A 3D primary human cell-based in vitro model of non-alcoholic steatohepatitis for efficacy testing of clinical drug candidates

Non-alcoholic steatohepatitis (NASH) is a progressive and severe liver disease, characterized by lipid accumulation, inflammation, and downstream fibrosis. Despite its increasing prevalence, there is no approved treatment yet available for patients. This has been at least partially due to the lack of predictive preclinical models for studying this complex disease. Here, we present a 3D in vitro microtissue model that uses spheroidal, scaffold free co-culture of primary human hepatocytes, Kupffer cells, liver endothelial cells and hepatic stellate cells. Upon exposure to defined and clinically relevant lipotoxic and inflammatory stimuli, these microtissues develop key pathophysiological features of NASH within 10 days, including an increase of intracellular triglyceride content and lipids, and release of pro-inflammatory cytokines. Furthermore, fibrosis was evident through release of procollagen type I, and increased deposition of extracellular collagen fibers. Whole transcriptome analysis revealed changes in the regulation of pathways associated with NASH, such as lipid metabolism, inflammation and collagen processing. Importantly, treatment with anti-NASH drug candidates (Selonsertib and Firsocostat) decreased the measured specific disease parameter, in accordance with clinical observations. These drug treatments also significantly changed the gene expression patterns of the microtissues, thus providing mechanisms of action and revealing therapeutic potential. In summary, this human NASH model represents a promising drug discovery tool for understanding the underlying complex mechanisms in NASH, evaluating efficacy of anti-NASH drug candidates and identifying new approaches for therapeutic interventions.

LDH release. Activity of lactate dehydrogenase (LDH) released from damaged cells into the supernatant was measured using the LDH-Glo Cytotoxicity Assay kit (Promega Corporation), according to manufacturer's protocol. Cell culture supernatants were diluted 1:5 in LDH storage buffer and the diluted samples (25 µl) were mixed with 25 µl detection reagent in white half area assay plates (Greiner Bio-One  41 was applied to remove unintended batch effects by using sva R package. Hierarchical clustering was performed using correlation distance metric and ward.D2 linkage method. Pre-ranked Gene Set Enrichment Analysis (GSEA) 42 was performed using the cluster Profiler R library 27,43 . DEA output was used as input for GSEA with genes ranked according to log2FC. Normalized Enrichment Score (NES) was used as an enrichment metric. GSEA was run on MsigDB database 44

Characterization of LEAN and NASH conditioned hLiMTs. Confirmation of maintenance of key cell
types in 3D liver microtissues over 10-day culture period. First, the morphology and presence of the relevant liver cell types for disease inductions were confirmed in NASH and compared with LEAN hLiMTs. Figure 1a shows that the NASH and LEAN hLiMTs contained the main liver cells types responsible for disease initiation and progression, namely, PHHs (albumin containing cells), LECs (detected using the cell type specific marker, CD31) and HSCs (detected using the cell type specific marker, vimentin). The presence of KCs was detected by staining for CD68 (a pro-inflammatory M1 macrophage marker) and CD163 (a pro-fibrotic M2 macrophage marker) 50 .
In addition, hematoxylin and eosin (H&E) staining revealed visible accumulation of lipids in NASH hLiMTs compared to LEAN hLiMTs. The treatment regimen used to induce NASH was not toxic to the hLiMTs, which maintained viability without signs of necrosis over the entire 10-day culture period (Supplementary Fig. S1). Figure 2 shows the steatosis-like phenotype and lipid accumulation in NASH treated hLiMTs. In addition to the increased accumulation of lipids in NASH compared to LEAN hLiMTs, there was evidence of ballooning of hepatocytes with displaced nuclei (Fig. 2a). NASH hLiMTs also contained significantly higher levels of TG compared to LEAN

NASH treatment in 3D liver microtissues results in inflammation. Cytokines and chemokines
shown to be upregulated in serum of patients with NASH 51 were also upregulated in NASH hLiMTs. NASHtreated microtissues have shown increased secretion of pro-inflammatory cytokines IL-6 and TNF-α ( Fig. 3a) and pro-inflammatory chemokines MCP-1, MIP-1α, IL-8 and IP-10 ( Fig. 3b). Conversely, the levels of these cytokines and chemokines were very low (MCP-1) or even below the lower limit of quantification (LLOQ) in LEAN hLiMTs. We assessed the cytokine and chemokines at day 5 due to their largest peak in secretion with sufficient sensitivity between LEAN and NASH conditions. Treatment with NASH-inducing stimuli leads to fibrotic phenotype in 3D liver microtissues. To show the effect of NASH treatment on fibrosis development and HSC activation, the protein expression of collagen type I and III was investigated using IHC. The NASH hLiMTs expressed higher protein levels of both collagen type I and III than LEAN hLiMTs (Fig. 4a). Active collagen synthesis was evaluated by measuring the levels of cleaved C-terminal pro-peptide, a product of procollagen I synthesis, in the supernatants. The synthesis and secretion of procollagen type I was ~ 10-fold higher in NASH than LEAN hLiMTs (Fig. 4b), which correlated well with the level of collagen type I protein expression. Moreover, collagen fibrillation was visualized by Sirius Red staining of histological sections, a well-established method to qualify pathological changes of collagen fiber content and structure in clinical tissues sections 52 . We assessed tissue sections under bright www.nature.com/scientificreports/ and polarized light and showed an increased collagen deposition and fibrillation in NASH compared to LEAN hLiMTs (Fig. 4c).

