3D bioprinted, vascularized neuroblastoma tumor environment in fluidic chip devices for precision medicine drug testing

As chip-material for our '3D bioprinting-on-a-chip platform' we use polymethylmethacrylate (PMMA). PMMA sheets are laser-engraved to achieve channel structures and specific surface patterning on macroscopic and microscopic level using an industry-standard laser engraving system (Epilog Laser, Cameo, Munich). PMMA chips are bonded to glass slides using a biocompatible adhesive tape (ARcare 90106NB, Adhesive Research Inc., Pennsylvania, USA). High-throughput laser engraving allows straightforward production of standardized PMMA slides that serve as hard shells for bioprinted tissue equivalents (figure 1). Cell-laden hydrogel is then printed into the chamber(s) of the chip with contactless micro-jet printing in a 3D Discovery BioSafety bioprinter (RegenHu, Villaz-Saint-Pierre, Switzerland). Tissues in 12 chips are fabricated in one run with approximately 3 min printing time per chip (supplemental movie M1). For the generation of hollow channels within the soft tissue, the fugitive thermo-sensitive support material Pluronic F-127 was extrusion printed in a way that channel structures through the tissue are directly connected to the laser-engraved channels on the chip. Pluronic F-127 liquefies when lowering the temperature below 15 °C. After full polymerization and shifting to lower temperatures, the liquefied support material is evacuated leaving hollow channels (figures 1(c) and (d)).

Zoom In Zoom Out Reset image size

First generations of chips contained channels on the PMMA bottom side, connectors on both sides of the chip and a central round cavity for the bioprinted tissue (figure 1(b)). These chips required more handling as the inlets to the 'tissue chamber' needed to be manually closed with drops of Pluronic F-127 before printing to prevent biomatrix blocking them during the printing process. However, with these chips we observed two difficulties: In some cases, medium diffused between hydrogel and glass, which detached the bioprinted tissue from the glass bottom. More importantly, upon longer culture (>2 weeks) the tissue frequently detached also from the walls of the chamber, which interrupted perfusion through channels and caused leakage (figure 1(b), right). This results from cell-mediated contraction forces in the developing tissue, which causes shrinkage of the tissue equivalent. We therefore developed a chip design, where PMMA chips are engraved from top at varying parameters (as visualized by different coloring in the print file (figure 1(c), left)) to achieve different depths and surface modifications. As chips do not need to be turned upside-down during engraving, this streamlines the manufacturing procedure and increases the precision of the engraving/cutting process. Inlets and supply channels are engraved at high laser intensity to achieve deep cavities in the PMMA and the camber-wall is cut with a ridged surface (one design example shown in figure 1(b)) to efficiently anchor the tissue to the PMMA chip. Surface areas of the chip were engraved around the central chamber and along the supply channels. For this patterning, the ratio between laser-pulse frequency and engraving speed was adjusted in a way that the laser generated a rough, burling-like surface for improved hydrogel anchoring to the PMMA.

To manufacture multi-layered tissue in these chips, a thin layer of bioink is printed first into the central chamber using microjet heads followed by UV-crosslinking. After polymerization, channel geometries are drawn with 35% Pluronic F-127 using a direct-dispensing print head and the areas in between and above the Pluronic-strands are then printed layer-by-layer via microjetting. Like in the central tissue chamber, Pluronic F127 strands in the supply channels are surrounded with bioink, thus building a tissue equivalent with living afferent and efferent supply vessels that are efficiently anchored to the PMMA. This also distributes forces that occur during tissue maturation and further prevents disruption of channels. After printing, the chips are exposed to violet light (5 mW cm⁻²) for 2 min to fully polymerize the bioinks. This improved chip design efficiently prevented detachment from walls and channel leakage after longer cultivation (figure 1(e)).

