If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Advanced in vitro tissue models better reflect healthy and disease tissue conditions in the body. However, complex tissue models are often manufactured using custom solutions and can be challenging to manufacture to scale. Here, we describe the automated fabrication of a cell-dense, thick (≤ 1 cm), human vascularized liver tissue model using a robotic biomanufacturing platform and off-the-shelf components to build, culture, and sample liver tissues hands-free without compromising tissue health or function. Fabrication of the tissue involved 3D bioprinting and incorporation of primary human hepatocytes, primary human non-parenchymal cells, and isolated fragments of intact human microvessels as vascular precursors. No differences were observed in select assessments of the liver tissues fabricated by hand or via automation. Furthermore, constant media exchange, via perfusion, improved urea output and elevated tissue metabolism. Interestingly, inclusion of adipose-derived human microvessels enhanced functional gene expression, including an enhanced response to a drug challenge. Our results describe the fabrication of a thick liver tissue environment useful for a variety of applications including liver disease modeling, infectious agent studies, and cancer investigations. We expect the automated fabrication of the vascularized liver tissue, at the point of use and using off-the-shelf platforms, eases fabrication of the complex model and increases its utility.
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
]. Many of these liver models are three dimensional, whether as liver cells assembled on to scaffolds (synthetic or cell-derived), multicellular liver organoids and spheroids, cellularized chips, or liver tissue mimics, and exhibit improvements in hepatocyte function as compared to 2D cultures [
]. Certainly, inclusion of multiple cell types in addition to hepatocytes, such as non-parenchymal stromal cells (NPCs), are creating more complex cell dynamics thought important in liver function [
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
]. Additional tissue complexity may be achieved by structuring the liver tissue space, creating different cellular sub-compartments, or creating different oxygen zones within the tissue. Additionally, interstitial fluid flow and dynamic culture perfusion has been shown to promote an organotypic morphology as well as boost drug response and gene expression [
]. Thus, combinations of three dimensionality, cellular complexity, tissue structuring, and perfusion are producing in vitro models that better capture native liver biology.
Optimization of in vitro 3D liver models, to replicate better long-term liver function, is an active area of study with many groups focusing on incorporating liver NPCs to reconstitute more of the liver tissue environment [
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
]. While hepatocytes themselves perform the major functions of the liver, NPCs are crucial for liver homeostasis, dynamic cell signalling, long-term hepatocyte function, pathological processes such as fibrosis and inflammation, and drug responses in vitro [
]. Several studies have focused on supplementing a hepatocyte-only in vitro liver model with endothelial cells to recapitulate a more organotypic hepatocyte morphology and increase in albumin and urea production. The focus generally has been on expanded culturing of isolated subsets of NPCs including sinusoidal endothelial cells, Kupffer cells, and stellate cells [
]. While subculturing can offer better repeatability and control over the proportion of individual cells included in the model, issues related to changes in phenotype and activation, and loss of cell viability may limit this strategy [
]. Crude NPCs preparations provide a more extensive recapitulation of the stromal cell population, however high source variability, as crude NPCs are isolated from different patients have been shown to bring in different cell populations, and limited culture survival need to be considered [
]. Regardless, most advanced in vitro models of liver tissue include efforts to reconstitute the liver stroma to complement the hepatocyte presence.
Complex, advanced tissue models can be challenging to fabricate, particularly at a scale useful in moderate and high throughput activities. To our knowledge an example of not only automated tissue fabrication, through bioprinting, but also long-term culture, media exchange and imaging has not yet been demonstrated. Furthermore, as tissue models become more complex, many aspects of the fabrication may involve specialized approaches, such as 3D bioprinting or cell aggregate picking/placing. Moreover, tissue constructs fabricated at a central facility need to be preserved and delivered, limiting the potential utility of the models for end-users located far from the facility. Commercial-level production of even simple tissue constructs (e.g. cells on a scaffold), much less tissues with more complex architectures and multiple components require more effective manufacturing solutions, preferably at point-of-use. While solutions are emerging for the large-scale expansion of cells and their incorporation into tissue constructs, no solution is yet in place to manipulate these fabricated tissue constructs through a production line or assembly into larger tissue systems at a manufacturing scale. Thus, the automated fabrication of tissue constructs at the point of use in a form readily configurable for different applications and/or experimental settings would significantly increase the utility of these more informative liver tissue systems [
Using an agile, robotic biomanufacturing platform developed by our group, we demonstrate the automated manufacturing of a complex human tissue as a proof-of-concept. The fabrication involves multiple steps including the 3D printing of fugitive materials to help shape the liver tissue. The platform involves a 6-axis robotic arm (BioAssemblyBot®), capable of multiple, automatic tool exchanges, with a modular tissue culture incubator and a confocal scanner uniquely integrated for the automated manufacturing of tissues and organs. The example use-case tissue fabricated is a 3D, thick (∼1 cm), human vascularized liver construct to model the liver tissue environment compatible with standard multi-well plate formats and configurable for static or through-tissue perfusion. The liver tissue is comprised of primary hepatocytes and a crude preparation of non-parenchymal cells. In addition, human, adipose-derived, microvessel fragments (haMVs) are added to supplement the NPC cell population. haMV are whole, intact microvascular fragments consisting of a collection of support and stromal cells. While sourced from adipose tissue haMVs have been shown to be capable of phenotype adaptation when placed into various culture environments [
Microvessel-like structures from outgrowth endothelial cells from human peripheral blood in 2-dimensional and 3-dimensional co-cultures with osteoblastic lineage cells.
]. We reason the addition of haMVs will provide more dynamic cell-cell signalling within the tissue environment and provide a microvascular niche environment to support stromal cell functions. Functional measurements of liver tissue constructs produced automatically are indistinguishable to liver tissue constructs made by hand.
