A novel perfusion bioreactor promotes the expansion of pluripotent stem cells in a 3D-bioprinted tissue chamber

hChaMPs were redesigned from the previously-reported structure [5] via Autodesk Meshmixer and Blender (figure 1(a)), and a g-code was generated in Slic3r. hChaMPs were fabricated (figure 1(b)) as previously reported [5]. One day prior to printing, gelatin methacryloyl (GelMA; University of Minnesota Bioprinting Facility) was reconstituted to 26.7% w/v at 60 °C in a solution containing 75% mTeSR1 (STEMCELL Technologies, cat#85850), 25% acetic acid (20 mM), and 1.3% w/v lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP, Allevi). Methacrylated collagen (ColMA; Advanced Biomatrix, cat#5198) was solubilized to 2% w/v in 20 μM acetic acid at 37 °C. Prior to printing, the GelMA and ColMA solutions were mixed (3:1 GelMA solution:ColMA solution, giving 20% w/v GelMA, 0.5% w/v ColMA) and neutralized with 1 M sodium hydroxide.

Zoom In Zoom Out Reset image size

hiPSCs (overexpressing cyclin D2 under the myosin heavy chain gene [26], female) were used in these studies. hiPSCs were maintained in mTeSR1 on Matrigel (Corning, cat#08-774-552) and were regularly passaged via ReLeSR (STEMCELL Technologies, cat#05872). For printing, hiPSCs at 60%–80% confluency were singularized with Accutase (Millipore Sigma, cat#A6964). The cell suspension was washed and centrifuged twice to remove any excess Accutase that may degrade the matrix-based scaffold. Cells were then resuspended at 30 million cells ml⁻¹ in an ECM solution, containing 187 μg ml⁻¹ mouse laminin/entactin (LN; Corning, cat#354259), 187 μg ml⁻¹ human fibronectin (FN; Corning, cat#356008 or R&D systems, cat#1918-FN-02M), and 10 μM Y-27632 ROCK inhibitor (Selleckchem, cat#S1049) in mTeSR1, and incubated at room temperature for 15 min. The cells in ECM were mixed 1:1 with the GelMA/ColMA mixture, giving the ink a final composition of 15 million cells ml⁻¹ in 10% w/v GelMA, 0.25% w/v ColMA, 93.75 μg ml⁻¹ LN and FN, 0.5% w/v LAP, and 5 μM Y-27632. hChaMPs were printed on an INKREDIBLE extrusion bioprinter (CELLINK). For high resolution with our low-viscosity bioink, printing was conducted in a support bath based on FRESH v2.0 [27]. The support bath was created in a biological safety cabinet to maintain sterility; a mixture of 52.5% sterile Milli-Q water and 47.5% ethanol was heated to 45 °C in a water bath created in a secondary container. The mixture was stirred at 400 rpm and covered with foil to minimize evaporation. Sodium hydroxide (1 mM) was added to raise the pH to 7, followed by addition of gelatin type B (Sigma Aldrich, cat#G9382, to 2% w/v) and gum arabic (Sigma Aldrich, cat#9752, to 1% w/v). The solution was mixed at 45 °C for 10 min before being cooled to 35 °C, followed by slow cooling to 25 °C (<1 °C min⁻¹). Temperature was then lowered to 10 °C and stirred for an additional 5 min. The mixture was transferred to conical tubes and centrifuged at 100 G for 4 min. Supernatant was replaced with cold, sterile Milli-Q water, the tubes were vortexed and centrifuged at 500 G for 5 min. The product was then diluted in cold, sterile 1X phosphate-buffered saline (PBS; Fisher Scientific, cat#BP3994), and stored at 4 °C until the day of printing. At this time, the bath was centrifuged at 1200 G for 3 min to remove excess PBS and UV irradiated in thin layers for 30 min.

