1. Introduction
The toxicological testing of drugs and chemicals using in vivo animal models is required for pre-clinical phases of drug development [
1] and in chemical safety evaluations [
2,
3,
4]. The challenges of translating in vivo animal data to human health decisions are well documented [
5], and the legislative encouragement to reduce or eliminate animal testing is increasing in both the European Union [
4,
6] and the United States [
7]. Consequently, the scientific community and regulatory bodies have embraced ambitious goals to develop so-called “new approach methods” (NAMs) for the efficacy and safety evaluation of chemicals and drugs [
8,
9,
10,
11,
12].
NAMs span a wide range of in vitro, in chemico, and in silico methods. In vitro methods include both traditional static 2D cell cultures and more complex models, such as microtissues (e.g., spheroids and organoids) and microphysiological systems (MPS). MPS have been in active development over the past decade, stimulated by major investments from government agencies, private donors, and investors [
13,
14,
15]. Many MPS, both academic and commercial, have been developed for a wide range of tissues [
16], as well as two- [
17] and multi-tissue models [
18,
19]. With the accelerated pace of MPS development, there has been an increase in interest in their qualification and use in regulatory and safety evaluations of chemicals and drugs [
9,
14,
20]. However, a number of concerns regarding the implementation of MPS in safety and efficacy testing pipelines have been expressed by many end users [
21,
22], leading to a concerted effort by the manufacturers, end users, and regulators to inform the scientific community of how these models should be tested to demonstrate robustness, reproducibility, and utility for certain decision contexts [
11,
23,
24,
25]. Furthermore, several recent studies have provided clarity on the anticipated utility of liver MPS by investigating their performance with different cell sources [
26,
27,
28]. These studies have also compared MPS to simpler culture methods, quantitatively assessing both the variability within and between studies. The aim of these investigations was to inform potential end users about the expectations, advantages, and challenges associated with MPS.
The liver is one of the organs where the development of alternative models capable of addressing contexts of use currently unmet by simpler models is of high interest [
29], and many diverse liver MPS have been published [
30]. The lack of consistency surrounding the characterization of the functionality of each of these models has been documented; suggestions on how these models should be tested before wide adoption have been offered by the end users [
31]. Both academic publications reporting the development of new MPS and technical notes provided by MPS vendors accompanying their products offer some functional data and information on drug effects/metabolism data for liver MPS. However, there has been limited independent testing conducted to address the concerns regarding the robustness and reproducibility of these models. Additionally, there is a need to determine the most appropriate decision contexts and conditions of use for these MPS [
20]. It is noteworthy that recent publications by the manufacturers of several liver MPS included data that shows performance across experiments and donors for several dozen compounds [
32] or compared prediction accuracy for liver injury in different study designs (i.e., one replicate per drug vs. a dose–response) for over a hundred drugs [
33]. Still, independent testing of several liver MPS to establish their functionality, reproducibility, robustness, and reliability not only adds useful information on what cell types and combinations may be most functional but also provides comprehensive data on the expected variability in these complex models and the challenges that end users may face [
26,
27,
28].
One of the established perfused liver MPS models is the PhysioMimix LC12, a 12-well device comprised of primary hepatocytes, with or without nonparenchymal cells (NPCs), seeded into 3D perfusable scaffolds to form liver sinusoid-like structures, with circulating cell culture medium that continuously re-oxygenate the cells. The current technology, based on the LiverChip design developed by the Griffith lab over 10 years ago [
34,
35], has been used in a number of toxicological and toxicokinetic studies [
36,
37,
38,
39,
40]. The current commercial version of this MPS, PhysioMimix LC12, has been collaboratively tested by the manufacturer and the US Food and Drug Administration [
41] and also independently at the Roche Innovation Center Basel [
42]. While previous studies provide robust and diverse data to assess the suitability of this model for drug pharmacokinetics and hepatotoxicity studies, each published report included a limited number of replicate experiments. As a result, quantitation of inter- and intra-experimental variability is difficult. Moreover, these studies solely utilized primary human hepatocytes (PHHs) with minimal or no inclusion of physiologically proportional levels of non-parenchymal cells (e.g., Kupffer cells) and did not typically include comparisons with other liver MPS models or conventional 2D systems.
To add to the previous body of literature regarding the performance of liver MPS and to provide critical new data on the robustness and reproducibility of the PhysioMimix LC12 for toxicological studies of the liver, we conducted multiple replicate studies lasting an average of 14 days in culture. These studies used either PHHs or human induced pluripotent stem cell (iPSC)-derived hepatocytes, with and without combinations of NPCs. We evaluated both basal function and metabolic capacity and characterized the model performance with various cell combinations, under intra- and inter-experimental replication, and compared them to multi-well plates and other MPS. Additionally, we conducted power analysis to determine the minimum number of independent replicates required to show statistical significance for changes in specific biomarkers related to hepatic function, drug metabolism, and immune-mediated hepatotoxicity.
