Galunisertib | Inhibitor Screening

Galunisertib modifies the liver fibrotic composition in the Abcb4Ko mouse model

Seddik Hammad1,2 · Elisabetta Cavalcanti3 · Julia Werle1 · Maria Lucia Caruso3 · Anne Dropmann1 · Antonia Ignazzi3 · Matthias Philip Ebert4 · Steven Dooley1 · Gianluigi Giannelli3

Received: 11 February 2018 / Accepted: 23 May 2018
© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
Transforming growth factor (TGF)-β stimulates extracellular matrix (ECM) deposition during development of liver fibrosis and cirrhosis, the most important risk factor for the onset of hepatocellular carcinoma. In liver cancer, TGF-β is responsible for a more aggressive and invasive phenotype, orchestrating remodeling of the tumor microenvironment and triggering epi- thelial–mesenchymal transition of cancer cells. This is the scientific rationale for targeting the TGF-β pathway via a small molecule, galunisertib (intracellular inhibitor of ALK5) in clinical trials to treat liver cancer patients at an advanced disease stage. In this study, the hypothesis that galunisertib modifies the tissue microenvironment via inhibition of the TGF-β pathway is tested in an experimental preclinical model. At the age of 6 months, Abcb4ko mice—a well-established model for chronic liver disease development and progression—are treated twice daily with galunisertib (150 mg/kg) via oral gavage for 14 consecutive days. Two days after the last treatment, blood plasma and livers are harvested for further assessment, including fibrosis scoring and ECM components. The reduction of Smad2 phosphorylation in both parenchymal and non-parenchymal liver cells following galunisertib administration confirms the treatment effectiveness. Damage-related galunisertib does not change cell proliferation, macrophage numbers and leucocyte recruitment. Furthermore, no clear impact on the amount of fibrosis is evident, as documented by PicroSirius red and Gomori-trichome scoring. On the other hand, several fibrogenic genes, e.g., collagens (Col1α1 and Col1α2), Tgf-β1 and Timp1, mRNA levels are significantly downregulated by galunisertib administration when compared to controls. Most interestingly, ECM/stromal components, fibronectin and laminin-332, as well as the carcinogenic β-catenin pathway, are remarkably reduced by galunisertib-treated Abcb5ko mice. In conclusion, TGF-β inhibition by galunisertib interferes, to some extent, with chronic liver progression, not by reducing the stage of liver fibrosis as measured by different scoring systems, but rather by modulating the biochemical composition of the deposited ECM, likely affecting the fate of non-parenchymal cells.
Keywords Galunisertib · Abcb4ko · Liver fibrosis · TGF-β pathway

Seddik Hammad, Julia Werle and Maria Lucia Caruso share first authorship.
Steven Dooley and Gianluigi Giannelli share senior authorship.
Introduction
Transforming growth factor-beta (TGF-β) is a potent fibro- genic cytokine, playing a major role in activating hepatic stellate cells to produce extracellular matrix (ECM) proteins

leading to liver fibrosis (Dooley and Ten Dijke 2012). Depo-

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00204-018-2231-y) contains supplementary material, which is available to authorized users.

⦁ Seddik Hammad ⦁ [email protected]
⦁ Gianluigi Giannelli ⦁ [email protected]
Extended author information available on the last page of the article
sition of ECM in the liver is a reversible process, although worsening towards cirrhosis frequently occurs and represents the most important risk factor for the development of hepa- tocellular carcinoma (HCC) (Kocabayoglu and Friedman 2013). Fibrous ECM is mainly composed of various types of collagens, elastins, fibronectins and laminins (Alberts et al. 2007). Alterations in the ECM composition are associated with cell signaling modulation in the liver (Williams et al.