The bioink was printed at 27 °C with a 27 G, 1-inch needle. For consistent print structures, important for the perfusion set-up, pressure was varied based on the drop rate of ink from the needle, as previously described [5]. With the ink used here for the given print structure, a drop rate of 3.1 s between drops from the needle was applied. Ink was crosslinked with blue light (405 nm) after printing for 20 s on each of six faces. The prints in the support bath were then moved to an incubator to melt the support bath, followed by three washes with warm, sterile Dulbecco's PBS (DPBS; without Ca²⁺ and Mg²⁺; Thermo Fisher Scientific, cat#14190144). hChaMPs were cultured overnight in mTeSR1 in 5 μM Y-27632, under static conditions to allow for cell attachment.

Bioreactors were designed in SOLIDWORKS with ports for interior hChaMP perfusion (In_i and Out_i) and ports for flow around the exterior of the hChaMP (In_e and Out_e) (figure 1(c)). Bioreactors were 3D-printed (figure 1(d)) with E-shell 300 (EnvisionTEC, cat#1500300) via a Perfactory IV digital light projector printer (EnvisionTEC).

For perfusion, a Masterflex L/S Standard Digital Drive (cat#07522-30) was used with Masterflex L/S 14 platinum-cured silicone tubing (1.6 mm ID, cat#96410-14). The media reservoir was a 50 ml conical tube with three holes in the lid, two for the inlet and outlet of flow and one for a 0.22 μm syringe filter (Whatman, cat#6780-2502) that allowed for sterile oxygen exchange in the media. The media reservoir was initially filled with double the volume of media (mTeSR Plus [STEMCELL Technologies, cat#100-0276] supplemented with 100 U ml⁻¹ penicillin, 100 ug ml⁻¹ streptomycin [pen-strep; Thermo Fisher Scientific, cat#15140122] for the first day of perfusion) than the system required in the tubing and bioreactor. This allowed for 50% media changes (mTeSR Plus, no pen-strep) on subsequent days by a simple swap of the conical tube to minimize the possibility of contamination and bubble formation that could potentially occur when draining the tubing. The static control was cultured in an upright T25 flask with an equivalent volume of media, also receiving 50% media changes daily.

Perfusion began one day after printing. Bioreactors were sterilized with a 30 min soak in 70% ethanol under UV light in a biological safety cabinet. Bioreactors were then washed three times with sterile 1X PBS. To prepare the hChaMP for perfusion (figure S1), the hChaMP was moved to a Petri dish with warm DPBS. Silicone tubing (15 mm long, 2 mm ID, 3 mm OD; uxcell, cat#a19011600ux0295) was inserted into both vessels of the hChaMP. The hChaMP with inserted tubing was moved to a clean, dry Petri dish, and sterile 20% GelMA (with 0.5% w/v LAP, dissolved in mTeSR1 and acetic acid as described above) was pipetted at the interface of the tubing and hChaMP while crosslinking with blue light. The hChaMP was moved to the bioreactor filled with DPBS and the other end of the tubing was inserted into the bioreactor channels using sterile, fine-tooth forceps (figure 1(e)). The 20% GelMA was added to both ends of each tubing to create a seal against leakage. The tubing for Out_i (Masterflex C-Flex tubing, 3.2 mm ID, cat#06422-04) was filled with DPBS and attached to the bioreactor to remove surface tension at Out_i which creates pressure differentials that hinder perfusion through the hChaMP. To check for leaks in internal perfusion, DPBS dyed with food coloring (filtered through a 0.22 μm filter) was injected at In_i with a syringe and needle. More 20% GelMA was applied if leaks existed at the tubing connections. Dyed DPBS was washed out with warm mTeSR Plus. The bioreactor was then attached to the perfusion set-up.

With the bioreactor attached to the primed tubing, a 30 mm silicone gasket was placed on the bioreactor and the lid of the bioreactor was attached with nuts and bolts. The pump was turned on to 3 ml min⁻¹ to remove remaining air for the bioreactor. The perfusion set-up (figure 1(f)) was moved to an incubator (37 °C, 5% CO₂), and flow was changed to intermittent, 15 s of flow at 3 ml min⁻¹ every 15 min. The 50% media changes were performed daily until the end of perfusion.

To create a computational model of fluid flow in the hChaMP, pressure under flow was measured at the inlet and outlet of the bioreactor. Briefly, an acellular hChaMP was inserted in the bioreactor and attached for flow, as described above. A hemostasis valve (Qosina, cat#80395) was placed at the base of the bioreactor inlet or outlet to allow insertion of a pressure catheter in the set-up. Pressure profiles were measured with an ADV500 PV system and a 1.4 Fr pressure catheter (Transonic) with data acquisition at a sampling rate of 5000 Hz.