Methods
Liver construct fabrication
Individual wells of a 96 well plate were coated with 100 ml of 3mg/ml pH neutralized, type I rat tail collagen (Corning) and dried. Vertical pillars of 40% Pluronic F-127 (Sigma) in Hank's Balanced Salt Solution (HBSS) were printed on top of the dried collagen with the BioBotBasic® (Advanced Solutions Life Sciences). Primary hepatocytes (Zen-Bio) and crude, primary non-parenchymal cells (Zen -Bio) were thawed into hepatocyte supplement pack media (Fisher Scientific) (Supplemental Table 1). Cells were counted, mixed, and spun down together in a 25% Percoll gradient (Sigma) at 100XG for 10 minutes. Cells were resuspended in 3mg/ml collagen I in DMEM at 20 million Hepatocytes/ml and 2 million NPCs/ml. When human adipose microvessels (haMVs; Advanced Solutions Life Sciences) were added to the construct, vessels were spun down separately at 400Xg for 4 minutes, resuspended in 3mg/ml collagen at either 100,000 (2 replicates) or 200,000 (2 replicates) vessels/ml, and then used to resuspend the hepatocyte/NPC pellet. 100ul of the collagen/cell suspension was pipetted around the printed pillars in each 96 well plate, care was taken to ensure the cell suspension did not go over the top of the printed pillars. After casting, plates were quickly placed into the incubator to gel the collagen. Following collagen gelation, hepatocyte growth media (Thermo Scientific) was pipetted on top. During the first day, media was changed 2X times to remove the Pluronic pillars. Subsequently, media was changed every 24 hours (+/- 4 hours) for 14 days. Media supernatants were frozen back at -80°C. On day 13, select constructs received 10µM rifampicin (Sigma) prepared in DMSO and diluted in culture media. 0.1% DMSO was added to untreated cultures as a vehicle control. On day 14 constructs, were cut vertically in half and ½ of the constructs were removed and snap frozen for genomic analysis. The other half was fixed with 10% Neutral buffered Formalin (NBF) overnight at 4°C.
Well plate perfusion
When perfusion was added to the human liver constructs, the same fabrication strategy was used as described above using the interior a 24 well cell culture insert instead of a 96 well plate. After collagen gelation at 37°C, the cell cultures inserts were moved to a manifold aligner (Advanced Solutions Life Sciences) and 1.1 ml of media was added to the bottom compartment with 100μl added to the insert interior. A perfusion manifold (PerFuseIt; Advanced Solutions Life Sciences) fitted with tubing and syringes was placed on the plate and perfusion started the same day as the tissues were established. Fluid flow at 10µl/hr was administered for 14 days using a syringe pump. The inlet tube connected to the syringe pump was attached to a hypotube in the manifold located to the top of the insert reservoir while the outlet tube was withdrawn from a hypotube in the bottom compartment. When media was collected at day 1, and 7, the insert-aligned plate with the inserts were transferred to fresh 24 well plates with fresh 1.1ml of media. The manifold and tubing were placed back on the top of the insert plate and perfusion was continued. The media below the inserts were collected and stored at -80°C for analysis. To ensure supernatants were only representative of a single day this process occurred both 24 hours before supernatants were collected to give a fresh media bath and then again on the day of interest. On day 14 cultures were cut in half and snap frozen or fixed in 10% NBF for analysis.
Automation
Automation of the sample occurred using the BioAssembly® Platform (Advanced Solutions Life Sciences). To set-up the automated workflow, robotic arm pathing was calibrated between the print stage, tilt station, INCell 6500 Imager (Cytiva), and the BioStorageBot™ (Advanced Solutions Life Sciences). Workflow tasks and operations were assembled into BioApps™ (Advanced Solutions Life Sciences), which managed the entire workflow. Hepatocytes, NPCs, and haMVs were thawed and resuspended in collagen as before. The cell/haMV/collagen mix was placed into the BioAssembly Platform and the remainder of the protocol including printing, cell seeding, media changes with media collection, and long-term incubation were performed by the robot for one week. After 1 week the cultures were removed, tissues cut in half, and either fixed or frozen for analysis.
En block staining
After fixation, constructs were washed 2X with 1X PBS and stained with UEA-1 (Vector Labs). Constructs were blocked with 5% BSA overnight at 4°C. UEA-1 was diluted 1-50 in 5% BSA and incubated overnight at 4°C followed by washing 2X with PBS and imaged via a confocal microscope (FV300; Olympus) or the INCell 6500 Imager (Cytiva). Confocal imaging took advantage of the autofluorescence of hepatocytes as well as the UEA-1 staining.
Histology
Fixed tissues were paraffin embedded and 5μm sections prepared (Saffron Scientific). For staining, sections were deparaffinized and rehydrated with graded xylene and EtOH washes followed by rinsing in running water. Antigen retrieval was performed by adding the samples to Tris-EDTA Buffer in a pressure cooker for 2 minutes. All samples were then washed with 1X PBS and blocked with 1.5% normal goat serum (Vector Labs) and 1:100 rabbit IgG (Vector Labs) in PBS for 30minutes. For colorimetric staining, samples were quenched with 0.3% hydrogen peroxide (FisherSci) in methanol for 30 minutes prior to blocking and an additional 15 minutes of blocking with both biotin and avidin (Vector Labs) was performed. The primary antibody (αSMA Dako or CD45 Abcam) was then added to the blocking serum and incubated overnight at 4°C. Negative controls were performed with a mouse IgG in 1.5% NRS. For fluorescence immunostaining, the secondary antibody was diluted 1:4000 in 1.5% NRS and incubated for 1 hour at RT. Samples were rinsed with PBS, counterstained with Hoechst (Fisher, 1:6000) then prolong gold (Invitrogen) was used to mount coverslips and seal the samples. For colorimetric staining, samples were incubated with Impress reagent (Vector Labs) for 30 minutes, rinsed with 1X PBS, then incubated with the DAB kit (Vector) according to the manufacturer instructions, and viewed under the scope to monitor the development time needed. Samples were rinsed in water for 5 minutes then coverslips were mounted and sealed with prolong gold. Sealed slides were images using an Olympus confocal or Nikon confocal system and number of stained cells/# of hepatocytes were recorded. When sections were stained with UEA-1 (Vector) deparaffination was performed then samples were blocked with 5% BSA in PBS for 30minutes. UEA-1 was diluted 1:50 in 5% BSA and incubated samples for 10 minutes at room temperature. Slides were then sealed and imaged using confocal microscopy.
