Characterization of liver disease regression in murine models of liver fibrosis and portal hypertension Authors: Philipp Königshofer, Ksenia Brusilovskaya, Benedikt S. Hofer, Oleksandr Petrenko, Benedikt Simbrunner, Katharina Bareiner, Kerstin Zinober, Katharina Regnat, Vlad Taru, Thomas Sorz-Nechay, Katharina Bonitz, Hubert Scharnagl, Tatjana Stojakovic, Michael Trauner, Stefan G. Kauschke, Larissa Pfisterer, Philipp Schwabl, Thomas Reiberger Published: 2026 ABSTRACT Characterization of toxin-induced mouse models of regressive liver disease, focusing on the detecting of a suitable therapeutic window during fibrosis and portal hypertension (PH) regression. C57BL/6JRj mice were studied when exposed to thioacetamide (TAA) or carbon tetrachloride (CCl4) for 8 or 12 weeks (8w/12w) and after one/two (+R1/+R2) week(s) of TAA/CCl4-free regression. Healthy controls received NaCl or olive oil (OO). Primary readouts included liver fibrosis (collagen proportionate area, CPA), portal pressure (PP), serum transaminases and hepatic fibrosis-related gene expression. The diseased models (8wTAA/12wTAA/8wCCl4/12wCCl4) showed significant fibrosis (CPA: 3.03/3.87/8.53/16.90%; vs. NaCl/OO: 0.87/1.15%) and PH (PP: 6.8/7.7/8.5/10.0 mmHg; vs. NaCl/OO: 5.5/5.1 mmHg). CPA significantly regressed in 8wTAA+R1 (2.60%; p = 0.01) and 12wCCl4+R1 (10.9%; p = 0.01), but not in 12wTAA+R1/8wCCl4+R1. CPA regression was significant in all +R2-groups (p < 0.001) but not in 8wCCl4+R2 (p = 0.30). Significant PP regression was found at 8wCCl4+R1 (5.8 mmHg; p = 0.004) and 12wCCl4+R1 (8.4 mmHg; p = 0.01), but not in the 8w/12w TAA models. All +R2-groups showed significant PP regression: 8wTAA+R2 (6.0 mmHg; p ≤ 0.001), 12wTAA+R2 (6.0 mmHg), 8wCCl4+R2 (6.0 mmHg, p = 0.004) and 12wCCl4+R2 (7.9 mmHg, p = 0.008). Serum AST/ALT levels at +R1/+R2 returned to levels of healthy controls. Hepatic Col1a1 and Timp1 gene expression peaked in 12w-groups and showed a gradual decrease at +R1/+R2: Col1a1 (12wTAA+R1/+R2: -14%/-29%, p = 0.002; 12wCCl4+R1/R2: -28%/-23%, p = 0.05) and Timp1 (12wTAA+R1/+R2: -41%/-61%, p ≤ 0.001; 12wCCl4+R1/R2: -43%/-40%, p ≤ 0.001). While liver disease regressed in all models/timepoints, both liver fibrosis (CPA) and PP remained increased in the 12wCCl4+R1/R2 settings – qualifying this model as suitable for pre-clinical testing of therapeutic strategies aimed at promoting liver fibrosis and PH regression.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-54627-5. Introduction Liver fibrosis results from repetitive or chronic liver injury1,2 and may progress to cirrhosis and portal hypertension (PH)3. Hepatic liver fibrosis is characterized by excessive accumulation of extracellular matrix (ECM)4 which is clinically assessed by semi-quantitative pathohistological scores on liver biopsies5 or continuous quantitative assessment of the collagen proportionate area (CPA)6. Liver fibrosis represents the main structural driver of PH. Accordingly, CPA correlates well with portal pressure (PP), yet this correlation is affected by the disease etiology7,8. PH significantly contributes to the development of severe complications, such as variceal bleeding or ascites9,10. While in the clinical setting, PH can be indirectly assessed via the hepatic venous pressure gradient (HVPG)11, in animal models PP can be directly measured via cannulation of the portal venous vascular bed12. Historically, extensive liver fibrosis was believed to be irreversible, yet evidence has accumulated, that removal of the underlying liver injury enables liver fibrosis regression, even in cirrhosis: Convicting data emerged from patients with viral hepatitis B13 or hepatitis C14, receiving antiviral therapeutics, and also from patients with metabolic dysfunction-associated liver disease (MASLD)15 or alcohol abstinence in ALD, who succeeded in changing their respective lifestyle habits. Importantly, the removal of the primary etiological factor is a prerequisite for the recompensation criteria defined by Baveno VII16 and supports the link of disease regression to recompensation17. In experimental settings natural liver fibrosis regression was observed in rat models after cessation of the disease-inducing agents such as thioacetamide (TAA)18 and carbon tetrachloride (CCl4)19, or reverting to a standard chow after a high fat diet20. TAA and CCl4 are hepatotoxic