The interior flow was resolved using fluid–solid-interaction (FSI) approach in Simvascular [28]. The wall was described using a neo-Hookean model with an elastic modulus of 1.96 kPa, per experimental results. The fluid was modeled as Newtonian with a viscosity of 1 cP. A waveform was used to describe the intermittent flow in the inlet, and a pressure matching the experimental results was imposed as boundary condition at the outlet. Simulations were run on the Minnesota Supercomputing Institute for ten cycles in order to stabilize the pressure in the system.

The exterior flow was resolved using Ansys Fluent, release 20.1, assuming rigid walls. Like the interior flow simulations, the fluid was described as Newtonian fluid with a viscosity of 1 cP, and a waveform was imposed as the inlet boundary condition, with the flow rate matching the outlet results from the interior flow simulations. The outlet was set as the zero-reference pressure.

After 7 d of culture, the bottom of the hChaMP (the apex) was removed for ribonucleic acid (RNA) extraction (see below). The remainder of the hChaMP was fixed overnight in 4% paraformaldehyde at 4 °C. The following day, hChaMPs were washed three times with 1X PBS. The hChaMP chamber (between the vessels and the removed apex) was cut into two portions to allow for sectioning from different regions of the structure. Samples were incubated in 30% w/v sucrose in PBS for 2 d at 4 °C prior to sectioning. To prepare the samples for freezing, samples were moved to a 1:1 mixture of 30% sucrose:O.C.T. Compound (Tissue-Tek, cat#25608-930) for 1.5 h, followed by a 3 h incubation in O.C.T. Compound in cryomolds. The samples were then frozen at −80 °C.

Samples were sectioned at a thickness of 10 μm using a Leica CM1900 cryostat. Slides were left at room temperature overnight and used or moved to −20 °C the following day. To image cell distribution through the hChaMP, sections were stained with 5 μg ml⁻¹ 4',6-diamidino-2-phenylindole (DAPI, Invitrogen, cat#D1306) for 8 min and mounted with a solution of DAPI/1,4-diazabicyclo [2,2,2] octane (DABCO; Sigma-Aldrich, cat#D27802) in glycerol and PBS. Slides were imaged as tilescans on a Leica DMi8 fluorescence microscope. To quantify colony density in ImageJ, edges of sections were traced using tilescan brightfield images, DAPI-labeled tilescans were made binary, and the Analyze Particles function was used to identify cell colonies ⩾1000 μm². Five sections were analyzed per hChaMP. Distribution was quantified by drawing ten lines across the thickness (from outside to inside of chamber) of each section on brightfield tilescans (where colonies could not be easily visualized) and then finding the Plot Profile of the binarized DAPI along each line. Profiles were analyzed in Microsoft Excel to determine occurrence within each fifth section across the tissue thickness. The average wall thickness was calculated as the average length of all lines.

For IF, slides were permeabilized with 0.2% Triton X-100 (Sigma-Aldrich, cat#T8787) for 1 h and blocked with a solution of 5% bovine serum albumin, 1% glycine, 2% goat serum, and 0.1% Triton X-100 (BGST) for 2 h. Primary antibodies (1 μg ml⁻¹ mouse anti-OCT3/4 [Santa Cruz Biotechnology, cat#sc-5279], 5 μg ml⁻¹ mouse anti-KI67 [BD Biosciences, cat#556003], and 2.5 μg ml⁻¹ rabbit anti-PAX6 [Invitrogen, cat#42-6600], figure S2) in BGST were incubated on the sections overnight at 4 °C. Sections were then washed twice with 0.2% Tween-20 (Millipore, cat#655204) in 1X PBS and once with 1X PBS before 2 h incubation in secondary antibody (4 μg ml⁻¹ AlexaFluor 647 goat anti-mouse [Invitrogen, cat#A21236] or AlexaFluor 647 goat anti-rabbit [Invitrogen, cat#A32733]). Samples were again washed with Tween-20 and PBS prior to being stained with DAPI and mounted with DAPI/DABCO as described above. The 20× images of colonies were acquired and the area of fluorescent markers was quantified in ImageJ. Specifically, DAPI images were made binary, and the Analyze Particles feature was used to trace colony borders as regions of interest (ROIs) for colonies ⩾1000 μm². Then, the corresponding IF image was made binary, and the area covered by the marker within each ROI was quantified and normalized to DAPI density in the corresponding ROI. For each marker, >30 colonies were analyzed per hChaMP.

