Development of in vitro cardiovascular tissue models within capillary circuit microfluidic devices fabricated with 3D Stereolithography printing

Abstract Human cardiovascular tissue and diseases are difficult to study for novel drug discovery and fundamental cellular/molecular processes due to limited availability of physiologically-relevant models in vitro.[1–3] Animal models may resemble human heart structure, however there are significant differences from human cardiovascular physiology including biochemical signaling, and gene expression.[4–6] In vitro microfluidic tissue models provide a less expensive, more controlled, and reproducible platform for better quantification of isolated cellular processes in response to biochemical or biophysical stimulus.[6–12] The capillary driven-flow microfluidic device in this study was manufactured with a 3D stereolithography (SLA) printed mold and is a closed circuit system operating on principles of capillary action allowing continuous fluid movement without external power supply. Human umbilical vein endothelial cells (HUVECs) and human cardiomyocytes (AC16) were encapsulated into a fibrin hydrogel to form vascular (VTM) and cardiac (CTM) tissue models respectively. To determine response to biophysical stimulus, the 3D cardiovascular tissue was directly loaded into the device tissue culture chambers that either had no microposts (DWoP) or microposts (DWPG) for 1, 3 and 5 days. The tissues were analyzed with fluorescent microscopy for morphological differences, average tube length, and cell orientation between tissues cultured in both conditions. In DWPG VTMs displayed capillary-like tube formation with visible cell alignment and orientation, while AC16s continued to elongate around microposts by day 5. VTM and CTM models in devices with posts (DWPG) displayed cell alignment and orientation after 5 days, indicated the microposts induced biophysical cues to guide cell structure and specific organization.


