Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index; 13 of the 35 sources its claims cite were checked for support.

Utility of spherical human liver microtissues for prediction of clinical drug-induced liver injury

Open access

Abstract

Drug-induced liver injury (DILI) continues to be a major source of clinical attrition, precautionary warnings, and post-market withdrawal of drugs. Accordingly, there is a need for more predictive tools to assess hepatotoxicity risk in drug discovery. Three-dimensional (3D) spheroid hepatic cultures have emerged as promising tools to assess mechanisms of hepatotoxicity, as they demonstrate enhanced liver phenotype, metabolic activity, and stability in culture not attainable with conventional two-dimensional hepatic models. Increased sensitivity of these models to drug-induced cytotoxicity has been demonstrated with relatively small panels of hepatotoxicants. However, a comprehensive evaluation of these models is lacking. Here, the predictive value of 3D human liver microtissues (hLiMT) to identify known hepatotoxicants using a panel of 110 drugs with and without clinical DILI has been assessed in comparison to plated two-dimensional primary human hepatocytes (PHH). Compounds were treated long-term (14 days) in hLiMT and acutely (2 days) in PHH to assess drug-induced cytotoxicity over an 8-point concentration range to generate IC 50 values. Regardless of comparing IC 50 values or exposure-corrected margin of safety values, hLiMT demonstrated increased sensitivity in identifying known hepatotoxicants than PHH, while specificity was consistent across both assays. In addition, hLiMT out performed PHH in correctly classifying hepatotoxicants from different pharmacological classes of molecules. The hLiMT demonstrated sufficient capability to warrant exploratory liver injury biomarker investigation (miR-122, HMGB1, α -GST) in the cell-culture media. Taken together, this study represents the most comprehensive evaluation of 3D spheroid hepatic cultures up to now and supports their utility for hepatotoxicity risk assessment in drug discovery.

Overview

CoChat Summary

TL;DR

In a 110-drug retrospective panel, 3D human liver microtissues (hLiMT) detected more cytotoxicity IC50 values and showed greater sensitivity than plated 2D primary human hepatocytes (PHH) for identifying clinical DILI-positive compounds, with similar specificity. The hLiMT model also supported exploratory liver-injury biomarker measurement in culture supernatant.

Key findings

  • hLiMT detected more IC50 values than PHH across the 110-drug set: 56/110 vs 40/110. Among DILI-positive compounds, no IC50 was determined for 33% (23/69) in hLiMT vs 54% (37/69) in PHH.
  • At fixed IC50 thresholds, hLiMT sensitivity was higher at every threshold: at 10 μM, 18.8% (13/69) vs 4.3% (3/69); at 100 μM, 60.9% (42/69) vs 33.3% (23/69). Specificity at 100 μM was 85.4% for both assays; PLR was 4.16 for hLiMT vs 2.28 for PHH, and NLR was 0.46 vs 0.78.
  • Using margin-of-safety (MOS = IC50/total plasma Cmax) thresholds, hLiMT again showed higher sensitivity: at 10× MOS, 36.2% (25/69) vs 20.3% (14/69), with specificity 97.6% for both and PLR 14.86 vs 8.32. At 100× MOS, sensitivity was 59.4% vs 40.6%, with specificity 80.5% vs 85.4%.
  • hLiMT better distinguished some structurally related pairs: bosentan and sitaxsentan had MOS values of 12.5× and 8.7×, while ambrisentan was not detected; PHH detected no IC50 for any of the three endothelin-receptor antagonists. Exploratory biomarkers α-GST, HMGB1, and miR-122 were measurable in hLiMT supernatant after chlorpromazine exposure.

Methods

  • A panel of 110 marketed drugs was classified into five DILI severity categories; categories 1–3 were considered DILI-positive (69 drugs) and categories 4–5 DILI-negative (41 drugs).
  • Cytotoxicity was assessed by ATP-based CellTiter-Glo: plated 2D PHH were treated for 48 h, while 3D hLiMT (primary human hepatocytes co-cultured with non-parenchymal cells) were treated for 14 days with re-dosing; IC50 values were generated over an 8-point concentration range and exposure-corrected MOS values were calculated using total plasma Cmax.
  • Predictive performance was evaluated with ROC analysis, fixed IC50 and MOS thresholds, Cohen’s kappa, and positive/negative likelihood ratios; reproducibility was tested across different NPC and hepatocyte lots, and exploratory biomarkers were measured in hLiMT culture media.