Urea analysis
Culture supernatants were analyzed for urea using a urea assay kits (Abcam and Sigma). Samples from static cultures were diluted 1:100 or 1:50 depending on the hepatocyte lot. Perfusion samples were not diluted for analysis. A standard curve was prepared with urea concentrations between 0 and 5 nmol/well. Standards and samples were mixed with supplied reagents and incubated at 37°C for 1hour. Absorbance at 570nm was measured using a plate reader (Biotek Synergy LX). A linear trendline was fit to the standard curve and used to report the urea concentration.
Media analysis
Culture supernatants were analyzed for lactate dehydrogenase (LDH), sodium pyruvate, and lactate using appropriate colorimetric kits (Sigma) according to the manufacturer's instructions. Briefly, a standard curve was prepared. Samples were diluted and combined with the respective reaction master mix and allowed to incubate for the desired timeframe then an absorbance reading was taken. For LDH, the samples were incubated at 37°C for 20 min. then a second reading was taken. Changes in LDH, sodium pyruvate, and lactate measurements were calculated and compared to the standard curve.
Gene expression
Snap-frozen liver constructs were thawed, and RNA isolated using an RNAeasy kit (Qiagen) according to the manufacturer's instructions. Briefly, thawed tissues were placed into lysis buffer consisting of 70% Ethanol, Buffer RLT, and b-mercaptoethanol. As controls, frozen haMVs were used. Sample was homogenized on ice using a pro 200 homogenizer. The lysate was centrifuged for 3 minutes at 14000xG. The pellet size was checked and if needed the homogenization process was repeated. Once the pellet was fully homogenized 1 volume of 70%EtoH was added to the lysis buffer and the sample was transferred to a RNAeasy spin column and centrifuged for 15seconds at 8000xG. The flow thorough was discarded, and the spin column was washed with Buffer RPE by centrifugation at 8000xG for 15seconds. The flow though was again discarded and additional Buffer RPE was added for an additional wash at 8000xG for 2 minutes. The empty column was centrifuged at 14000xG for 1 minute for remove any excess liquid. Then the spin column was transferred to an empty collection tube and 30μl of RNAse free water was added. The column was spun for 1minute at 8000xG to collect the isolated RNA. RNA was stored in -80 for long-term storage. Next RNA was converted to cDNA using Invitrogen Superscript IV VILO Mater Mix Kit (Fisher). Briefly, sample RNA was incubated with 10X ezDNAase Buffer, ezDNAase enzyme, and nuclease free water at 37°C for 2 minutes. Then Superscript IV VILO mastermix, and additional nuclease free water was added. As a control, blanks were made up with each cDNA conversion in which Supercript IV VIOL NoRT control was added in place of the superscript IV VILO mastermix. Samples and blanks were annealed by incubating at 25°C for 10minutes then RNA was reverse transcribed by incubating at 50°C for 10minutes followed by an incubation at 85°C for 5 minutes to inactivate the enzyme. A RT-PCR reaction was run with each sample by incubating the converted cDNA was SYBER select mater mix, and the forward and reverse primers (Supplemental Table 2). Real time PCR was run using an CFX96 Real Time System (Bio Rad). The PCR reaction was initiated by an incubation at 95°C for 5 minutes, followed by 35 cycles of incubation at 95°C for 1 minutes, 60°C for 1minute, 72°C for 1 minute. The plate was read following each cycle. The reaction ended with an incubation at 72°C for 10 minutes. Data was analyzed by normalizing cycle times for each PCR product to cycle times for GAPDH. The GAPDH normalized cycles were again normalized to liver samples without haMVs present. When GAPDH normalization occurred between static and perfusion samples, splicing factor, arginine/serine-rich 4 (SFRS4) was used to normalize the data.
Statistics
Statistics were performed using SigmaPlot 11.0 within an experimental group and the remaining experimental repeats were seen to follow the same trend. A student T-test was used to evaluate differences between two groups with only one variable. When two or three variables were present, a respective two- or three-way ANOVA with a Holm-Sidak post-hoc method was used for analysis. When the data failed normality and equal variance tests, a Mann Whitney Rank Sum test was used.
Static experiments were performed on four separate occasions with 3-6 replicates per experimental group. Rifampicin treatments were performed on two separate occasions with 3-6 replicated per experimental group. Perfusion and Automation experiments were performed on one occasion with 3-6 replicates per experimental group.
