figure 3, and Supp. Videos 4–5).

Figure 7. Magnetically assisted capillary templating improves the viability of cardiomyocyte-laden tissues.

(A) Fluorescent microsphere (Ø = 4 μm) perfusion through acellular 4 mg mL⁻¹ collagen hydrogels. (B) Fluorescent images of EC networks seeded sparsely and densely, then embedded in collagen gel and imaged 48 hours after lipase-mediated microfiber sacrifice. (C) Single slice confocal images microvessels expressing podocalyxin localized along apical/lumenal surfaces and VE-cadherin localized to cell-cell junctions; * denotes open lumenal space. (D) Fluorescent microsphere (Ø = 4 μm) perfusion through GFP EC-seeded channels in 10 mg mL⁻¹ fibrin hydrogels. (E) Cell viability quantification and (F) fluorescent images of Phospho-Histone H2A.X- and DAPI-stained iPSC-CM-laden fibrin gels cultured for 5 days following PCL degradation. n ≥ 7 fields of view. All data are mean ± SD. *P < 0.05.

To test whether magnetically assisted capillary templating and coculture with ECs could support the high metabolic demands of cardiomyocytes, induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) were incorporated within 10 mg mL⁻¹ fibrin gels at high density (40,000,000 cell mL⁻¹). After 5 days in culture following PCL degradation, cell viability was assessed by staining for Phospho-Histone H2A.X, which is a marker of DNA damage. Hydrogels lacking predefined channels, regardless of EC presence, resulted in low iPSC-CM viability (Figure 6E–F). The inclusion of acellular microchannels or alternatively admixed ECs led to an increase in iPSC-CM viability. Interestingly, highest iPSC-CM viability occurred in tissue with endothelialized microchannels. These findings suggest the importance of not only channels for nutrient delivery but also endothelial cells which reflects current literature concerning the relationship between endothelial cells and cardiomyocytes. Specifically, ECs are known to secrete multiple paracrine signals, such as nitric oxide, neuregulin-1, prostaglandin E2, prostacyclin, and parathyroid hormone-related peptide that have previously been reported to support adult cardiac tissue function.^[42] Nitric oxide is primarily produced by ECs and has been shown to support CM contractility.^[42,43] ECs also secrete numerous proteins and growth factors, such as parathyroid hormone-related peptide, to modulate CM structure and function.^[42] Although these secretory factors were not independently investigated in this study, our results suggest that the presence of endothelial cells on the channels play an important role in cardiomyocyte function that may influence cell processes and viability.

Further, UEA-1 positive ECs largely remained localized to the channels created by the PCL networks (Figure 6F). As HUVECs do not readily develop vasculature in high density fibrin hydrogels via vasculogenic self-assembly,^[44] our results suggest magnetic patterning of endothelial structures can facilitate vascularization in high density hydrogels. Using this approach, inclusion of EC-lined channels in cardiac tissues improved tissue survival and enables fabrication of larger cardiac tissue models. Although not explored here, we anticipate that manipulating the design of the lattice could additionally impact the organization and anisotropy of the parenchymal tissue.

2.4. Limitations and outstanding challenges

Although we believe the approach presented here represents an advance and novel approach to engineering capillary-scale vasculature, several challenges and outstanding obstacles should be acknowledged. The techniques utilized here for creating PCL microfiber lattices involve electrostatic attraction, which results in inconsistencies in microfiber deposition and positioning, thereby presenting challenges for fabricating more complicated lattice designs. Additionally, aggregation of FMPs within the PCL solution occurred frequently, like due to their permanent magnetization and mutual attraction. These aggregates resulted in locally heightened cell attachment and overall non-uniform cell seeding of lattice microfibers. The capillary-scale of patent channels and microvessels formed via this method also presented challenges for hydrostatic pressure-mediated perfusion, which has emerged as a preferred means of perfusion due to the relative ease of implementation. In future work, strategies to homogenize the distribution of FMPs within PCL microfibers as well as integrate microchannels with microfluidic devices for flow-controlled perfusion will be essential to the fabrication and application of more complex capillary beds.

