Introduction

Hemp extracts derived from Cannabis sativa L. have grown in popularity and are being consumed orally in a variety of ways, including drugs, supplements, or foods. They are being sold in the form of broad-spectrum extracts with multiple phytocannabinoids, or as pure cannabinoids, such as cannabidiol (CBD), cannabinol (CBN), cannibichromene (CBC), and cannabigerol (CBG). Hemp contains over 100 cannabinoids, of which CBD is the most prevalent1. CBD has been studied in clinical trials as an anticonvulsant, analgesic, antioxidant, anti-inflammatory, anxiolytic, and sleep aid2. Other cannabinoids, such as CBN, CBC, and CBG, are also gaining popularity for use in health and wellness products and have been studied for multiple effects such as appetite stimulation3, anxiety reduction4, blood pressure regulation5, and pain relief6. Based on the Congressional Research Service (CRS) statutory definition of hemp, these extracts differ from marijuana based on less than 0.3% delta-9 tetrahydrocannabinol (Δ9-THC) on a dry weight basis7.

The safety of CBD as a pure compound or as a component of hemp extract has been investigated in numerous animal and human studies. The liver is the target organ of CBD toxicity as discovered in both animal8,9,10 and human studies11,12,13,14,15,16,17,18,19,20, but a limited number of studies of CBD found no adverse effects on liver functions. A 14-day and 90-day oral toxicity study of CBD reported no treatment related adverse effects in Sprague Dawley rats at any treatment dose and determined an oral no observed adverse effect level (NOAEL) of 150 and 140 mg/kg-bw/day, respectively21. A placebo-controlled human study in 20 patients with Crohn’s disease found no effects on liver function (based on blood tests for liver enzyme activity) after daily consumption of 20 mg highly purified CBD in olive oil for eight weeks22. In addition, CBD in foods and supplements delivering 50.3 ± 40.7 mg/day and consumed by healthy adults for a minimum of 30 days did not appear to alter the prevalence of elevated liver enzymes compared to the general adult population in the United States23.

However, a purified form of CBD, EPIDIOLEX®, has been evaluated in several double-blind, placebo-controlled, randomized clinical trials for treating seizures from Dravet syndrome12,15,16, Lennox-Gastaut syndrome14,19, and tuberous sclerosis complex (TSC)18. More than a three-fold increase in alanine transaminase (ALT) and aspartate aminotransferase (AST) was seen in these studies at doses starting at 10 mg/kg bw/day in a 14–16-week dosing period. In addition to AST and ALT, two of these studies showed increased gamma-glutamyl transferase (GGT) levels18,19. These elevations in ALT, AST and GGT are indicators of liver toxicity24,25. EPIDIOLEX® at doses of 1500 mg/day (~ 20 mg/kg bw/day) has also been found to increase ALT, GGT, AST, and alkaline phosphatase (ALP) after 2–4 weeks in several healthy subjects17,20. In another study, 4 out of 59 participants experienced elevated liver enzymes (> threefold higher than the upper limit of the normal ranges) after receiving oral CBD (99.6% purity) at 300 mg/day (150 mg twice daily) for 4 weeks11. In toxicological studies of EPIDIOLEX®, increased liver weights and elevated liver enzymes were noted in CD-1 mice, HSd Han Wistar rats, and beagle dogs10. Hepatotoxicity was also reported in animal studies of CBD-rich hemp extracts where increases in liver weight and liver-to-body weight ratio were observed8,9,26 as well as hypertrophy of liver cells around the central vein8. Similarly, in male B6C3F1 mice, elevation of serum ALT, AST, and total bilirubin as well as up-regulation of genes associated with oxidative stress (Hmox1, Nqo1, and Txnrd1) and an increased level of oxidized glutathione in the blood were reported9. Overall, there are several indications that EPIDIOLEX® may cause liver injury, especially at higher doses.

Several regulatory agencies have conducted safety assessments on CBD over the past few years. The Australian Therapeutic Goods Administration (TGA) has authorized low-dose CBD-containing products, up to a maximum of 150 mg/day, for use in adults, to be supplied over the counter through pharmacies. Other regulatory agencies such as Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (UK COT), European Food Safety Authority (EFSA), and U.S. Food and Drug Administration (FDA) have expressed concern for safety data gaps, including liver toxicity, based on potential effects seen with the high dose studies with the CBD drug EPIDOLEX®.

Given the concern for potential hepatotoxicity of CBD, there could also be possible concerns with the safety of other structurally related cannabinoids used in food and supplements. There is little evidence to support safety for oral consumption of these cannabinoids, including CBN, CBG, and CBC, and therefore their potential for hepatotoxicity is unclear. A study of water-soluble CBN in male ICR mice dosed at up to 4 mg/kg bw/day via oral gavage for 14 days did not find statistically significant changes in AST, ALT, bilirubin, or liver weight27. A 14-day study evaluating CBN in MCT oil in Sprague Dawley rats found no difference in liver weight or enzymes between any CBN dose (up to 100 mg/kg bw/day) and control groups28. A human study using a water-soluble CBN dosed up to 4 mg/day for at least seven days for sleep improvement reported no serious adverse events, although liver function tests were not performed29. A 14-day study in Sprague Dawley rats evaluating the safety of CBG found no statistically significant effects related to the liver up to 70 mg/kg bw/day30. A statistically significant increase in ALP for males and females in the high dose group (140 mg/kg bw/day) was observed, although the authors determined this was not toxicologically relevant due to its low magnitude and the values being within normal reference values. A seven-day human clinical study evaluating 100 mg CBG combined with 70 mg CBD for its effects on recovery from delayed onset muscle soreness reported no serious adverse events, although liver function tests were not performed31. A 14-day study administering CBC orally to Sprague Dawley rats found no treatment related adverse effects related to liver weight or enzymes up 100 mg/kg bw/day28. More research is needed to address these safety data gaps for CBD and other cannabinoids consumed in foods and supplements.