Results and discussion
Thick, human liver tissue environment supplemented with human microvessels
We developed a fabrication strategy (Fig. 1) to build mesoscale liver tissue using primary hepatocytes, a crude preparation of non-parenchymal liver cells, and isolates of fragmented adipose microvessels (haMVs) [
] to supplement the tissue environment. Isolated primary hepatocytes have emerged as the gold standard for in vitro liver toxicity as they recapitulate variation in human population and have been shown to be the closest match to human in vivo cell dynamics [
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
]. Additionally, inclusion of liver non-parenchymal cells and endothelial cells, as either crude preparations or subpopulations, in a liver tissue model contribute to more robust toxicity assessments [
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
]. Finally, as an additional source of vascular cells, we employed fragments of intact microvessels isolated from human adipose (haMVs). Isolated microvessel fragments retain their native vessel structure and cellular composition [
Angiogenic potential of microvessel fragments is independent of the tissue of origin and can be influenced by the cellular composition of the implants.
]. We reasoned inclusion of the haMVs would provide a blood vascular component and introduce any cells populating the perivascular niche that would possibly contribute to the liver tissue environment. This combination of cells and the fabrication approach produced a liver tissue construct that 1) contained a variety of liver cell types, 2) created dense zones of hepatocytes interspersed with regions of collagen, and 3) maintained liver-like functions for at least 2 weeks.
Fig. 1Fabrication of the liver tissue construct (A) schematic of the liver tissue fabrication strategy. A digital model of the fugitive pillars in a well is used to direct 3D printing of fugitive material. After printing, hepatocytes, NPCs, and haMVs from primary sources are mixed with matrix and dispensed around printed pillars. Printed pillars are then washed away, and the final tissue construct is formed. (B) A phase image (top view) of the final fabricated liver tissue. (C) Projected confocal scan image of an intact fabricated liver construct showing the autofluorescent hepatocytes (green) and UEA-1 labeled microvessels (orange). (D) Confocal image of a construct histology section showing hepatocytes (green) and UEA-1 labeled microvessels (purple). (E) Urea levels over time within the tissue supernatant. (F) LDH activity in tissue supernatants over time. (G) Day 14 hepatocyte density measured from tissue sections ** P ≤ 0.001. All data and graphs shown represent experiments with both 100,000 and 200,000 vessels/ml. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
]. Nearly all the experiments were performed with a single lot of hepatocytes to provide more repeatability when comparing the impact of the human adipose microvessels in the liver tissue model across all experiments. The flexibility of the model and its fabrication is such that a variety of hepatocyte sources, including stem cell-derived hepatocytes, multiple primary human hepatocyte lots and primary hepatocytes from different species could be accommodated with this model. We reasoned using a crude preparation of NPCs instead of isolated subpopulations would better capture the full spectrum of cell types present within native liver tissue. While segregation and subculturing provide for more repeatable approaches due to cell homogeneity, such approaches are limited by phenotypic deviations as passage number increases in addition to the loss of minor NPC populations [
To address possible complications related to nutrient exchange and cell necrosis, we employed a 3D printing strategy to structure the liver tissue constructs wherein media-tissue exchange spaces were created by printing pillars of a fugitive hydrogel around which the cell and haMV compositions are deposited (Fig. 1A). The approach is such that multiple, pillar-like spaces of varying diameters are 3D printed in custom, off-the-shelf multi-well plates or cell culture inserts. Here, 6 pillars (400 μm Ø, 5 mm in height) were evenly spaced with a central pillar surrounded by 5 pillars per well of a 96-well plate (Fig. 1A and Sup. Fig. 3). The number and arrangement of the pillars was chosen to generate tissue spaces no thicker than ∼ 300 μm, the maximum thickness considered possible to prevent necrosis in vitro [
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
] (Sup. Fig. 2). Following printing of fugitive pillars, the liver cell/haMV cell slurry was mixed with a small amount of type I collagen and cast around the pillars (Fig. 1A). After incubation to “set” the tissue, the pillars were washed away leaving behind open spaces extending top to bottom of the liver tissues. The resulting liver tissue constructs were comprised of hepatocytes distributed uniformly throughout the tissue space, without evidence of necrotic zones including those areas away from the open spaces at 14 days of culture (Fig. 1B). Endothelial cells, often in microvessel structures, were also relatively uniformly distributed, residing in spaces between hepatocytes in the tissue (Fig. 1C and D).
In all cases, urea production decreased from day 1 to day 7 and again from day 7 to day 14 (Fig. 1E). LDH release levels, a measure of cell death, were high on the day after fabrication and decreased to near zero levels by day 7 and 14, indicating good cell recovery after fabrication and stability over 14 days (Fig. 1F). The inclusion of haMVs significantly reduced early LDH release by the cells of the construct when compared to the no MV samples. Hepatocyte density showed no significant difference between constructs containing haMVs and those without (Fig. 1G). Combined, these findings indicate the formation and stable maintenance of a hepatocyte-rich tissue environment for at least 14 days.
Our motivation for a thicker construct was to enable studies and assays requiring a more substantial and complex liver tissue space such as in evaluating malaria sporozoites cycling in the liver [
]. A key aspect of our approach to improve liver tissue function is structuring the cell-rich tissue bed to increase surface to volume ratios facilitating diffusion to the tissue interior. This was accomplished by 3D printing pillars of a fugitive material and casting cells/haMVs around those pillars. Spacing between pillars, at approximately 300 microns, was chosen based on oxygen diffusion distances reported for other in vitro liver models (e.g., spheroids). While we didn't measure tissue oxygen concentrations, the absence of frank necrosis indicates that there was sufficient oxygenation to maintain hepatocyte viability. However, the pillar spacing could easily be modified to have variations is the geometry of the tissue compartment and allow for modelling oxygen zonation within this printed tissue.