2014). In particular, laminin (Ln)-332 (previously named Ln-5) in a TGF-β context is implicated in HCC progression (Giannelli et al. 2005). A central role is displayed by the hepatic stellate cells in the tumor stroma that, after acti- vation by different factors including TGF-β, are acting as cancer-associated fibroblasts (CAFs) and secrete Ln-332 (Santamato et al. 2011), which in turn downregulates epithe- lial E-cadherin and translocates β-catenin into the nucleus, thereby facilitating EMT (Giannelli et al. 2005). The pres- ence of Ln-332 is an important factor in creating a favora- ble environment for the progression of HCC, by increasing the proliferative activity of cancer cells and correlates with the worst clinical outcome (Bergamini et al. 2007; Fransvea et al. 2009; Giannelli et al. 2014). In addition, there is accu- mulating evidence that TGF-β is a crucial EMT modulator in hepatocytes and HCC cells (Franco et al. 2010). Further- more, intermediate states of EMT and myofibroblast fate- determining events have recently been identified as potential triggers of organ fibrosis progression (Nieto et al. 2016). On the other hand, tissue inhibitors of matrix metalloproteases, including TIMP-1, ensuring ECM turn over finely tune the proteolytic remodeling of ECM by matrix metalloproteases and activating components stored within the fibrotic tissue in an inactive form, such as TGF-β (review by Bissell 2001). TGF-β signaling starts with binding of the active TGF-β ligand to the TGF-β receptor II (TβRII), followed by heter- odimerization with TβRI/ALK5. The activated TGF-βRII/ ALK5 complex initiates C-terminal phosphorylation of intracellular proteins Smad2 and Smad3. Most importantly, phosphorylated Smads translocate into the nucleus and acti- vate the downstream targets (Dooley and Ten Dijke 2012). Therefore, disease stage-specific molecular intervention in the TGF-β signaling pathway is a promising strategy to interfere with disease development and progression, which has been proven in vitro (George et al. 1999; Okuno et al. 2001; Dituri et al. 2013) and in vivo (Qi et al. 1999; De Gouville et al. 2005; Kim et al. 2006; Giannelli et al. 2014; Cao et al. 2017). A small molecule kinase inhibitor, galuni- sertib (LY2157299), directed against the ALK5 receptor, is in clinical use (e.g., NCT02423343 for HCC, Rodon et al.
2015).
Mice deficient in canalicular phospholipid flippase (Mdr2/Abcb4 knock out; hereafter named Abcb4ko) pre- sent a reproducible model of spontaneous chronic biliary liver disease progression associated with inflammation and fibrosis (Lammert et al. 2004; Popov et al. 2005). The model is based on the deletion of the Mdr2 gene, a mouse ortholog of the human MDR3 (ABCB4) gene. Lack of this gene causes disturbances in bile constituents that induce liver damage, biliary fibrosis and sclerosing cholangi- tis (Fickert et al. 2004). More importantly, the Abcb4ko mouse shows histological features of human primary scle- rosing cholangitis (Mauad et al. 1994). Given the lack of
antifibrotic therapies and the profibrogenic role of TGF-β, we tested galunisertib to revert fibrosis progression and regress liver disease severity in Abcb4ko mice.

Materials and methods
Animal experiments

Abcb4ko mice (on balb/c background) were maintained in a specific pathogen-free environment. Mice were housed in a fixed 12 h dark/light cycle and fed with normal chow and water, ad libitum. All experiments were performed in accordance with ethical standards, the Declaration of Helsinki, and according to national and international guidelines, and prior approval was obtained from the authors’ institutional review boards. Mouse age, gender and numbers are summarized in Table S1. In the first experiment, mice were housed for 12, 26, 28, 36, 44, 58, 72 and 84 weeks in identical conditions. In the second experiment, 24-week-old mice were treated twice a day with galunisertib for 14 consecutive days. Mice received either galunisertib (TβRI/ALK5-specific inhibitor; Sell- eckchem, Munich, Germany) or the vehicle compound via a gastric tube, or remained untreated, hereafter referred to as galunisertib, vehicle or untreated. The vehicle com- pound, in which 1 g of galunisertib is dissolved consists of
2.67 ml DMSO (Sigma-Aldrich, 41639, 500 ml, Missouri, USA), 8 ml PEG400 (Sigma-Aldrich, 202398, 500 g, Mis- souri, USA), 2.67 ml EtOH 70% (Roth, K928.4, Karlsruhe, Germany), 8 ml saline (0.9% Braun 250 ml, 13352450, Melsungen, Germany) and 5.3 ml 0.01 M HCl (10 M HCl Fluka, Cos 7647-01-0, diluted with distilled water, Mis- souri, USA). Two days after the last administration, mice were anaesthetized, then blood and livers were harvested for further analysis. In detail, the left liver lobes were col- lected in 4% paraformaldehyde (Roth, P087.1, Karlsruhe, Germany) and embedded in paraffin for histopathological investigations. Caudate and right liver lobes were snap- frozen for hydroxyproline, biochemical and molecular biology analyses.

