3D Bioprinting of Collagen-based Microfluidics for Engineering Fully-biologic Tissue Systems

Microfluidic and organ-on-a-chip devices have improved the physiologic and translational relevance of in vitro systems in applications ranging from disease modeling to drug discovery and pharmacology. However, current manufacturing approaches have limitations in terms of materials used, non-native mechanical properties, patterning of extracellular matrix (ECM) and cells in 3D, and remodeling by cells into more complex tissues. We present a method to 3D bioprint ECM and cells into microfluidic collagen-based high-resolution internally perfusable scaffolds (CHIPS) that address these limitations, expand design complexity, and simplify fabrication. Additionally, CHIPS enable size-dependent diffusion of molecules out of perfusable channels into the surrounding device to support cell migration and remodeling, formation of capillary-like networks, and integration of secretory cell types to form a glucose-responsive, insulin-secreting pancreatic-like microphysiological system.


Collagen Bioink Preparation
Unless stated otherwise, a 23 mg/mL acidified collagen bioink was utilized for all prints and prepared as previously described (15).Briefly, sterile 35 mg/mL neutral collagen bioink (LifeInk 200,Advanced Biomatrix,5278) was diluted in a 2:1 volume ratio with 0.24M acetic acid (VWR, or sterile 35 mg/mL acidified collagen bioink (LifeInk 240, Advanced Biomatrix, 5267) was diluted in a 2:1 volume ratio with sterile DI H2O and mixed back and forth 40 times between two mated syringes.For the preparation of fluorescent bioinks, acidified collagen bioinks were mixed with 10-20 µL of 500 µg/mL human fibronectin (Corning,356009) fluorescently conjugated to either 488,555,and 633 NHS Esters (50,51).In multi-material printing experiments, the final concentration of the fibronectin within the collagen bioink was 50 ug/mL.In all cases, syringes containing acellular bioink were centrifuged at 3000 g for 5 min at room temperature to remove any air bubbles generated during the preparation process.The bioink was then transferred to a Hamilton glass syringe for printing.

Cellular Bioink Preparation
Vascular and pancreatic bioinks were prepared for cellular bioprinting experiments following similar protocols.For the vascular bioink, HUVECS (passage 4-6) and MSCs (passage 2-4) were cultured following the previously described procedure and lifted using a trypsin-EDTA solution.Trypsin was neutralized using trypsin neutralizing solution with 7.5 μM bivalirudin (Cayman Chemical, 23035) as a thrombin inhibitor at a 1:2 ratio.The cells were pelleted at 200 g for 5 minutes and then resuspended in 1 mL Hank's Balanced Salt Solution (HBSS, Gibco, 14175-095). 1 X 10 6 MSCs and 9 X 10 6 HUVECs were transferred into a 1 mL BD syringe.This syringe was then centrifuged at 190 g for 5 minutes and the supernatant was aspirated until approximately 100 μL remained.165 μL of 120 mg/mL fibrinogen (Millipore Sigma, 341573) and 65 μL 5% xanthan gum (dissolved in HBSS) were loaded into a separate 500 μL Hamilton gastight syringe.The 1 mL and 500 μL syringes were connected with a female luer lock adapter and the fibrinogen, xanthan gum, and cells were mixed 50 times.The vascular bioink was centrifuged at 300 g for 3 minutes in the 1 mL syringe to remove bubbles and transferred to the 500 μL syringe.The final vascular bioink consisted of 30 X 10 6 cells/mL, 60 mg/mL fibrinogen, and 1.0% (w/v) xanthan gum.The pancreatic bioink was prepared similarly to the vascular bioink with the following adaptations.MIN6 cells (passage 9-15) were lifted using trypsin solution for 5 min, and 10 X 10 6 MIN6 cells were added to the 1 X 10 6 MSCs and 9 X 10 6 HUVECs in a 1 mL syringe.The final pancreatic bioink consisted of 60 X 10 6 cells/mL (30 X 10 6 MIN6/mL, 27 X 10 6 HUVEC/mL, 3 X 10 6 MSC/mL), 60 mg/mL fibrinogen, and 1.0% (w/v) xanthan gum.