Characterization of non-parenchymal cell population
It is widely accepted that the endothelial cells (ECs), Kupffer cells (KCs), and hepatic stellate cells (hSCs) of the liver, play an important role in long term culture maintenance and hepatocyte function. We analyzed our engineered tissues for the presence of ECs, KCs, and hSCs cells using UEA-1, αSMA, and CD45 in tissue both with and without the inclusion of haMVs in addition to the crude preparation of NPCs. Very few endothelial cells (UEA-1+) were observed in liver tissue constructs that did not include the haMVs at the time of fabrication (Fig. 2A and B). In contrast, endothelial cell numbers, as dispersed and clustered with a microvessel structure, were increased when haMVs were included. Interestingly, alpha actin-positive cells were also elevated in the constructs made with haMVs, presumably reflecting perivascular cells as many were associated with the incoming haMVs (Fig. 2C and D). However, there were also individual alpha-actin positive cells dispersed throughout the tissue space (Fig. 2C). CD45+ cell numbers were also elevated in liver constructs made with haMVs (Fig. 2E and F).
Fig. 2NPC composition of liver tissue constructs. Images of sections with and without microvessels stained with (A) UEA-1 (C) α-SMA cell (E) CD45. (B, D and F) Assessments of UEA-1+, α-SMA+, CD45+ cells endothelial cells (C) densities in 14-day liver tissue constructs from confocal images of 5 μm thick histology sections. In all cases, the autofluorescent hepatocytes are shown in green and the specific cell marker is shown in red. Scale Bars indicate 60 μm. ** P ≤ 0.001. All data and graphs shown represent experiments with both 100,000 and 200,000 vessels/ml. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
Increases in stromal cells numbers in haMV-containing constructs undoubtedly reflects the cells added with the haMVs, such as endothelial cells and perivascular cells. However, using twice the starting haMV numbers (from 100K to 200K per ml) did not double the number of stromal cells, as would be expected if the increases in NPC numbers were due solely to haMVs. Moreover, as the haMVs are intact haMVs, mesenchymal cells residing in the perivascular niche such as pericytes and stem cells [
] are also being included in the liver tissue constructs via the no haMVs. These cells are in addition to the endothelial cells intrinsically part of the microvessels and known to improve function and activity in other liver models [
], are contributing to a more stable stromal tissue environment in the liver constructs, possibly preserving cell numbers (hepatocyte and NPCs) in the constructs. Furthermore, the cells of isolated microvessel fragments cultured in 3D produce growth factors and cytokines [
]. The lower LDH levels at the 1st day of static culture with the haMVs is consistent with this. Additionally, the absence of inflammatory cytokines (Figs. 3, 4 and Sup. Fig. 2) suggests low to no inflammation in the tissue constructs. While we didn't specifically assess for M1 vs. M2 myeloid markers, the low level of inflammatory cytokines suggests a relative absence of M1 macrophages. Regardless of the mechanism, the inclusion of haMVs improved the function of the fabricated human liver tissues.
Fig. 3Gene expression assessments in fabricated liver tissue constructs. Real time-PCR was used to measure the relative expression of (A) CYP3A4, (B) MRP2, and (C) ALB, from liver tissue constructs after 14 days in culture without or with a rifampicin challenge. In all cases, expression was normalized to constructs without haMVs (no Mvs) and with vehicle. (E) Agarose gels of corresponding PCR products for CYP3A4, MRP2, and ALB and the inflammation related genes IL-1, and IL-6 with (+) and without (-) haMVs. * P < 0.05. ** P ≤ 0.02. ***P ≤ 0.009. ****P ≤ 0.001. Most data and graphs shown represent experiments with both 100,000 and 200,000 vessels/ml, Rifampicin challenge data represents data with only 100,000 vessels/ml.
Fig. 4Through-tissue perfusion of reformatted liver tissue constructs. (A) Schematic of the perfusion set-up showing the cell culture insert, liver tissue construct, and the fluid tubes of the overlying perfusion manifold. (B) Top view of a phase image of the perfused liver tissue at day 14. Measurements of (C) urea levels, (D) LDH activity, (E) lactate levels, and (F) pyruvate levels over time of tissue supernatants from statically grown and perfused constructs without and with haMVs. *P < 0.001; **P ≤ 0.005; ***P = 0.002; ****P = 0.02; and *****P = 0.007. Additionally, significant differences were observed between day 1 samples as compared to day 7 (p ≤ 0.007) and 14 (P < 0.03). (G) Relative GAPDH expression measured by real time-PCR normalized to SFRS4. All data and graphs shown represent experiments with 200,000 vessels/ml.
Primary hepatocytes: current understanding of the regulation of metabolic enzymes and transporter proteins, and pharmaceutical practice for the use of hepatocytes in metabolism, enzyme induction, transporter, clearance, and hepatotoxicity studies.
]. Indeed, CYP3A4, a member of the P450 system, was upregulated in liver tissue constructs following exposure to rifampicin for 4 days (Fig. 3A). However, this response was only observed in those constructs containing the haMVs. Consistent with other reports, expression of MRP2 and ALB genes was also increased following exposure to rifampicin in constructs made with and without haMVs (Fig. 3B and C). A significant increase was seen in the expression of the drug transporter MRP2 in day 14 cultures containing haMVs, as compared to those without (Fig. 3B). All liver tissues showed baseline expression of ALB, however no difference in ALB expression was observed between tissues with and without haMVs included (Fig. 3C). In baseline constructs, we were not able to detect the expression of the inflammatory cytokines IL-1 and IL-6 regardless of whether haMVs were present or not (Fig. 3D).
It's unclear as to the underlying mechanism(s) of why constructs without the haMVs have low CYP3A4 expression, even following a challenge with rifampicin. We suspect, given the added cellular complexity brought into the liver tissue environment by the haMVs, that the liver constructs lacking the haMVs do not have the proper cell composition to properly promote the P450 system. For example, the cells of the haMV can modulate tissue cytokine dynamics [
]. The haMVs are equivalent to a supplementing further with NPC populations beyond what was already included. In this case, the number of NPCs included without haMVs appears insufficient to support robust P450 expression.