Hollow channels through a fibroblasts-containing tissue equivalent were filled and incubated with an acidic collagen type I solution (0.3 mg ml⁻¹, Arthro Kinetics Biotechnology, Krems, Austria) for 1 h to coat the inner walls of the conduits with a thin layer of collagen. This pre-coating improves the cell adhesion of endothelial cells to the inner conduit wall (supplemental figure S1 available online at stacks.iop.org/BF/14/035002/mmedia). To generate an endothelial cell layer we used human umbilical vein cells (HUVEC) which were immortalized and fluorescent-labeled by infection with hTert- and EYFP-encoding retroviruses. Coating the conduit walls with endothelial cells by flipping the chips every 15 min upside-down resulted in an inhomogeneous cell distribution on the channel walls. Less endothelial cells adhered to the roof- and side-walls of the channels, as shown in confocal images in figure 2(b). We therefore developed a programmable, 3D-orbital shaker to improve and standardize the conduit wall-coating with collagen and endothelial cells (figure 2(a)). The tissue chips are fixed in a laser-manufactured acryl-platform magnetically anchored to the orbital shaker plate and are rotated along the x-axis with defined stops every 90° and slowly agitated along the y-axis. This allows the cells to adhere also to the side- and roof-walls and results in a uniform endothelial cell coating of perfusion channels (figure 2(b) and supplemental movie M2).

Zoom In Zoom Out Reset image size

To test how the confluent endothelial cell layer affects vessel permeability, we injected a dextran-rhodamine B conjugate into the vessels and imaged the diffusion rate of the red-fluorescent dextran into the surrounding hydrogel by time laps fluorescence microscopy. In control chips without endothelial cells dextran markedly diffused into the hydrogel matrix within 30 min of incubation, whereas in channels coated with endothelial cells, the diffusion into the matrix was significantly reduced. In these channels, the margin of red fluorescent staining mostly lines up with the green fluorescent endothelial cell layer. The results demonstrate that the endothelial cells on the inner walls of the perfusion channels fulfill a barrier function similar to the endothelium in living organisms (figures 2(b) and (c)). Of note, endothelial cells only form an intact barrier, if the matrix of the tissue equivalent contains either fibroblasts or MSCs.

The diameter of the printing needle used for extruding the sacrificial support material is the limiting parameter for channel size and network complexity. When using a 27 G-needle the minimal diameter of the printed channels is approximately 200 µm. Dense, 3D microvascular structures below this resolution can be achieved either by photolithography or via triggering spontaneous self-organization of endothelial cells. Thereby, an interconnected micro-vessel network with an open lumen is formed which provides the tissue with nutrients and oxygen supply. Various hydrogel compositions were tested to increase endothelial cell viability and formation of micro-vessel networks. A base hydrogel of 5% GelMA was mixed with the additives alginate (5%), functionalized alginate-YISGR (5%), chitosan (1.25 mg ml⁻¹ ), hyaluronic acid (2.5 mg ml⁻¹ ), fibrinogen (2.5 mg ml⁻¹) or fibronectin (0.25 mg ml⁻¹) and was 3D bioprinted with 10⁶ HUVEC/hTert-EYFP cells/ml (data not shown). The viability and vessel formation potential of endothelial cells within 3D bioprinted tissue constructs depends on the co-cultivation with fibroblasts or MSCs: HUVEC/hTert-EYFP cells 3D bioprinted in 5% GelMA and 2.5 mg ml⁻¹ fibrinogen in a monoculture underwent cell death within 4 d (figure 3(a)). In contrast endothelial cells in a co-culture with primary adipose-tissue derived stem cells (ADSCs) in a ratio of 1:0.75 were viable 4 d after printing and began to form vessel networks at day 7 which matured into larger vessels on day 11 (figure 3(b)). Vessels formed by endothelial cells are stabilized by other cell types, most importantly by pericytes and smooth muscle cells. These additional cell types can either be directly supplemented [22] or differentiate at optimal conditions in situ from MSCs [23]. It is under debate, whether a subpopulation of MSCs represent pericyte-like cells [24]. The hydrogel composition of 5% GelMA and 2.5 mg ml⁻¹ fibrinogen provided an ideal microenvironment to allow cell migration, differentiation and vascular remodeling. After 14 d of culture, the tissue on-a-chip model was still stable with intact, permeable channels. To assess maturity of the vessel networks we stained HUVEC cells cultured in 2D (figure 3(c)) or grown in bioprinted tissue sections for 12 d (figure 3(d)) with antibodies directed against human CD31 and human VE-cadherin. We observed specific staining at cell-to-cell contacts in 2D cultures and in 3D tissue equivalents suggesting that microvessels undergo maturation in bioprinted tissue. To assess the biocompatibility of the bioink formulation also for other types of endothelial cells we tested human dermal microvascular endothelial cells (HDMECs). As shown by confocal imaging and 3D reconstruction, also this type of endothelial cells forms 3D microvessel structures with apparent open lumen (supplemental figure S2).