FRESH Support Bath Generation
Cellularized CHIPS were printed using a sterile support bath (LifeSupport, FluidForm) prepared according to the manufacturer's instructions.When printing collagen-based bioinks the support bath was rehydrated with a 2:1 mixture of cold 100 mM 4-(2-hydroxyethyl)-1piperazineethanesulfonic acid (HEPES), pH 7.4 (Corning, 60-034-RO) and serum free DMEM (Gibco,.When printing fibrinogen-based bioinks, the support bath was rehydrated with cold 100 mM HEPES, pH 7.4, and 1 U/mL thrombin (Millipore Sigma, T4648).Acellular CHIPS were printed using a FRESH support bath generated using a complex coacervation method as previously described (15).Briefly, FRESH v2.0 support bath (15) was made by dissolving 3.0% (w/v) gelatin type B (Fisher Scientific, G7-500), 0.3% (w/v) gum arabic (Sigma Aldrich, G9752), and 0.125% (w/v) Pluronic® F-127 (Sigma Aldrich, P2443) in a 50% (v/v) ethanol solution at 45 o C. Soon after dissolving, the pH of the solution was adjusted to 5.65 using 1M hydrochloric acid.The solution was sealed and stirred overnight at room temperature.The following morning the slurry was centrifuged at 300 g for 2 min, the supernatant discarded, replaced with DI H2O and shaken to wash the particles.The slurry was recompacted by centrifuging at 750 g for 3 min, the supernatant discarded, and replaced with DI H2O.This washing step was repeated a total of 3 times.After the final round of washing the slurry was resuspended to a final concentration of 100 mM HEPES, pH 7.4 and stored at 4 o C. Prior to printing, the slurry was placed in a vacuum chamber at room temperature for 30 min followed by centrifugation at 2000 g for 5 min.The supernatant was discarded, and the slurry was transferred into the print container of choice.

CHIPS Design
CHIPS were created in computer-aided design (CAD) software (Autodesk Inventor; Autodesk Fusion 360).All models were exported as STL files prior to printing.Overall design features and dimensions varied between experimental conditions and requirements.Perfusable networks were designed as void space within the printed CHIPS.All STL files used in the manuscript can be found at https://3dprint.nih.gov/users/awfeinberg.3D renders for manuscript preparation were generated within the rendering environment and exported as PNG files.Assembly videos of components were made within the animation environment and exported as .avior .mp4files.

FRESH 3D Bioprinting of Collagen Type I
Collagen type I was FRESH printed as previously described (15).All STL files were sliced using slicing software (Ultimaker, Cura; PrusaResearch, PrusaSlic3r; Slic3r) to produce G-code files.For 80 and 150 µm ID needles, a layer height of 32 and 60 µm was used, respectively.Chips were printed at 23-70 mm/s with 2 perimeters, 4 top and bottom layers, and 35% infill.All constructs were printed at room temperature (22 o C).Upon completion, constructs were incubated at 37 o C for at least 30 min to melt the support bath.The molten support bath was exchanged with warm print storage solution (PSS) consisting of consisting of 1X PBS, 50 mM HEPES, pH 7.4, and 1% (v/v) penicillin-streptomycin (Life Technologies, 15140-122).Acellular CHIPS were optionally sterilized by 10 min UV ozone treatment followed by overnight incubation in PSS to remove residual gelatin.

Brightfield and Stereoscopic Imaging
For all brightfield and stereoscopic images we utilized a Leica M165FC microscope with a 1X Plan Apo lens, fully adjustable base with darkfield capability, and a Prime 95B CMOS camera.Additionally, images of printed CHIPS and various equipment utilized were taken with either a Soney A5 camera equipped with a Laowa 24mm f/14 probe lens, or an iPhone 14 pro.Image contrast adjustment and resizing was performed in FIJI ImageJ or Adobe Photoshop.

OCT Imaging and 3D Gauging of CHIPS
The full 3D structure of CHIPS were imaged using optical coherence tomography (OCT) to non-invasively assess lumen patency for quality control and in-process monitoring as previously described ( 22).Briefly, OCT images were acquired using a Vega 1300 nm OCT system (Thorlabs, VEG210C1) mounted onto the bioprinter using an objective (OCT-LK4 objective) with an imaging depth of 11 mm and 13 µm lateral resolution.XYZ voxel sizes were acquired at 16.22 x 16.22 x 8.11 µm, respectively.Once scanned, XY, XZ, and YZ planes of the 3D images were analyzed to qualitatively assess the print for patency, significant defects, and potential blockages within the printed networks that would compromise flow through the CHIPS.Computational 3D gauging was performed using a combination of FIJI ImageJ, Imaris (Bitplane v9.5), 3D Slicer, and Cloud Compare software.Raw OCT images were denoised and scaled to account for refractive index of the imaging medium.Processed images were exported as TIFF stacks and imported to either Imaris or 3D slicer for segmentation of the internal perfusable networks (52).The segmented internal networks were exported as .STL files in 3D Slicer and imported into 3D builder to center the model at the origin.Both the segmented .STL and original .STL used for printing were imported into Cloud Compare 3D point cloud registration software.A standardized process was implemented to align the models to their bounding box centers and perform fine registration (22,52).A custom LUT centered around 0 was implemented to map positive (red) and negative (blue) deviations from the intended .STL onto the segmented .STL model.Gauging data and rendered 3D images of the deviations were exported for further analysis, display, and graphing.