In future studies a native, dynamic microvasculature could be established with the haMVs. It is conceivable to combine fluidics, possibly via the existing channels, with the microvessels in the liver constructs to establish a perfused vascular element to the model. Further work is needed to understand better the requirements for such an approach. We envision these fabricated liver tissues to have utility in a variety of applications including liver disease modelling, drug target identification and screening, and more complex modelling addressing liver tissue dynamics.
Metabolic changes in liver tissue constructs exposed to tissue perfusion
Despite the absence of necrotic zones, urea production declined over the two-week period in static cultures. However, this was not due to continued cell loss as indicated by LDH release, which remained very low during the 2-week cultures following an initial spike in LDH release at day 1. Urea production positively correlates with the metabolic state of hepatocytes [
], suggesting that during the 2 week period the remaining hepatocytes are likely energetically constrained, compromising urea production. We reasoned that reconfiguring the model to accommodate media flow through the liver tissue constructs permits a constant delivery of fresh media, potentially minimizing stagnant diffusion zones and removing this energetic constraint. With this in mind, we reconfigured the vascularized liver tissue constructs in a format for perfusion using off the shelf solutions. The same bioprinting-based, fabrication strategy was used to fabricate liver tissues on top of culture insert membranes. Using a commercial perfusion manifold, media was delivered to the wells (at 10μl/hr) in a manner causing fluid to move through the tissues in the insert for 14 days (Fig. 4A). Culture perfusion did not change the overall gross morphology of the liver tissue constructs (Fig. 4B). As expected, urea production was stimulated with the perfusion, with production levels remaining relatively constant during the entire 2-week period, in contrast to the drop in urea production that occurred in the static constructs (Fig. 4C). LDH levels at day 1 were significantly lower with perfused tissues than with static tissues. Samples without haMVs showed a lower day 1 LDH level than samples with haMVs (Fig. 4D). Interestingly, perfused constructs produced both lactate and pyruvate at a significantly higher rate than statically cultured constructs. In addition, perfused constructs without haMVs produced higher lactate and levels than constructs with haMVs (Fig. 4D and E). This was accompanied by an increase in GAPDH gene expression (Fig. 4G). As expected, no IL-1 and IL-6 expression by perfused liver tissues was detected, demonstrating that adding perfusion into the system did not promote tissue inflammation (Fig. 4H).
As expected, media perfusion significantly increased urea synthesis and maintained steady urea levels over the two-week period. Multiple studies have shown that interstitial levels of fluid flow in both 2D and 3D hepatocyte cultures increases urea production and long-term maintenance of hepatocyte viability [
]. However, lactate and pyruvate generation, and GAPDH expression were increased with a perfusion through the liver tissue in addition to increased urea production. While upregulation of glycolysis in hepatocytes can reflect a hypoxic condition [
], the concomitant upregulation of pyruvate with the lactate increase suggests that perfusion promoted a more metabolically active hepatocyte in the fabricated liver tissues, likely from more glucose availability. Whether this is due to continuous media exchange and/or providing a higher media:tissue ratio, which necessarily occurred in the perfusion setup, is not clear. Interestingly, inclusion of haMVs appeared to influence the metabolic status of the liver tissues in the perfused condition, as well. In perfused cultures, liver constructs with haMVs exhibited a lower ratio of lactate:pyruvate (∼1/3 of that for cultures without haMVs). While lactate metabolism in hepatocytes is complicated [
], it's tempting to speculate that this change in ratio may reflect a switch in metabolism favoring oxidative metabolism over glycolysis or, perhaps, promotion of gluconeogenesis. Clearly, more in-depth studies are needed to determine the basis for these metabolic dynamics in perfused constructs and when haMVs are included in the liver tissue constructs. Regardless, the perfusion findings further suggest that the liver tissue constructs are capable of dynamic changes in hepatocyte metabolism, depending on the model format. Importantly, the fluid flow paths described within these studies involves 3 compartments for potential fluid movement in the perfused version of the liver model: 1) media through the printed channel spaces, 2) media through the membrane support of the culture insert, and 3) liver interstitial spaces. Given the extremely low flow rate and pressure head, fluid movement through the tissue interstitium is likely small to negligible, as diffusion and not convective flow is predominant. The near zero Reynolds number for fluid moving from inside to outside the insert, (Supplemental Material) supports this conclusion. Also, consistent with the extremely low flow rate and near-zero Reynolds number, fluid movement in all 3 compartments is not turbulent.
Automated fabrication of liver tissue constructs
Working from the by-hand fabrication protocol, we developed an automation workflow for the fabrication, culturing, and sampling of the liver tissue constructs. To this end, we employed the robotic arm based BioAssembly® Platform to establish automated protocols for different tasks in the liver tissue construction and use. These protocols were then implemented by the Platform in a single workflow via a BioApp™ (our workflow controls software). For this, the hepatocytes, NPCs, haMVs, and collagen were pre-mixed by-hand and placed into the clean fabrication environment of the robot along with wash buffer and media containers, and a 96 well plate (Sup. Video 1). The environment was closed, the liver fabrication BioApp™ protocol called up, and the workflow started. As with the by-hand procedure, the automated workflow included a fabrication step (3D printing and tissue casting) (Fig. 5B), fugitive pillar wash out, media exchanges (Fig. 5C), incubation (Fig. 5D), and imaging using a high content analysis scanner. Automated media changes occurred daily as part of the workflow with media reservoirs replaced by the user, as needed. At no point during the seven days of culture did the plate of liver tissue constructs get removed or handled by a person. The automation workflow produced liver tissue constructs comparable to those made by hand, the only difference being the pillar location was modified to accommodate robotic pipetting of buffers and media (Fig. 5A and B). Analysis of culture supernatants collected during the 7-day workflow by the robot revealed that liver tissue constructs produced urea and released LDH comparably to those constructs made by hand (Fig. 5D and E). Similarly, the expression of ALB, MRP2, and CYP3A4 in the constructs was not different between the two fabrication methods (Fig. 5F).