Zoom In Zoom Out Reset image size

Cross sections of the vascularized tissue equivalent were imaged by scanning electron microscopy (SEM) to analyze channel architecture and the ultrastructure of the matrix (figure 3(e)). Open pores were observed within the channels which indicate an open lumen formation of the microvessel structures in the 3D bioprinted tissue channels (figure 3(f)). The matrix of vascularized tissue models after cultivation for 21 d was intensively remodeled in comparison to cell-free hydrogel constructs, which presented with smooth and uniform cross-sections (supplemental figure S3).

Next, we investigated how the microvessel network responds to pro- and anti-angiogenic factors. For these experiments we used the angiogenic growth factor VEGF-165 (40 ng ml⁻¹) and the anti-angiogenic drug bortezomib (5 nM), a reversible proteasome inhibitor with anti-angiogenic effects [25–27]. At a concentration of 5 nM, bortezomib did not impair the viability of endothelial cells in 3D tissue as visualized by fluorescence microscopy (figure 4(a) at day 5 and supplemental figure S4) and measured by resazurin cell viability assays (supplemental figure S7). A faster onset and more pronounced vessel formation was observed in tissue models cultured with VEGF165 compared to controls (figure 4(a)). Upon VEGF165 treatment, early vessels were formed already at day 5. A single dose with 5 nM bortezomib at day 2 (after printing) almost completely abrogated vessel formation at day 5, but did not permanently prevent micro-vascularization as small vessels were visible at day 8 and vessel networks formed at day 12 (figure 4(a)). Co-treatment with VEGF165 accelerated the recovery from bortezomib significantly.

Zoom In Zoom Out Reset image size

To better quantify vessel formation, the interconnected total length (total master length) and total length of the formed vessels, segments and branches was analyzed and quantified with the Angiogenesis Analyzer plugin for ImageJ [28] (figure 4(b)). VEGF165 treatment significantly increases the total master segment length at day 5 and both, total master segment length and total length at day 8 compared to controls (figure 4(c) and supplemental figure S5). After 12 d, the vessel-supporting effect of VEGF was not evident anymore and similar total master segment length and total lengths were observed in controls and VEGF-treated tissues. The treatment with bortezomib significantly reduced the total master segment length at day 5, day 8 and day 12. The total length was inhibited at day 8 and day 12 by bortezomib treatment alone, but was not significantly different to controls when VEGF165 was combined with bortezomib (figure 4(c)). In tissues treated with bortezomib alone, microvessels appeared after 12 d, thus demonstrating that the ability of the tissue to form microvessel networks was not permanently hampered. This suggests that bortezomib at the used concentration does not eliminate endothelial cells or MSCs present in the bioprinted tissue.