Limitations

  • The paper states that hLiMT failed to properly classify approximately 40% of DILI-positive compounds; no cytotoxicity was observed for 23/69 DILI-positive drugs, and 28/69 DILI-positive compounds fell above a 100× MOS threshold. Many misclassified compounds were associated with low-incidence or idiosyncratic DILI.
  • The paper notes that the hLiMT spheroids contain only approximately 1000 cells, which could limit sensitivity for secondary supernatant endpoints, although exploratory biomarkers were detected. It also states that it remained unclear whether enhanced predictive value was due to culture complexity, treatment duration, or other factors.
  • Evident from the text: the ROC comparison using only compounds with IC50 values was biased by the different numbers of IC50 values obtained between assays, so the authors relied on fixed practical thresholds; these thresholds are somewhat arbitrary and assay performance depended on the compound set and concentration ranges achievable in vitro.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1007/s00204-017-2002-1

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 3D hLiMT were more sensitive to overall drug-induced cytotoxicity than 2D PHH. Cytotoxicity IC50 values for 110 drugs in PHH treated for 48 h (open symbols) and hLiMT treated for 14 days (closed symbols) for compounds classified as DILI+ve (DILI severity categories 1–3) (a) and DILI–ve (DILI severity categories 4–5) (b), respectively. Data represent IC50 value estimate for each compound listed on the x-axis. For compounds where no IC50 value converged, these were plotted at the top and classified as not determined (ND)
Fig. 1 3D hLiMT were more sensitive to overall drug-induced cytotoxicity than 2D PHH. Cytotoxicity IC50 values for 110 drugs in PHH treated for 48 h (open symbols) and hLiMT treated for 14 days (closed symbols) for compounds classified as DILI+ve (DILI severity categories 1–3) (a) and DILI–ve (DILI severity categories 4–5) (b), respectively. Data represent IC50 value estimate for each compound listed on the x-axis. For compounds where no IC50 value converged, these were plotted at the top and classified as not determined (ND)CC BY · PubMed Central
Fig. 2 Optimized receiver operating curve (ROC) for total plasma concentration (C max) alone as predictor of clinical hepatotoxicity. Optimized ROC curve for total plasma concentration (C max) for 110 drugs associated with and without clinical hepatotoxicity. ROC curve was generated from total plasma C max data for the test set and an optimized threshold (in bold) was determined
Fig. 2 Optimized receiver operating curve (ROC) for total plasma concentration (C max) alone as predictor of clinical hepatotoxicity. Optimized ROC curve for total plasma concentration (C max) for 110 drugs associated with and without clinical hepatotoxicity. ROC curve was generated from total plasma C max data for the test set and an optimized threshold (in bold) was determinedCC BY · PubMed Central
Fig. 3 Exposure-corrected cytotoxicity (MOS) of 110 marketed drugs stratified across the five DILI severity categories. A compound was considered to be DILI+ve if classified in the following: DILI severity category 1 (Severe clinical DILI), severity category 2 (High clinical DILI concern, cases of liver failure), and severity category 3 (low clinical DILI concern, isolated and infrequent cases of DILI). Conversely, a drug was considered DILI−ve if classified as DILI severity category 4 (enzyme elevations in clinic) and severity category 5 (No DILI). a Schematic of the four quadrants identifying true and false positives and true and false negatives across the 5 DILI severity categories in relation to optimized MOS threshold. b Plated primary human hepatocytes, dotted line threshold of 50× MOS, c hLiMT, dotted line threshold of 50× MOS. DILI severity category is indicated in brackets with the number of compounds that did not have an IC50 value converge per DILI severity category listed at top. d, e Comparison of the MOS values for structurally related hepatotoxic and non-hepatotoxic compounds in hLiMT (d) and PHH (e) for catechol-O-methyltransferase (COMT) inhibitors (tolcapone and entacapone), endothelin-receptor antagonists (sitax(s)entan, bosentan and ambrisentan), insulin sensitizers (troglitazone and rosiglitazone), and sedatives/antidepressants (nefazodone and buspirone). Filled circles margin of safety (MOS) = IC50/C max value; open circles not detected (ND) value and filled square <IC50/C max value
Fig. 3 Exposure-corrected cytotoxicity (MOS) of 110 marketed drugs stratified across the five DILI severity categories. A compound was considered to be DILI+ve if classified in the following: DILI severity category 1 (Severe clinical DILI), severity category 2 (High clinical DILI concern, cases of liver failure), and severity category 3 (low clinical DILI concern, isolated and infrequent cases of DILI). Conversely, a drug was considered DILI−ve if classified as DILI severity category 4 (enzyme elevations in clinic) and severity category 5 (No DILI). a Schematic of the four quadrants identifying true and false positives and true and false negatives across the 5 DILI severity categories in relation to optimized MOS threshold. b Plated primary human hepatocytes, dotted line threshold of 50× MOS, c hLiMT, dotted line threshold of 50× MOS. DILI severity category is indicated in brackets with the number of compounds that did not have an IC50 value converge per DILI severity category listed at top. d, e Comparison of the MOS values for structurally related hepatotoxic and non-hepatotoxic compounds in hLiMT (d) and PHH (e) for catechol-O-methyltransferase (COMT) inhibitors (tolcapone and entacapone), endothelin-receptor antagonists (sitax(s)entan, bosentan and ambrisentan), insulin sensitizers (troglitazone and rosiglitazone), and sedatives/antidepressants (nefazodone and buspirone). Filled circles margin of safety (MOS) = IC50/C max value; open circles not detected (ND) value and filled square <IC50/C max valueCC BY · PubMed Central
Fig. 4 Reproducibility of the IC50 values from independent experiments following incubation of the 3D hLiMT to test compounds for 14 days. Cytotoxic IC50 values for subset of compounds were plotted for hLiMT were prepared using either: the same cryopreserved primary human hepatocyte lot (IZT) but four different NPC lots: filled square JJB; filled circle RHV; filled triangle ZAR; open circle QCU (a) or the same NPC lot but different lots of cryopreserved primary human hepatocytes (IZT, OFA, SSR or EBP) (b)
Fig. 4 Reproducibility of the IC50 values from independent experiments following incubation of the 3D hLiMT to test compounds for 14 days. Cytotoxic IC50 values for subset of compounds were plotted for hLiMT were prepared using either: the same cryopreserved primary human hepatocyte lot (IZT) but four different NPC lots: filled square JJB; filled circle RHV; filled triangle ZAR; open circle QCU (a) or the same NPC lot but different lots of cryopreserved primary human hepatocytes (IZT, OFA, SSR or EBP) (b)CC BY · PubMed Central
Fig. 5 POC study demonstrating the utility of hLiMTs to detect biomarkers of hepatocellular injury in vitro. Representative changes in biomarker levels are depicted following exposure to chlorpromazine for 14 days (a) or 5 days (b and c). a In relation to decreases in total cellular ATP (filled circle) as determined on day 14. Filled square α-GST levels were determined on days 5, 9, and 14 and the values combined to give a fold change relative to control. b, c In relation to decreases in total cellular ATP as determined on day 5. Filled circle cell viability (% control); b open triangle HMGB-1 ng/ml, c open inverted triangle miRNA-122 relative expression. Data are from single experiments in triplicate
Fig. 5 POC study demonstrating the utility of hLiMTs to detect biomarkers of hepatocellular injury in vitro. Representative changes in biomarker levels are depicted following exposure to chlorpromazine for 14 days (a) or 5 days (b and c). a In relation to decreases in total cellular ATP (filled circle) as determined on day 14. Filled square α-GST levels were determined on days 5, 9, and 14 and the values combined to give a fold change relative to control. b, c In relation to decreases in total cellular ATP as determined on day 5. Filled circle cell viability (% control); b open triangle HMGB-1 ng/ml, c open inverted triangle miRNA-122 relative expression. Data are from single experiments in triplicateCC BY · PubMed Central