2. Materials and Methods
2.1. Overall Experimental Design
This study consisted of multiple rounds of experiments that probed the performance of different lots of PHHs and iPSC-derived hepatocytes, cell combinations, and toxicological contexts of use. Various combinations of cell types were seeded into the PhysioMimix LC12 and cultured for a duration of 8–28 days. Typical experimental designs for different contexts of use are shown in
Figure 1. Exact experimental designs, detailed study protocols, and all experimental data are available (see
Supplemental Table S1 for hyperlinks). After cells were seeded, media changes were performed every 1 to 3 days, with a collection of effluent for biochemical assays to assess biomarkers of liver cell function. The latter was examined using albumin and urea secretion as markers of hepatic function and LDH activity as a marker of cell viability. CYP3A4 activity was measured using a live, luminescent assay with Luciferin IPA as a marker for metabolic capacity.
These experiments examined various utilization scenarios, including the exploration of diverse combinations of cell types (
Supplemental Table S2). The testing of various combinations of cell types was used to determine if immortalized cell lines may be sufficient as NPCs and to assess potential differences in hepatic function between monocultured and co-cultured PHHs and iPSC-derived hepatocytes within the PhysioMimix LC12. Two main types of human liver parenchymal cells were used: PHHs (from 5 different donors) and iPSC-derived hepatocytes. Both monoculture and co-culture (with NPCs) conditions were tested. For co-culture with PHHs, either (i) THP-1 cells differentiated with phorbol 12-myristate 13-acetate into macrophage-like cells (referred to herein as THP-1 cells); (ii) Kupffer cells; (iii) Kupffer cells and hepatic stellate cells (HSC); or (iv) Kupffer cells, HSC, and liver sinusoidal endothelial cells (LSEC) were seeded. For the co-culture of iPSC-derived hepatocytes, cells were co-seeded with THP-1 and HMEC-1 cells.
Metabolic capacity was assessed across the various cell type conditions, including both types of hepatocytes. Midazolam (MDZ) was tested as a representative substrate for CYP3A4 [
43]. Cells were treated with MDZ for up to 24 h on different days of culture, with sampling at 2, 6, and 24 h after treatment. The collected media was processed using high-performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS) to determine the concentration of both the remaining parent compound (MDZ) and metabolites formed (1′-OH MDZ and 1′-OH MDZ glucuronide). We also tested the metabolism of equimolar mixtures (1 or 5 µM) of 20 pesticides that were previously used to evaluate the metabolic competence of different in vitro liver models [
44]. Cells were treated on day 6 of culture, and media were collected for analysis 48 h thereafter. Samples were analyzed using either gas chromatography (GC)-MS/MS or HPLC-MS/MS to determine whether parent molecule concentrations diminished, consistent with the metabolic activity of the cells.
For each optimized cell type combination, the variability and reproducibility of the functional and metabolic parameters were evaluated via repeated studies. Both intra- and inter-experimental variability was evaluated across time in culture. The coefficient of variation was calculated and compared to the benchmark of 30% or less proposed for liver MPS [
31]. Power analyses were performed to determine the minimum number of replicates required to achieve statistical differences in hepatic function and metabolism over time.
2.2. Chemicals and Materials
Midazolam (19391, CAS#59467-70-8) and 1-hydroxymidazolam (10385, CAS#59468-90-5) were purchased from Cayman Chemical (Ann Arbor, MI, USA); 1-hydroxymidazolam beta-D-glucuronide (CAS#81256-81-7) was kindly provided by Sanofi (Paris, France). Lipopolysaccharide (LPS, Cat#L2630) and Trovafloxacin mesylate (Cat# PZ0015) were purchased from Millipore Sigma (St. Louis, MO, USA). The twenty pesticides that were used in this study as analytes or standards (
Supplemental Table S3) were purchased from Millipore Sigma, or Chem Service (West Chester, PA, USA). Analytical chemistry reagents methanol (Cat No.: 646377), acetonitrile (Cat No.: 34998), pentane (Cat No.: 34956), diethyl ether (Cat No.: 309966), and distilled water with 0.1% formic acid (Cat No.: 576913) were purchased from Millipore Sigma.
The microfluidic tissue chips used in this project were purchased from either CNBio (PhysioMimix LC12, Cambridge, UK), Mimetas (OrganoPlate
® 2-lane 96, Leiden, The Netherlands), or Nortis Bio (SCC-001, Bothell, WA, USA). A PhysioMimix LC12 plate has the footprint of a traditional 96- or 384-well plate and contains 12 individual devices. Each chip contains a media reservoir and a culture well consisting of a plastic retaining ring that holds a scaffold and filter in place. Cells were seeded onto and cultured within the scaffold. The plate was attached to a pneumatic docking station, which has micropumps that produce recirculating media flow, which is adjustable using the PhysioMimix LC12 controller. The OrganoPlate
® 2-lane 96 plate contains 96 devices, each of which contains one gel channel and one perfusion channel [
33] to enable the culture of a perfused tubule adjacent to the extracellular matrix without a membrane. The “liver acinus MPS” (LAMPS) model was based on the Nortis SCC-001 device that contains a central growth area where cells and extracellular matrix are sequentially layered [
45]. Black-walled, clear-bottom, tissue culture-treated 96-well plates (3603, Corning, Corning, NY, USA) and 384-well plates (3764, Corning) were used for 2D cell culture experiments.