As our main goal is to standardize and personalize bioprinted tissue-on-chips models for precision medicine applications involving immune cells in future, we next tested whether MSCs differentiated from iPS cells also promote spontaneous vessel formation in bioprinted tissue equivalents similar to primary ADSCs. In contrast to fibroblasts that are sufficient to keep endothelial cells in the matrix alive, but do not promote vessel formation (supplemental figure S6), MSCs apparently differentiate into vessel-associated cell types such as pericytes and smooth muscle cells that are essential for development and maintenance of microvessel structures. In this respect, it was also demonstrated before that in principle iPSCs can be differentiated into complex vascularized tissue consisting of multiple cell types [29]. We used a simpler protocol and differentiated iPSCs into an MSC-like population according to a protocol by Hynes et al [30]. The tri-lineage differentiation potential of the iPS-MSC population was compared to primary ADSCs of the same passage. Both iPS-MSC and ADSCs retained the functional differentiation potential to form adipocytes, osteoblasts and chondrocytes within 3 weeks of differentiation (figures 5(a) and (b)). Endothelial cells (HUVEC/hTERT-EYFP, 10⁶ cells ml⁻¹) were 3D bioprinted in a GelMA (5%)—fibrinogen/factor XIII (2.5 mg ml⁻¹) hydrogel either alone or in a co-culture with ADSCs or iPS-MSCs (both at passage 15) at a ratio of 1:0.75. Whereas endothelial cells died in monoculture, in co-culture with ADSCs and iPS-MSCs vessels were formed (figure 5(c)). iPS-MSCs demonstrated a higher potential for interconnected vessel network formation than passage-matched ADSCs (figure 5(d)).

Zoom In Zoom Out Reset image size

In a next step, we investigated how the bioprinted soft tissue interacts with neuroblastoma tumor spheroids that were embedded during the printing process. Recently, the Kurreck lab described a 3D bioprinted tumor model that combines a neuroblastoma cell line with a renal microenvironment for drug testing [31]. In our experiments, we generated spheroids from patient-derived high-stage neuroblastoma tumors (STA-NB15 cells) that were retrovirally infected to express red-fluorescent mCherry protein (NB15/mCherry). These cells were grown in a low-attachment V-bottom 96-well plate to form spheroids. Spheroids were resuspended in 10 µl collagen type I solution (3 mg ml⁻¹) and transferred into the area between channels of bioprinted tissue chips during the printing process. Thereby, neuroblastoma tumors are embedded into vessel-promoting bioink, which serves as a 3D tumor-environment. Within 6 d of culture the endothelial cells formed a vessel network around the tumor which matured into larger vessels until day 12 and persisted until day 16 (figure 6(a)). As demonstrated by confocal microscopy, dense vessel structures not only formed outside the neuroblastoma spheroid but some sprouting vessels also invaded the spheroid (optical sections from top to bottom (figures 6(d)) and 3D model from confocal z-stack (figure 6(e) and supplemental movie M3). We observed two different fates of spheroids during post-printing culture: many spheroids did not significantly increase in size during two weeks of culture, although newly formed vessels grew towards and in part into these spheroids (figures 6(a) and (d)). This is similar to encapsulated, more benign tumors in patients. The second phenotype mimicked an early metastatic tumor, as red fluorescent tumor cells left the spheroid around day 8 and migrated in part along vessel structures (white arrows) into microvessel knots around the spheroid. In these tissues, the tumor lost its spheroid shape, grew and infiltrated the surrounding tissue (figure 6(f)).

Zoom In Zoom Out Reset image size

It is not clear, which mechanism defines whether a spheroid remains spheroidic or if it enters into a metastatic phenotype—this is currently under investigation in our lab. In both phenotypes, however the neuroblastoma spheroids attracted microvessels from the surrounding soft tissue and mimicked mechanisms of tumor-angiogenesis and metastasis. Thus, this neuroblastoma model in fluidic chip devices represents a novel platform for studying the interplay between tumor tissue and its environment.