Verification

Claims the paper makes

9 Valid · 1 Not valid · 15 Not checked · 80 Own results and interpretations

55 more claims

Create a free account to see all 105 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 32
  • Probably not a substantial claim. 15

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

1 more claim about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Cited by

See all 328

Metadata

Authors
Proctor, W. R., Foster, A. J., Vogt, J., Summers, C., Middleton, B., Pilling, M. A., Shienson, D., Kijanska, M., Ströbel, S., Kelm, J. M., Morgan, P., Messner, S., & Williams, D.
Affiliations
  • Genentech (US)
  • AstraZeneca (United Kingdom) (GB)
Corresponding author
Foster, Alison J.
Venue
Archives of Toxicology
Published
2017-06-13
Volume
91
Issue
8
Pages
2849-2863
ISSN
  • 0340-5761 (ISSN-L)
  • 1432-0738
OpenAlex
W2625311319
Cited by
328
References
52
Citations by year
2026: 30 · 2025: 49 · 2024: 33 · 2023: 29 · 2022: 33 · 2021: 33 · 2020: 49 · 2019: 35 · 2018: 31 · 2017: 6
FWCI
35.17
Type
journal article
Language
en
Open access
Locations
  • Archives of Toxicology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • Apollo (cc-by)
Funders
AstraZeneca
Topics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Pharmacology › Drug-Induced Hepatotoxicity and Protection
  • Health Sciences › Veterinary › Small Animals › Animal testing and alternatives
  • Health Sciences › Medicine › Hepatology › Liver physiology and pathology
Keywords
drug-induced liver injury, hepatotoxicity prediction, primary human hepatocytes, drug cytotoxicity, IC50, safety margin, liver injury biomarkers, miR-122, HMGB1, alpha-glutathione S-transferase
MeSH terms
Humans, Liver, Predictive Value of Tests, Time Factors, Drug Design, Biomarkers, Risk Assessment, Inhibitory Concentration 50, Hepatocytes, Drug Discovery, Chemical and Drug Induced Liver Injury, Drug-Related Side Effects and Adverse Reactions

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.