2.3. Cells and Cell Culture Reagents
All cells used in these studies are listed in
Supplemental Table S2. Primary human hepatocytes were obtained from ThermoFisher (HMCPIS, Lots #HU8373 and #HU8300, Waltham, MA, USA), Lonza (HUCPG, Lots #HUM183231 and #HUM182531, Basel, Switzerland), and LifeNet Health (Lot #212782, Virginia Beach, VA, USA). Primary human hepatocyte plating media consisted of William’s E media (A1217601, ThermoFisher) supplemented with thawing and plating supplements (3.6% cocktail A, 5% FBS, 10 nM dexamethasone (CM3000, ThermoFisher) and was used to plate and seed cells. Primary human hepatocyte maintenance media consisted of William’s E media with primary hepatocyte maintenance supplements (4% cocktail B, 1 nM dexamethasone; CM4000, ThermoFisher) and was used for cell culture from day 1 until the end of each experiment.
Human iPSC-derived hepatocytes were purchased from FujiFilm-Cellular Dynamics International (C1023, iCell Hepatocytes 2.0, Lot# 103934, Santa Ana, CA, USA). iPSC-derived hepatocyte plating media consisted of RPMI 1640 (11835030, ThermoFisher) supplemented with 2% B-27 supplement (17504044, ThermoFisher), 2% iCell Hepatocytes 2.0 Medium Supplement (M1024, FujiFilm-Cellular Dynamics International), 100 nM dexamethasone (265005, Millipore Sigma), 25 µg/mL gentamicin (15710072, ThermoFisher), and 20 ng/mL oncostatin M (295-OM-010, R&D Systems, Minneapolis, MN, USA), and was used for pre-differentiation and plating of cells according to the manufacturer’s protocol. iPSC-derived hepatocyte maintenance media consisted of RPMI 1640, 2% B-27, 2% iCell Hepatocytes 2.0 Medium Supplement, 100 nM dexamethasone, and 25 μg/mL gentamicin, and was used for cell culture from day 1 until the end of each experiment. Primary human Kupffer cells (NPC-AD-KC, Lot #2118082) and primary human stellate cells (HSC, NPC-AD-SC-P0, Lot #2118082p0) were obtained from LifeNet Health (Virginia Beach, VA). Primary human liver sinusoidal endothelial cells (LSEC, ACBRI 566, Lot #566.01.01.01.1T) were obtained from Cell Systems (Kirkland, WA, USA). Kupffer cells were cultured in Advanced DMEM (12491015, ThermoFisher), 5% fetal bovine serum, and 3.6% Cocktail A (CM3000, ThermoFisher). HSCs were cultured in DMEM (11965092, ThermoFisher), 10% fetal bovine serum, and 1% antibiotic/antimycotic cocktail (15240062, ThermoFisher). LSEC were cultured in Complete Classic Medium with Serum and CultureBoost™ (4Z0-500, Cell Systems) with 1% CultureBoost, and 0.2% Bac-Off® supplements (4Z0-643, Cell Systems). HSC and LSEC purity were verified to be >95% for both cell types (data not shown) using fluorescence-activated cell sorting using anti-human α-SMA (NBP2-34522AF488, Novus Biologicals, Centennial, CO, USA) and CD-31 (ab215912, Abcam, Cambridge, UK) primary antibodies. Samples were analyzed using a Cytek Amnis CellStream Flow Cytometer (Cytek Biosciences, Fremont, CA, USA), and excitation lasers (488 and 647 nm) were used to image single cells. A-SMA (488 nm) and CD-31 (642 nm) channels were used for data collection. Color compensation was not necessary as no overlapping emission spectra were observed.
THP-1 monocytes and HMEC-1 endothelial cells were obtained from ATCC (Manassas, VA, USA). THP-1 monocytes were cultured in RPMI (30-2001, ATCC) with 10% FBS (30-2020, ATCC), 1% Pen-strep (15140148, ThermoFisher), and 2 mM L-glutamine (25030149, ThermoFisher). THP-1 monocytes were differentiated [
45] into mature, adherent macrophages by treatment with 200 ng/mL phorbol 12-myristate-13-acetate (356150050, ThermoFisher) for 48 h before seeding. HMEC-1s were cultured in MCDB 131 medium (10372019, ThermoFisher) with 10% FBS, 2 mM L-glutamine, 100 units/mL penicillin-streptomycin, 1 µg/mL hydrocortisone (H0888, Millipore Sigma), and 10 ng/mL epidermal growth factor recombinant human protein (PHG0314, ThermoFisher).
2.4. PHH and NPC Culture Methods in the PhysioMimix LC12 and 96-Well Plates
The day when PHHs were seeded into PhysioMimix LC12 plates was defined as day 0 (
Figure 1). PHHs were cultured in the PhysioMimix LC12 following the manufacturer’s protocols. PHHs were thawed into CHRM (CM7000, Thermo Fisher), centrifuged at 100×
g for 10 min, resuspended in plating media, and then seeded into the PhysioMimix LC12 scaffold at a seeding density of 475,000–600,000 cells/chip. The flow was initiated at 1 µL/s.