The apex of the hChaMPs were gently removed with a razor blade and flash froze in a liquid nitrogen tank for future RNA extraction. Samples were lysed and RNA isolated using the PureLink RNA Mini Kit (Invitrogen, cat#12183018A). Quality and quantity of RNA were quantified on a Cytation3 imaging reader (BioTek) using Gen5 Software. cDNA was created with the SuperScript IV VILO Master Mix (Invitrogen, cat#11756050), according to manufacturer's instructions. Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was performed with PowerSYBR Green Master Mix (1X, Applied Biosystems, cat#4367659) on a Roche LightCycler 96 (pluripotency markers, 60 °C annealing) or Eppendorf Mastercycler ep realplex² (germ layer markers, 55 °C annealing) for 30 cycles. Expression of the markers were normalized to GAPDH. Primer sequences (Integrated DNA Technologies, table S1) were found from literature, and primer pairs were chosen based on specificity for the gene of interest in primer-BLAST [29] and low possibility of dimerization according to Integrated DNA Technologies' OligoAnalyzer Tool. hiPSC-derived germ layer tissues for positive controls were derived as described in the supplemental information.

Three hChaMPs per condition were tested. Sample group comparisons were analyzed in JMP Pro 16.0. P-values were calculated for a null hypothesis of unequal means (two-tailed) with a 95% confidence interval. Student's t-tests were used to calculate the p-values comparing static and perfused hChaMPs, whereas the multiple conditions from RT-qPCR were compared with analysis of variance (ANOVA) tests and post-hoc Tukey–Kramer honestly significant difference (HSD) tests.

The hChaMP structure designed for perfusion is shown in figure 1(a) (video S1). The structure was modified from the previously reported structure [5] to act as a single cardiac ventricle, allowing for future modeling of ventricular function (i.e. pumping) with applied fluid flow and mechanical valve actuation. The hChaMP, bioprinted via extrusion printing (figure 1(b)), is formed from a bioink containing hiPSCs at 15 million cells ml⁻¹, gelatin methacryloyl (GelMA), methacrylated collagen, laminin/entactin, and fibronectin.

Due to the delicate nature of the ECM-based scaffold, with a storage modulus of 1.79 ± 0.13 kPa (figure S3), a bioreactor was designed to support the hChaMP throughout perfusion (figures 1(c)–(e)). Short pieces of tubing are used as a connection between the hChaMP and the bioreactor. The tubing is inserted into the hChaMP and sterile 20% GelMA is added and photocrosslinked at the connection to seal the tubing in place. The bioreactor contains a compartment to stably position the hChaMP, with two channels for insertion of the tubing attached to the hChaMP. The bioreactor was designed to properly guide fluid through the interior of the hChaMP and then around the exterior (figure 1(c) and videos S2–S3) to maximize nutrient exchange. To reduce fluid resistance through the hChaMP, Out_i was made larger than In_i. The flow from Out_i leads to the In_e, which enters the bioreactor compartment below the hChaMP and exits Out_e from the lid above the hChaMP (figure 1(c) and video S3).

Upon completion of a bioreactor design that provided both interior and exterior flow, flow rates for perfusion were considered. Flow rates were optimized to minimize the shear stress experienced by cells in the hChaMP while also providing adequate nutrients throughout the entirety of the chamber. A flow rate of 3 ml min⁻¹ was used, as rates below this did not allow fluid to reach the bottom of the chamber before exiting the chamber, creating dead zones of no flow (figures 2(a) and S4). However, continuous flow at this rate or higher damaged the hydrogel; thus, intermittent perfusion was tested. Infrequent perfusion, 1 min every 30–60 min, caused little cell growth, presumably due to the lack of oxygen exchange within the bioreactor and hChaMP and limited waste removal. Therefore, intermittent perfusion at 15 s every 15 min was employed. This duration of flow allowed for full media changes within the interior of the hChaMP, and the frequency supported cell survival within the tissue.