NASH treatment induces disease-relevant gene expression changes in human 3D liver microtissues.
To characterize the NASH phenotype on a molecular level, we performed whole transcriptomic profiling on days 5 and 10 of NASH induction, using a targeted sequencing approach (Fig. 5). Principal component analysis of all samples revealed distinct clustering of treatment groups on both days, indicating the induction of distinct phenotypes (Fig. 5a). In LEAN condition differences between day 5 and 10 were observed. These differences were predominantly observed for hallmarks of adipogenesis (Fig. 5b). This was related to an ongoing adaption of the cells to the LEAN culture condition after completion of the model manufacturing. The hierarchical clustering of samples based on gene sets for hallmarks of NASH (i.e. free fatty acid metabolism, inflammatory response and collagen synthesis), showed a distinct clustering for NASH and LEAN hLiMTs, confirming the activation of pathways associated with NASH (Fig. 5b) www.nature.com/scientificreports/ lism (e.g. FASN, LDLR, PPARα), inflammation (e.g. IL1B, CCL2, CXCL8) and fibrosis (e.g. COL1A2, TIMP1, FN1) were analyzed and found to be up-or down-regulated on both days 5 and 10, with larger fold changes and significance levels observed on day 10 ( Supplementary Fig. S2c,d). These results are in accordance with the biochemical findings, and further support the progression of the disease. Differentially expressed gene analysis and GSEA were performed ( Fig. 5c and Supplementary Fig. S2a,b). GSEA revealed a strong regulation of metabolic and inflammatory pathways, further confirming the induction of the NASH phenotype. There was a regulation of many metabolic pathways, including CYPs, which is in line with studies showing a dysregulation of CYPs in NASH patients 53 .
Proof-of-concept studies: recapitulation of clinical and mechanistic findings. Anti-steatotic effect of Firsocostat (ACCi). The anti-steatotic and anti-fibrotic effects of Firsocostat were investigated by treatment of hLiMTs with NASH conditioned medium in the presence and absence of compounds. TG levels in NASH hLiMTs were significantly (*p < 0.05) decreased by 0.5 and 10 μM Firsocostat (Fig. 6a) and comparable with those in LEAN hLiMTs. Confocal images of Nile Red-stained hLiMTs confirmed a concentration-dependent decrease in the accumulation of lipids by Firsocostat in NASH treated tissues (Fig. 6b). Firsocostat at 10 µM, significantly decreased (*p < 0.05) the accumulated lipids to a lower level than that in untreated NASH hLiMTs (Fig. 6c). The secretion of procollagen type I was mildly decreased in a concentration-dependent manner by Firsocostat, although this was only observed as a trend (Fig. 6d). Firsocostat did not affect the secretion of the six measured pro-inflammatory cytokines/chemokines (data not shown); however, whole transcriptomic profiling of 10 μM Firsocostat-treated NASH hLiMTs on day 10 indicated down-regulation of pro-inflammatory pathways (Fig. 6e,f). There was also an upregulation of metabolism-related genes by Firsocostat (Fig. 6e,f). Interestingly, hierarchical clustering based on gene sets associated with hallmarks of NASH ( Supplementary Fig. S3b) showed distinct clustering for adipogenesis, inflammatory responses and collagen formation.