Introduction
Human heart tissues and cardiovascular diseases are a challenge to study for novel drug discovery and fundamental cellular/molecular processes due to the limited availability of physiologically-relevant models in vitro. [1][2][3] While animal models have been used to study heart structure, previous research has demonstrated notable differences from human cardiac and vascular physiology including vascular ow rate, biochemical signaling, and gene expression. [4][5][6] In vitro 3D tissue models with micro uidic culture systems provide an inexpensive, more controlled, and reproducible platform for better quanti cation and evaluation of cellular processes exposed to biochemical or biophysical stimulus. [6-12] Incorporating chips over the course of 25 days and resulted in angiogenic sprouting and network formation, as well as active perfusion and vascularization. Additionally, to supporting the organoid and vasculature, 3D printing technology can create biocompatible, long-term culture, and adaptability for devices conformable to co-culture of different cell types and vascularization.
Veldhuizen et. al demonstrates a micro uidic device designed to co-culture three cell types and promote anisotropy of resultant cardiac tissues. [12] hiPSC-derived cardiomyocytes (CMs) were cocultured with human cardio broblasts (CFs), encapsulated within the hydrogel, and placed into the device with staggered microposts to form a 3D tissue. Co-cultured tissues were allowed to grow up to 3 weeks and tissues formed in the device displayed mature cellular organization, production of proteins, and upregulation of genes, where microposts served as microenvironment cues that induced cell lengthening and alignment observed in human myocardium. Tissue function increased with synchronicity of spontaneous beating and calcium transients.. Capillary circuit micro uidic devices function without the help of external pumps, valves, or support, and liquid movement is driven by capillary forces determined by the geometry and surface chemistry of microchannels through the micro uidic device. [30,31] Our study presents a novel 3-step method and capillary driven-ow micro uidic platform to develop mechanically-responsive 3D cardiovascular tissue models in vitro. The method implemented in this study consisted of three major steps: (1) Micro uidic device design, manufacture, and validation, (2) 3D Cardiovascular tissue model development, and (3) Sample characterization with uorescence microscopy and computational image analysis (Fig. 1). The capillary-ow micro uidic device presented in this study was manufactured with a 3D stereolithography (SLA) printed mold and is a closed circuit system that operates on the principles of capillary action which allows continuous uid movement without the need for external power supply. Human umbilical vein endothelial cells (HUVECs) and human cardiomyocytes (AC16s) were encapsulated into a brin hydrogel and cultured in the micro uidic devices for 1, 3, and 5 days. The vascular tissue models and cardiac tissue models demonstrated statistically signi cant differences in cell alignment and cell orientation in samples housed in devices with microposts (Devices with posts, grid, DWPG) relative to devices without microposts (Device without posts, DWoP) and standard transwell inserts. The capillary driven-ow micro uidic device presented in this study was manufactured with a 3D stereolithography (SLA) printing and polymer casting method. The device is a closed circuit system that operates on the principles of capillary action which allows continuous uid movement without the need for external power supply. In these studies, the device remained open, while inside a petri dish; however, a lid can easily be designed to encase the entire device. Micro uidic capillary pumps incorporate different geometries of microstructures and surface properties to generate capillary pressure and self-regulated liquid delivery. [30] Our capillary-ow micro uidic device was designed in Solidworks CAD Software (Solidworks Corporation) with the Young-LaPlace and Navier-Stokes equations for capillary uid ow with dimensional constraints dictated by the printing resolution of a Formlabs Form3B SLA printer and Formlabs V4 Clear Resin.
The micro uidic device designs consisted of two different main chamber designs: no microposts and grid microposts (DWoP and DWPG, respectively) (Fig. 2). The formation of the microposts were based on the design of simple capillary pumps called "tree lines" and "hexagons." Tree lines are straight lines with equal vertical and horizontal spacing, mimicking a grid formation and referred to as DWPG. The DWPG have equal spacing vertically and horizontally. Hexagonal shaped microposts were integrated into the main chamber in their micropost arrangement (grid) to evaluate cell alignment and orientation in a microenvironment with and without microposts (Fig. 1). Overall device dimensions include: size of microposts (0.25 mm), vertical and horizontal spacing of microposts (0.25 mm), microchannel width (2 mm), and chamber height (2.5 mm).
Based on the design considerations, capillary pressure and ow rate were numerically calculated. Capillary pressure occurs at the liquid-air interface within a microchannel as a result of surface tension of the liquid and the curvature formed by the wettable contact angle. [30] The Young-Laplace equation outlines the relationship between contact angle, microchannel size, and capillary pressure (Eq. 1). [30] [Eq. 1] Where is the capillary pressure, is the surface tension of the liquid, is the channel height, is the channel width, is the contact angle of the liquid with the top microchannel wall, is the contact angle of the liquid with the bottom microchannel wall, is the contact angle with the left microchannel wall, and is the contact angle with the right microchannel wall.
The contact angle on the microchannel walls is equal for devices built from a single material. [30] A contact angle of was used for the micro uidic devices in this work based on previous studies. [32] Surface tension of water at room temperature ( ) and the height and width of the microchannels of the devices were used to calculate capillary pressure.
The Navier-Stokes equation assumes a laminar, steady state ow, and absence of gravitational effects to evaluate the ow rate ( ) of a liquid in a microchannel. The equation is as follows: [30] [Eq. 2] Where is the microchannel height, is the microchannel, is the difference in capillary pressure across the microchannel, is the uid dynamic viscosity, and is the length of liquid in the microchannel. The height and width of the microchannels and dynamic uid viscosity of liquid water at room temperature were used to calculate the ow rate.
3D Stereolithography printing of micro uidic device mold A Formlabs Form3B SLA printer with Clear V4 resin (Formlabs) was used to print the molds of the micro uidic devices. The micro uidic devices were designed in Solidworks, inverted as molds and uploaded to PreForm 3D Printing Software (Formlabs). Printing supports were added and the print job was initiated.
After printing, post-processing techniques were followed as recommended by the manufacturer for Clear V4 resin (Formlabs). Prints were removed from the build platform and submerged in the Form Wash (Formlabs) with fresh isopropyl alcohol (IPA) for 10 minutes.
[33] The prints were air dried in the Form Wash rack. the Devices were added into Form Cure (Formlabs) and UV cured at 60 for 15 minutes. After the cure, ush cutters (Formlabs) were used to carefully remove the supports from the molds. [33] c. Polymer casting process to manufacture micro uidic device The micro uidic devices were fabricated with Ostemer 322, a clear UV-curable resin. [34] All work with Ostemer 322 was completed inside of a chemical fume hood. The approximate volume needed to ll the molds for the device were calculated (~ 20 mL). Ostemer bottles were referenced for speci c mixing ratios for component A and B (A:B, 1.09:1, respectively). For 20 g, the following equation was used to calculate how much of each component was needed: Component B was measured rst, followed by Component A. A wooden stirrer was used to mix both components to ensure a homogenous mixture and centrifuged for 3 minutes at 1300 g to remove air bubbles.
The mold was cleaned with tape to remove any debris or dust and placed on a piece of aluminum foil. The Ostemer was slowly poured into the mold and air bubbles were removed with a pipette tip. The mold was placed under a UV lamp (i.e. Formlabs UV cure machine) and cured at 60 for 2-3 minutes intervals, checking for a exible sample and allowed 3-5 minutes to cool down. The device was removed from the mold and placed in a furnace at 90 for an hour.
d. Capillary uid ow validation experiments The micro uidic devices, DWoP and DWPG, underwent initial uid ow experiments to determine capillary-ow. A Canon PowerShot SX620 HS camera was placed on a tripod positioned above the micro uidic device. The camera was set to video and clicked the record button once the micro uidic device is set in frame (Supplementary Videos 1-2). The steps for adding liquid into the micro uidic device are listed below.
Trypan Blue (Fisher Scienti c) and PBS (Gibco) at a ratio of 0.1 to 10, respectively were mixed thoroughly. Trypan Blue (100 ) and PBS (10 mL) were measured into a 15 mL conical tube and mixed. A standard 1000 pipette tip was cut carefully using scissors to t the inlet of the micro uidic device. The cut pipette tip was placed in the inlet of the device. 1000 of the Trypan Blue and PBS mixture was released into the cut pipette in the inlet of the micro uidic device with a new, uncut pipette tip. Another 1000 was measured and released into the inlet. Trypan Blue and PBS were allowed to ow from the inlet to the outlet of the device, without any external aid. The experiment concluded when the outlet was completely lled. The same steps were repeated for DWPG (Supplementary Videos 1 and 2).