Manual Perfusion of Branching Vascular Bed CHIPS
Perfusion of the branching vascular bed CHIPS was demonstrated by manually perfusing a concentrated ddH2O solution of blue food coloring through the top inlet of the construct with a 10 mL plastic BD syringe and a 20-gauge needle.The perfusion rate was modulated to achieve filling of the branching network and exit from the bottom outlet.Micromanipulators and helping hands devices were implemented to stabilize the perfusion without moving the printed construct.Perfusion was performed with the vascular bed construct submerged in a 50 mM HEPES buffer, pH 7.4.

VAPOR Assembly
Vasculature and perfusion organ-on-a-chip reactor (VAPOR) was designed with computeraided design (CAD) software (Autodesk, Inventor; Autodesk, Fusion 360) and printed from Biomed Clear (Formlabs, RS-F2-BMCL-01) resin on a Form 3B (Formlabs, RS-F2-BMCL-01).Stainless steel M3 hex nuts (McMaster Carr, 94150A325) were inserted into the nut cutouts in the bioreactor main body.A glass coverslip was sealed into the lid by pipetting on 100 µL of resin followed by pressing a 22 x 22 mm glass coverslip (VWR, 48366-227) into the lid which was then sealed in place by baking in a UV oven.Custom gaskets were cut from 1.5 mm thick silicone sheeting (McMaster-Carr, 5787T93) and pressed into a cutout in the lid.

Bioreactor Perfusion System Assembly
A 100 mL glass bottle (Cole Parmer, EW-34523-00) was used as a media reservoir.Holes for tubing lines were drilled into the cap and 1/16" ID silicone tubing (Cole Parmer, EW-95802-02) was pulled through to ensure airtightness.Two additional holes were drilled for air filtration and media exchange.A peristaltic pump (Ismatec, EW-95663-34) using 1.42 mm ID peristaltic tubing (Cole Parmer, EW-95663-34) was connected to the media reservoir tubing.Autoclavable bubble traps (Darwin Microfluidics, LVF-KBT-L-A) with 1/16" barb adapters (Darwin Microfluidics, CIL-D-646) were placed after the pump and then connected to stopcocks on the reactor.Tubing exiting the reactor's perfusion channels and lymph outlet returned to the media reservoir.

Bioreactor Perfusion of CHIPS
For sterile operation 3D printed parts were sonicated for 30 min in sterile filtered 70% ethanol, dried for 1 hour in a biosafety cabinet, and sterilized by 15 min UV ozone treatment.All remaining parts such as peristaltic tubing, media reservoir and bubble traps were autoclaved.For air filtration, a 0.2 µm pore size filter (VWR, 28145-501) was screwed onto the media reservoir.The system was assembled in a biosafety cabinet on an incubator shelf.All flow paths were purged with media to avoid entrapment or perfusion of air bubbles on or in the tissue.CHIPS were then transferred into the VAPOR chamber, gently pressed onto the barbs and sealed by bolting on the lid.CHIPS were perfused from 60 -1000 µL/min.The system was then inserted into a 37 o C incubator for perfusion culture.Cellularized CHIPS containing vascular bioink (HUVEC and MSC) were perfused with endothelial cell media while pancreatic CHIPS were perfused with a 50:50 ratio of endothelial cell and MIN6 media.(53)pH-Sensitive Dye Perfusion Serpentine CHIPS were placed in VAPOR reactors and perfused at 100 µL/min at each inlet.Colorimetric video acquisition was performed with a Sony A5 camera equipped with a Laowa 24mm f/14 probe lens.During perfusion, one flow path consisted of an acidic phenol red solution adjusted to pH 6.5 using hydrochloric acid (HCl).The second flow path consisted of a basic PBS buffer solution adjusted to pH 11 using sodium hydroxide (NaOH).Initial perfusion was performed with a pulsatile roller pump (Masterflex 77202-60) to stimulate mixing along the serpentine network length.The change in phenol red color from the acidic yellow to basic magenta was quantified using the Color Profiler (ImageJ (53), downloaded and installed from https://imagej.net/ij/plugins/color-profiler.html) for a segmented line then traversed the length of the serpentine network.Values for Magenta (White-Green) and Yellow (White-Blue) were extrapolated from RGB intensity to determine the ratio of Yellow:Magenta along the path during different perfusion states.The ratio was graphed as a function of path length along the serpentine network and color coded to match the yellow and magenta values according to the pH indicator values for phenol red.