Zoom In Zoom Out Reset image size

Of note, in cases where increased pressure was applied to the hChaMP in the bioreactor, leakage would consistently occur at the tubing-hChaMP interface. This pressure limitation restricts the utility of the model for downstream applications, as the hChaMP is not able to reach higher pressures when there are leaks at the interface. Therefore, methods were explored to improve the adherence of the soft tissue to the hydrophobic tubing. With pretreatment of 3-(Trimethoxysilyl)propyl methacrylate (supplemental methods, figure S5(a)), the tubing showed improved bonding to GelMA added at the tubing-hChaMP interface. This allowed the interface to withstand higher pressures, and the hChaMP itself was found to reach a maximum pressure of 16.8 ± 0.6 mmHg before rupture (figures S5(b), (c) and video S4). This is double the average pressure the system experiences with the peristaltic pump (∼8.2 mmHg). However, this maximum, or 'burst', pressure was found to be lower than the maximum pressure applied by the peristaltic pump (∼27.1 mmHg). This discrepancy is attributable to the oscillatory pressures of the peristaltic pump, where rapid fluctuations of low and high pressures cause expansion and contraction of the hChaMP (video S3). With the burst test, pressure was slowly applied directly to hChaMP walls, causing continuous expansion of the structure (figure S5(b) and video S4) until the wall became too weak to hold more fluid.

To approximate the effects of fluid flow on the hChaMP, particularly pressure and shear stress at the hChaMP walls, computational simulations were generated. Pressure profiles were measured at the hChaMP inlet and outlet using a pressure catheter inserted into a hemostasis valve at the base of In_i or Out_i. Measured maximum, minimum, and average of the pressure profiles were used as the input for hChaMP flow in the computational model. SimVascular svFSI solver, an open-source software tailored for cardiovascular simulations [28], was used to solve for the internal flow profiles in the hChaMP. The software has been previously validated against in vitro models and has demonstrated its effectiveness in describing flow dynamics [30]. With this software and the 3D model of the hChaMP, fluid flow profiles matched those found from the in vitro experiment (figures 2(b), S4 and video S2), confirming the ability of the model to replicate the system. This allowed for computation of the generated shear stress on the inner wall of the hChaMP, a critical parameter in stem cell culture.

Pressures within the hChaMP were found to be fairly uniform, where variations throughout a single pump cycle matched experimental values, with an average around 4.5–5.3 mmHg and a maximum of 25.5 mmHg (figure 2(c)). Pressure was negative at its minimum, reflecting what is seen during the pulsatile cycling of the pump. This variability in pressure did not appear to cause issues on the integrity of the hChaMP after several days of perfusion. Wall shear stress was less variable, with an average value around 1.0 dyn cm⁻², and a maximum of 2.7 dyn cm⁻² through most of the hChaMP chamber, although drastically increasing near Out_i as fluid exited the chamber, reaching up to 15.0 dyn cm⁻² (figure 2(c)).

Owing to the intricate geometries present in the external flow, Ansys Fluent was used to solve for the external flow profiles. Fluent serves as a commercial alternative to the open-source SimVascular and has been extensively utilized to demonstrate fluid flow patterns in the cardiovascular system, including in vitro models [31]. Solving for the external flow profiles enabled computation of wall shear stress on the exterior wall of the hChaMP. Pressures and wall shear stresses were much lower on the exterior, likely due to the large compartment size of the bioreactor and a pressure drop from Out_i to In_e, leading to lower flow rates around the hChaMP (figures 2(d)–(f)).