e. Fluid ow nite element analysis (FEA)
Fluid ow within the capillary circuit device was further validated with nite element analysis (FEA) using COMSOL Multiphysics software (Fig. 3). The computational simulation study parameters were modeled after experimental results from uid ow experiments using Trypan Blue diluted in Phosphate Buffered Solution (PBS) to compare ow velocities between DWoP and DWPG in the devices made with Ostemer 322. The procedure for FEA with COMSOL Multiphysics steps simulation began with importing the 3D model of the micro uidic device and selecting stationary Laminar Flow study. Liquid water at room temperature (RT) was selected for the simulation. The inlets and outlets were added, where the inlet velocity was assigned as 1.87 mm/s and 1.39 mm/s respectively based on initial uid ow experiments. Commercially available brin hydrogel was used to develop the cardiovascular tissue in this study. Fibrinogen solution was prepared by dissolving 75% clottable brinogen (Fibrinogen from bovine plasma, Sigma Aldrich) in thrombin bovine (Thrombin, Bovine, Sigma Aldrich) in 1% bovine serum albumin (BSA, Fisher Bioreagents). Fibrinogen (25 mg) was mixed in sterile PBS warmed to 37 and crosslinked with thrombin after addition of cells. HUVECs (ATCC, CRL-1730) between passages 4-6 were used to develop the vascular tissue model (VTM). HUVECs were cultured on 6-well plates coated with 0.2% Gelatin Type B (Sigma) with EGM-2 media (Lonza) and were passaged every 2-3 days. Once at 80% con uency, HUVECs are dissociated using Trypsin (0.25% Corning).