Color Dye Perfusion
Laminar flow within Serpentine CHIPS was demonstrated by perfusing either red or blue food coloring while maintaining equal flow rates of 100 µL/min in both channels.Vessel patency and dye diffusion into the bulk of stacked or 3D helical channel CHIPS was demonstrated by perfusing red and blue food coloring through separate channels.Colorimetric video acquisition was performed with a Sony A5 camera equipped with a Laowa 24mm f/14 probe lens.Extended time lapse imaging was acquired with a GoPro Hero 5 camera mounted to a tripod.

FITC-Conjugated Dextran Perfusion
Dual parallel channel CHIPS were perfused at 100 µL/min with 0.1 mg/mL 3, 10, 40 or 70 kDa dextran.Dextrans were conjugated with fluorescein isothiocyanate (FITC) (Thermofisher Scientific, D3305; D1821; D1844; D1823).Dual parallel channel CHIPS' second vessel was perfused with 1X PBS.CHIPS were perfused from 1 -72 hours.Time lapse images were recorded on an epifluorescent stereomicroscope (either Nikon SMZ1000; or Leica M165FC) using a FITC filter, an X-Cite lamp (Excelitas), and a Prime 95B Scientific CMOS camera (Photometrics) with an image being taken every minute.To quantify dextran diffusion, fluorescence intensity over time was measured using ImageJ (National Institutes of Health) (53).Six regions of interest (ROIs) were selected at increasing distances away from the FITC and PBS channels and fluorescence intensity over time was calculated relative to the intensity at the initiation of perfusion while accounting for background signal.

Microbead Perfusion
Fluorescent polystyrene microbeads 10 µm in diameter were perfused at 100 µL/min at a concentration of 3.6 X 10 3 beads/mL.Beads had either 580/605 (red) (Thermofisher Scientific, F8838) or 505/515 (yellow-green) (Thermofisher Scientific, F8836) excitation/emission wavelengths.Beads were perfused through dual parallel CHIPS in either the same or opposite directions, and videos were recorded on stereofluorescence microscopes with a TexasRed filter set similar to dextran perfusions.Particle tracking and bead velocimetry were performed in Imaris 9.5.1 (Bitplane) using spot detection and tracking algorithms.

Perfusion of Dual Parallel Channel CHIPS with Afterload Pressure
Dual parallel channel CHIPS were perfused as previously described at 100 µL/min with 0.1 mg/mL 40 kDa FITC-conjugated dextran and 1X PBS.Each reservoir contained 20 mL of solution.Pressure within the CHIPS' dextran channel was increased by raising the height of the dextran reservoir to produce an additional 5 or 10 mmHg of afterload.To assess the diffusion of dextran from the source channel into the systemic PBS circulation, 50 µL samples were taken from the PBS reservoir bottle at 0 and 24 hours.The relative concentration of FITC-conjugated dextran compared to the source reservoir was then assessed by spectrophotometric analysis (Molecular Devices, SpectraMax i3x).
To assess the effect of afterload on molecular diffusion through CHIPS, time lapse images of perfusion with 5 mmHg of afterload (HP) were recorded as previously described and compared to perfusion with no additional afterload pressure (NP).The recordings were overlaid, and the fluorescence signal of the HP time lapse was divided by the NP time lapse after accounting for background signal.A vertical profile analysis was performed down the center of the HP/NP time lapse in ImageJ at various time points to further visualize the impact of increased afterload on diffusion into the peripheral regions of CHIPS.

Multi-Material Needle Alignment
In order to align multiple needles, we created a custom dual camera optical alignment system.Briefly, two 1X, 40mm WD CompactTL™ Telecentric C-mount Lens (Edmund Optics #63-745) were mounted to Alvium 1800 U-500 (Allied Vision) USB cameras.A custom 3D printed alignment plate and XY positioning system allowed for focus adjustment to achieve parfocality.To image the bottom needle tip to obtain the XY position and needle diameter, a mirror (Thorlabs ME2S-G01) was mounted at a 45º angle.The second camera was mounted perpendicular to the XY camera to view the side profile of the needle tip for Z-height alignment.A custom LabView program was written to simultaneously view the XY and Z positions.Each extruder was then moved to the center of the field of view for each camera to measure the relative XYZ offsets between each needle.The offset positions were stored as additional global software coordinate systems for use during multi-material printing using the Aerotech CNC operator's interface.
Multi-Material FRESH Printing 3D models were prepared using Fusion 360 (Autodesk) for multi-material printing by creating individual nested components for each material within the desired location of the CHIPS.
Each component was exported as an STL part file and imported into Cura 5.2 (Ultimaker) slicing software.The main components were centered around the XYZ origin and offset according to their designed spacing based on the original 3D model location.A separate material profile was created for each bioink within Cura to permit the assignment and indexing of the respective bioinks to one of the three extruders.Additionally, the creation of individual bioink specific profiles enabled component-specific color visualization of the CHIPS within the slicing software.Custom start and end G-code was specified for each extruder tool profile to recall the stored position offsets determined during the alignment process and prime the extruder between tool changes.