Fig. 5Automated manufacturing of human liver tissue constructs. (A and B) Top view gross images of 6 liver tissue constructs fabricated using the automation platform. (B–D) Still images from a video (see Supplemental Material) depicting the automated liver tissue fabrication workflow showing 3D printing fugitive pillars in a 96 well plate, washing and media change activities of liver tissue constructs, and transferring the plate containing the liver tissue constructs to the modular, automation-compatible incubator. (E) Urea levels and (F) LDH activity in tissue supernatants of liver tissue constructs fabricated manually or automatically. (G) Relative CYP3A4, MRP2, and ALB gene expression between manually and automatically fabricated liver tissues. *P < 0.001; **P < 0.05. In all cases, an F-test indicated equal variance between the manual and automated experimental groups. All data and graphs shown represent experiments with 200,000 vessels/ml,
In the automation, the same model fabrication and use protocols performed “by hand” were recapitulated by the platform including pipetting, aspirating, plate movements, etc. The entire 87 step workflow including 3D printing of the fugitive mold, dispensing of the cell/haMV/matrix mix, incubating the tissue constructs in an integrated incubator, washing out the fugitive material, culturing the tissues over a week, daily media exchanges (including collecting culture media into separate plates for analysis), and moving the plate of tissues to the scanner was performed entirely by the robotic platform. Importantly, tissues fabricated by the platform were functionally identical to those made by hand. Therefore, this automation solution can, as demonstrated with the liver tissue model, execute complicated tissue and tissue model fabrications, including those involving 3D printing. Furthermore, because the solution is self-contained, tissue model fabrication and assaying are readily performed in-house without an impact to tissue function and scalable using off-the-shelf, multi-well plates and resources.
Nishimura et al. have demonstrated the automated culture of tissue sheets using a closed culture system in a cell culture incubator. Their system demonstrates the ability to create tissue sheets functionally similar to manual fabrication and includes an automated imaging approach to observe the tissues in culture. The culture medium in this study was exchanged via a sophisticated array of tubing, in contrast the automation within this study which used a robotically controlled pipette tool, similar to what is typically used within manual cell culture. Advantages to the BioAssemblyBot® (BAB) workcell described in this study include lower waste/void volume, lower run preparation time as the tubing arrays do not need to be manipulated for each run, and greater flexibility with experimental design and media/drug delivery as one media bottle source is not feeding the entire culture system. Additionally, our system allows for 3D printing and culture within standard well plate formats while the system described in this study used custom culture inserts to hold their tissues. Finally, while automated imaging is performed in both studies, our study utilized a florescent cell scanner while Nishimura et al. used a phase contrast microscope [
Another example of automated tissue fabrication is Hou et al. who have developed a system to automate organoid fabrication and analysis in a high throughput manner [
]. This system uses off the shelf 384-1536 well plates and is highly efficient in dispensing cells to form spheroids and automating the drug discovery process at high throughput. While a very effective system, the complexity of the tissue that created is limited and the entire automation process uses multiple instruments and robotic platforms to function. The automated liver tissue described in our study is a more complex tissue, involving multiple fabrication steps, than that described by Hou et al. and the BAB workcell is more compact in terms of space as one instrument, the BAB, is performing operational tasks. While the BAB workcell can also perform the tasks necessary for organoid workflows, throughput is modest compared to the platform described by Hou et al.
Conclusion
We demonstrate two innovations for in vitro tissue modelling in the current study. The first is the successful automation of an entire workflow used to fabricate, culture, and image a tissue model involving multiple, varied steps in the protocols. This automation was performed using an all-in-one platform centered around a multi-axis robotic platform and integrated peripherals within reach of the arm. In the demonstrated liver tissue use-case, the cells, reagents, and plastics were loaded into the platform, the automation protocol was initiated, and the tasks were executed automatically, without intervention except for replenishing fresh media and removing collection plates for subsequent assays. The other innovation is the use of isolated, human adipose microvessel fragments (haMVs) to incorporate vascular elements into the tissue space. Others have used endothelial cells with and without mural cells in liver modeling [
], and the non-parenchymal cell preparations we used do contain liver endothelial cells. Interestingly, inclusion of the haMVs improved liver model outcomes as compared to an absence of haMVs, despite the presence of the endothelial cells of the non-parenchymal cell preparation in these non-haMV containing constructs. Additionally, the presence of the haMVs was associated with apparent changes in hepatocyte metabolism when liver tissue cultures were perfused. Whether these differences in the presence of haMVs are due to additional endothelial cells from the added haMVs and/or the other cell types present within each microvessel fragment, including macrophages, is not clear. Regardless, we reason that reconstitution of the tissue space with as many of the stromal cell types as possible is important in capturing relevant tissue biology in vitro and improving model outcomes.
Author contributions
S.M.M, set up, fabricated, and cultured, tissue constructs. H.A.S assisted with long term culture. E.S. performed early prototyping experiments of the liver model design. S.M.M., J.S., M.Y., and M.C. performed transcript and immunostaining analyses. S.M.M., J.S., and B.H. configured and performed the automated fabrication experiments. S.M.M. and J.B.H. conceptualized the project, designed experiments, and interpreted findings. S.M.M. and J.B.H. drafted the manuscript with contributions from authors. J.B.H. coordinated the project.
Data and materials availability
All data associated with this study are present in the paper or the Supplementary Materials.