Microphysiological systems (MPS), also known as Organ-on-a-Chip technology, are a class of complex in vitro models that simulate the properties of tissue microenvironments. In recent years, MPS are emerging as an alternative to traditional animal testing for chemical toxicity assessment32,33,34,35. These systems allow for an evaluation of toxic responses to organ-specific human cells in an integrated physiological system. The Emulate Quad Culture Liver-Chip S1 recreates the complex liver microenvironment with multicellular architecture creating a tissue-tissue interface, and microfluidic fluid-flow34,36. The predictive validity of the Human Liver-Chip was demonstrated in a study evaluating 27 known hepatotoxic and non-toxic small molecule drugs with a sensitivity of 87% and a specificity of 100%36. The Center for Food Safety and Applied Nutrition (CFSAN), a division of the United States FDA, also evaluated the utility of a Liver-Chip model for hepatotoxicity risk assessment by testing known hepatotoxic compounds (usnic acid, benzbromarone, tamoxifen, and acetaminophen (APAP)) and compounds that have no reported human cases of liver toxicity (dimethyl sulfoxide (DMSO), theophylline, and aminohippurate), and showed that the Liver-Chip accurately predicted toxicity with high sensitivity and specificity while demonstrating low variability between biological replicates as measured through the various readouts32. Thus, Liver-Chips can maintain an in vivo-like phenotype for longer periods than traditional 2D static cultures while having the potential to accurately predict human relevant toxicity in comparison to animal models, thereby reducing the high costs and demand for animal models in accordance with the 3Rs Collective36.

This study aimed to evaluate and compare the hepatoxicity of CBD, CBN, CBG, and CBC using the Human Liver-Chip as an alternative to animal testing and other in vitro models. The morphological effects, hepatic functions, and cell viability were examined upon the exposure to each cannabinoid at three different exposure concentrations. Furthermore, the potential mechanisms of action for hepatotoxicity of the different cannabinoids were also explored by monitoring mitochondria function, inflammatory responses, and oxidative stress in the MPS.

Materials and methods

Liver-chip culture

The Liver-Chips have two independent microfluidic channels separated by a porous membrane. The channels allow for unidirectional flow of cell-specific media, and effluent can be sampled at desired time intervals. Liver-Chips (Emulate Inc., Chip-S1®, Boston, MA) (Fig. 1a) were cultured in this study according to standard Emulate protocols (EP226 Rev. B, Emulate, Inc.)34,36,37. On day −1, Liver-Chips were surface-activated with 0.5 mg/mL ER-1® Activation Reagent (Emulate Inc., 10461, Boston, MA) reconstituted in ER-2® Activation Reagent (Emulate Inc., 10462, Boston, MA) and irradiated with UV light for 15 min. After activation, the membrane was coated with ECM components – 100 µg/mL Collagen Type I (Corning, 354249, Tewksbury, MA) and 25 µg/mL Fibronectin (Thermo Fisher, 33010018, Waltham, MA) in DPBS and stored at 4 °C overnight. On day 0, Liver-Chips were warmed for 1 h at 37 °C. Primary human hepatocytes from one donor (Gibco, Grand Island, NY) were seeded to the top channel at a density of 3.5 × 106 cells/mL. Liver-Chips were incubated for 4 h at 37 °C and 5% CO2, and then gently washed with hepatocyte seeding media to remove any unattached cells and debris and then returned to 37 °C. On day 1 the hepatocytes were overlayed with 2.5% Matrigel™ (Corning, 354234, Tewksbury, MA) in hepatocyte maintenance media and stored in an incubator overnight. On day 2, liver-specific NPCs, including liver sinusoidal endothelial cells (LSECs), Kupffer cells and stellate cells (each of a single source from a different donor), were seeded to the bottom channel. LSECs (Cell Systems, Kirkland, WA) were seeded at a density of 3 × 106 cells/mL, Kupffer cells (Samsara, San Diego, CA) were seeded at 2 × 106 cells/mL, and stellate cells (iXCells, San Diego, CA) were seeded at 0.1 × 106 cells/mL. LSECs were expanded four days prior to seeding in a T-75 flask pre-treated with Attachment Factor™ (Cell Systems, 4Z3-500, Kirkland, WA), whereas Kupffer and stellate cells were seeded directly from the vial after thawing. NPC seeding media was the same as NPC maintenance media, except supplemented with 10% fetal bovine serum. Both channels were gently washed with cell-specific media 4 h after seeding and left to incubate overnight at 37 °C and 5% CO2. On day 3 the Liver-Chips were connected to Pods™ (Fig. 1b) filled with 4 mL respective media and placed in the Zoë (Emulate, Boston, MA), to initiate media flow over the cells37. The Liver-Chips were maintained for 3 days under flow at 30 µL/h in both channels (Fig. 1c).

Fig. 1
figure 1

Study design and test system to assess cannabinoid hepatotoxicity. a) Liver-Chip schematic showing the PDMS-based Chip containing two independent channels – the top channel (1) with hepatocytes (3) and the bottom channel (8) with non-parenchymal cells (NPCs) such as liver sinusoidal endothelial cells (7), Kupffer cells (6) and stellate cells (5). The channels are separated by a porous, ECM-coated membrane (4). b) The Pod contains four independent reservoirs that either provide fresh media to the Chips (inlet) or collect media that has perfused over the Chips (outlet). Each inlet or outlet correspond with either the top or bottom channel. Media flow from the inlet to the outlet per channel is modeled. c) Liver-Chips were established by seeding hepatocytes to the top channel and NPCs to the bottom channel two days later. After allowing the hepatocytes 6 days to mature on the Chip, dosing began on day 0. Effluent collection from Liver-Chip was on days 1, 3, 5 and 7 post-dosing. This effluent was used for assays such as albumin, LDH, ALT, AST, and cytokines. At the end of the study (day 7 post-treatment), all the Chips were stained and live imaged for oxidative stress and mitochondrial function.