Tissue Clearing
After immunofluorescent staining, tissues were optically cleared using Benzyl Alcohol/ Benzyl Benzoate (BABB).Samples were first serially dehydrated by 1 hour incubation each in 10%, 25%, 50%, 75%, 90%, and 100% (v/v) ethanol solutions.Samples were then transferred into fresh 100% ethanol solution and incubated overnight at 4 o C. Samples were then optically cleared by incubation in BABB for at least 1 hour prior to imaging.

Confocal Imaging
All fluorescence confocal imaging was performed on a Nikon A1R HD MP multiphoton microscope equipped with a 4× (NA = 0.20) plan apochromat objective, 16× (NA = 0.80) long working distance water immersion objective, a 25× (NA = 1.10) plan apochromat water immersion objective, 4 visible light internal detectors, 4 visible laser lines (405, 488, 561, 633 nm), a motorized Prior Z-deck stage, Piezo Z Nosepiece, and Insight X3 DeepSee multiphoton laser (Spectra Physics).Large overview tile scans and 3D z-stack images of tissues were acquired using the 4× (NA = 0.20) plan apochromat (Nikon) objective with NIS Elements software.3D rendering and image processing was performed in Imaris (v9.5, Bitplane).For cleared tissues, a custom machined aluminum chamber was constructed to permit imaging through a large coverslip window and immobilization of the tissue within the BABB solution.

Fluorescence Image Analysis
Advanced 3D fluorescence image analysis and animations were generated in Imaris 10.0 (Oxford Instruments).Specifically, the machine learning based vascular segmentation wizard was utilized to quantify the migratory network density and diameter within the vascular CHIPS.Colocalization analysis for the Actin channel with CD-31 channel in the pancreatic CHIPS was performed on regions of interest using the Coloc-2 plugin within FIJI Image-J.Pearson's correlation coefficients and manders M1 and M2 values were recorded.

Glucose-Stimulated Insulin Secretion
Glucose stimulated insulin secretion (GSIS) was conducted using a static incubation approach in which FRESH bioprinted pancreatic CHIPS were removed from the bioreactor.The MIN6-containing pancreatic CHIPS were exposed to serial incubations in 4 mL low glucose (LG) (1.67 mM) and high glucose (HG) (16.7 mM) Krebs Buffer similar to previous work (36,54,55).A pre-incubation period in LG Krebs buffer was followed by serial incubations in fresh LG followed by HG Krebs buffer for 1.5 hours each.GSIS was performed in duplicate.Samples collected from each incubation phase were stored at -80°C for subsequent analysis of insulin concentration.Insulin content was analyzed using a mouse insulin ELISA (Mercodia) with each sample assayed in duplicate.

Statistics and Data Analysis
Statistical analysis was performed with Prism 10 (Graphpad) using appropriate tests based on experimental conditions and data.For comparison of GSIS insulin concentrations, an unpaired t-test was performed.Statistical significance was based on a P < 0.05 (*) with lower P-values being denoted as P < 0.01 (**).Non-significant P-values were denoted as ns.
Figures and visuals were constructed using Illustrator version 28.1 and Photoshop version 25.3.1 (Adobe).Supplemental videos and time-lapse images were edited in FIJI ImageJ and compiled in Premiere Pro version 24.1 (Adobe).
fig.S2.FRESH printing CHIPS designed for VAPOR perfusion and internal network quantification.(A) A CHIPS model midway through FRESH printing within a gelatin support bath.(B and C) Top (B, helical) and bottom (C, stacked) view of CHIPS immediately after FRESH printing.(D) A stereomicroscope image of a Stacked CHIPS model after release from the FRESH support bath.(E to G) OCT quantification of measured versus expected channel 5 dimensions for Serpentine (G), Stacked (H), and 3D Helical (I) CHIPS.