Clinical chemistry

Blood was collected from retro bulbar plexus and centri- fuged at 14,000 rpm at 4 °C for 3 min. Plasma was pipet- ted in a new reaction tube and stored at − 20 °C. Then, liver enzymes alanine aminotransferase (ALT), aspartate aminotransferase (AST) and alkaline phosphatase (ALP) were measured in blood plasma using a Hitachi automatic analyzer.

Assessment of hepatic fibrosis

We analyzed 85 formalin-fixed paraffin-embedded tissue sections (FFPE) from the Abcb4ko mouse model at 12, 26, 28, 36, 44, 58, 72 and 84 weeks (Table S1). Then, we analyzed 19 FFPE tissue sections from Abcb4ko mice at 26 weeks that were subdivided in three groups: galunisertib, vehicle-treated and untreated (Table S1). Hematoxylin & Eosin (H&E, Merck, Darmstadt, Germany), PicroSirius red (PSR, Missouri, USA) and Gomori-trichrome (GT) stain- ing (Gomori 1950) were performed to assess liver fibrosis stages. Staging of inflammation and fibrosis were separately evaluated in H&E and GT stained slides by two independent pathologists. Fibrosis was staged with both METAVIR scor- ing (Bedossa and Poynard 1996) and the PBC stage system (Hiramatsu et al. 2006). With the METAVIR system, fibro- sis is scored as follows: score 0, no fibrosis; score 1, portal fibrosis without septa; score 2, fibrous septa portal-to-portal; score 3, fibrous septa portal-to-portal and portal-to-central; and score 4, surviving hepatocytes arranged in rounded regenerative nodules. The other evaluation system adopted was the PBC system in which three items are considered for disease staging: (1) the degree of fibrosis, (2) bile duct loss and (3) cholestasis. For every item, the pathologist assigns a score from 0 to 3. The final stage is obtained by adding the scores for the three items; 0 corresponds to stage I, no or minimum progression; 1–3 to stage 2, mild progression; 4–6 to stage 3, moderate progression; and 7–9 to stage 4, advanced disease. In addition, PSR positive areas were quan- tified from 15 images per mouse in the galunisertib experi- ment using ImageJ software (http://rsbweb.nih.gov/).
Immunohistochemistry (IHC)

IHC was used to study the expression of fibronectin (Fn), laminin-332 (Ln-332), β-catenin, pSmad2, CK19, CD45, F4/80 and Ki-67 in FFPE liver tissue sections. Liver tissues were sectioned and incubated with the following antibod- ies: anti-fibronectin (Invitrogen, MA5-11981, mouse, dilu- tion 1:500, Massachusetts, USA), anti-Ln-332, γ2 chain (Abcam, ab210959, mouse, dilution 1:500, Cambridge, England), anti-β-catenin (Agilent, M3539, mouse, dilution 1:200, Massachusetts, USA), anti-pSmad2 (Cell Signaling Technology, CST#3108, rabbit, dilution 1:100, Massachu- setts, USA), anti-CK19 (Proteintech, 14965-1-AP, rabbit, dilution 1:100, Illinois, USA), anti-CD45 (BD biosciences, 550539, rat, dilution 1:100, New Jersey, USA), anti-F4/80 (BioRad, MCA497, rat, dilution 1:2000, California, USA) and anti-Ki-67 (Cell Signaling Technology, CST#12202, rabbit, dilution 1:200, Massachusetts, USA). Slides were processed using immunohistochemical staining protocols according to Hammad et al. (2014) and Vartak et al. (2016). In the same staining batch, some slides were stained with
secondary antibodies only and used as a negative control (Fig. S3). Images of stained liver tissue were processed with ImageJ software to obtain the mean DAB (DAB tablets, Sigma-Aldrich, Missouri, USA) signal intensity value. Fif- teen fields were randomly chosen and examined at 200-fold magnification; the number of stained cells for each intensity score was counted.
Hepatic hydroxyproline determination

Hydroxyproline (HYP) was determined colorimetrically in triplicates from snap-frozen liver lobes as described in Cho and co-workers (2000). Briefly, approximately 100 mg of tissue from the caudate liver lobe was homogenized and hydrolyzed in 2 ml 6 N HCl at 110 °C for 16 h. Hydroxypro- line content was then measured photometrically at 558 nm. Based on relative hepatic HYP (per 100 mg of frozen liver), total hepatic HYP was calculated (total liver, as obtained by multiplying liver weights with relative hepatic HYP).
Isolation of mRNA, reverse transcription and qPCR