The HUVEC line, (ATCC^® CRL-1730™) and HDMEC,(ATCC PCS-110-010™) were cultured with Ham's F-12 K (Kaighn's) medium (Thermo Fisher Scientific, Waltham, USA) supplemented with 30 µg ml⁻¹ endothelial cell growth supplement (ECGS), and 100 µg ml⁻¹ heparin. Human fibroblast cells (ATCC^®, PCS-201-010) were cultured in DMEM, high glucose (Thermo Fisher Scientific, Waltham, USA). The neuroblastoma cell line STA-NB15 [44, 45] was cultured in RPMI1640 (Lonza, Basel, Switzerland). STA-NB15 cells were isolated from diagnostic left-over material at the St. Anna Children Hospital, Vienna, with written consent from patient's parents. All media contained 10% fetal bovine serum (GIBCO BRL, Paisley, UK), 100 U ml⁻¹ penicillin, 100 mg ml⁻¹ streptomycin and 2 mM L-glutamine (Lonza,Basel, Switzerland) at 5% CO₂.

Primary ADSCs were isolated from adipose tissue from abdominoplasties at the Department of Plastic and Reconstructive Surgery, Medical University of Innbruck. The study was approved by the ethics commission with the reference number AN2014-0244. ADSCs were cultured up to passage number 15 in DMEM/F12 (Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12, Thermo Fisher Scientific, Waltham, USA) supplemented with 2.5% FBS, 10 ng ml⁻¹ rhEGF, 5 ng ml⁻¹ rhFGF2 (ImmunoTools, Friesoythe, Germany) and 10 µg ml⁻¹ insulin.

Human iPS cells were provided by the Helmholtz Center Munich, iPSC Core Facility, Neuherberg, Germany. The iPS cell lines HMGU#1 and HMGU#8 were cultured in StemMACS™ iPS-Brew XF (Miltenyi Biotec, Bergisch Gladbach, Germany). Cells were passaged 1:10 every 3–5 d with Accutase^® solution and 10 µM ROCK inhibitor Y27632 (STEMCELL Technologies, Köln, Germany) for the first 24 h on Geltrex^™ (Thermo Fisher Scientific, Waltham, USA) coated six-well tissue culture plates.

All cultures were routinely tested for mycoplasma contamination using the VenorR GeM-mycoplasma detection kit (Minerva Biolabs, Germany).

Culture of Phoenix^TM packaging cells for helper-free production of amphotropic retroviruses and retroviral infection was described previously [46–48]. The retroviral vector pLIB-mCherry-iresPuro was used to retrovirally infect NB15 cells (NB15/mCherry). The vector was created by cloning the mCherry coding sequence from pQCXIN-mCherry-Tubulin-iresNeo (donated by S Geley) into the EcoR1 and BamH1 sites of the retroviral vector pLIB-MCS2-iresPuro [45]. The retroviral vectors pLIB-hTert-CMVpuro, pLIB-ECFP-iresPuro [49], and pQCXIN-EYFP-iresNeo (donated by S, Geley) were used for the generation of HFF/hTert-ECFP and HUVEC/hTert-EYFP cells, respectively.

Differentiation of iPS cells (HMGU#1 and HMGU#8) into mesenchymal-like stem cells (iPS-MSC) was done according to the protocol of Hynes et al [30]. For differentiation, we used minimum essential medium eagle alpha modification supplemented with 10% FBS, 100 μM L-ascorbic acid phosphate, 1 mM sodium pyruvate (Thermo Fisher Scientific, Waltham, USA), 1x MEM NEAA (Thermo Fisher Scientific, Waltham, USA), 100 U ml⁻¹ penicillin, 100 mg ml⁻¹ streptomycin and 2 mM L-glutamine. 80% confluent iPS colonies in a six-well tissue culture plate were detached with 0.5 ml EDTA solution (Thermo Fisher Scientific, Waltham, USA) for 5 min at 37 °C. Cells were collected with 2 ml medium and centrifuged at 300 g for 3 min. The pellet was carefully re-suspended in 2 ml medium and transferred into gelatin-coated six-well plates at a ratio of 1:1. After 14 d of differentiation cells were collected with Accutase^TM solution and split in a ratio of 1:3 into 0.1% gelatin-coated tissue culture dishes. After the first passage, uncoated dishes were used for serial passaging. After five passages an MSC-like population was obtained and characterized by functional three lineage differentiation into adipocytes, osteoblasts and chondrocytes.