For co-culture conditions, THP-1 monocytes and HMEC-1 endothelial cells were seeded either on day 0 or 1 at a seeding density of 60,000–160,000 cells/chip or 60,000 cells/chip, respectively. Primary KCs, HSCs, and LSECs were seeded either on day 0 or 1 at a seeding density of 100,000 cells/chip, 35,000 cells/chip, and 70,000–140,000 cells/chip, respectively. Cell seeding ratios were determined based on previous experiments [
26,
27] or reports of the physiological cell ratios in the human liver [
30]. Media were collected and exchanged every 1–3 days by aspirating and replacing media from the media reservoir (1.8 mL/chip).
For the PHH 2D model, 96-well plates were first coated with 200 μg/mL collagen I (CB354249, Corning Inc.) and 50 μg/mL fibronectin (F1141, Sigma-Aldrich, St. Louis, MO, USA) and incubated at 37 °C, 5% CO2 for 1 h. Thawed PHHs were collected by centrifugation (100× g, 10 min), resuspended in plating media, and seeded at a density of 50,000–60,000 cells/well. After incubation overnight, the media were changed to maintenance media. The total well volume was 100 μL, and plates were incubated at 37 °C, 5% CO2, with media refreshed every 1–2 days.
For evaluation of MDZ metabolism, cells were exposed on days 4–16 of culture to 5 µM MDZ (dissolved in methanol for the final concentration of 0.16%
v/
v). After drug addition, media were collected every 2, 6, and 24 h. After 24 h, the drug treatment media were replaced with fresh maintenance media. For the experiment with pesticide mixtures, we selected chemicals from the Agency for Toxic Substances and Disease Registry Substance Priority List, which contains compounds that are commonly detected at Superfund sites and are known to be hazardous to human health [
46]. A molar-equivalent mixture was created by combining all 20 pesticides and diluting them to a final concentration of 1 or 5 μM each in dimethyl sulfoxide (DMSO, final concentration 0.5%
v/
v). Media was collected 48 h after treatment for the analyses detailed below. For evaluation of immune-mediated toxicity with Trovafloxacin and LPS treatment, PHHs + THP-1 monocytes or PHHs + KCs were exposed on day 6 of culture to 25 or 100 µM of Trovafloxacin, with or without LPS. After 48 h, the drug treatment media were collected for biochemical assays detailed below.
2.5. iPSC-Derived Hepatocytes and NPC Culture Methods in the PhysioMimix LC12 and 96-Well Plates
Before seeding into the PhysioMimix LC12 platform, iPSC-derived hepatocytes were pre-differentiated according to the manufacturer’s protocols as follows. iPSC-derived hepatocytes were thawed, cells counted (average viability 85.1 ± 5.5%), and seeded at a density of 2.5 × 10
6 viable cells/well on a 6-well plate pre-coated with type 1 collagen (657950-005, Greiner Bio-One North America, Monroe, NC, USA) in plating media and cultured for 4 h, after which unattached cells were removed with a media change. iPSC-derived hepatocyte cells were pre-differentiated for 5 days with daily changes in plating media [
26]. The differentiated clusters were collected by centrifugation (200×
g, 3 min) and resuspended in maintenance media, and then seeded into the PhysioMimix LC12 scaffold at a seeding density of 900,000 cells/chip. The day when the iPSC-derived hepatocytes were seeded into PhysioMimix LC12 plates was defined as day 0 (
Figure 1). After incubation for 1 h, flow was initiated (1 µL/s). For co-culture conditions, THP-1 monocytes and HMEC-1 endothelial cells were seeded either on day 0 or 1 at a seeding density of 90,000 cells/chip based on previous studies [
33]. Media were collected and exchanged every 1–3 days by aspirating and replacing media from the media reservoir (1.8 mL/chip).
For the iPSC-derived hepatocytes 2D culture experiments, a 384-well plate was coated with 200 μg/mL collagen I (CB354249, Corning Inc.) and 50 μg/mL fibronectin (F1141, Sigma-Aldrich, St. Louis, MO, USA) and incubated at 37 °C, 5% CO2 for 1 h. Differentiated clusters were collected as previously described, resuspended in maintenance media, and seeded at a density of 100,000 cells/well. The total well volume was 100 μL, and plates were incubated at 37 °C, 5% CO2, with media refreshed every 1–2 days.
2.6. iPSC-Derived Hepatocytes Culture Methods in the OrganoPlate® 2-Lane 96
iPSC-derived hepatocytes were cultured in the OrganoPlate
® 2-lane 96 (Mimetas, Leiden, The Netherlands) using a previously established protocol [
26]. Day 0 was defined when cells were seeded into OrganoPlate
® 2-lane 96 plates. Differentiated clusters were collected as previously described and resuspended into 3.33 mg/mL collagen (Cultrex 3-D Culture Matrix Rat Collagen-I, 3447-020-01, R&D Systems; 5 mg/mL type 1 collagen, 1 M HEPES, 37 g/L sodium bicarbonate at a ratio of 4:1:1, respectively) at a density of approximately 8.0 × 10
6 cells/mL. The iPSC-derived hepatocyte/collagen suspension (2.5 μL/device) was injected into the inlet of the gel channel in each of the 96 devices on the plate using an electronic pipettor. The plate was then incubated at 37 °C, 5% CO
2, for 15 min to allow for collagen polymerization. For co-culture conditions, a mixture of HMEC-1 endothelial cells at 40 × 10
6 cells/mL and THP-1 monocytes at 3 × 10
6 cells/mL was prepared in iPSC-derived hepatocyte maintenance media, and 2.5 μL was injected into the inlets of the perfusion channel of each device using an electronic pipettor. The plates were incubated at a 70° angle at 37 °C, 5% CO
2 to allow attachment of the iPSC-derived hepatocytes in the gel channel. After a 15 min incubation, 50 µL of iPSC-derived hepatocyte maintenance media was added into medium inlets and outlets of the perfusion channel, and the plates were incubated at a 70° angle at 37 °C, 5% CO
2 for an additional 45 min. The plates were then placed on the perfusion rocker platform (Mimetas) set to cycle every 4 min to a maximum angle of approximately 15° to induce gravity-driven media to flow through the perfusion channel. The media was collected and exchanged every 1–2 days (
Figure 1) by aspirating and replacing media from medium inlets and outlets (50 µL each).