compounds. Their use is established in experimental liver fibrosis models for studying disease mechanisms of hepatic fibrogenesis21 and of PH22 and for pre-clinical drug testing23,24. Both the TAA and CCl4 model qualify for experimental studies focusing on liver disease regression, as the toxic stimulus can easily halted25. However, the course of liver fibrosis regression and dynamics of PH have not been systematically evaluated, considering disease severity grades and regression period. Thus, we aimed to characterize TAA and CCl4 mouse models at different time points of peak fibrosis and of disease regression duration in regard to hepatic fibrosis and PP as primary readouts. The goal was to identify a model with an optimal ‘therapeutic window’ for preclinical testing of drugs aimed at promoting regression of liver disease. Methods Animal models and study design In total, 124 male C57BL/6JRj mice (obtained from Janvier Labs) with 10 weeks of age, were used in this study. According to the liver fibrosis modelling agent, they were distributed into TAA and CCl4 groups, which were further subdivided according to the fibrosis induction duration, being either 8 or 12 weeks to reach different disease severity levels. In each of these four different settings, 8 mice were allocated to a healthy control (NaCl/OO), a diseased control (TAA/CCl4), or a regression group adding one (+R1) or two (+R2) weeks of profibrogenic agent free time before the experimental endpoint (Fig. 1A). All mice were housed and handled according to the standards of care and maintenance at the Core Facility Laboratory animal breeding and husbandry (Medical University of Vienna), including ad libitum access to standard laboratory chow and water, housing under standardized housing conditions (day/night cycle 12:12 h; room temperature 22 ± 2 °C, humidity 55 ± 10%) and had a two-week acclimatization period prior to the experiment start, where groups were randomized considering an equal body weight (BW) distribution. Animals were weighed 3x/week using a precision balance (KERN & Sohn, model 572 − 39), equipped with an animal weighing (averaging) function, that compensates for movement by calculating a stable mean value from fluctuating measurements. Fig. 1(A) Experimental timeline of each animal model sorted by induction agent of liver disease (TAA vs. CCl4) and induction period (8 vs. 12 weeks). (B) Liver fibrosis baselines (healthy vs. diseased) of each regression model. (C) Representative images of the final morphometry readout of CPA. (D) Portal hypertension baseline (healthy vs., diseased) of each regression model. (E) Body weight development of the study groups timelines - induced with either NaCl/OO or TAA/CCl4. (Significant p values are stated within each graph. Statistical tests used: One-way ANOVA corrected by Dunnett, or Kruskal-Wallis corrected by Dunn tests were used for group comparisons.) Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (CPA) collagen proportionate area, (PH) portal hypertension, (PP) portal pressure, (w) week. (A) Experimental timeline of each animal model sorted by induction agent of liver disease (TAA vs. CCl4) and induction period (8 vs. 12 weeks). (B) Liver fibrosis baselines (healthy vs. diseased) of each regression model. (C) Representative images of the final morphometry readout of CPA. (D) Portal hypertension baseline (healthy vs., diseased) of each regression model. (E) Body weight development of the study groups timelines - induced with either NaCl/OO or TAA/CCl4. (Significant p values are stated within each graph. Statistical tests used: One-way ANOVA corrected by Dunnett, or Kruskal-Wallis corrected by Dunn tests were used for group comparisons.) Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (CPA) collagen proportionate area, (PH) portal hypertension, (PP) portal pressure, (w) week. TAA was injected intraperitoneally (i.p.) in 0.9% saline (NaCl) at a dose of 150 mg/kg BW and CCl4 was gavaged (p.o.) as a 20% v/v solution in olive oil (OO) at a dose of 2 mL/kg BW three times a week (Mo/Wed/Fri). At start of the induction period of 8 (8w) or 12 weeks (12w) in total, the dose of TAA/CCl4 was gradually increased over 14 and 7 days, respectively, to avoid excessive toxicity related BW drops. The healthy controls received only the vehicle substances, being NaCl (i.p.) or OO (p.o.), respectively. Groups