Adipocyte differentiation was performed with the StemPro^® adipogenesis differentiation kit and osteoblast differentiation with the StemPro^® osteogenesis differentiation kit (Thermo Fisher Scientific, Waltham, USA) according to manufacturer protocols. For adipogenesis 25 000 cells/well and for osteogenesis 10 000 cells/well of iPS-MSCs or primary ADSCs were seeded in triplicate in 96-well plates and cultured for 3 weeks with differentiation or control medium. Medium was changed every 3–4 d. Cells were fixed with ROTI^®Histofix (Carl Roth, Karlsruhe, Germany) for 15 min at RT and washed with PBS. For adipocyte staining a stock solution of 0.5% oil-red-O (ORO) dye in 100% 2-propanol was diluted 3:2 with deionized H₂O (dH₂O) and sterile filtered via a 0.2 µm syringe filter. Cells were stained with ORO for 15 min at RT and washed with dH₂O twice. 0.2% alizarin red dye was dissolved in dH₂O and the pH was adjusted to 4.2. Osteoblasts were stained for 1 min at RT and washed five times with PBS. Chondrocyte differentiation was performed according to the protocol of Solchaga et al [50]. After three weeks of differentiation chondrocyte pellets were fixed with ROTI^®Histofix (Carl Roth, Karlsruhe, Germany) for 30 min at RT and were embedded in HistoGel^TM (Thermo Fisher Scientific, Waltham, USA). Samples were dehydrated in a graded ethanol series (50%, 70%, 90% and 100%, 5 min each) and then embedded in paraffin. 10 µm sections were de-paraffinized with ROTI^®Histol (Carl Roth, Karlsruhe, Germany), rehydrated to water and stained with sterile-filtered 0.1% toluidine blue dye in dH20 for 2 min. Stained cells were imaged with a Zeiss Axiovert200M microscope (Zeiss, Oberkochen, Germany).

GelMA functionalization was adapted from a protocol from Loessner et al [13]. Briefly, 30 g gelatin (gelatin from porcine skin, type A,) was dissolved in 300 ml dH₂O at 50 °C. Eighteen grams methacrylate were added dropwise during constant stirring and gelatin was functionalized for 3 h at 50 °C in the dark. Then the solution was centrifuged for 5 min at 3.600 g at 25 °C. The supernatant was diluted 1:1 with pre-warmed dH₂O and dialyzed with a MWCO 6–8.000 Spectra/Por^TM dialysis membrane (Fisher Scientific U.K. Limited, Loughborough, UK) against water for 7 d with daily water exchange at 50 °C in the dark. The pH was adjusted to 7.4 with 1 M NaHCO₃ and the GelMA solution was sterile filtered with a 0.4 µm aPES bottle top filter (Thermo Fisher Scientific, Waltham, USA). The GelMA solution was snap-frozen with liquid nitrogen, lyophilized and stored at −80 °C for long-term storage up to one year.

Chips were laser cut from 2 mm PMMA (Laser Cutter Mini, Epilog Laser, Colorado, USA). The chip design shown in figure 1(b) was created with CorelDraw Graphics Suite software (Corel Corporation, Ottawa, Ontario, Canada). Different colors define cutting or engraving procedures (red lines: cutting; green, blue and magenta: engraving settings are summarized in table 1). The chips were bonded to glass coverslips (26 × 21 mm, Menzel, Thermo Fisher Scientific, Waltham, USA) with biocompatible adhesive tape (ARcare 90106NB, Adhesive Research Inc., Pennsylvania, USA) [51], sonicated for 20 min, boiled twice in dH₂O for 5 min and sterilized in 70% ethanol overnight.