2.7. PHH Culture Methods in a Human Microfluidic four-Cell LAMPS Model
PHHs were cultured in microfluidic tissue chips (SCC-001, Nortis Bio) using a previously established protocol [
27]. Device chambers were coated with a mixed solution of 50 μg/mL fibronectin and 200 μg/mL collagen I in PBS and incubated for 1 h at room temperature. Once the devices were coated, PHH (TU-HU8373) was injected at a density of 2.75 × 10
6 cells/mL (150 μL/chip) in PHH plating media (see above) and incubated overnight (37 °C, 5% CO
2). The following morning, a 2.5 mg/mL solution of pH 7.2 rat tail collagen/10 mM HEPES/HBSS was injected into the devices (150 μL/chip). Chips were then inverted and incubated for 1 h at 37 °C. After this short incubation, chips were re-inverted and incubated overnight at 37 °C.
The next morning, device tubing was prepared as follows: inlet tubing was prepared by cutting 10-inch sections of polyetheretherketone (PEEK) tubing (1569XL, IDEX, Lake Forest, IL, USA), and fitting 1 inch of c-flex tubing (06422–00, Cole-Parmer, Vernon Hills, IL, USA) to each end with a catheter blunt (SC 20/15, Instech, Plymouth Meeting, PA, USA) on one end to interface with the inlet of the chip (the other c-flex section would fit over a needle blunt on the media syringe). Outlet tubing was prepared by cutting 2-inch sections of PEEK tubing and fitting a 1-inch section of c-flex tubing to one end with a catheter blunt to interface with the outlet of the chip. The other end of the PEEK tubing was placed into a glass vial with a pre-slit cap. All tubing was autoclaved and primed with cell culture media prior to chip connection. PHH maintenance media was perfused using syringe pumps (Fusion 200, Chemyx, Stafford, TX, USA) at a rate of 15 μL/h for 9 days. Each day, 360 μL of effluent was collected for analysis.
2.8. Biochemical Assays
Cell culture media was collected every time media was exchanged (
Figure 1) and stored frozen at −80 °C until analyses for various biomarkers, including albumin, urea, lactate dehydrogenase (LDH), and human interleukin-6 (IL-6). The ELISA assays for albumin (E88-129, Bethyl Laboratories, Montgomery, TX, USA), urea (EIABUN, ThermoFisher), LDH (ab102526, Abcam, Cambridge, UK), and IL-6 (D6050, R&D Systems) were conducted using manufacturer’s instructions. For CYP3A4 activity, the P450 Glo 3A4 with Luciferin-IPA assay (V9001, Promega, Madison, WI, USA) was used; this is a live cell assay, and it was conducted according to the manufacturer’s instructions as follows. The Luciferin-IPA substrate was diluted 1:1000 in maintenance media and added to chips or wells 1 h prior to reading luminescence using a plate reader (SpectraMax iD3, Molecular Devices, San Jose, CA, USA). After the assay, the media was replaced.
2.9. Immunocytochemical Staining
At the conclusion of the cell culture period, scaffolds were removed from the PhysioMimix LC12 plate and placed in a 24-well plate containing phosphate-buffered saline (PBS). After washing with PBS twice, scaffolds were fixed by adding 4% formalin solution for 10 min. Scaffolds were washed again with PBS to remove the formalin and treated for 10 min with 0.1% Triton X-100 (BP151, ThermoFisher) in PBS to permeabilize cells. Cells were washed again with PBS and incubated for 60 min in 2% bovine serum albumin (A8806, Millipore-Sigma) in PBS blocking solution. An anti-CYP3A4 polyclonal antibody (18227-1-AP, ThermoFisher), anti-ZO-1 monoclonal antibody (ZO1-1A12, ThermoFisher), anti-Human Albumin antibody (CL2513A, Cedarlane, Burlington, NC, USA), anti-ABCB11/BSEP antibody (ab155421, Abcam), and anti-MRP2 antibody (MAB4150, Sigma Aldrich) were added to label CYP3A4 metabolic enzymes, tight junctions, albumin proteins, and transporter proteins, respectively. Plates were incubated overnight at 4 °C and goat anti-Mouse IgG secondary antibody (ab150115, AF647, Abcam) and goat anti-Rabbit IgG secondary antibody (ab150079, AF647, Abcam) were added to visualize CYP3A4, ZO-1, BSEP, Albumin, and MRP2. Plates were incubated for 1.5 h at room temperature. All scaffolds were counterstained with FITC-phalloidin (F432, ThermoFisher) to visualize the cytoskeleton and incubated for 0.5 h at room temperature. Scaffolds were then mounted on a glass slide using ProLong Gold Antifade mountant with DAPI (P36935, ThermoFisher), which counterstained cells with DAPI to visualize nuclei. Cells were imaged using the ImageXpress Micro (Molecular Devices).