Anti-inflammatory and anti-fibrotic effect of Selonsertib (ASK1i). The anti-inflammatory and anti-fibrotic effects of Selonsertib were investigated by incubating hLiMTs in NASH induction medium in presence and absence of the drug. As expected, there was no effect of Selonsertib on tissue TG levels (data not shown). However, the anti-fibrotic effect of Selonsertib was evident as a concentration-dependent and significant decrease the synthesis and secretion of procollagen type I (Fig. 7a). Anti-inflammatory activity of Selonsertib was reflected as a concentration-dependent decrease in the secretion of pro-inflammatory cytokines: TNF-α and IL-6 ( Fig. 7b) and chemokines: MCP-1, MIP-1α, IL-8, IP-10 (Fig. 7c).
Whole transcriptomic profiling revealed that Selonsertib treated (2 μM) hLiMTs (Fig. 7d,e) resulted in a strong down-regulation of pro-inflammatory pathways and decreased expression of key pro-inflammatory markers on day 10, in alignment with the biochemical analyses. According to hierarchical clustering based on gene sets associated with hallmarks of NASH, there was a distinct clustering for adipogenesis, inflammatory responses and collagen formation ( Supplementary Fig. S3c).
Prevention of fibrosis in NASH hLiMTs by ALK5i. The anti-fibrotic effects of ALK5i were investigated by incubating hLiMTs in NASH induction medium in presence and absence of the drug. There was a concentration-dependent decrease in the synthesis and secretion of procollagen type I by ALK5i (Fig. 8a). Furthermore, decreased deposition of collagen fibrils was visualized under bright field and polarized light on histological sections stained with Sirius Red (Fig. 8b).
As expected, Alk5i had no effects on TG accumulation, however an increase of cytokine/chemokine release was seen (data not shown), which can be explained by a reduction of TGF-β mediated anti-inflammatory effects 54 .
The results of the biochemical analyses were further confirmed by whole transcriptomic profiling of ALK5i (0.5 μM)-treated NASH hLiMTs on day 10. There was a strong down-regulation of pathways associated with collagen synthesis and ECM production (Fig. 8c), as well as a significant decrease in key fibrotic genes, such as COL1A2, COL3A1, ELN and TIMP3 (Fig. 8d). Hierarchical clustering based on gene sets associated with hallmarks of NASH revealed a distinct clustering for adipogenesis, inflammatory responses and collagen formation ( Supplementary Fig. S3d).

Discussion
There is an unmet need for development of human-relevant NASH models that can recapitulate key hallmarks of the disease, facilitate mechanistic studies, and allow rapid assessment of various modalities or combinatorial treatments. Here we demonstrated a complex 3D liver model suitable for NASH disease modeling. We have characterized a novel primary 3D human liver model that incorporates relevant liver cell types for investigating NASH. The maintained presence of the liver cells critical to NASH: PHHs, LECs, KCs, and HSCs, were confirmed throughout the treatment period by positive staining of cell-specific markers. To generate a phenotype that mimics NASH, the non-parenchymal cells were required to ensure inflammatory state and HSCs to facilitates liver fibrosis. The different cell types were individually added to the model, which facilitates the potential to explore how the ratio of the different cell types affects the disease phenotype. The PHHs represent a pool of ten donors, whereas the KC/LECs and HSCs derived from single donors. The protocol allowed for LEAN as well as NASH hLiMTs; the latter by supplementing the medium with elevated sugars, FFA, and a pulse of LPS. This complex and clinically-relevant stimuli resulted in hLiMTs displaying all three major clinical hallmarks of NASH. Both micro-as well as macrovesicular steatosis were evident through Nile Red staining of lipids, accumulation of TGs, and displacement of nuclei. Induction of inflammatory pathways was detected by a substantial increase of cytokine and chemokine secretion into the media. Recent review articles have highlighted a key subset of the www.nature.com/scientificreports/ cytokines/chemokines observed in the clinical disease 51 and most of them were increased in our model. A significant amount of fibrosis was achieved and represented by both secretion of procollagen type I, as well as increased collagen deposition and fibrillation. Not only were biochemical endpoints (TG, cytokines/chemokines, and PCI) achievable in these LiMTs, but they also exhibit an impressive dynamic range, which allows for ranking of compound effects on all three hallmarks of NASH. Whole transcriptomic profiling also clearly distinguished between LEAN and NASH treated hLiMTs. Significant up-regulation of genes related to inflammation and fibrosis, as well as down-regulation of genes related to lipid metabolism, reaffirmed the biochemical processes and progression known to occur in NASH. The gene expression of a range of cytokines/chemokines such as interleukin-1β (IL-1β), chemokine (C-X-C motif) ligand 6 (CXCL6), C-C motif chemokine ligand 20 (CCL20), was increased by the loading of cells with FFAs and LPS ( Supplementary Fig. S2d). Furthermore, the fibrosis biomarkers such as tissue inhibitor of metalloproteinases 1 (TIMP-1), collagens, fibronectin and elastin ( Supplementary Fig. S2d) were upregulated on gene expression level 2,55 .