5% (wt/vol) GelMA and 0.5% LAP were dissolved in 50% of the final volume in cell culture medium at 37 °C for 3 h. After dissolving, the hydrogel was further soaked overnight at 4 °C. Prior to printing, the hydrogel was incubated for 3 h at 37 °C. Fibrinogen and thrombin (TISSEEL, Baxter Healthcare GmbH, Vienna, Austria) were thawed at RT and diluted in sterile PBS with 40 µM CaCl₂. Fibrinogen was added to the hydrogel to a final concentration of 2.5 mg ml⁻¹. HUVEC/hTert-EYFP and primary ADSCs were added to the hydrogel to obtain 10 × 10⁶ cells ml⁻¹ and 7.5 × 10⁶ cells ml⁻¹ respectively (Ratio 1:0.75). The cell-laden hydrogel was printed with a micro-jet printhead at a pressure of 20 kPa and a feed-rate of 15 mm s⁻¹ directly into the chip (CF300H, 3D Discovery BioSafety, RegenHu, Villaz-Saint-Pierre, Switzerland). Each printed layer was cross-linked for 5 s with the integrated light curing system (365 nm, 360 mW). Channel structures were printed via direct-dispensing with the thermo-sensitive material Pluronic F-127. 3.5 g Pluronic F-127 dissolved in 10 ml sterile 0.9% NaCl at 4 °C overnight are printed with a 20 G, 1" needle (Gonano Dosiertechnik, Breitstetten, Austria) at 20 °C with 180 kPa pressure and 2 mm s⁻¹ feed-rate. Neuroblastoma tumor spheroids (NB15/mCherry) were grown in a 96-well V-bottom microplate to a diameter of 400–500 µm and manually placed between the printed channels with 10 µl collagen type I solution (Arthro Kinetics Biotechnology, Krems an der Donau, Austria). Next, three layers of cell-laden hydrogel were printed on top of the channels. The chip and the inlets were covered with an additional layer of cell-free hydrogel (5% GelMa, 0.5% LAP) and finally cross-linked for 2 min with a custom-build LED array (390–400 nm, 6500 µW cm^–2). The fibrinogen component in the hydrogel was cross-linked with 3 ml of a 2 U ml⁻¹ thrombin solution at 37 °C for 20 min. The chips were cultured in 2.5 ml EGM2 medium (PromoCell, Heidelberg, Germany) at 37 °C, 5% CO₂ in six-well plates with medium change every second day.

Images were collected with an Axiovert200M microscope equipped with filters for ECFP (Ex: BP436/20, Em: BP480/40), EYFP (Ex: BP500/20, Em: BP535/30) and mCherry/RFP (Ex: BP550/32Bl-HC, Em: LP590) and diverse objectives ranging from 63× oil to 5× (Zeiss, Oberkochen, Germany). Live confocal microscopy was done on an inverted microscope (Zeiss Observer.Z1; Zeiss, Oberkochen, Germany) in combination with a spinning disc confocal system (UltraVIEW VoX; Perkin Elmer, Waltham, MA, USA) with a 10× objective or an Operetta CLS High Content Analysis System (Perkin Elmer, Waltham, MA, USA) with a 20× objective. Image analyses were done with Axiovision Software (Zeiss, Oberkochen, Germany).

Cells were seeded into eight-well μ-slides with glass-bottom (Ibidi, Gräfelfing, Germany) coated with 0.1 mg ml⁻¹ collagen (Corning, New York, USA) and cultured until confluence. Cells were fixed with 4% ROTI^®Histofix (Carl Roth, Karlsruhe, Germany) for 15 min and permeabilized with 0.1% TritonX-100. After blocking with 2% bovine serum albumin and 1% fetal bovine serum, CD31-antibody (1:1600, Cell Signaling, Boston, MA, USA) or VE-cadherin antibody (1:200, LifeSpan BioSciences, Washington, USA) were added, incubated overnight and after a washing step detected with anti-mouse or anti-rabbit-AlexaFluor488 FITC-conjugated secondary antibody (1:1500, Invitrogen, Waltham, USA), respectively. Nuclei were stained with 100 nM Hoechst33342 dye (Sigma-Aldrich, St. Louis, USA).