3. Conclusion

In this work, we introduce a novel approach to templating capillary-scale microvasculature for supporting 3D tissues via magnetically assisted cell seeding. To demonstrate that broad adoptability of this method, we modified two commonly used techniques for producing polymeric microfibers to generate magnetic lattices. To efficiently endothelialize these highly porous lattices, ECs were endocytically loaded with SPIONs and subsequently magnetically attracted to FMPs embedded within lattice microfibers. Importantly, SPION magnetization of ECs did not impair viability or adherens junction assembly. Incorporating engineered capillaries within dense iPSC-CM-laden tissue, we show that the endothelialized microchannels improve tissue survival. As compared to previously established methods for engineering capillary beds including angiogenesis and vasculogenic assembly, this top-down approach can yield readily scalable capillary beds over shorter time scales. We anticipate this method will have broad utility in engineering large, dense, 3D tissues that contain metabolically active parenchymal cells.

4. Experimental Section/Methods

Reagents:

All reagents were purchased from Sigma-Aldrich and used as received, unless otherwise stated

Lattice support device fabrication:

Lattice support device molds were custom designed in Solidworks and 3D printed using a Formlabs Form 3 SLA printer with Grey v4 resin. Printed molds were treated with trichloro(1H,1H,2H,2H-perfluorooctyl)silane (TPS) to enable de-molding upon replica casting with Sylgard 184 polydimethylsiloxane (PDMS, 20:1 base:crosslinker). PDMS casts were additionally functionalized with TPS and used as stamps to emboss PDMS onto plasma-treated coverslips. To promote adhesion of collagen to the embossed PDMS lattice supports and the devices were treated with 5 v/v% (3-aminopropyl)trimethoxysilane in 100% ethanol for 24 hours, then dried and treated with 0.5 v/v% glutaraldehyde in Milli-Q water for 30 minutes. After treatment, devices were rinsed and dried in a vacuum oven overnight to remove residual glutaraldehyde.

Lattice fabrication via fiber electropulling (FEP) technique:

Dry stocks of ferromagnetic neodymium iron boron (NdFeB) microparticles (FMPs) (Magnequench) were prepolarized by exposure to a 1.0 T magnetic field pulse. Polycaprolactone (PCL; Mn 80,000) was solubilized in chloroform in concentrations ranging from 10 to 25 w/v% and prepolarized FMPs were suspended at 4 w/v%. Solutions were loaded into a 1 mL syringe and collected using a modified version of a previously published dry spinning method where microfibers are pulled from solution via a rotating collection mandrel.^[22,23] PCL solution was expelled at 0.2 mL hr⁻¹ via syringe pump under a 3 kV voltage gradient. A grounded, rotating collector plate was loaded with the lattice support devices and translated at linear speeds of 28 cm s⁻¹, unless otherwise noted, to draw out solid PCL microfibers from solution. A linear actuator was used to translate the spinneret at a rate of 1 mL hr⁻¹ laterally along the mandrel axis to generate aligned arrays of microfibers. To generate multilayered microfiber lattices, lattice support devices were rotated 90 degrees in between sequential rounds of deposition.

Lattice fabrication via solution electrowriting (SEW) technique:

PCL (Mn 80,000) was solubilized in chloroform in the concentration range from 10 to 30 w/v%, while the solutions containing FMPs were prepared as for the FEP method. For microfiber visualization, CellTracker Red CMTPX Dye (ThermoFisher) was added to the solution at 25 ~g mL^-1. A repurposed fused deposition modeling (FDM) 3D printer (Ender 3 V2, Creality, China) was used for solution electrowriting (SEW) (alternatively termed near field electrospinning). The FDM printer’s extruder was replaced with a custom-built holder for a blunt 22G needle, while the printer’s bed was replaced with an insulated copper plate serving as a charged collector surface. The needle was grounded and connected to the 1 mL syringe to extrude the polymer solution at 0.9 mL h⁻¹ feed rate, while the collector surface was connected to the high voltage power supply (Gamma High Voltage Research, USA). Fiber spacing and placement into lattices in SEW were controlled using custom gcode files. PCL fibers with and without magnetic particles were produced using a voltage gradient ranging from 2.5 to 5.5 kV (depending on polymer solution concentrations) and collected at a translation speed of 2500 mm min^-1. For cell-seeding studies, lattices were treated with 10 ng mL⁻¹ human plasma fibronectin (ThermoFisher) for 24 hours at 4°C to facilitate cell adhesion.

Lattice magnetization and magnetic characterization:

PCL lattices doped with FMPs were magnetized using a 1 T magnetic field pulse generated in a magnetorheology module in a TA Instruments Discovery Hybrid Rheometer HR30. To rewrite and align magnetic domains and define specific magnetization orientations, lattice microfibers were exposed either parallel or perpendicular to the imposed magnetic field. The magnetic properties of the NdFeB doped fibers were characterized at room temperature in a Lake Shore 7400 vibrating sample magnetometer. Magnetic hysteresis loops were acquired using a scanning field within the range of ±15 kOe to specific orientations, and paramagnetic backgrounds were subtracted. The magnetic remanences were measured at zero magnetic fields from the hysteresis loops.