In order to recreate NASH disease progression within a short exposure time, optimal medium conditions have been critical. NAFLD is closely linked with obesity and diabetes 56 , and is therefore the basis of the stimuli used in the NASH hLiMT model described here. NASH is thought to be initiated by fatty acid accumulation and endotoxin derived inflammation, resulting in a cascade of inter-linked processes that cause oxidative stress, fatty acid oxidation, mitochondrial dysfunction and, ultimately, a pro-inflammatory state that leads to fibrosis 12,56 . We have modeled these processes in the short NASH induction phase by incubating the hLiMTs under diabetic conditions, lipotoxic and inflammatory stress, rather than with the inflammatory cytokines themselves. Others have investigated in vitro NASH models using strong stimuli like TGF-β and showed that simply adding this potent pro-fibrotic mediator resulted in poorer fibrotic phenotype compared to that induced by supplementing the medium with FFA 13 . Overstimulation of in vitro models might affect cell viability, as well as reduce the response of the model to compound effects. Although LPS is also considered a strong stimulus, it reflects the clinical situation as it is measured in the blood or NASH patients, likely a result of gut microbiota alterations and bacterial translocation 57 . Others have used a very high concentration of palmitic acid (0.5 mM) to mimic NASH in multicellular InSphero 3D InSight™ 3D liver microtissues 23 ; however, such high concentrations are not observed clinically. In their model, although high palmitic acid concentrations induced high toxicity and tissue damage, as well as several markers of the inflammatory (e.g., IL-8) and pro-fibrotic process (e.g., TIMP-1, PDGFRβ, collagen I and III), palmitic acid did not induce lipid accumulation or a steatosis-like phenotype.
In addition to the stimuli used in an in vitro NASH model, the cell type and culture format are also crucially important. The most commonly used 2D human cellular models (e.g. immortialized cells, primary hepatocyte monocultures) fall short due to diminished longevity and the absence of critical hepatocyte-NPC interactions required for NAFLD progression 58 . Some researchers have combined immortalized cells e.g., HepG2 and LX2, to represent hepatocytes and stellate cells, respectively 59 . However, the use of immortalized cells does not allow for investigation of donor variations and introduces uncertainties with respect to their genotype 14 . HepG2 cells have the disadvantage of lacking sufficient metabolic capacity, while HepaRG cells represent an alternative hepatocyte type that exhibits a more comparable metabolizing capacity to PHHs 60 . Although they are also immortalized cells from a single donor, HepaRG cells have been used to investigate mechanisms involved in NASH, either in the absence of NPCs in 2D culture 19 or in combination with other liver cell types 13,61 . These models tend to only represent one part of the NASH pathogenesis e.g., steatosis 19 or fibrosis only 61 . As an alternative to cell lines, some researchers have incorporated extra-hepatic primary cells, such as HUVECs 13 to the other liver cell types in the 3D tissue model; however, these might not fully mimic liver specific processes and cell-cell interactions, which is crucial to understand initiation and progression of a complex disease like NASH. It has been shown that spheroidal 3D tissue model composed of mixture of transformed cell lines: HepaRG and HUVEC as well as primary KC and HSC, developed lipid accumulation, inflammation and fibrosis upon treatment with FFA 13 . Recently has been reported that human hepatic 3D spheroid system consisting of PHH cultured with free fatty acids and insulin mimic the steatotic conditions in a reversible manner 62 . Similar spheroidal model of PHH in co-culture with crude fractions of NPC from several matched or non-matched donors, has been shown to display a fibrotic phenotype either spontaneously, primarily observed in patatin like phospholipase domain containing www.nature.com/scientificreports/ 3 (PNPLA3) mutant donors, or after challenge with FFA, as determined by COL1A1 and αSMA expression 63 . In this model, however, no inflammatory process as a hallmark of NASH was observed.