Supplemental material
Rat vascularized liver tissue construct
To demonstrate flexibility of the assay rodent hepatocytes (Lonza) and rodent NPCs (Lonza) with human adipose microvessels (haMVs, Advanced Solutions Life Sciences) were used. In the case of static cultures, the same fabrication strategy as described for human liver constructs was used with hepatocytes seeded at 10M cells/ml and NPCs at 1M cells/ml. Constructs were cultured for 1 week prior to the drug treatment. On day 8 of the culture, acetaminophen (Sigma) was resuspended in hepatocyte maintenance media (FisherSci) at 30mM, 3mM and 0.3mM for high, media and low does, respectively and added to the culture supernatants. Dextromethorphan (Alpha Aesar) was also resuspended in hepatocyte maintenance media (FisherSci) and 100μM, 10μM, and 1μM for high, medium, and low doses, respectively and added to the static cultures. Cultures were treated for an additional 1 week with the respective drugs.
Assay Flexibility
Using the same fabrication strategy for human constructs, primary rat hepatocytes were used to create static liver constructs in a 6 well plate to demonstrate the ability to scale the size of the tissue constructs (Supplemental Figure). In this format, more pillar spaces were created to provide more opportunities for media-tissue exchange. Interestingly, in some vascularized rat liver tissues fabricated in 96 well plates, lines of packed hepatocytes spontaneously developed between pillar spaces within 7 days (Supplemental Figure). Additionally, exposure of the rat liver constructs to acetaminophen resulted in significant darkening of the constructs in a dose-dependent manner (Supplemental Figure). Furthermore, the ability to create different pillar configurations, including a central horizontal channel, while maintaining tissue integrity was explored (Supplemental Figure).
Horizontal Perfusion Format
An alternate perfusion strategy was added to the cultures using a commercially available 12 well perfusion manifold (PerFuseIt™, Advanced Solutions Life Sciences). A channel pattern with inlet and outlet ports (Supplemental Figure) was printed followed by fitting the manifold to the top of the printed structure such that drop down tubes from the manifold aligned with the ports. Hepatocytes/NPCs/collagen were mixed at 10million hepatocytes/ml and 1million NPCs/ml and cast around the printed structures and the manifold tubes. The tissue was incubated for 1 hr at 37°C followed by the addition of 1ml of media to the top of the tissue. Media was changes 2X to remove the channel material. After 24 hours, media was perfused through channel system via the manifold at 20μl/hr. Cultures were perfused for 1 week without drug treatments. On day 7 low and medium, does of acetaminophen were added to the perfusate and tissues were cultured for an additional week and gross images taken of the constructs (Supplemental Figure).
Supplemental Figure Legend. Example configurations of mesoscale liver tissue constructs using rat primary hepatocytes, NPCs, and microvessels. A) Gross top view of fabricated rat liver tissue in a 96 well format at day 14. Lines of cells spontaneously developed between open spaces (arrows). B) Gross images of rat liver tissues in 96 well plates treated with different doses of acetaminophen and dextromethorphan. C) Gross top view of fabricated rat liver tissue in a 12 well format treated with acetaminophen for 1 week. D) 3D digital model used to print a horizontal arrangement of channels with access ports with fugitive material. E) Phase images of the liver tissue formed around the channels after washing out the fugitive material.
Sup. Fig. 2: A&B. H&E sections of liver tissue C. ELISA results of conditioned media for IL2.
Sup. Fig. 3: 3D design used to print sacrificial pillars. Individual pillars are 400µm in diameter and 5mm in height. A. Pillars position relative to each other. B. pillars centered within a 96 well plate.
Supplemental Video Legend. Video (3X speed) of different aspects of the automated fabrication process.
CRediT authorship contribution statement
Sarah M. Moss: Writing – original draft, Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Writing – review & editing. Jillian Schilp: Data curation. Maya Yaakov: Data curation. Madison Cook: Data curation. Erik Schuschke: Conceptualization. Brandon Hanke: Methodology, Software. Hannah A. Strobel: Investigation. James B. Hoying: Conceptualization, Methodology, Formal analysis, Writing – original draft, Funding acquisition.
Declaration of Competing Interest
Advanced Solutions Life Sciences manufactures and sells the Angiomics®haMVs, the BioAssemblyBot®, the BioStorageBot™, and the PerFuseIt™ manifolds used in the study. J.B.H. is an equity holder in Advanced Solutions Life Sciences. Other authors don't have any competing interest.
Acknowledgments
Research was sponsored by the Office of the Secretary of Defense and was accomplished under Agreement Number W911NF-17-3-003. We thank the Center of Integrated Biomedical and Bioengineering Research (CIBBR) at the University of New Hampshire, supported via the Center of Biomedical Research Excellence (COBRE) program through the National Institutes of Health (NIH), for use of their imaging resources.
Recent advances in 2D and 3D in vitro systems using primary hepatocytes, alternative hepatocyte sources and non-parenchymal liver cells and their use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.
Microvessel-like structures from outgrowth endothelial cells from human peripheral blood in 2-dimensional and 3-dimensional co-cultures with osteoblastic lineage cells.
Primary hepatocytes: current understanding of the regulation of metabolic enzymes and transporter proteins, and pharmaceutical practice for the use of hepatocytes in metabolism, enzyme induction, transporter, clearance, and hepatotoxicity studies.
Angiogenic potential of microvessel fragments is independent of the tissue of origin and can be influenced by the cellular composition of the implants.
13680-4&id=gr1.jpg){kind=link}
13680-4&id=gr2.jpg){kind=link}
13680-4&id=gr3.jpg){kind=link}
13680-4&id=gr4.jpg){kind=link}
13680-4&id=gr5.jpg){kind=link}