Full size image

Compound treatment

Forty-eight Liver-Chips were used in this study to investigate the toxicity of CBD, CBN, CBG and CBC in comparison to the positive control acetaminophen (APAP). 5 M Acetaminophen (APAP, Millipore Sigma, St. Louis, MO) was prepared in DMSO. CBC (KND Labs, CBC-009, Lakewood, CO) was received as a 1 g/mL oil, and diluted to 10 mg/mL in DMSO. CBD, CBN, and CBG (KND Labs, KND-406, CBN-3012, CBG-011, Lakewood CO) were available in powder form and reconstituted to 10 mg/mL in DMSO. All the cannabinoid test concentrations were prepared from their respective 10 mg/mL stock solutions. The vehicle control group was 0.1% DMSO, and APAP was evaluated at 1, 3 or 10 mM in triplicates. CBD, CBG, or CBC were each tested at 0.24 µM, 3 µM, or 4.7 µM in triplicates.

The CBD concentrations selected for this study were based on the known hepatotoxicity of 20 mg/kg bw/day for Epidiolex10 and probable safe dose of 1 mg/kg bw/day evaluated by TGA38. A mid-dose of 10 mg/kg bw/day was added and the desired exposure concentrations of 15, 187, and 293 ng/mL (48, 595, and 932 nM) were selected based on the Cmax of CBD found in the plasma from similar doses of CBD in previous studies39. The concentrations of CBN, CBG, and CBC were matched to these same concentrations calculated for CBD to allow for a direct comparison.

Compound distribution analysis

The Liver-Chip is fabricated with polydimethylsiloxane (PDMS) due to its transparency, elasticity, oxygen permeability, and biocompatibility40. Small hydrophobic molecules can absorb into PDMS, thereby impacting compound bioavailability41. Based on the partition coefficient (logP), polar surface area, and molecular-weight, cannabinoids were determined to have a high risk of compound absorption. Hence, the test articles were evaluated for their potential absorption and adsorption in the abiotic Chip-S1® Stretchable Chip at days 0, 1, 2 and 3 post-dosing (Emulate Inc., EP129 v1.0)42.

CBC and CBN have similar physicochemical properties and identical polar surface areas (29.46 Å2), and CBG is more like CBD (40.46 Å2), therefore only CBD and CBN’s absorption to the Chip were evaluated. CBG and CBC were assumed to have compound distribution comparable to CBD and CBN, respectively. Twelve acellular, non-porous chips, each containing two independent micro-fluidic channels (Emulate, Inc., NPS01, Boston, MA), (n = 3 chips per group) were connected to Pods (Emulate Inc., Boston, MA). Pods contain inlet and outlet reservoirs for both the channels on the Chips. The hepatocyte/top channel contained hepatocyte maintenance media, and the non-parenchymal cell (NPC)/bottom channel contained NPC maintenance media (Emulate Inc., EP226 Rev. B)37. The compounds to be dosed were added to their respective media in the Pod inlets. The Pods were connected to Zoë® Culture Modules for two regulate cycles to reduce air bubbles in the system, where the media flowed over the chips at a flow rate of 30 µL/h at 37 °C and 5% CO2 for 24 h.

To assess if the absorption occurs from Pods and/or Chips, 48 nM CBD or CBN were dosed in both hepatocyte and NPC inlet channels and flowed at 400 µL/h. Media was collected from the Pod inlets and outlets at 24, 48 and 72 h. For the Pod only condition, the Chips were removed and replaced with the temporary reservoir for 2 h where the flow rate was increased to 1000 µL/h. Media from the inlet and outlet, and media that flowed into the reservoir during this time was collected for liquid chromatography-mass spectrometry (LC–MS) analysis (24–26 h, 48–50 h, 72–74 h post-initial dosing).

The samples were prepared by protein precipitation using 25 µL of sample mixed with 100 µL of precipitation solvent composed of 70% acetonitrile, 30% isopropanol, and 0.1% formic acid containing 250 ng/ml of internal standard (D3)-deuterated CBD. The samples were mixed and then centrifuged to remove protein and other insoluble materials. An equal volume of supernatant and water were mixed to generate the samples for analysis. Reverse phase liquid chromatography was used to separate sample components and selected reaction monitoring on a Sciex 6500 + Qtrap mass spectrometer (Sciex, Framingham, MA, USA) was used to selectively detect the cannabinoids.

Separation and detection were accomplished using an Eksigent Ekspert LC200 microflow UPLC System (Sciex, Framingham, MA, USA) interfaced to a Sciex Qtrap 6500 + mass spectrometer operated in positive ion mode and unit resolution in both quadrupoles. Electrospray ionization was used with the IonDrive ion source outfitted with a 65 µm ID stainless steel electrospray capillary. Ion source temperature was set to 250 °C, Gas 1 and Gas 2 were set at 50 psi, curtain gas was set at 40 psi, and the spray voltage was set to 5000 V. The linear reverse phase LC gradient ranged from 15%−95%, mobile phase B was run at 75 μL/min through an Eksigent HALO C18, 50 × 0.5 mm, 2.7 µm particle size column (Sciex, Framingham, MA, USA) over 2.5 min. Column temperature was set to 60 °C. Mobile phase A was water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid.

The method was developed and validated following the US FDA Bioanalytical Method Validation – Guidance for Industry, CDER, CVM and FDA, May 2018 for the analysis of CBD in cynomolgus monkey plasma. A method bridging study was conducted to change sample matrix from cynomolgus monkey plasma to chip buffers. The study consisted of one day of accuracy and precision, selectivity, sensitivity and carryover evaluations using CBD, with the same acceptance criteria applied to both cannabinoids. Accuracy (as compared to nominal concentrations) and precision (reported as coefficient of variance) were both within ± 20% for the analytical method. Carryover was less than 1% and a detection range of 4–40,000 nmol/L was established. Compound distribution in both the chip and Pod was calculated using Emulate EC001 v1.1 calculator42.