Cell culture:

Human umbilical vein endothelial cells (HUVECs; Lonza) or normal human dermal microvascular endothelial cells (MVECs; Lonza) were cultured on tissue culture plastic (TCP) dishes with standard endothelial growth medium-2 (EGM2; Lonza). Induced pluripotent stem cells (iPSCs) containing a GFP fusion tagged titin reporter (Allen Institute) were differentiated into cardiomyocytes (iPSC-CMs) using a previously established protocol.^[45] Briefly, iPSCs were cultured in mTeSR1 media (StemCell Technologies) and differentiated in RPMI 1640 media supplemented with 2 v/v% B27 minus insulin (ThermoFisher) and 1 v/v% GlutaMAX (ThermoFisher, 100x). Differentiation was initiated with the addition of 12 μM CHIR99021 at day 0 and followed by 5 μM IWP4 at day 3. iPSC-CMs were purified in RPMI lacking glucose and glutamine (Biological Industries) supplemented with 4 mM D/L-lactate and subsequently maintained in RPMI 1640 supplemented with 2 v/v% B27 (ThermoFisher) and 1 v/v% GlutaMAX. To generate a GFP-expressing ECs, cells were infected with lentivirus transducing pLJM1-EGFP (Addgene plasmid #19319). Lentivirus was produced in HEK293Ts using polyethylenimines-based transfection of 3rd generation viral packaging and transgene plasmids.

Cell magnetization:

Polyvinylpyrrolidone (PVP) coated 8 nm super paramagnetic iron oxide nanoparticles (SPION; US Research Nanoparticles, Inc.) or NdFeB particles (Magnequench) were suspended in EGM2 at varying concentrations (50, 100, 200, 400, or 1000 ~g mL⁻¹) and filtered through 10 μm sieves (pluriSelect) to remove large particle aggregates. Suspensions of magnetic particles in media were added to passage 3 – 8 ECs at 90% confluency and allowed to associate with cells for 16 hours prior to harvest.

Lattice seeding and culture:

Magnetic particle-loaded ECs were washed twice with 1x phosphate-buffered saline (PBS) to remove non-endocytosed and non-adherent magnetic particles and detached from TCP with 0.05 v/v% Trypsin/EDTA solution. Cells were suspended in EGM2 and seeded onto fiber lattices at 200,000; 200,000,000; or 5,000,000 cells mL^-1. Cells were allowed to adhere to the PCL lattice for 24 hours prior to backfilling lattices with 4 mg mL⁻¹ rat tail type I collagen (Corning), prepared as previously described, or 10 mg mL⁻¹ fibrin.^[11] Briefly, collagen gels were prepared on ice with a reconstitution buffer (10 mM HEPES, 0.035 w/v% sodium bicarbonate, M199) and titrated to a pH of 7.6 with 1 M NaOH and Milli-Q water. For hydrogels encapsulating EC-seeded fibers without addition of parenchymal cells, gels were hydrated in EGM2 and media was replaced daily. Lipase was added at 1 U mL⁻¹ for 24 hours the day following encapsulation. Fibrin gels were prepared by mixing fibrin precursor solutions containing 10 mg mL⁻¹ fibrinogen from bovine plasma and 1 U mL⁻¹ bovine thrombin. For co-culture studies, iPSC-CMs were encapsulated in fibrin gels at 40,000,000 cells mL⁻¹ following pre-seeding with 5,000,000 cells mL⁻¹ ECs. Hydrogels containing iPSC-CMs were hydrated in RPMI 1640 media containing 2 v/v% B27 and 1 v/v% GlutaMAX supplemented with 5 ~M Y-27632, 2 v/v% FBS, 25 ng mL⁻¹ phorbol 12-myristate 13-acetate (PMA), 50 ng mL⁻¹ vascular endothelial growth factor (VEGF; Peprotech), and 0.05 mg mL⁻¹ aprotinin.^[46] Hydrogel-encapsulated, cell-seeded lattices were maintained for an additional 24 hours prior to PCL microfiber degradation. To enzymatically degrade PCL lattices encapsulated in CM-containing hydrogels, 0.5 U mL⁻¹ lipase from Pseudomonas sp. was added to culture medium for 48 hours. Lipase was added to RPMI 1640 media containing 2 v/v% B27 and 1 v/v% GlutaMAX supplemented with 2 v/v% FBS, 25 ng mL⁻¹ PMA, 50 ng mL⁻¹ VEGF, 0.05 mg mL⁻¹ aprotinin, 250 mM HEPES, and 0.37 w/v% sodium bicarbonate. Lipase media was replaced after 24 hours. PCL was degraded after 48 hours of lipase exposure and media was replaced daily with RPMI 1640 media containing 2 v/v% B27 and 1 v/v% GlutaMAX supplemented with 2 v/v% FBS, 25 ng mL⁻¹ PMA, 50 ng mL⁻¹ VEGF, 0.05 mg mL⁻¹ aprotinin for the following 5 days.