From pieces of the right middle liver lobes, mRNA was iso- lated with the InviTrap® Spin Universal RNA Mini Kit from Stratec (1060100300, Birkenfeld, Germany). In a peqStar machine, cDNA was produced from oligo(dT)18 primer (SO132, Thermo Scientific, Massachusetts, USA), dNTP Mix (R0191, Thermo Scientific Massachusetts, USA), and RevertAid H Minus Reverse Transcriptase (EP0451, Thermo Scientific, Massachusetts, USA), and then used for real-time (rt)-PCR (5× HOT FIREPol® EvaGreen® qPCR Mix Plus (ROX), 08-24-00020, Solis BioDyne, Tartu, Estonia) in a StepOne machine. Primers were Col1α1 (forward: GGA GAGAGCATGACCGATGG, reverse: AAGTTCCGGTGT GACTCGTG), Col1α2 (forward: 5-AGTCGATGGCTGCTC CAAAA-3, reverse: 5-AGCACCACCAATGTCCAGAG-3),
Timp1 (forward: GGCATCTGGCATCCTCTTGT, reverse: ACTCTTCACTGCGGTTCTGG), Tgf-β1 (forward: AGG GCTACCATGCCAACTTC, reverse: CCACGTAGTAGA
CGATGGGC). All primers were purchased from Eurofins Genomics (Ebersberg, Germany).
Statistical analysis

Correlations between fibrosis score and mouse age were sta- tistically analyzed by the Chi-square test, test z for propor- tions and the Kruskal–Wallis rank test. All P values were determined by two-sided tests and P values < 0.05 were con- sidered significant. Statistical analyses were performed with StataCorp. 2007 Stata Statistical Software: release 10 (Stata- Corp LP, College Station, TX, USA). Parametric data were analyzed using the two-tailed unpaired t test (Mann–Whit- ney test) and the means and SD were calculated.

84 (n = 11)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
2 (18.2) 0.01
5 (45.4) 0.14
4 (36.3) 0.02
Results
–
Development and progression of liver fibrosis in Abcb4ko mice

Table 1 METAVIR staging of liver fibrosis in 12 to 84-week-old Abcb4ko mice
Fibrosis stage
Mouse age (weeks)
12 (n = 12)
n (%)
26 (n = 6)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
1 (16.6) 0.02
4 (66.7) 0.04
1 (16.6) 0.32
28 (n = 10)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
6 (60.0) 0.76
4 (40.0) 0.24
0 (0.0) –
36 (n = 10)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
6 (60.0) 0.76
4 (40.0) 0.24
0 (0.0) –
44 (n = 12)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
3 (25.0) 0.03
8 (66.6) 0.007
1 (8.3) 0.32
58 (n = 11)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
7 (63.6) 0.88
4 (36.4) 0.30
0 (0.0) –
72 (n = 12)
n (%) Pf vs.
12 weeks
0 (0.0) 0.14
4 (33.3) 0.10
8 (66.6) 0.007
0 (0.0) –
F0 F1 F2 F3 F4
2 (16.6)
8 (66.6)
2 (16.6)
0 (0.0)
–
–
–
–
–
–
–
–
–
–
–
–
–
–
Correlation analysis is based on test z vs. 12-week-old mice
Pf is calculated based in test z for proportions
P value = 0.003 according to Chi-square test
To trace chronic liver disease dynamics in Abcb4ko mice, blood and livers were subjected to standard fibrosis inves- tigation using METAVIR and PBC scoring on H&E, GT and PSR stained livers from mice aged between 12 and 84 weeks. METAVIR is more suitable for evaluating active hepatic injury and fibrosis, while PBC is more specific for primary biliary cirrhosis (PBC) scoring, and thus a perfectly suitable index for biliary fibrosis of Abcb4ko mice. In 12-week-old mice, we observe a preva- lence of histologic lesions in stages F1 (66.67%) and stage 2 (83.34%) at METAVIR and PBC scoring, respectively (Table 1, Table S2). Particularly in F1, periportal fibrosis and enlargement without fibrotic septa is visible (Fig. 1a, b). Only 40% of Abcb4ko mice at this age show F2 with portal-to-portal bridging septa (Fig. 1c). Twenty-eight- week-old mice similarly present with F1 (60%) and F2 (40%) fibrosis using METAVIR scoring, whereas 90% of the livers are dedicated stage 2 in PBC scoring (Fig. 1b, c; Table 1, Table S2). At 36 weeks, there is no significant progress in fibrosis scoring using METAVIR as compared to 28-week livers. PBC scoring results in 70% stage 2 and 30% stage 3 fibrosis (Fig. 1c; Table 1, Table S2). In 44-week-old mice, F2 is prevalent (66.67%) using METAVIR, whereas all livers present with stage 2 with PBC scoring. Major morphological peculiarities in this disease stage are neo-cholangiogenesis (ductular prolif- eration), and periductular and pre-tombstone (onion skin pattern) fibrosis (Figs. 1c, 2; Table 1, Table S2). At the last progression stage (84 weeks), the prevalence of F3 (36.36%) and stage 3 (45.4%) increased compared to the F2 and stage 2 scores (Figs. 1c, 2; Table 1, Table S2). This disease stage is associated with increasing numbers of portal-to-portal septa, appearance of portal-to-central septa, progressive interlobular bile duct loss and intracel- lular cholestasis (Figs. 1d, 2; Table 1, Table S2). Particu- larly at 84 weeks, Abcb4ko mice present with portal-to- central septa, pseudo-nodule formation due to thickening of the fibrous tissue, alterations in the biliary epithelium, ductular ectasia and chronic inflammation with giganto- cellullar elements around cholesterol crystals (Figs. 1d, 2; Table 1, Table S2). Neither F4 fibrosis nor cirrhotic nodules are observed in the tested mice. However, a sig- nificant correlation between liver fibrosis progression was revealed by METAVIR (Table 1, P < 0.003), PBC scoring (Table S2; P < 0.004) and mouse age (weeks). In line with the tissue-based analyses, blood plasma liver enzymes ALT, AST and ALP were determined. Surprisingly,