Based on the compound absorption results (Table 1), all dosing concentrations of cannabinoids in the subsequent experiments were increased to account for 80% absorption to the chips and Pod. This increase resulted in the final concentrations noted above for the cannabinoids of 75, 935, and 1465 ng/mL (0.24, 3, and 4.7 µM). All cannabinoid dosing stocks were made 1000X the target dosing concentration in DMSO and diluted in channel-specific media, resulting in 0.1% DMSO in all groups. The solubility of APAP does not exceed 5 M, so the 10 mM dose was made from a 5 M stock solution. Liver-Chips were dosed continuously by adding 4 mL media in both the hepatocyte and NPC Pod inlets. The flow rate was increased to 400 µL/h to account for PDMS absorption. Due to Pod size (maximum volume of 4 mL), each dose was replaced every 8 h, for seven days of treatment.

Table 1 Experimental Design: A Study layout depicting the treatment groups, dosing concentrations, timepoints, and endpoints. All the compounds were continuously dosed in both channels at 400 µL/h for seven days.
Full size table

Assays and analysis

Morphological analysis

Brightfield images were captured on the Revolve ECHO (ECHO, San Diego, CA) at days 1, 3, 5, and 7 post initial dosing with a 10X objective and a 170% digital zoom. Both hepatocyte and NPC channels were imaged, and cells were scored blindly based on apparent cytotoxicity. The scoring system values were based on percentage of dead cells as follows: < 1% dead cells (0), < 5% (1-minor toxicity), < 20% (2-mild toxicity), < 50% (3-moderate toxicity), and > 50% (4-severe toxicity).

Effluent-based assays

On days 1, 3, 5, and 7 post-treatment, 9.6 mL effluent was collected from the hepatocyte and NPC channel outlets of all Liver-Chips. The 9.6 mL effluent was concentrated to 0.72 mL using 3 kDa spin columns (Thermo Fisher, 88,526, Waltham, MA) per manufacturers protocols. Hepatocyte channel effluent was analyzed for albumin, ALT, AST, LDH and cytokines.

Albumin (Abcam, ab179887, Waltham, MA), ALT (Abcam, ab234578, Waltham, MA), AST (Abcam, ab263881, Waltham, MA), LDH (Promega, J2381, Madison, WI), and cytokines (IL-6, IP-10, IL-8, TNF⍺, IL-10, IL-1β, IFNγ, MCP-1) (Meso Scale Discoveries, K15067L-2, Rockville, MD) were measured from concentrated effluent samples. Albumin was assayed according to vendor protocols at a 1:500 dilution. ALT and AST were assayed from undiluted samples according to vendor protocols except for the standard curve which was assayed two dilutions below the recommended protocol. LDH was measured according to vendor protocols at a 1:5 dilution. All cytokines were measured according to vendor protocols at a 1:2 dilution. The dilution preparation and/or aliquoting for the Albumin, ALT, and AST assays was performed using the Microlab VANTAGE Liquid Handling System (Hamilton). Albumin, ALT, AST and LDH assays were read using the SynergyNeo Microplate Reader (BioTek, Winooski, VT). The cytokine measurements were taken with the QuickPlex SQ 120MM (Meso Scale Discoveries, Rockville, MD).

Live immunofluorescent imaging

On day 7 post-treatment, Liver-Chips were disconnected from Pods, washed with 1X DPBS, and stained for reactive oxygen species (ROS) with CellROX (1:500, Thermo Fisher, C10422, Waltham, MA) and mitochondrial function with MitoTracker (1:5000, Thermo Fisher, M7510, Waltham, MA) in both channels. Two drops NucBlue (Thermo Fisher, R37605, Waltham, MA) per mL staining solution was also added to visualize nuclei. The stains were made in cell-specific media (without serum), and 150 µL was added to each channel. The Liver-Chips were incubated at 37˚C for 30 min in the dark. Images were captured on the Opera Phenix™ Plus High-Content Screening System and processed in Harmony v4.9 Imaging and Analysis Software (PerkinElmer, Waltham, MA). Images were acquired with the following parameters: 20X Water NA 1.0 objective, DAPI-100 ms exposure, CellROX Deep Red-120 ms exposure, MitoTracker Orange-100 ms exposure, 100% laser power, 12 FOV’s with 5% overlap, 100 µm Z-stack volume in 5 µm increments, 37˚C and 5% CO2.

Statistical analysis

Results from replicate samples were reviewed, and the values exceeding 20% coefficient of variance were considered as outliers and excluded from the dataset. Additionally, Liver-Chips that experienced low or no flow of effluent had their results removed from the dataset. These excluded data points were replaced by the median values imputed from the remaining data points of corresponding Liver-Chips and subsequently used in the dataset for statistical analysis.

Statistical analyses were performed using R version 4.3.2 and statistical packages: ggpubr, openxlsx, rstatix, and tidyverse. Replicate sample results from each Liver-Chip were averaged to obtain one mean value per Liver-Chip per timepoint. In addition, pairwise comparisons within each timepoint and dose level were also conducted using independent T-tests. The vehicle control and APAP were included for these analyses in addition to the four cannabinoids. Bonferroni correction was used to adjust for multiple comparisons. Statistical significance is reported as compared to the vehicle control when P value is < 0.05.

Results

Evaluating cannabinoid-chip interactions and exposure dose confirmation

Compound distribution analysis

The absorption of cannabinoids to the Chips and Pods was evaluated using a non-porous, acellular Chip by comparing the differences in compound concentrations between the inlet and outlet of the Chips (Fig. 2). CBD showed approximately 20% recovery suggesting 80% of the compound was absorbed by the Chips + Pods. However, when the compound concentrations were compared in the Pod only group without the Chip, there was 30% recovery, suggesting 70% absorption to the Pod. Therefore, most of the CBD absorption occurred upstream, prior to reaching the chip and if dosed as is, the maximum potential cellular exposure was 20% or less. CBN showed similar results, where approximately 60% and 70% absorption were observed in the Pod and Chip + Pod groups respectively. Thus, the maximum CBN potential cellular exposure was 30% or less.