and experimental timelines are summarized in (Fig. 1A). Health monitoring, including BW measurements, was performed 3 times per week including a detailed health score to ensure model safety (see Supplemental Digital Content). Euthanasia procedures were performed by cervical dislocation according to state-of-the-art practice in laboratory animal science, in line with the humane endpoint defined in the ethical approval. Hemodynamic measurements Hemodynamic readouts were performed under general anesthesia and followed a strict time management. Animals received ketamine (100 mg/kg) and xylazine (7.5 mg/kg) i.p., to reach surgical-plane anesthesia, which was confirmed by a toe-pinch test. After laparotomy was performed under a microscope (VisiScope SZT250, VWR, Radnor, USA), the portal vein got exposed and the PP was measured following a direct cannulation of the portal vein with a 26G needle (BD Microlance 3; 0.45 × 16 mm, REF:303300; Becton Dickinson S.A., Spain). The needle was attached to a polythene tubing (0.4 mm ID; 800/100/140; Smiths Medical, Minneapolis, USA), which was connected to a pressure transducer (MLT0380/D ADInstruments, Colorado Springs, USA). PP was recorded for 1 min using a multichannel recorder (ML870 PowerLab 8/30 ADInstruments, Colorado Springs, USA) and under stable hemodynamic conditions (Fig. 3B, C). Subsequently, blood was withdrawn to measure serum levels of aspartate transaminase (AST), alanine transaminase (ALT), bilirubin, bile acids, and glucose. After euthanasia of the mice, spleen and liver were weighed. The lateral left liver lobe was used for CPA assessment, the right and left median liver lobe for molecular analyses and the superior right lateral lobe in its entirety for evaluation of hepatic hydroxyproline levels. Fig. 2(A) Level of portal pressure as direct portal hypertension marker at respective endpoint of each study groups timeline, (B,C) measured by canulation of the portal vein under standardized surgical plane anesthesia. (D) Liver/body weight ratio as basic marker of liver disease and (E) Spleen/body weight ratio as indirect marker of portal hypertension. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (BW) body weight. (A) Level of portal pressure as direct portal hypertension marker at respective endpoint of each study groups timeline, (B,C) measured by canulation of the portal vein under standardized surgical plane anesthesia. (D) Liver/body weight ratio as basic marker of liver disease and (E) Spleen/body weight ratio as indirect marker of portal hypertension. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (BW) body weight. Morphometric analyses of liver fibrosis The left lateral liver lobe was cut along its main arterial vessel to achieve a representative area of cross sections for the whole lobe in at least 2 liver tissue slices. Histological slides of these cross sections were stained with a picro-sirius red (PSR; Morphisto, Nr. 13425) / 0.1% fast-green (FG; REF: 104022, C.I.42053, Sigma-Aldrich, St.Louis, USA) combination and slides were digitalized in high resolution of 40x magnification using a digital research slide scanner (Slideview VS200; OLYMPUS). Using quantitative morphometry software (HALO, V3.3.2541.184, Indica Labs) liver tissue was analyzed for PSR- compared to FG-stained areas, which was considered as collagen proportionate area (CPA). Molecular analysis of liver tissue Genes related to liver fibrogenesis and regression were selected to follow the disease course, including collagen type 1 alpha 1 (Col1a1), transforming growth factor beta (Tgfb), alpha-smooth muscle actin (Acta2/ commonly referred to alpha-SMA), and tissue inhibitor of metalloproteinases (Timp1). Log fold changes were calculated after qRT-PCR against Peptidylprolyl isomerase A (PPia) as housekeeping gene. Details on method and materials in Table-S1. Statistics GraphPad Prism version 10.1.1 (GraphPad Software, Inc., San Diego, CA, USA) was used for all statistical analyses and the significance level was set at 0.05 (95% confidence interval). Data sets of each study group were tested for normal distribution (Shapiro-Wilk test). Subsequently, either Welch`s t-test or Mann-Whitney test with Welch`s correction was used for comparison of healthy control with the peak disease groups or the +R1, or the +R2 groups, such as for the comparison within