The NASH model described in the current study involves the formation of hLiMTs composed of multi-donor PHHs in co-culture with primary HSCs and KCs and LECs. All these liver cell types are crucial for disease development, and enhance the in vivo translational relevance of the NASH model. Through the 3D incorporation/ communication between human relevant primary cells derived from the liver, and donor variability minimized by using ten multi-donor PHHs lot, we are able to overcome the limitations seen in immortalized cell lines and individual donor lots of primary cells. Compound efficacy testing in a multi-donor PHH lot represents a larger average population, preventing the potential biased conclusion from a single human donor. To overcome the problems associated with allogeneic reaction of mixing multi-donor lot of PHH with multi-donor lots of KCs, LEC and HSC only single donors of NPCs were used. For the initial establishment the 3D NASH hLiMTs, several KCs and LECs lots were tested and the ones with no allogeneic reaction, but with good inflammatory and fibrosis response were chosen for further model development. In contrast to flow-based microphysiological systems (MPS) with limited throughput capacity 64,65 , the 96-well format of the hLiMT model is scalable and enables the testing of numerous compounds, as well as combinations of compounds. Furthermore, current models with MPS and hemodynamics conditions have been shown to contain only PHHs, KCs (macrophages) and HSCs, but lack the important LECs 22,64,65 . LEC have been shown to pay pivotal role in provoking fibrosis through stromal cell-derived factor I receptors CXCR7/CXCR4 and fibroblast growth factor receptor 1 (FGFR1) 2 . Similarly to our NASH model, the other also reported an in vitro spheroidal NASH model, which incorporates PHHs, KC, LEC and HSC. A NASH-like phenotype has been induced by addition of medium containing FFA and TNF-α for treatment period of 10 days. In contrast to studies presented here, the authors reported only low level of steatosis, inflammation and early fibrosis. No increased extracellular collagen fibers deposition and formation was demonstrated on histological level 55 .
Perhaps the most important aspect of research model systems is the ability to recapitulate compound effects observed in the clinic. A prime example of predictivity is the use of the 3D liver model for toxicity assessment, which has been validated by pharma with a set of more than 100 compounds 66 . Once more clinically proven drugs are available for testing, a similar validation study on the NASH model can further underline its value in recapitulating the disease. Nevertheless, advanced model systems can help to further understand MoAs of compounds and deconvolute processes like cellular crosstalk and cell matrix interaction. As a first step towards such predictive models, we performed proof-of-concept experiments to analyze the effect of clinical drug candidates in our model system and compared them to results from the clinic. Firsocostat has been shown not only to decrease the de novo lipogenesis, but directly to suppress TGF-β-induced activation of HSC 6 . The anti-steatotic effects of Firsocostat were evident as a reduction of TG in NASH hLiMTs. This effect was confirmed by image-based quantification of lipids, correlating with the effect observed by magnetic resonance imaging-estimated proton density fat fraction (MRIPDFF) diagnostics in patients taking part in a clinical study 49 . Interestingly, Firsocostat did not decrease lipids to LEAN control level. The rationale for this observation could be that in our model, cells are constantly kept at a nutrient rich "fed" state, thus keeping beta-oxidation at an already low level. Firsocostat has been shown to have anti-fibrotic effects, which was nicely recapitulated in our model as a dose-dependent decrease in procollagen type I production. Interestingly, while Firsocostat did not affect the secretion of the six measured pro-inflammatory cytokines/chemokines, there was a down-regulation of other markers such as cytokines CCL19, CCL20 and CCL21 at the gene expression level 25 . In this case, the gene expression profile provided additional information on the MoA of Firsocostat, which is advancing in clinical combination studies with a Farnesoid X receptor (FXR) agonist and glucagon-like peptide-1 receptor (GLP-1R) agonist highlighting the importance of combined and multiparametric endpoints 67 . The maximum serum concentration (C max ) in patients treated with Firsocostat has been shown to be 5.4 ng/ml (9.5 nM) at 2-3 h post-dosing with the compound 68 . This concentration is lower that the effective tested concentrations of Firsocostat (0.5 µM and 10 µM) in NASH hLiMT. Other studies also demonstrated an effect of 0.5 µM and 1 µM Firsocostat on inhibition of fibrosis in LX-2 cells and primary