Fig. 2
figure 2

Cannabinoid distribution analysis on an acellular, poreless, Emulate Chip-S1. The recovered concentrations of CBD and CBN from the top and bottom channels were independently analyzed by LC-MS over the course of 74 h. Recovered concentration represents the portion of compound available and not absorbed into the system. The Pod absorbed more highly than the Chips, resulting in ~ 70% and ~ 60% loss of CBD and CBN, respectively. The chip additionally absorbed ~ 10%, regardless of compound. All data represents Mean ± SD (n = 3 Chips per group).

Full size image

Due to similar physicochemical properties and identical polar surface areas, CBC and CBN were assumed to have similar absorption, while CBG and CBD were assumed to have similar compound distribution. Accordingly, the dosing concentrations for all the cannabinoids tested were adjusted based on their compound absorption to allow for accurate cellular exposure. To achieve a desired exposure concentrations of 15, 187, and 293 ng/mL (48, 595, and 932 nM), the Liver-Chips were dosed with 75, 935, and 1465 ng/mL (0.24, 3, and 4.7 µM) of cannabinoids.

Evaluating the dose-responses of cannabinoid hepatotoxicity

Morphological analysis

All the Liver-Chips were observed under brightfield microscopes to assess the hepatocyte and NPC health and morphological changes (Fig. 3a and Table 2). On day 0, before treatment, the hepatocytes in the top channel demonstrated their prototypical cuboidal and binucleated morphology with distinct bile canaliculi formation, and the NPCs in the bottom channel were characteristically polygonal in shape and formed a confluent monolayer. After treatment, the vehicle control demonstrated healthy morphology for the study duration, whereas the Liver-Chips treated with APAP showed time- and dose-dependent cell death in hepatocytes and NPCs. Hepatocytes demonstrated minor toxicity at 3 and 4.7 µM CBD on day 7, while NPCs demonstrated healthy morphology for the study duration at all doses. CBN demonstrated minor toxicity in hepatocytes on day 5 at 0.24 and 3 µM, with mild toxicity observed at 4.7 µM. Hepatocytes demonstrated mild toxicity at all doses on day 7 while NPCs demonstrated minor toxicity in response to 4.7 µM CBN. CBG demonstrated no changes in morphology for hepatocytes, but minor toxicity at 4.7 µM on days 5 and 7 for NPCs. Hepatocytes displayed a dose- and time-dependent cytotoxicity in response to CBC with minor toxicity noted at 3 and 4.7 µM doses by day 3, mild toxicity in the same doses at day 5, and finally mild toxicity at 3 µM and moderate toxicity at 4.7 µM on day 7. In addition, minor toxicity to the NPCs was shown in the 3 µM dose on day 7 and the 4.7 µM dose on days 3 and 5 followed by mild toxicity on day 7.

Fig. 3
figure 3figure 3

Assessing health and toxicity of cannabinoids on Liver-Chips. a) Representative bright-field images of hepatocyte and NPC morphology on days 1 and 7 post-treatment. Blue arrows signify healthy morphology whereas white arrows signify toxicity. The vehicles maintained healthy morphology for the entirety of the study, whereas APAP caused a severe time- and dose-dependent toxicity. Of the cannabinoids, CBC caused the most toxicity, and CBG or CBD caused the least toxicity. b) Hepatotoxicity as measured by LDH release on days 1, 3, 5, and 7 post-treatments from top channel effluent. APAP caused an expected increase in LDH, and 4.7 µM CBC caused an increase on day 3, although it was not statistically significant. 4.7 µM CBD treatment increased LDH on day 7 post-treatment, and although not statistically significant, the magnitude of release suggests biological relevance. For the cannabinoids, low dose is 0.24 µM, mid dose is 3 µM, and high dose is 4.7 µM. All data represents Mean ± SD. T-test with Bonferroni correction for multiple comparisons (n = 3 Chips per group). *p < 0.05, **p < 0.005, ***p < 0.0005 and ****p < 0.0001.

Full size image
Table 2 Morphology Observation Summary:​ The bright-field images of hepatocytes and NPCs from all the Chips were morphologically assessed for health and assigned a score of 0–4. Two independent reviewers assessed and blind-scored four images per chip for all the 48 Chips at four time points (average of n = 3 per group, 4 images/Chip).
Full size table

Endpoints evaluating hepatocyte function

LDH release

Exposure to the positive control, APAP, resulted in significantly higher LDH levels at the 10 mM concentration from day 3 onward (Fig. 3b). Liver-Chips treated with 4.7 µM CBD demonstrated an increase on day 7, which was not statistically significant due to the large variability within the group, but may be biologically relevant due to magnitude of release. The 4.7 µM CBC treatment demonstrated a statistically significant increase, compared to vehicle control, at day 7. CBC at 4.7 µM demonstrated an upward trend on days 3 and 5 that was not statistically significant again due to variability. CBN only showed a statistically significant transient increase in 0.24 µM CBN group on day 5 which returned to baseline by day 7. CBG had no effect at any dose or time point.

Albumin secretion

A dose- and time-dependent decrease in albumin levels was observed in APAP treatment groups, with 10 mM APAP resulting in no albumin production by day 7 post-treatment (Fig. 4a). Liver-Chips treated with 0.24 µM CBD exhibited a statistically significant decrease in albumin on days 3 and 7, although the release measured on day 3 was not biologically relevant as it is within the physiologically relevant range (20–105 µg/day/million cells) for this model. No statistically significant changes in albumin levels were observed from treatment with CBG or CBN at any dose. CBC-treated Liver-Chips demonstrated a dose- and time-dependent decrease in albumin production with statistically significant decreases seen in the high dose on days 3 and 7 and the mid dose on day 7.