healthy and peak disease groups (using # as significance indicator). One-way ANOVA corrected by Dunnett, or Kruskal-Wallis corrected by Dunn tests were used for group comparisons between regression groups +R1/+R2 with peak diseased groups (8w/12wTAA & 8w/12wCCl4) as the reference group (using * as significance indicator). Correlation coefficients (Rho) were calculated by Spearman’s correlation. Written results are presented as mean ±standard deviation of the mean. All graphical data are presented as mean ±standard error of the mean (SEM) for reasons of presentation. Significance indicators are used in the following manner: (*/# p ≤ 0.05; **/## p ≤ 0.01; ***/### p ≤ 0.001). Ethics Procedures involving animal subjects have been approved by Federal Ministry of Education, Science and Research in Austria by identification number BMBWF-V/3b/2020 − 0.398.007. All experiments were performed in accordance with the guidelines of the Laboratory Animal Care and Use Committee (ACUC) of the Medical University of Vienna and Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines, next to the 3Rs26 and FELASA guidelines27. Results Baseline characteristics of used animal models Group randomization based on initial body weight at start of the model timeline resulted in comparable (n.s.) baseline weights across all groups of the NaCl/TAA model (23.5 ± 0.9 g) and the OO/CCl4 model (20.0 ± 0.9 g). Healthy control groups showed no significant difference between each other regarding fibrosis levels (CPA: 8wNaCl: 0.89 ± 0.14%; 12wNaCl: 0.87 ± 0.20%; 8wOO: 1.55 ± 0.89%; 12wOO: 0.83 ± 0.16%) or the degree of PH (PP: 8wNaCl: 5.39 ± 0.29 mmHg; 12wNaCl: 5.51 ± 0.42 mmHg; 8wOO: 4.92 ± 0.78 mmHg; 12wOO: 5.19 ± 0.20 mmHg). Model characteristics including all gathered data are all summarized in detail in Table-S2. Peak disease groups showed ascending levels of severity in liver fibrosis according to disease induction duration: 8wTAA: 3.03 ± 0.18%, 12wTAA: 3.87 ± 0.36% (8wTAA vs. 12wTAA, p ≤ 0.001), 8wCCl4: 8.53 ± 5.04%, 12wCCl4: 16.9 ± 3.90% (8wCCl4 vs. 12wCCl4, p = 0.02) (Fig. 1B); with CPA being significantly increased compared to the respective healthy groups: 8wTAA vs. 8wNaCl (p ≤ 0.001); 12wTAA vs. 12wNaCl (p ≤ 0.001); 8wCCl4 vs. 8wOO (p = 0.01); and 12wCCl4 vs. 12wOO (p ≤ 0.001) (Fig. 3A). In line, the degree of portal hypertension rose with disease duration in both settings: 8wTAA: 6.80 ± 0.35 mmHg; 12wTAA: 7.71 ± 0.57 mmHg (8wTAA vs. 12wTAA, p ≤ 0.001); 8wCCl4: 8.45 ± 1.28 mmHg; 12wCCl4: 9.99 ± 1.17 mmHg (8wCCl4 vs. 12wCCl4, p = 0.16) (Fig. 1D) and were all significantly higher as the respective healthy controls: 8wTAA vs. 8wNaCl (p ≤ 0.001); 12wTAA vs. 12wNaCl (p ≤ 0.001); 8wCCl4 vs. 8wOO (p ≤ 0.001); and 12wCCl4 vs. 12wOO (p ≤ 0.001) (Fig. 2A). Fig. 3(A) Collagen proportionate area and (B) Level of hepatic hydroxyproline as liver fibrosis marker at respective endpoint of each study groups timeline. (C) Representative images of the liver to compare the macroscopic view, the picro-sirius red/fast green stained liver tissue and the final morphometry readout of CPA across a whole liver lobe section at the 8week model of NaCl/TAA, (D) the 12week model of NaCl/TAA, (E) the 8week model of OO/CCl4, (F) and the 12week model of OO/CCl4. (Significant p values are stated within each graph. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (PSR/FG) picro-sirius red/fast green. (A) Collagen proportionate area and (B) Level of hepatic hydroxyproline as liver fibrosis marker at respective endpoint of each study groups timeline. (C) Representative images of the liver to compare the macroscopic view, the picro-sirius red/fast green stained liver tissue and the final morphometry readout of CPA across a whole liver lobe section at the 8week model of NaCl/TAA, (D) the 12week model of NaCl/TAA, (E) the 8week model of OO/CCl4, (F) and the 12week model of OO/CCl4. (Significant p values are stated within each graph. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks, (PSR/FG) picro-sirius red/fast green. Cessation of disease-causing agent leads to histological fibrosis regression CPA levels showed a significant regression of liver fibrosis at the 8wTAA regression model after +R1 (8wTAA+R1: 2.60 ± 0.28%; p = 0.01) and +R2 (8wTAA+R2: 2.34 ± 0.21%; p ≤ 0.001) compared to its peak at 8wTAA (3.03 ± 0.18%). In the 12wTAA regression model, fibrosis regression was only significant in the +R2 setting (12wTAA+R2: 2.90 ± 0.30%; p ≤ 0.001) compared to 12wTAA (3.87 ± 0.36%) (Fig. 3 A). In the 8wCCl4 regression model no significant liver fibrosis regression was seen, while in the 12wCCl4 regression setting (16.90 ± 3.89%) significant fibrosis regression was measurable within one week (12wCCl4+R1: 10.90 ± 3.05%; p = 0.01) and two weeks (12wCCl4+R2: 5.43 ± 1.44%; p ≤ 0.001). Despite visible improvements in CPA content, no regression group returned to a healthy state of liver fibrosis. Representative histological images of PSR/FG-stained liver tissue are shown in Fig. 3C–F. In contrast, the hepatic hydroxyproline levels showed no significant difference between any of the groups (Fig. 3B). When studying relative organ weights, we noted that liver/BW ratio decreased to healthy values after one (8wTAA+R1, p = 0.01) and two weeks (8wTAA+R2, p = 0.01) of regression compared to 8wTAA, while in the 12wTAA setting only the two-week regression reached a significant reduction (12wTAA+R2 vs. 12wTAA, p = 0.007) (Fig. 2D). Again, the 8wCCl4 regression model showed no significant difference in liver/BW ratio during the regression periods. Conversely, in the 12wCCl4 regression setting a significant increase of liver/BW ratio was notable after one (p = 0.03) and two weeks (p = 0.005) of regression, respectively. This observation was attributed to higher absolute liver weights in the 12wCCl4+R1 and 12wCCl4+R2 groups as compared to 12wCCl4, while in the other settings the absolute liver weight remained similar between the groups (Fig. S2A). Fibrosis regression improves portal hypertension Portal pressure decreased significantly following two weeks of regression in the TAA models lasting 8 (8wTAA+R2 vs. 8wTAA: 6.01 ± 0.22 vs. 6.80 ± 0.35 mmHg; p ≤ 0.001) and 12 weeks (12wTAA+R2 vs. 12wTAA: 6.04 ± 0.52 vs. 7.71 ± 0.57 mmHg; p ≤ 0.001), respectively (Fig. 2A), while no significant differences were visible after only one week of regression in either TAA setting. In contrast, a significant decrease of portal pressure was already detectable after one week of regression in both of the CCl4 models, as shown in 8wCCl4+R1 (5.84 ± 0.24 vs. 8.45 ± 1.28 mmHg, p = 0.004) and in 12wCCl4+R1 (8.35 ± 0.79 vs. 9.99 ± 1.17 mmHg, p = 0.01). Two weeks of regression confirmed the significant decrease against the diseased control (8wCCl4+R2: 6.02 ± 0.92 mmHg, p = 0.004; 12wCCl4+R2: 7.89 ± 1.16 mmHg, p = 0.008), but did not add to the PP reduction. Notably, disease regression in the 8wCCl4 model led to PP values comparable to the healthy control group (8wCCl4+R1/+R2 vs. 8wOO, p = ns). In addition, the spleen/BW ratio, has been measured, as this is a clinically meaningful PH surrogate marker. Indeed, a trend towards a spleen size reduction was seen in all regression groups, but only with statistical significance in 8wCCl4+R1 (p = 0.02 vs. 8wCCl4) and 12wCCl4+R2 (p = 0.05 vs. 12wCCl4) (Fig. 2E). Necroinflammation reacts quickly to cessation of disease-causing agents Serum levels of AST and ALT were significantly decreased in all regression groups (Fig. 4A, B), as the glucose levels recovered to healthy values to significant extent (Fig. 4C), compared to respective peak disease groups. Bile acids were increased to a mild level in the TAA model, while regression period allowed a significant/full reduction to normal values. In contrast, bile acid levels rose sharper in the CCl4 model, yet without any significant changes after the regression period (Fig. 4D). Bilirubin levels were only significantly increased in the regression model of 12wCCl4 (p < 0.01) compared to 12wOO. However, bilirubin levels were significantly decreased in 12wCCl4+R1 (p = 0.02) compared to 12wCCl4 (Fig. S2B). Fig. 4Blood serum levels of each respective study group or model timeline: (A) Aspartate transaminase (AST), (B) Alanine transaminase (ALT), (C) Glucose level and (D) Bile acids. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks. Blood serum levels of each respective study group or model timeline: (A) Aspartate transaminase (AST), (B) Alanine transaminase (ALT), (C) Glucose level and (D) Bile acids. Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+R1) including a regression period of 1 week, (+R2) including a regression period of 2 weeks. Correlation between fibrosis and portal pressure is attenuated at regression TAA animals at peak disease showed significant correlation between CPA and PP (8wTAA & 12wTAA: Rho = 0.556; p = 0.03) – such as all CCl4 animals (8wCCl4 & 12wCCl4: Rho = 0.807; p ≤ 0.001.) The TAA regression groups showed as well significant correlation (Rho = 0.437; p = 0.01) while the CCl4 regression groups lacked in significance (Rho = 0.250; ns). Peak disease and regression animals exhibited distinct regression trajectories (Fig. S3A, B). Hepatic fibrosis-related gene expression during regression A significant regression of Col1a1 was shown at +R1 compared to the peak disease groups in 8wTAA+R1 (-22%, p = 0.03) and 12wCCl4+R1 (-28%, p = 0.006), and ultimately in all +R2 groups: 8wTAA+R2 (-37%, p ≤ 0.001), 12wTAA+R2 (-29%, p = 0.002), 8wCCl4+R2 (-26%, p ≤ 0.001) and 12wCCl4+R2 (-23%, p < 0.05). Yet Col1a1 expression was still elevated in all regression groups, as compared to the respective healthy controls (Fig. 5A). Expression of Timp1 showed similar data (Fig. 5B), with a significant reduction in all +R1 and +R2 groups: 8wTAA+R1/+R2 (-45%/-58%, both p ≤ 0.001), 12wTAA+R1/+R2 (-41%/-61%, both p ≤ 0.001), 8wCCl4+R1/+R2 (-35%/-45%, both p ≤ 0.001) and 12wCCl4+R1/+R2 (-43%/-40%, p ≤ 0.001). Fig. 5Liver fibrosis related gene expression levels of bulk liver tissue: (A) Collagen type alpha 1 (Col1a1), (B) tissue inhibitor of metalloproteinases (Timp1), (C) Transforming growth factor beta (Tgfb) and (D) Alpha-smooth muscle actin (Acta2/ commonly referred to alpha-SMA). Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+ R1) including a regression period of 1 week, (+ R2) including a regression period of 2 weeks. Liver fibrosis related gene expression levels of bulk liver tissue: (A) Collagen type alpha 1 (Col1a1), (B) tissue inhibitor of metalloproteinases (Timp1), (C) Transforming growth factor beta (Tgfb) and (D) Alpha-smooth muscle actin (Acta2/ commonly referred to alpha-SMA). Statistical indicators: Significant differences between regression groups +R1/+R2 and peak diseased groups are indicated on top of the graph bar using *: p ≤ 0.05; **: p ≤ 0.01; ***: p ≤ 0.001, while significant comparisons compared to healthy control are indicated under the graph using: #: p ≤ 0.05; ##: p ≤ 0.01; ###: p ≤ 0.001. Abbreviations: (NaCl) sodium chloride, (OO) olive oil, (TAA) thioacetamide, (CCl4) carbon tetrachloride, (+ R1) including a regression period of 1 week, (+ R2) including a regression period of 2 weeks. Expression of Tgfb showed only a significant decrease in the regression groups of the 8wTAA model (8wTAA+R1: p ≤ 0.001; 8wTAA+R2: p ≤ 0.001) compared to peak disease (Fig. 5C). In the other three models, regression did not affect Tgfb expression patterns. The expression of Acta2 was mainly statistically inconclusive, with a slightly decrease across all regression groups as compared to the diseased controls (Fig. 5D). Evaluation of therapeutic window during 2 weeks of liver disease regression For pre-clinical drug testing aiming at promotion of liver fibrosis regression and PH reduction, three conditions should be fulfilled: (i) At peak disease, both liver fibrosis (CPA) and PH (PP) should be significantly different to healthy controls (p < 0.001), but also (ii) after a certain natural regression time-period these parameters should remain at significantly elevated levels (p < 0.001 vs. healthy controls) to allow assessment of a therapeutic “acceleration/promotion” effect. (iii) Ideally, the model identified by condition 1 and 2 includes a 2-week regression period to maximize the therapeutic window duration. Thus, based on our observations, the ideal model and settings that fulfilled these requirements are the 12-week CCl4 model after two weeks (+ R2) of regression (PP/CPA: 12wCCl4 vs. 12wOO (p < 0.001/p < 0.001); 12wCCl4+R2 vs. 12wOO (p < 0.001/p < 0.001) (Table S3). Discussion Investigation of liver disease regression and liver regeneration are key players for a better understanding of hepatic recompensation, including the broad variety of underlying mechanism17. In this study, we investigated various settings of two liver fibrosis models (i.e., TAA and CCl4) at different peaks of liver disease severity (i.e., after either 8 or 12 weeks of induction) and at different time points of liver disease regression (i.e., respectively after one or two weeks of regression, respectively). While