Fig. 1 Staging of liver fibrosis dynamics in 12–84-week-old Abcb4ko mice HE, GT and PSR stained tissue. Two different methods, META- VIR and PBC, were applied for liver fibrosis scoring. Scoring results are summarized in Table 1 and Table S2. a Non-fibrotic livers are observed in 16.67% of 12-week-old mice. b F1 (stage 1) fibrosis characterized as periportal fibrosis accompanied by enlargement of

enzyme levels were only slightly increased in 26-week- old mice compared to 12-week-old mice (Fig. 2e–g). Therefore, 24-week-old mice were chosen for therapeu- tic targeting of the TGF-β signaling pathway. In sum- mary, time-resolved histological investigations confirm that Abcb4ko mice indeed represent a preclinical model
portal spaces (yellow arrow). c F2 (stage 2) fibrosis with several por- tal–portal bridging septa (yellow arrow). (d) F3 (stage 3) fibrosis, as recorded in mice at advanced ages (72 and 84 weeks); several portal– portal bridging septa (yellow arrow), as well as portal–central bridg- ing septa (white arrow) are present. Scale bars are 200 µm. Images are representatives of 6–12 mice per group

that pathomorphologically phenocopies progressive bil- iary fibrosis as evident in human patients with primary sclerosing cholangitis.

Hematoxylin&Eosin
a

d
Picro-Sirius red
Cytokeratin-19

e
80 0

AL T [ U/ l]
60 0

40 0

20 0

80 0

AS T [ U/ l ]
f
60 0

40 0

20 0

1250

g
1000

AL P [U / l]
75 0

50 0

25 0

0
12 26 28 36 44 58 72 84
Mo us e a ge (W eeks)

0
12 26 28 36 44 58 72 84
Mo us e a ge (W eeks )

Fig. 2 Time-resolved analyses of ductular reactions and blood liver enzyme levels in Abcb4ko mice. Representative images of liv- ers stained with H&E, PSR and CK19 are presented. a, b Typical examples of periductular fibrosis (onion skin pattern) without bile duct epithelial cell (BDEC) proliferation (white arrows). PSR posi- tive areas (yellow arrows). c At this stage, so-called “pre-tombstone aggregate” portal lymphoid structures are present (yellow arrow). d
BDEC proliferation, as already obvious from H&E and confirmed by CK19 positive staining (yellow arrows). Scale bars are 100 µm. e–g Liver enzymes ALT, AST and ALP as analyzed in a time-resolved manner. Data are means ± SD of 6–12 mice per group. N.S.: P > 0.05;
*P < 0.05; **P < 0.01; ***P < 0.001 compared with 12-week-old mice

a
pSmad2
Untreated

Vehicle

Galunisertib

pS m a d2 +
n o n- p a re nc hy m a l n uc le i pe r fie ld [n ]
b 200 # # c
pS m a d2 + he pa to cyte nu c lei pe r fie ld [n ]
175