Fig. 4
figure 4figure 4figure 4

Liver-Chip biomarker assessment. a) Evaluating hepatocyte health by measuring albumin secretion on days 1, 3, 5 and 7 post-treatments with CBD, CBN, CBG, CBC or APAP in comparison to the vehicle control. Albumin release from the vehicle Chips was observed to be within the physiologically relevant range (≥ 20 µg/day/million cells) for the duration of the study. Values above the dotted line represent health. APAP and CBC caused time- and dose-dependent decreases in albumin secretion, whereas all other treatments trended more similarly to the vehicle. b) Evaluating hepatocyte injury by measuring AST release on days 1, 3, 5, and 7 post-treatments with cannabinoids – CBD, CBN, CBG, CBC, or APAP in comparison to the control. APAP and CBC caused time- and dose-dependent increases in AST secretion, whereas all other treatments trended more similarly to the vehicle. c) Evaluating hepatocyte injury by measuring ALT release on days 1, 3, 5, and 7 post-treatments with cannabinoids – CBD, CBN, CBG, CBC, or APAP in comparison to the control. Values above the dotted represent injury. APAP caused a time- and dose-dependent increase in ALT release, whereas all other treatments cause ALT release like the vehicle. The low dose of CBN and CBG caused increases in ALT secretion on day 5, however, these values returned to baseline by day 7. For the cannabinoids, low dose is 0.24 µM, mid dose is 3 µM and high dose is 4.7 µM. All data represents Mean ± SD. T-test with Bonferroni correction for multiple comparisons (n = 3 Chips per group). *p < 0.05, **p < 0.005, ***p < 0.0005 and ****p < 0.0001.

Full size image
AST release

10 mM APAP treatment caused significant increases in AST release on days 3, 5, and 7 (Fig. 4b). No significant increases in AST were observed for CBD or CBG treated hepatocytes at any doses, whereas a statistically significant decrease in AST was observed for the mid and high dose of CBD. At 4.7 µM, CBC resulted in a statistically significant increase of AST relative to the control on days 1 and 7. Liver-Chips treated with CBN showed no increase in AST release except for an anomaly of a significant increase at the low dose on day 5. There was also a statistically significant decrease in AST observed in the high dose on day 7.

ALT release

10 mM APAP treatment caused an increase in ALT from day 3 onwards (Fig. 4c). A significant increase in ALT release was seen in the hepatocytes treated with 0.24 µM CBN on day 1. ALT also increased on day 5 for all cannabinoids at the low dose, and it reached statistical significance for CBN and CBG in comparison to the control. ALT level rose above the normal range in the vehicle control on day 7, while its levels were statistically significantly lower in the chips treated with the mid and high doses of CBD and CBG. The ALT levels in the chips treated with CBC and CBN were not different from the control on day 7.

Inflammatory cytokine release

As shown in Fig. 5, APAP treatment did not increase cytokine production, rather showed decreases in IL-8 and MCP-1 on day 7. Treatment with 4.7 µM CBD caused a sharp increase in IL-6 (Fig. 5a), IL-8 (Fig. 5b), and MCP-1(Fig. 5c) on day 7, but these changes were not statistically significant compared to the control due to large intra-group variability at this dose. CBC also caused increases in IL-6, IL-8, and MCP-1 on day 3 although the results were not statistically significant due to large intra-group variability. Treatment with 3 and 4.7 µM CBG or CBN resulted in a statistically significant decrease in IL-8 production on day 7. The levels of IFN-γ, IL-1β, and TNF-α were found to be below the limits of detection and data is not presented.

Fig. 5
figure 5figure 5figure 5

Liver-Chip inflammatory cytokine assessment. a) IL-6, b) IL-8 and c) MCP-1 release were assessed from Liver-Chip effluent on days 1, 3, 5 and 7 post-treatments. Although not statistically significant, 4.7 µM CBC and CBD treatment caused increased cytokine release on days 3 and 7, respectively. APAP treatment decreased cytokine release, suggesting severe cell death. All other treatments trended like the control. For the cannabinoids, low dose is 0.24 µM, mid dose is 3 µM and high dose is 4.7 µM. All data represents Mean ± SD. T-test with Bonferroni correction for multiple comparisons (n = 3 Chips per group). *p < 0.05, **p < 0.005, ***p < 0.0005 and ****p < 0.0001.

Full size image

Oxidative stress

Imaging analysis (Fig. 6a) showed a qualitative dose-dependent increase in ROS staining in the hepatocytes and NPCs treated with APAP. There were no significant differences in ROS for hepatocytes treated with CBD, CBN, or CBG when compared to the vehicle control, while treatment with CBC led to a dose-dependent increase in ROS after seven days. The NPCs treated with cannabinoids demonstrated increases in ROS accumulation compared to the vehicle control by day 7 for all three dose levels, with the most pronounced increase at the lowest dose of CBN.

Fig. 6
figure 6figure 6

Qualitative immunofluorescent images of cannabinoid treatment on Liver-Chips. a) Oxidative stress assessment of Liver-Chips on day 7 post-treatment. All Chips were stained for CellRox (red), and nuclei were stained with DAPI (blue). Scale bar represents 100 µm. Dose-dependent increases in ROS and decreases in NPC count were observed in response to CBC or APAP treatment. All other treated hepatocytes had ROS expression like the vehicle, whereas all cannabinoid treated NPCs had increased ROS expression as compared to the vehicle. b) Mitochondrial function assessment of Liver-Chips on day 7 post-treatment. All Chips were stained for Mitotracker (yellow), and nuclei were stained with DAPI (blue). Scale bar represents 100 µm. APAP, CBN or CBC caused dose-dependent decreases in Mitoracker signal hepatocytes and NPCs. CBG also caused decreases in NPC expression of Mitotracker, whereas all other treatments trended like the control.