2.10. Sample Preparation for the Analytical Chemistry Analyses
After treatment with MDZ or pesticide mixtures (
Supplemental Table S3), 50 µL media samples were mixed with 100 µL of 0.1 or 1 µM internal standard (Caffeine-13C3, C-082, Millipore Sigma;
Supplemental Table S3) solution in acetonitrile on ice, respectively. Samples were vortexed and centrifuged (12,000×
g, 10 min, room temperature) to allow for protein precipitation. The supernatant was collected from each sample, dried under vacuum, and reconstituted in 50 µL aqueous mobile phase. The reconstituted samples were transferred to glass autosampler vials with 200 µL inserts and stored at −20 °C prior to HPLC-MS/MS analysis. After treatment with pesticide mixtures, 50 μL media samples were spiked with 10 μL of 10 μM internal standards, mixed with 50 μL of methanol and 200 μL of pentane: diethyl ether (1:1,
v/
v), vortexed, and then centrifuged at 600×
g for 5 min. Supernatants were transferred to an amber vial for GC-MS/MS analysis.
2.11. Determination of Midazolam and Metabolites and Pesticide Metabolism Using HPLC-MS/MS
For the determination of MDZ and its metabolites using HPLC-MS/MS, 8 µL of each sample was automatically injected and chromatographed on a ZORBAX SSHD Eclipse Plus C18 column (3.0 × 50 mm, 1.8 µm, 959757-302, Agilent, Santa Clara, CA, USA) using a 1290 Infinity II LC (Agilent) with a guard column (2.1 × 5 mm, 1.8 µm, 821725-901, Agilent). Column temperature was set at 40 °C and flow rate to 0.4 mL/min. Chromatographic conditions started at 90% mobile phase A (water with 0.1% (v/v) formic acid) and 10% mobile phase B (acetonitrile with 0.1% (v/v) formic acid) for 1 min, increased to 80% B by 3 min, to 95% B by 4 min, and then returned to and stayed at initial conditions at 5 to 8 min for equilibration prior to the next run. MS/MS analyses were performed using an Agilent 6470 triple quadrupole MS (Agilent) in positive ion mode with an electrospray ionization source. Capillary voltage, sheath gas pressure, and sheath gas temperature were set at 3500 V, 40 psi, and 350 °C, respectively.
For the determination of pesticide metabolism, chromatographic conditions were as previously reported in [
47], and HPLC-MS/MS analysis was performed using the instrument and columns mentioned above. Analytical response was acquired in positive or negative ion modes. Chromatographic conditions started at 98% mobile phase A (water with 0.1% (
v/
v) formic acid) and 2% mobile phase B (methanol with 0.1% (
v/
v) formic acid) for 1 min, increased to 80% B by 3 min, to 95% B by 4 min, and then returned to initial conditions at 5 min and held for an additional 3 min for a total run time of 8 min per sample at a flow rate of 0.4 mL/min. For negative ion compounds, the LC gradient and flow rate were the same as in positive mode, except that mobile phase A was water and mobile phase B was acetonitrile.
2.12. Determination of Pesticide Metabolism GC-MS/MS
Detection of analytes was achieved using a 7890B GC and 7010B GC/MS triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, CA, USA), as detailed in [
47]. Samples (1 μL) were injected in splitless mode, and the analytes were separated with a VF-5 ms GC column (60 m × 250 μm × 0.25 μm, Agilent Technologies) and ionized with an electron ionization source. The column head pressure was set at 21.5 psi with a constant flow rate of 1.2 mL/min using helium gas. Initial column temperature was held at 70 °C for 5 min, increased to 150 °C at 50 °C/min, ramped to 280 °C at 4 °C/min, and then held for 15 min. The total run time was 42.1 min. The injector temperature was set at 250 °C. The ion source and auxiliary transfer line temperatures were 300 °C. Electron multiplier voltage was set at 1884 V. Nitrogen gas was used as the collision gas for all MS/MS experiments, and the pressure of collision gas was set at 16.8 psi.
2.13. Statistical Analyses
General descriptive statistical analyses were conducted using GraphPad Prism 9.2 (San Diego, CA, USA). Statistical significance (
p < 0.05) was tested with one-way ANOVA with Dunnett’s multiple comparisons test, as indicated in Figure legends. Multi-factor analysis of variance was used to decompose variation in the measurement from different days, experiments, and other factors. Power analyses for sample size needed to reach statistical significance with 80% statistical power were estimated using (i) the two-sided paired
t-tests and (ii) two-sided two-sample
t-tests with different variances for paired and independent samples, respectively. MATLAB (v. R2018a), SAS (v. 9.4), and PASS (v. 15) software packages were used to estimate the effect size and calculate the required sample size [
48,
49].