3D bioprinted tissue was fixed with 4% ROTI^®Histofix (Carl Roth, Karlsruhe, Germany) for 16 h. Samples were washed in PBS and dehydrated in a graded ethanol series (50%, 70%, 90% and 96%, 45 min each and 100% for 16 h). Samples were transferred to ROTI^®Histol (Carl Roth, Karlsruhe, Germany) for 2 h and then embedded in paraffin. 8 µm sections were de-paraffinized with ROTI^®Histol, rehydrated and washed in PBS. Antigens were unmasked by cooking the samples with 10 mM citrate buffer (pH6) in a microwave for 20 min. Slides were blocked and permeabilzed with 1% bovine serum albumin, 2% fetal bovine serum and 0.1% Triton X-100 for 45 min. Auto fluorescence was blocked with filtered 0.1% Sudan Black B/ 70% ethanol solution for 30 min. After washing with PBS, cells were incubated with CD31-antibody (1:200, Cell Signaling, Boston, MA, USA) or VE-cadherin antibody (1:100, LifeSpan BioSciences, Washington, USA), washed and incubated with anti-mouse or anti-rabbit-AlexaFluor488 FITC-conjugated secondary antibody (1:200, Invitrogen, Waltham, USA), respectively. Nuclei were stained with 100 nM Hoechst33342 dye (Sigma-Aldrich, St. Louis, USA) and embedded with fluorescence mounting medium (Dako North America, California, USA).

3D tissue chips were printed with 15 × 10⁶ HFF/hTert-ECFP cells/ml in the cell-laden hydrogel as described before. After cultivation for 7 d channels were pre-coated with 0.3 mg ml⁻¹ collagen type I (Arthro Kinetics Biotechnology, Krems an der Donau, Austria) in 20 mM acetic acid for 1 h at 37 °C with a custom-built orbital shaker. The channels were washed with PBS and coated with 30 µl HUVEC/hTert-EYFP cells (30×10⁶ cells ml⁻¹) in EGM2 medium for 90 min in the orbital shaker. The chips were cultured in EGM2 medium with medium change every second day. The barrier function was tested after 4 d by perfusion with 0.25 mg ml⁻¹ 70 kDa dextran-rhodamine B conjugate. Channels were filled with 30 µl dextran-rhodamin solution and the diffusion into the matrix was imaged in the following 30 min by fluorescence microscopy. The increase of fluorescence intensity in the matrix was quantified using Axiovision Software (Zeiss, Oberkochen, Germany) and expressed as a percentage compared to uncoated control chips.

3D bioprinted tissue models were washed with PBS and fixated for 24 h at 4 °C in 3 ml of 2.5% glutaraldehyde. The models were dehydrated with a graded ethanol series (50%, 70%, 90% and 100%, 5 min each) and left to dry out for 1 h in a biosafety cabinet. The dried models were placed on aluminum pins and were sputtered with gold (Agar Sputter Coater, Agar Scientific Ltd, Stansted, GB, UK) for 1 min and analyzed by SEM (JSM-6010LV, JEOL GmbH, Freising, Germany).

Fluorescence images were processed in Fiji software, a distribution of ImageJ [52]. Total master length, branching and total length of vessel networks were quantified with the Angiogenesis Analyzer plugin for ImageJ [28]. Data were analyzed in GraphPad Prism 8 software. Standard deviation (SD) and Student's t-test (two-tailed, with criteria of significance: *p < 0.05; **p < 0.01, ***p < 0.001 and ****p < 0.0001) were calculated when applicable.