HUVEC Culture and Dissociation
The trypsin was neutralized by adding Endothelial Growth Medium-2 (EGM2, Lonza) and the cell suspension was collected and centrifuged at 1300 rpm for 3 minutes at 23 (Sorvall, ST 8R Centrifuge, Thermo Scienti c). Supernatant was aspirated carefully to not disturb the cell pellet and resuspended in 500 fresh EGM2. A small amount of the cell suspension (10 ) was removed and dispensed into a hemocytometer to count the cells with a microscope.

d. Cell Encapsulation within Fibrin Hydrogel
Prior to encapsulation, the thrombin in 1% BSA was prepared for crosslinking. The cells were prepared for encapsulation by following dissociation steps listed above in "HUVECs Dissociation" and "AC16s Dissociation." The cell pellet was resuspended in 500 of media EGM2 for HUVECs and DMEM/F12 for AC16s. At this step, the cells were ready for encapsulation.

With a new 1000
pipette tip, the cell suspension was mixed and cells were collected at densities of 500,00 cells/mL (HUVECs) and cells/mL (AC16s). The cells were added into the brinogen and PBS mixture and mixed for a homogenous mixture.

e. Cell-Hydrogel loading into Micro uidic Device
Micro uidic devices were fabricated at least 1 day prior to cell encapsulation into the brin hydrogel. A 100 mL beaker was sprayed with 70% ethanol and lled with 50 mL of 70% ethanol. Micro uidic devices were added into the beaker and fully immersed in 70% ethanol for 15 minutes. At the end of the 15 minutes, the micro uidic devices were allowed to completely dry before adding the cell-hydrogel.
When devices were sterilized and dry, 1000 of the cell-brinogen-PBS mixture was collected and slowly released into the tissue culture chambers of the micro uidic device. To crosslink the hydrogel, 150 of thrombin in 1% BSA was added directly to the hydrogel. The cell-hydrogel in micro uidic devices were incubated for 10-15 minutes at 37 and 5% . The cells were supplemented with 1000 of media (EGM2 for HUVECs and DMEM for AC16s) after cross linking. The micro uidic devices were placed into a 100 mm petri dish (Fisherbrand Petri Dish, 100 mm) and then into the incubator for 1, 3, and 5 days.
Transwell inserts for a multi-well plate were used as a control in this study to better assess cell morphology and behavior without exposure to Ostemer 322 and microenvironment biophysical cues (microposts). The cell-brinogen mixture (700 ) was added to each Transwell and crosslinked with 150 of Thrombin in 1% BSA. The multi-well plate was incubated for 10-15 minutes at 37 and 5% . After incubation, maintenance media (500 ) was added.
For maintenance of VTM and CTM (cells encapsulated within brin hydrogel), spent media was removed from the micro uidic devices with a pipette tip 24 hours after initial loading and supplemented with media every other day (day 3 and day 5).

Computational Fluorescence Image Analysis -Cell Orientation
Cell orientation in VTM and CTM samples were quanti ed with ImageJ OrientationJ plugin by measuring the gradient structure in uorescence images. [35] Images were placed in ImageJ, evaluated using Gaussian function, and exported into a histogram. Axis ranges were adjusted to include 0 to 90 preferred orientation in the images. Angles begin at 0 in the east direction and the orientation is measured counter clockwise ( Supplementary Fig. 1).

Computational Fluorescence Image Analysis -Network formation
VTM and CTM samples were evaluated for tube formation and quaniti ed using ImageJ Angiogenesis Analyzer plugin, which measures the tube length and diameter in uorescence images. [36,37] First, the Factin and DAPI channels (red and blue channels, respectively) were merged together to form 1 image. This image was then changed to a binary image using ImageJ and the Angiogenesis Analysis plugin measured the binary image.

e. Statistical Analysis
All quantitative measurements were performed and values are expressed as mean ± standard deviation (SD). Two-way ANOVA with post-hoc Tukey tests were used to compare average length and p < 0.05 was used to assess statistical signi cance using Graph Pad Prism Software. micro uidic devices (DWoP and DWPG) (Fig. 3). At the inlet, the ow splits evenly into both channels leading into the main chamber of the devices housing the VTMs and CTMs. As ow reaches the outlets of the device, there is an increase in uid ow allowing the uid to continue traveling into the second chamber, indicating both are subjected to the same conditions.