Full size image

Mitochondrial function

The mitochondrial function images on day 7 are presented in Fig. 6b. A dose-dependent decrease in mitochondrial function was observed in APAP-treated chips. No changes in mitochondrial activity were observed in the hepatocytes treated with CBD or CBG. However, a dose-dependent decrease in mitochondrial activity was observed in the hepatocytes treated with CBN or CBC. No changes were observed in NPCs treated with CBD. The NPCs treated with CBN, CBG, or CBC all displayed a dose-dependent decrease in mitochondrial activity.

Discussion

The toxicity of cannabinoids has been evaluated in animal models such as rats, mice, dogs, etc.8,9,43. The safety of a CBD drug Epidiolex® has been limitedly assessed in some human clinical studies in addition to its comprehensive toxicological testing in animal models. While CBD effects on the liver are well characterized, the evidence on the effects of other cannabinoids on human health is limited. In addition, the metabolic and pharmacokinetic differences between animals and humans further complicates our understanding of cannabinoid toxicity. Furthermore, there is an increasing demand for evaluating alternative methods to preclinical safety and toxicity testing which is emphasized by the 90% failure rate of drugs in clinical trials44. Since January 2022, at least eleven clinical trials have reported drug induced liver injury and some deaths45. Fialuridine (FIAU) was deemed non-toxic in rats, but showed severe hepatotoxicity and death when advanced to human clinical trials. When replicated on an MPS34 no toxicity was observed on the rat Liver-Chip whereas the human Liver-Chip accurately modeled hepatotoxicity, further suggesting how MPS can compensate for animal models’ lack of human relevance.

Gao et al. used new approach methodologies (NAMs) based on bioinformatic analysis of high throughput transcriptomic data to conduct a concentration response transcriptomic analysis of hemp extracts and its constituents such as CBD, CBC, CBG, and CBN. They showed that each compound impacted a distinct biological pathway/function, but all the cannabinoids caused oxidative stress and impaired liver metabolism46. In the present study, another in-vitro alternative, the microfluidic Liver-Chip, was used as an alternative to animal models to compare and evaluate the toxicity of cannabinoids CBD, CBN, CBG, and CBC. This was done by assessing the cellular morphology and effluent-based readouts such as albumin, LDH, ALT, AST, and cytokines. To further explore the mechanism of toxicity, Liver-Chips were evaluated for oxidative stress and mitochondrial activity. To our knowledge, this is the first study to use Organ-Chips to evaluate toxicity of four different cannabinoids in comparison to a known hepatotoxin, APAP. One study compared the toxicity of a cannabidiol-rich cannabis extract in C57BL6/J mice and the human Liver-Chips, and demonstrated Liver-Chips had no statistically significant changes in the elevated ALT or decreased intrahepatic and increased effluent levels of miR-12247. However, a robust, dose-dependent induction of Cytochrome P450 (CYP) enzymes was observed in the mice and human Liver-Chips, suggesting the presence of metabolically active hepatocytes between both models. Additionally, CYP enzyme (CYP2B, CYP3A) activity was evaluated by administering CYP probe substrates to hepatocytes, and the CYP activity of those cultured on the Liver-Chip was found to be comparable to that of freshly isolated hepatocytes34. Since metabolism is a primary function of the liver, demonstrating robust CYP enzyme responses highlights the Liver-Chip’s ability to support this essential hepatic function.

APAP is a hepatotoxin known to cause drug-induced liver injury (DILI) through glutathione depletion resulting in oxidative stress and mitochondrial dysfunction48. In the present study, APAP caused an expected time- and dose-dependent cell death in hepatocyte and NPC morphology, decreases in albumin secretion, and increases in LDH, ALT, and AST levels, demonstrating a clear liver toxicity. APAP did not increase pro-inflammatory cytokine production but did increase ROS and reduce mitochondrial activities. These findings demonstrated the Liver-Chip accurately modeled APAP toxicity. Similar results were also shown by Jang et al., who demonstrated glutathione depletion, declines in ATP as well as albumin, and an increase in oxidative-stress-related injury markers on the same Liver-Chip model34.

Ewing et al. evaluated a CBD-rich cannabis extract and its interactions with APAP in mice and showed a low dose of CBD + APAP produced liver injury and mortality, while at the higher dose of CBD + APAP there was lower incidence of hepatotoxicity49. They attributed this paradoxical effect to the rapid re-synthesis of glutathione since the total hepatic glutathione levels in the two CBD doses without APAP were similar, but the hepatic glutathione was statistically significantly lower in the low dose CBD + APAP compared to the low dose CBD alone. Our study found that CBD treatment at 3 and 4.7 µM caused some cytotoxicity to the hepatocytes on day 7 as evidenced by morphological changes but did not adversely affect the NPC viability. Similarly, 4.7 µM CBD caused an increase in LDH on day 7 that was not statistically significant due to variability in the LDH levels of the different Liver-Chips but could suggest the onset of toxicity due to the magnitude of release. The liver biomarkers trended like the vehicle control, which suggests the hepatocytes were still functional and the toxicity may be minimal at the doses tested. The 4.7 µM CBD dose caused a pro-inflammatory response as measured through increases in IL-6, IL-8, and MCP-1 on day 7 but did not exhibit oxidative stress nor a decrease in mitochondrial function, indicating a potential onset of toxicity mediated by inflammation. Other researchers also showed a lack of increased ROS activity in hepatocytes treated with CBD aligning with their known antioxidant properties50,51,52,53. Together, these results suggest that CBD could be toxic at higher concentrations acutely or for extended durations. This is congruent with most literature on CBD efficacy and safety in neurological conditions, which demonstrates CBD is well tolerated54. Some adverse events such as ALT and AST elevations have occurred in clinical trials with a dose of 10–20 mg/kg of CBD therapy in combination with other drugs such as Valproate and Clobazam. However, the elevated plasma levels returned to normal either spontaneously during treatment or when the dose of CBD or other drugs were reduced14.