Cell magnetization, viability, and proliferation assays:

To assess magnetic particle loading into ECs, cells were passaged at a 1:8 surface area ratio into a 6-well TCP dish placed on top of a custom 3D printed plate holder housing 4 neodymium magnets (Figure 4A). Magnetization efficiency was quantified by the number of cells adhering above the magnets compared to the number of cells adhering halfway between each magnet (5 mm). To determine cell viability in 2D, culture media was supplemented with Hoechst (1:1000; ThermoFisher) and propidium iodide (1:1000; ThermoFisher) for 20 minutes. To determine cell viability in 3D, hydrogels were stained after fixing. All cultures were then washed twice in PBS prior to fixation. Cell viability was determined from the percentage of propidium iodide or Phospho-Histone H2A.X positive nuclei of all DAPI- or Hoechst-stained nuclei. For proliferation assays, magnetized cells were seeded onto 18 mm coverslips or passaged twice prior to seeding onto coverslips. To quantify proliferating cells, EdU was supplemented in culture medium for the last 24 hours of culture. After fixation, EdU was fluorescently labeled following the manufacturer’s protocol (ClickIT EdU, Life Technologies).

Bead perfusion:

To assess lumenization of the vessels created after PCL microfiber degradation, the vessels were perfused with 4.18 ~m fluorescent spheres (1:1000, Bangs Laboratory) diluted in PBS. Bead solution was added to one of the media wells after fixing the sample, the other media well was left empty to drive fluid flow through the patent channels.

Scanning electron microscope (SEM) imaging:

HUVECs cultured without magnetic particles and with SPION or NdFeB particles were fixed with 4% paraformaldehyde (PFA). Samples were then dehydrated in graded concentrations of ethanol (30%, 60%, 90%, 100%) for 1 hour each. Terminal dehydration was performed in hexamethyldisilane which was evaporated in a vacuum chamber. After gold sputter coating using SPI-Module Carbon/Sputter Coater, SEM imaging was performed with a Thermo Fisher Nova 200 Nanolab SEM.

Fluorescence, Staining and Microscopy:

Samples were fixed with 4% PFA for 1 hour at room temperature. To visualize the actin cytoskeleton and nuclei, samples were stained with FITC phalloidin (ThermoFisher) and DAPI (1:1000, ThermoFisher) for 1 hour for 2D samples or 8 hours for 3D samples at room temperature. For immunostaining, samples were additionally permeabilized in PBS containing Triton X-100 (5 v/v%), sucrose (10 w/v%), and magnesium chloride (0.6 w/v%) and blocked in 4% bovine serum albumin. Samples were then incubated in mouse anti-VE-cadherin (1:1000, Santa Cruz), mouse anti-podocalyxin (1:500, R&D Systems), Phospho-Histone H2A.X (1:200; #9718 Cell Signaling), or DAPI (1:1000) for 8 hours followed by Alexa-conjugated anti-mouse or anti-rabbit secondary antibodies for 8 hours each at room temperature. For UEA-1 staining, hydrogels were not permeabilized and blocked before staining with UEA-1 DyLight 649 (1:200; Vector). Fluorescent imaging was performed with a Zeiss LSM 800 laser scanning confocal microscope. Z-stacks were acquired with a 10x objective, unless stated otherwise. All images are presented as maximum intensity projections unless otherwise stated.

Statistical analysis:

Statistical significance was determined by one-way or two-way analysis of variance (ANOVA) with Tukey test for multiple comparisons or two-sided Student’s t-test where appropriate. Significance indicated by * p ≤ 0.05, with alpha = 0.05. Sample size is indicated within corresponding figure legends and all data are presented as mean ± standard deviation. For two- group comparisons, a two-tailed Student’s t-test was performed. GraphPad Prism v. 9.3.1 was used for data analysis and plotting. Sample size (n) and P-value are specified in the text of the paper or in figure legends.

Data availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.