both models differ in their inducing agent metabolization and the resulting liver injury type28,29, they represent well-established animal models of advanced liver fibrosis and portal hypertension22. In neither model or setting we observed safety signals and body weight developments were stable, considering that a stepwise increase of the TAA and CCl4 dosing schedule was followed in the first induction weeks. During the regression period, animals showed significant/considerable body-weight increases, reaching levels close to healthy controls. The peak fibrosis severity was milder in the TAA models than in the CCl4 models, however, fibrosis severity as assessed by CPA showed a greater variation (i.e., standard deviation) in the CCl4 groups than in the TAA groups. Severity of liver fibrosis remains a key prognostic factor in patients with liver disease and in turn, regression of fibrosis, e.g. in non-alcoholic steatohepatitis (NASH) is associated with a reduction of the risk of liver related complications30. Regression of liver fibrosis occurred in nearly all performed models within two weeks (except of the 8wCCl4 model) as evidenced by significantly decreased CPA values. The lack of significance in liver fibrosis regression in the 8wCCl4 model even at the R2 timepoint may be due to the slightly higher variation in CPA in this model. However, the lack of a significant CPA (fibrosis) decrease at the R1 timepoint after 8w-TAA induction, points towards biological differences distinct for this TAA-induced fibrosis pattern, as we observed only minimal variation in fibrosis severity (i.e., CPA) within different animals at the same TAA model timepoints. Interestingly, hepatic hydroxyproline levels as another liver fibrosis surrogate parameter showed no significant difference across all groups, however this unexpected finding was only caused by the lack of significant increase - despite clear numerical trends - at the peak fibrosis timepoints. While hepatic hydroxyproline levels seem a valuable complementary fibrosis readout, the direct quantitative histological fibrosis assessment via CPA at whole liver lobe level proved to be the more representative and accurate liver fibrosis assessment. Further, the lack of significant differences in hydroxyproline content may reflect the limited sensitivity of this assay to detect moderate and distinct structural changes in collagen fiber maturity, especially when spatial heterogeneity is expected; as well as differences between biochemical and histological readouts, such as sampling variability and fiber maturity/collagen crosslinking. The observation of increase in liver weight (i.e., liver-to-body weight ratio) during early fibrosis regression in the 12wCCl4 model may be explained by changes in liver morphology associated with CCl4-induced fibrosis, including enlarged and nodular architecture, which gradually remodel during regression31. Further, hepatocellular regeneration is initiated rapidly upon cessation of injury, whereas ECM degradation and resolution of inflammation may follow delayed kinetics, resulting in a transient imbalance favoring tissue mass expansion3. Finally, since this effect was most pronounced in the model with the highest peak liver disease severity, the magnitude of the regenerative response seems linked to the severity of prior injury, with more advanced fibrosis eliciting stronger cellular activation and tissue remodeling32–suggesting that a critical threshold of prior injury may be required. Regarding PH regression, the TAA models showed a steady decrease in PP from peak level to a significant lower level after one and two weeks of regression, respectively. In contrast, the PP in the CCl4 models were already significantly decreased after one week, yet not so much further reduction was seen subsequently in the +R2 setting. This faster initial decline in PP in the CCl4 model may indicate a more pronounced dynamic/inflammatory PH component in this model, compared to the TAA setting. In the TAA model, the regression of CPA and PP go parallel hand in hand after cessation of TAA. In contrast, in the CCl4 model PP decreased rapidly in the first regression week to a level expected from the remaining ‘static‘ PH component (i.e., liver fibrosis) and with a further (less pronounced) decrease of PP following liver fibrosis regression at