150

125

100

25
22 5 # #

20 0

17 5

15 0

12 5

10 0

0 0

Fig. 3 SMAD2 phosphorylation upon 2 weeks galunisertib admin- istration in 26-week-old Abcb4ko mice. a Representative images of (nuclear) pSMAD2 immunostaining from untreated, vehicle and galunisertib-treated livers. Scale bars are 100 and 50 µm for ×20 and ×40 magnification, respectively. b, c Separate quantification of pSMAD2-positive nuclei in hepatocytes and non-parenchymal liver
cells. Galunisertib significantly decreases pSMAD2 levels in both cell types, as compared with vehicle and untreated mice. Data are means ± SD of 5–7 mice per group. N.S.: P > 0.05; **P < 0.01 com- pared with untreated livers; * replaced by # for comparison between the indicated bars

b
Hy d r ox yp ro line [ µ g /g ]
7

PS R + ar ea [% ]
6 1000

5
800

4
600
3
400
2

1 200

0 0

1. 5

Co l1 α 1
[ Fo ld ch an ge ]
1. 0

0. 5

0. 0

1. 4

[ Fo ld c h a n ge ]
1. 2

1. 0

0. 8

Co l 1 α 2
0. 6

0. 4

0. 2

0. 0

4.0

[ Fo ld ch an ge ]
3.5

3.0

2.5

2.0

Tim p 1
1.5

1.0

0.5

0.0

1.5

Tg f - β 1
[ F o ld c ha ng e]
1.0

0.5

0.0

◂Fig. 4 Staging of liver fibrosis upon galunisertib treatment using H&E, GT and PSR staining in 26-week-old Abcb4ko mice. a, b Rep- resentative images of PSR and GT staining from untreated, vehicle
and galunisertib-treated livers. Scale bars are 200 µm. b Quantifi- cation of PSR-positive areas in 15 images per mouse using ImageJ. Compared with untreated livers, galunisertib-treated livers show no clear impact on fibrosis. c Liver tissue hydroxyproline levels. d–g mRNA levels of fibrogenic genes Col1α1, Col1α2, Timp1 and Tgf-β1 across different groups. Data are means ± SD of 5–7 mice per group. N.S.: P > 0.05; *P < 0.05; **P < 0.01; compared with untreated livers,
* replaced by # for comparison between the indicated bars

Galunisertib modulates the TGF‑β pathway by reducing Smad2 phosphorylation
without affecting proliferation, or macrophage and leucocyte infiltration

To prove efficacy of galunisertib-mediated ALK5 inhibi- tion in vivo, we investigated intracellular Smad phospho- rylation in 24-week-old Abcb4ko mice upon exposition to galunisertib for 2 weeks. pSMAD2-positive nuclei were separately quantified in parenchymal (hepatocyte nuclei) and non-parenchymal cells, and both numbers were sig- nificantly reduced after galunisertib treatment (P < 0.01) compared to vehicle or untreated livers (Fig. 3a–c). Next, we investigated proliferative activity as Ki-67 positivity and found no significant impact of galunisertib on hepatocytes and non-parenchymal cells (Fig. S1a–c). Similarly, no sta- tistical difference is obvious in F4/80 (macrophage marker) and CD45 (leucocyte marker)-positive cell numbers between galunisertib and vehicle-treated mice (Fig. S1a, d, e). We conclude that galunisertib is able to modulate TGF-β signal- ing in different liver cell types, but without obvious effects on proliferation and inflammation.
Targeting the TGF‑β pathway has no obvious effect on the amount of ECM deposits in Abcb4ko mice