HSC 29 . This difference in the compound concentrations found in the patient plasma and the   69 , anti-inflammatory effects were evident as a decrease in the secretion of pro-inflammatory cytokines and chemokines, as well as at the gene expression level, with a strong down-regulation of pro-inflammatory pathways. The anti-fibrotic mechanism the most relevant clinical endpoint for improving patients' health condition was determined in the hLiMTs as a decrease in procollagen type I secretion 69  www.nature.com/scientificreports/ for 48 weeks, there was not a significant anti-fibrotic effect observed in NASH patients with bridging fibrosis F3 or compensated cirrhosis F4 70 . In first-in-human clinical studies, the EC 50 in human whole blood was determined to be 56 ng/mL (0.13 μM) 71 , which is lower than the concentrations tested in the NASH hLiMTs model (2 and 10 μM). Thus, the concentrations tested here are in the pharmacologically relevant clinical range. This suggests that our NASH model most likely reflect the F2-F3 clinical stages of fibrosis, where we see a prominent anti-fibrotic effect of Selonsertib. The clinical C max of Selonsertib has been shown to be 40.5 or 3640 ng/ ml (0.09-8.17 µM) dependent on the applied compound dose 1 or 100 mg, respectively 71 . These clinical C max concentrations are very closed to the effective tested concentrations of Selonsertib (2 µM and 10 µM) in NASH model. The clinical Cmax of Selonsertib has been shown to be between 40.5 and 3640 ng/mL (0.09-8.17 µM) dependent on the applied compound dose 1-100 mg 71 . These clinical C max concentrations are very closed to the effective tested concentrations of Selonsertib (2 µM and 10 µM) in NASH hLiMT. Other studies also previously demonstrated an effect of 1 µM Selonsertib on inhibition of fibrosis and inflammation in palmitate-induced 3D NASH microtissue models 23 . ALK5i is not in clinical development but was used as a model compound to manipulate TGF-β signaling and the downstream impact on HSCs and fibrosis. There was a clear effect of ALK5i observed on the fibrotic phenotype of the NASH model, including a reduction of procollagen I release and deposition of collagen, as well as a down-regulation of fibrotic pathways and genes. The findings from ALK5i treatment indicate the presence of endogenous TGF-β signaling in the model, most likely induced by the lipotoxic and inflammatory stress. TGF-β has a stronger affinity to the matrix and it is not present in a free form. Therefore, determination of the level of secreted form of TGF-β in the supernatants of matrix rich NASH samples is difficult 72 . These findings also demonstrated that the model can help to elucidate the MoA of a candidate drug and its potency effects on NASH hLiMTs. Since the blocking and/or resolution of fibrosis is perhaps the most critical goal of NASH therapeutics, these data support the use of the 3D NASH model as a suitable pre-clinical drug discovery tool to assess anti-fibrotic effects of compounds.
In conclusion, we have developed a 3D NASH model, which incorporates all relevant primary human liver cell types, shown previously to play a role in the mechanisms of disease induction mediated by complex cellular crosstalk 14 . NASH hLiMTs can recapitulate the key hallmarks of NASH and in proof-of-concept studies, we have recapitulated the effects of drug candidates on lipid accumulation, inflammation and fibrosis in this disease model.
Despite these very promising results, further development of the NASH model is needed, since some aspects of the disease and targets for drug therapy are not present, e.g., peripheral immune cell infiltration. Since this is an important step in disease progression, the implementation of peripheral blood mononuclear cells may provide an additional complexity that helps identify new therapies. Moreover it will allow to assess the later stages of NASH progression and eventually model the reversion of the disease where peripheral blood macrophages play a relevant role for tissue remodeling 73 .
A second aspect is to be able to recapitulate more advanced stages of fibrosis and employ quantification methods that can be aligned with clinical diagnostic markers. To this end, we are currently working on testing N-terminal procollagen type III peptide (PIIINP) as an additional in vitro marker that is also used in clinics 74 , as well as the enhanced liver fibrosis (ELF) test 75 . Still the gold standard for assessment of NASH in the clinic is based on histology using liver biopsy slices. Therefore, the development of phenotypic quantification of fibrosis also using histology tissue slices in the NASH model is of high priority. For assessment of the translational relevance of the NASH hLiMTs model to the in vivo situation comparison experiments will be performed using liver tissue slices of clinical and rodent NASH.