Cell Orientation
Fluorescence images stained for F-actin cytoskeleton and nucleus visualization (Phalloidin and DAPI) from each time point and condition were analyzed using OrientationJ ImageJ plugin ( Supplementary  Fig. 1). Figure 6 summarizes percent cell orientation at 0° (longitudinal axis along tissue culture chamber) and 90° (transverse axis along tissue culture chamber). HUVECs in the VTMs did not demonstrate any signi cant differences in alignment along the 0° and 90° axis (Fig. 6(A) and (B)), this non-preferential alignment is demonstrated in the uorescence microscopy images. In contrast, AC16 cardiomyocytes in CTMs statistically signi cant differences in orientation at 0° for DWPG where alignment decreased from Day 1 to Day 5 ( Fig. 6(C) The AC16 cardiomyocytes also demonstrated statistically signi cant differences in orientation at 90° for DWPG on Day 5 ( Fig. 6(D)).These results demonstrate HUVECs are forming stable capillary-like networks with non-preferential orientation while AC16 cardiomyocyte structural organization is affected by physical cues in the microenvironment.  Fig. 3). Day 3 and Day 5 DWoP for both cell types, did not have a preferred orientation (Supplementary Fig. 3). Day 1 and 3 HUVECs samples in DWPG did not have a preferred orientation (Supplementary Fig. 4). Day 1 AC16s DWPG did not show a preferred orientation.
∘ ∘ ∘ ∘ ∘ HUVECs Day 5 DWPG showed preferred orientation at 60 . Day 3 and 5 AC16s samples in DWPG showed preferred orientation around 90 ( Supplementary Fig. 4). These results indicate cells are expanding their networks and orientations around the grid microposts.

Average Network Length
Average network length for day 1, 3, and 5 were quanti ed using ImageJ Angiogenesis Analyzer and the data was analyzed using GraphPad Prism. For VTM samples, there was no statistically signi cant difference between Transwell inserts, DWoP, and DWPG. However, for CTM samples, there was a statistical signi cance between Day 3 DWPG and Day 5 DWPG (Fig. 7).

Discussion
Our novel method and capillary-ow micro uidic device demonstrates the development of a mechanically-responsive, dynamic culture system for 3D cardiovascular tissue models with the use of stereolithography printing and polymer casting method. Fluorescence microscopy images and computational analysis demonstrated morphological differences between tissues cultured in DWoP vs.
DWPG. In DWPG VTMs displayed capillary-like tube formation with visible cell alignment and orientation, while AC16s continued to elongate around microposts by day 5 which indicated the microposts induced biophysical cues to guide cell structure and speci c organization. Computational analysis of uorescent images resulted in statistical differences in alignment at 0 and 90 degrees for CTMs, as well as, total network length. Quantifying alignment at 0 degrees displayed a statistical increase in preferred alignment at 0 degrees for Day 1 CTMs DWPG. Alignment at 0 degrees displayed a statistical decrease in preferred alignment at 0 degrees for Day 5 CTMs in DWPG. At 90 degrees, there is a statistical increase in preferred alignment at 90 degrees for Day 5 CTMs in DWPG, indicating AC16s in CTMs preferred alignment toward 90 degrees instead of 0 degrees. There was a statistical difference in total network length in DWPG CTMs on day 3 to DWPG CTMs on day 5. Future work will include expansion to more physiologically-relevant human cardiovascular tissue models by including human induced pluripotent stem cells, co-cultures, and engineered biomaterials for in vitro preclinical drug testing and fundamental studies of real-time cardiomyocyte-endothelial cell-extracellular matrix interaction.  Figure 1 Work ow implemented in this study including: micro uidic device design and fabrication, 3D Cardiovascular tissue model development, and sample characterization with uorescence microscopy and computational image analysis.