4. Discussion
A large number of liver MPS models have been developed in an attempt to create more physiological and translatable non-animal approaches that can be used in pre-clinical studies of drug metabolism and toxicology to improve benefit/risk assessment [
29,
30,
52]. The primary decisive factors to model liver tissue have been recently described [
31] and include (i) the creation of models that are multi-cellular in nature, allow for proper zonation, and can be used to study disruptions in bile acid homeostasis, (ii) evaluation of potential liver toxicity (including innate immune-mediated), and (iii) characterization of drug metabolism and pharmacokinetics, especially for slowly biotransformed compounds. Despite the diversity in the offerings of MPS and other more traditional and simpler in vitro liver models [
29,
57], a consensus on what models are useful, robust, reproducible, and fit for at least some regulatory decision making has yet to emerge. There is a clear shift from the use of liver MPS for hypothesis-driven research to their application in the laboratories of the potential end users, namely the pharmaceutical industry [
14]. Therefore, building confidence in the wide adoption of these new models is needed; however, the approaches to how this could be accomplished vary. Some have taken a path to demonstrate their model’s utility via testing of a large number of liver-toxic compounds and establishing “performance” against human liver toxicity data [
32,
33], while others have taken a lab-to-lab comparison approach to focus on reproducibility, robustness, and determination of the potential challenges with model performance to inform future users on the most optimal study designs, conditions of use, and set overall expectations [
27,
41]. Indeed, both approaches are needed and address different considerations for the ultimate human translation and wide adoption of MPS.
The studies detailed herein are of the latter type; we aimed to independently evaluate a relatively mature liver MPS, the PhysioMimix LC12. This perfused scaffold-based model has over a decade-long track record of publications from both the developer lab [
34,
35,
37,
39,
40,
50] and various end users of this technology after it was commercialized [
36,
38,
41,
42]. While we relied on the manufacturer’s protocols, the conditions of use for the PhysioMimix LC12 were adjusted based on the considerations of the members of the TEX-VAL Tissue Chip Testing Consortium, a collaboration that brings together pharmaceutical and consumer products companies, the trade association of chemical manufacturers, and government agencies who collectively work to qualify a number of MPS [
20]. The Consortium aims to not only conduct qualification experiments on a number of commercially available platforms, but also to create a transparent record of the process and make all data available by depositing it into a publicly accessible database [
58].
In addition, the data presented here is directly responsive to the regulatory agencies’ calls for new MPS and other non-animal models [
11,
12,
24,
25]. Building scientific confidence in the regulatory use of new methods is an evolving field that is informed by authoritative bodies such as the 2023 US National Academies Report “Building Confidence in New Evidence Streams for Human Health Risk Assessment: Lessons Learned from Traditional Toxicity Tests” [
59] and the 2005 OECD Guidance Document on the Validation and International Acceptance of New or Updated Test Methods for Hazard Assessment [
60]. Thus, we interpret the findings of our study in the setting of the specific decision contexts and frame the qualification-relevant results around both PECO [relevant human population (P), exposure (E), comparator (C), and outcome (O)] criteria and critical elements of the qualification framework proposed by the National Academies [
59]. With respect to the decision contexts, our studies aimed to characterize the reproducibility, robustness, and human translational relevance of the PhysioMimix LC12 model in the contexts of (i) varying conditions of use (i.e., cell types and culture duration), (ii) drug metabolism and pharmacokinetics, and (iii) immune-mediated toxicity. This MPS model’s PECO statement elements were P = specific types of PHHs, iHeps, and NPCs; E = drugs and chemicals dissolved in media or solvents; C = vehicle-control or pre-treatment conditions; and O = imaging, biochemical, and analytical chemistry readouts. We evaluated this model’s internal validity (systematic error due to study design or execution limitations), external validity (whether the results from tested MPS reflect the expected changes in humans), experimental variability, and biological (between donor) variability. Finally, the transparency of this qualification exercise was supported not only by the data presented herein and in the
Supplemental Files but also by the full access to the protocols and raw data for this large compendium of studies that is freely accessible online (
Supplemental Table S1).
Our study provides important information for establishing the internal validity of this liver MPS, i.e., determining the extent to which systematic error can influence the interpretation of the results. Previous studies have demonstrated that while drug binding to the materials of the PhysioMimix LC12 model is not likely to be of concern [
42], media evaporation during open-system recirculation needs to be taken into consideration during kinetic modeling. A very important internal validity challenge is the seeding efficiency within each scaffold and the resulting differences between scaffolds in one plate or parallel plates. We found that consistency in seeding of the scaffolds is difficult to achieve, especially when co-cultures with multiple NPCs are used. Media movement-induced sheer stress and re-oxygenation of the media are critical design elements that distinguish this MPS from other models or static hepatocyte or spheroid cultures [
34,
61]; thus, maintaining even media flow through the scaffold is critical. Indeed, uncontrollable variability in cell density within and among scaffolds in the PhysioMimix LC12 model is a confounding factor that is a challenge of this type of MPS. It has been estimated that only about 50% of the cells originally seeded (~600,000/scaffold for PHHs) remain within the culture system after 2 days of perfusion [
42]; however, a wide range (39–67%) was reported by that study. Because accurate cell number is needed for calculating kinetic parameters, it was suggested that “sentinel” scaffolds be reserved on each plate to evaluate seeding density via protein measurements [
42], a suggestion that may be difficult to implement if there is substantial variability between scaffolds on the same plate, as reported here and by others [
41]. Imaging of the scaffolds at the end of each experiment is used to visualize tissue morphology inside each scaffold to determine whether cells are retained. However, scaffolds are non-translucent, which precludes one from determining if a large proportion of the cells are located on top of the scaffold rather than inside the channels. Another observation that was made in many of our studies is that air bubbles become trapped under some scaffolds and may impede the stable circulation of the media through the system. Collectively, our finding of the high intra- and inter-experimental variability in basal liver function parameters, even with the same PHH donor cells, should be contemplated by the end users when considering study designs and throughput. A very high demand for the number of cells to be used and the limits on how many plates can be handled simultaneously by each PhysioMimix controller or can be realistically handled by one operator place this MPS (in its current configuration) into a low-throughput category of the available liver in vitro models.