Interestingly, pressure on the exterior depended on the location analyzed. The underside, where the fluid first contacts the hChaMP, had a higher maximum pressure than the upper side; yet the difference is fairly small, with a maximum pressure 1.13 × 10⁻⁴ mmHg for the underside and 1.05 × 10⁻⁴ mmHg for the upper portion. Similar to the pressures found for interior flow, the minimum pressure on the exterior of the hChaMP is slightly negative, but on the magnitude of 10⁻⁶ mmHg. Wall shear stress was also much lower than on the interior, ranging from values on the order of 10⁻⁷–10⁻⁵ dyn cm⁻² (figure 2(f)). Overall, interior wall shear stresses were found to be in a similar range as, and exterior shear stresses much lower than, values found in common bioreactors for stem cell expansion, which have been reported around 0.7–2.5 dyn cm⁻² [32–34]. Moreover, stem cells embedded in hydrogels were found able to withstand stresses greater than 30 dyn cm⁻² [22], suggesting the shear stresses approximated for our set-up are reasonable for hiPSC expansion.

To determine the effect of fluid flow on hiPSC proliferation and distribution, hChaMPs were inserted into the bioreactor for perfusion 1 d following printing, allowing time for cell attachment to the structure (figure 3(a)). Each subsequent day, conical tubes serving as the media reservoir were replaced with fresh media. Media in the tubing, corresponding to roughly half of the total volume in the system, was not drained to prevent the introduction of contamination or air bubbles into the system. Thus, a 50% media change with mTeSR Plus was performed daily. As biofabrication techniques in tissue engineering become more complex, the risk of contamination increases. Yet, the use of antibiotics on stem cells has generally been avoided, even for short-term treatment [35]. Here, penicillin-streptomycin was added on the first day of perfusion and gradually diluted with subsequent media changes. In this context, short-term exposure to penicillin-streptomycin had no impact on proliferation and potency of singularized hiPSCs in 3D culture (figure S6).

Zoom In Zoom Out Reset image size

hChaMPs were cultured under flow for 6 d, for a total culture time of 7 d post-printing, before being fixed and cryosectioned for analysis. DAPI staining (figure 3(b)) was used to determine the size and location of viable cell colonies. In previous work (figure S7) we found that cells that were not part of colonies were not viable, as they had likely died during singularization required for fabrication and remained entrapped in the engineered tissue. Based on these criteria, we found the perfused structures exhibited a 1.9-fold increase in hiPSC colony density relative to controls in static media (figure 3(c)). Sectioned tissues were examined to verify there was no effect of flow on hChaMP integrity. Static and perfused hChaMPs maintained consistent wall thickness, around 1.2–1.4 mm (figure 3(d)).

Moreover, the distribution of hiPSC colonies within the walls was improved in perfused hChaMPs relative to hChaMPs in static cultures (figure 3(e)). The majority of hiPSCs in the perfused hChaMPs were located in the middle of the tissue wall, with less variation in the number of colonies among the different regions of the wall. On the other hand, the static control had more colonies located near the outer portion of the hChaMP, with lower distribution throughout the tissue. Interestingly, both perfused and static hChaMPs had a lower percentage of colonies in the outermost portion of the hChaMP relative to the next inner portion. This is possibly due to cell loss during support bath removal and media changes or perfusion.

To test whether shear stress from fluid flow caused undesired differentiation of the hiPSCs, IF staining for OCT3/4 was performed on cryosections of hChaMPs (figure 4(a)). Qualitative and quantitative analysis of colonies showed nearly all cells remained pluripotent, and pluripotency was comparable to the static control. Moreover, cells were highly proliferative, with positive expression of KI67 throughout the colonies (figure 4(b)). This suggests that hiPSCs retained their pluripotent and self-renewing characteristics upon application of intermittent perfusion.

Zoom In Zoom Out Reset image size

Next, RT-qPCR was used to quantify pluripotency markers, as well as markers from each germ layer, to confirm our results on the transcript level (figure 4(c)). Expression of OCT4, NANOG, and SOX2 remained high in both static and perfused relative to the hiPSC control from 2D culture. Overall, germ layer markers were not detected, or were found at non-significant levels relative to hiPSCs or static hChaMPs, with one exception. Ectoderm marker PAX6 was found to be higher in cells from the perfused hChaMP relative to 2D hiPSCs (p=0.044). However, IF staining showed PAX6 was not expressed at the protein level in the hChaMPs (figure 4(d)).