CBC showed the most toxicity out of all the cannabinoids tested. CBC treatment caused declines in hepatocyte and NPC morphology and increases in LDH earlier than CBD, although not statistically significant until day 7. Dose-dependent decreases in albumin and increases in ALT and AST by day 7 post-treatment also suggest specific loss of function and injury to the hepatocytes. There were increases in IL-6, IL-8, and MCP-1 secretion on day 3 post-treatment, which could suggest the early onset inflammation drove some of the toxicity observed at later timepoints. The increased ROS accumulation and decreased mitochondrial function at the 4.7 µM CBC dose coincide with the high levels of toxicity observed through other endpoints. Jointly, these results indicate that CBC may induce hepatotoxicity earlier and at lower doses than CBD by targeting all liver cell types similarly. Taiwo et al. evaluated the safety of three doses of CBC for seven days in male Wistar rats and showed an elevation in the activity of antioxidant enzymes such as catalase, superoxide dismutase, and glutathione peroxidase while ALT, AST, and ALP were significantly decreased at 10 mg/kg dose. However, histological evaluation of kidney and liver revealed interstitial nephritis and periportal hepatocellular necrosis along with an increase in inflammation55. While there are species differences between rats and humans, both models accurately predicted the inflammation of CBC indicating, this could cause serious toxicity depending on the dose.

Morphologically, CBN demonstrated mild toxicity to the hepatocytes at all doses and minor toxicity to the NPCs at the highest dose only. However, CBN showed no differences in albumin production while increases in LDH, ALT, and AST were observed for only the lowest dose (0.24 µM) on day 5 post-treatment. CBN (3 or 4.7 µM) also reduced pro-inflammatory cytokines such as IL-8, but a dose-dependent reduction in mitochondrial function by day 7 indicates that, like CBD, prolonged treatment with CBN may potentially cause cellular damage.

CBG appeared to be the least toxic cannabinoid tested in this study, as no changes were observed morphologically or from any effluent and imaging readouts. Structurally, CBG is the most distinct of the 4 cannabinoids tested as it is the only cannabinoid with 1 ring structure56. This difference in structure may have some effect on the different responses observed. CBG did not change pro-inflammatory cytokines and showed reduced levels of IL-8 production in the mid and high dose on day 7 post-treatment. Similar results were shown by Borrelli et al., who demonstrated CBG’s protective properties in a murine model of inflammatory bowel disease by regulating IL-1β, IL-10, and IFN-γ levels, while inhibiting inducible nitric oxide synthase expression57. In another mouse study by Aljobaily et al., a low dose of CBG was found to have a therapeutic effect in reducing hepatic steatosis, fibrosis, and inflammation58.

The available literature on the oxidation potential of cannabinoids other than CBD is scarce. If available, it mostly focuses on hepatocytes, so it is worth noting that in the current model, treatment with cannabinoids over seven days induced ROS formation in the NPCs at all three concentrations. Despite the well-documented antioxidant properties of CBD in various cells/systems, a pro-oxidant effect of CBD at high doses was also reported in the literature53,59. CBD-induced ROS formation in the NPCs observed in this study appears to be consistent with its pro-oxidant activity in immune cells reported in the literature. CBD was found to induce ROS production and apoptosis in human CD14 + monocytes60 and elevate ROS levels in human THP-1 monocytes61 and primary mouse splenocytes62 in vitro. CBC, CBN, and CBG may behave similarly and display antioxidant or prooxidant activity depending on the cell type, doses, and/or experimental conditions.

Similarly, our study demonstrated that pharmacologically relevant concentrations of CBD (0.24, 3, 4.7 µM) did not adversely affect mitochondrial membrane potential of the NPCs, but found that CBC, CBG, and CBN at 4.7 µM clearly decreased mitochondrial activity. The decline was most evident in the CBC-treated NPCs as evidenced by the observation of increased cell death in the NPCs. This mitochondrial dysfunction could be attributed to direct impact of these cannabinoids on mitochondrial membrane potential and the synthesis of ATP. Furthermore, the increase in intracellular oxidative stress caused by these cannabinoids could exacerbate the toxic effect. Nonetheless, more research with mechanistic detail is needed to better understand the role of these cannabinoids on mitochondrial functions and oxidative stress in various types of cells. The Liver-Chip can successfully differentiate the toxicities of CBD, CBG, CBC, and CBN as well as demonstrate that different cannabinoids act through different mechanisms of action such as reduced mitochondrial function, oxidative stress, inflammation, or cellular toxicity.

Due to significant absorption of cannabinoid molecules to PDMS, the dosing concentrations and flow rates had to be adjusted. Due to the increased flow rates, the effluent samples had to be concentrated before running assays. While this study along with Gao et al. sets the foundation for using NAMs in safety and toxicity evaluation of cannabinoids, there is a lot more to be evaluated before implementing this data for clinical use.

Conclusions

Our study confirmed the utility of the Liver-Chip in hepatotoxicity evaluations as it accurately reproduced the known toxicity of APAP. The results showed the low toxicity potential for CBG, CBD, and CBN at lower doses and found mild toxicity for CBD and CBN as the dose and duration increased. CBG was shown to have the least potential for toxicity in this system at these doses for the duration of the study. CBC appeared to be the most hepatotoxic of all the tested cannabinoids, causing hepatotoxicity earlier and at lower doses. The findings of this study provide valuable evidence of the hepatotoxicity profile of cannabinoids, although further studies with additional hepatocyte donors, expanded dosing range and longer duration of treatment are needed to identify a clear dose–response curve and confirm the safety of long-term use of these cannabinoids at specific doses for the various potential health benefits already being studied.