For considerations of establishing external validity or the extent to which the results from this MPS model can be applied (generalized) to the intended population for drug development/safety, we provide liver physiological parameter data for comparisons with other human liver MPS models. Common benchmarks of the physiology of liver in vitro models include the level and sustainability in the production of albumin and urea and the activity of major metabolic enzymes [
31]. Quantitative comparisons to other studies in this liver MPS and to the expected human levels show that this model could achieve basal physiological function at the lower range of the in vivo human parameters. While our study did not account for the cell loss and normalized to the nominal (i.e., seeded) numbers, we found that the levels in the first week of culture are comparable to those reported for this liver MPS by others [
36,
39,
41,
42]. It is worth noting that the amount of albumin and urea produced in the PhysioMimix LC12 was comparable to those obtained in either LAMPS (using the same donor PHH) or in OrganoPlate
® 2-lane 96 (seeded with iPSC-derived hepatocytes). This collectively indicates that PhysioMimix LC12 may not hold a unique advantage in terms of basal synthetic liver function and that maintaining basal function over time is also a challenge.
Previous studies in the PhysioMimix LC12 that included time-course data for albumin and urea production suggested that their production is sustained for 1 or 2 weeks; however, we found that regardless of the donor or co-culture conditions, these biomarkers gradually diminished with time, declining most notably after 2 weeks in culture. This may be due to the differences in PHH donors, but this finding is informative to future users who should consider limiting studies in this model to 14 days. We report similar observations for CYP3A4 activity. However, we found that MDZ metabolism remained high over 14 days and was comparable to other published studies. Because almost all previous studies in the PhysioMimix LC12 focused on investigations of drug metabolism and clearance, our findings suggest that this model may be most suitable for studies of pharmacokinetics over days to weeks, a necessity for compounds with slow metabolism and low clearance that cannot be evaluated in traditional hepatocyte suspension cultures.
Our data show that high experimental variability is to be expected in this MPS. By quantifying variability over many replicate experiments, as well as comparing the variability in different conditions of use, we provide realistic estimates of the replicates that need to be considered for any future study design. While not unexpected because of the complexity of the model, the end users also need to determine what decision context they are pursuing when using the PhysioMimix LC12. For example, the variability in the physiological parameters was far greater than that observed when measuring drug metabolism. However, the typical “3 replicates” study designs may not be sufficient, even for the most reproducible endpoints, because of the difficulties in controlling cell seeding and other technical challenges that may arise during experiments that extend over 2–3 weeks.
Our study also offers information about biological variability. It is well known that both the physiological and metabolic function of primary liver parenchymal cells from different donors varies greatly [
54] and that the performance of various cell lines or iPSC-derived hepatocytes is also far from the liver in vivo [
62]. Suggestions have been made that co-cultures of PHHs with NPCs may improve the performance of in vitro models or that the use of pooled PHHs from different donors may be a sensible strategy. However, the field is still searching for a solution, and various MPS may offer only partial benefits [
63,
64].
In this respect, we found that co-cultures with NPCs in the PhysioMimix LC12 offer only a limited benefit but further increase the complexity and cost of the experiments for the biological response spaces addressed in this study. Specifically, the inclusion of innate immune cells, such as macrophages [
65,
66], or other non-parenchymal cells, such as hepatic stellate cells [
67], has been suggested as a pre-requisite to the utility of liver MPS as models for drug-induced liver injury or cholestasis. While it is without a doubt that co-cultures are necessary to replicate a complex whole liver cell–cell interaction, many MPS are still challenging with respect to model setup and equipment needs, even when hepatocyte monoculture is used. We found that while adding macrophages, either THP-1 or Kupffer cells, may yield some benefits and not erode overall performance, the addition of more complex combinations of NPCs requires additional optimization, and its benefits and reproducibility are yet to be demonstrated. One previous study showed LPS-mediated potentiation of trovafloxacin hepatotoxicity in PhysioMimix LC12, where PHHs were co-cultured with Kupffer cells [
41], albeit the results across laboratories and donors were variable. Our study showed that depending on the phenotyping method, such potentiation could be observed in cytokine release in experiments with Kupffer cells but not THP-1 cells.