R2. Consequently, distinct types and severities of liver injury seem to cause differences in fibrosis-related ‘static‘ components of PH both in progressive and regressive liver disease8. Findings at correlation of CPA and PP suggest that, during fibrosis regression, the coupling between structural changes (fibrosis) and functional parameters (portal pressure) is attenuated, supporting the notion that the dynamic component of vascular dysfunction becomes less pronounced in this phase. A rapid cessation of liver injury following removal of TAA and CCl4 exposure was indicated by the ‘normalization’ of serum AST/ALT levels – similar to levels observed in the healthy control groups – already at the R1 regression timepoints. In addition, also the blood glucose levels – likely representing the physiological hepatic function of gluconeogenesis – returned to similar levels as observed in the healthy controls already at R1. Interestingly, serum bile acids levels demonstrated a difference between the type of model (TAA vs. CCl4), with higher peak levels measured in the CCl4 groups. In addition, bile acid levels returned to normal in the TAA models at R1 and R2, while showing only a partly regression in the CCl4 groups at R1 and R2 time points. Thus, the different patterns of bile acid levels suggest a delayed and potentially persistent impact of liver injury on bile acid metabolism in regressive liver disease following CCl4 liver injury but not following TAA-related liver injury. Importantly, we have previously demonstrated a link between disturbed bile acid signaling to liver disease and PH severity in patients with cirrhosis33. On a molecular level, hepatic gene expression of Col1a1 and Timp1 was significantly reduced within two weeks of regression but remained significantly increased compared to healthy controls. These findings are in line with the persistently higher degree of liver fibrosis (i.e., CPA) versus healthy control groups. Interestingly, the regulation of Tgfb and Acta2 gene expression was less consistent along the regression timeline – suggesting that fibrogenic pathways regress with distinct kinetics with Col1a1 and Timp1 reflecting more dynamic markers of active ECM remodeling during liver disease regression. Important limitations of this study need to be discussed: Neither the TAA nor the CCl4 model do reflect all complex molecular pathomechanisms observed in human liver disease, as e.g. in alcohol-related liver cirrhosis that usually arises over years or even decades in patients. Liver disease in animal models is induced within a short time frame, at a different age, and obviously in a different species resulting in a different severity level of liver fibrosis and PH that does not recapitulate all clinical aspects of PH level observed in patients. Furthermore, the hemodynamic characterization of the PH syndrome was performed only by PP measurement, not capturing other parameters of the PH-associated hyperdynamic circulation. Additionally, there are also limitations of hemodynamic assessment related to anesthesia, and the type and depth of anesthesia used during hemodynamic assessment may impact systemic, splanchnic and portal hemodynamics. Specifically, anesthesia may lower systemic arterial and portal venous pressure by inducing systemic vasodilation, reducing cardiac output, and attenuating sympathetic nervous tone. Thus, following a standardized anesthesia protocol seems mandatory to create the required framework for statistical group comparisons. In conclusion, our characterized models seem suitable for experimental studies aiming at investigating different molecular aspects and timepoints of liver disease regression. For preclinical studies focusing on perpetuation of mild liver fibrosis, the 8w or 12w TAA model seems well suited due to low fibrosis variation (i.e., CPA levels with low standard deviation). The TAA-induced fibrosis severity level may be further increased by adaptation of dosage, or prolongation of the induction timeline. On the other hand, the 12wCCl4 model seems best qualified for pre-clinical studies on pharmacologic interventions aiming at promoting liver fibrosis regression, with specifically the 2-weeks regressing period still showing persistent PH and fibrosis, thus offering a suitable therapeutic window. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1