To test damage reducing and antifibrotic effects of interfer- ence with the TGF-β signaling pathway, blood and livers were analyzed further. Galunisertib administration showed no classic signs of toxicity, as illustrated by body and liver weight results (Fig. S2a–c). Liver enzymes ALT, AST and ALP show no significant alteration after galunisertib treat- ment, when compared with other groups (Fig. S2d–f). His- tological analyses are performed with H&E (Fig. S2g), GT and PSR-stained liver samples harvested from Galunisertib, vehicle-treated and untreated Abcb4ko mice. Galunisertib- treated mice and control groups similarly show F2 and F3 scores (Fig. 4a). Also, quantification of PSR-positive areas displays similar amounts of deposited collagen in all treat- ment groups (Fig. 4b). In line with these results, optical microscopy of GT-stained tissues does not provide clearly improved fibrosis between treated and untreated groups,
although the percentage of F2 fibrosis livers is slightly higher in galunisertib-treated mice, when using this stain- ing method (Table 2). These data are also reflected in the hydroxyproline analysis, which shows no significant differ- ences between the treatment groups (Fig. 4c). Otherwise, qPCR analysis of the fibrosis genes Col1α1, Col1α2, Timp1 and Tgf-β1 (Fig. 4d–g) reveals a significant downregulation of mRNA in the galunisertib group. Taken together, these results indicate that galunisertib-respective TGF-β signal- ing inhibition does not remarkably alter ECM deposition and histologically visible fibrosis, although mRNA levels of some typical fibrosis genes are significantly decreased.
Targeting TGF‑β modulates ECM composition in Abcb4ko mice

The genetically induced damage in Abcb4ko mice progresses to HCC. Furthermore, we know from previous studies that TGF-β may modulate hepatic stellate cell (HSC)/myofi- broblast fate towards cancer-associated fibroblasts (CAFs) in liver cancer, which is accompanied by significant ECM remodeling. We, therefore, investigated the effect of galuni- sertib on CAF-ECM and EMT components, which we have previously shown as controlled by TGF-β in HCC (Gian- nelli et al. 2016). Indeed, fibronectin, laminin (Ln)-332 and β-catenin are all strongly expressed in Abcb4ko mice at the age of 26 weeks (Fig. 5). Thereby, β-catenin is mainly expressed in hepatocytes located along the periendothelial membrane and in the cytoplasm of non-parenchymal cells, galunisertib treatment dramatically abrogates expression of these tumorigenesis facilitating TGF-β targets (P < 0.05 for fibronectin and Ln-332, and P < 0.01 for β-catenin), as compared with vehicle-exposed livers. (Fig. 5a). These data reveal that in Abcb4ko mice, TGF-β has rather a matrix remodeling than clear pro-fibrotic effect, and further indicate that galunisertib may interfere with TGF-β-driven tumori- genesis instead of inhibiting fibrosis.

Discussion
TGF-β activates HSCs that are the major source of ECM in liver fibrosis and cirrhosis. In this study, we wanted to confirm that blocking the TGF-β pathway could improve liver fibrosis by decreasing ECM deposition in Abcb4ko mice. This well-established genetic model presents with chronic biliary liver disease comprising early disease stages with periportal fibrosis and inflammation, up to periportal and septal bridging fibrosis and finally HCC (Mauad et al. 1994). Classical signs of biliary liver disease dynamics can be observed age dependently in these mice, as dissected in the present report with METAVIR scoring (Bedossa and Poynard 1996) and PBC staging (Hiramatsu et al. 2006).

Table 2 METAVIR scoring of liver fibrosis in Abcb4ko mice upon galunisertib administration

F0
n (%) F1
n (%) F2
n (%) F3
n (%) F4
n (%)
Untreated (n = 6) 0 0 1 (16.7) 4 (56.7) 1 (16.7)
Vehicle (n = 5) 0 0 2 (40) 3 (60) 0
Galunisertib (n = 7) 0 0 4 (57) 3 (43) 0
Fig. 5 Expression of ECM proteins upon galunisertib administration ▸ in 26-week-old Abcb4ko mice. a Representative images of fibronec- tin, laminin 332-γ2 and β-catenin immunostaining from untreated,
vehicle and galunisertib-treated livers. Scale bars are 100 µm. b–d quantification of fibronectin, laminin 332-γ2 and β-catenin-stained positive areas as percentage of the total investigated area. Upon gal- unisertib treatment, positive areas were decreased compared with vehicle and untreated groups. Data are means ± SD of 5–7 mice per group. N.S.: P > 0.05; *P < 0.05; **P < 0.01 compared with untreated

mice, * replaced by # for comparison between the indicated bars
This scoring system is based on GT staining, as shown in Fig. 4a

We herewith underscore their usefulness in preclinical drug testing for future clinical trials, as it was already stated previ- ously (Fickert et al. 2004; Popov et al. 2005; Ikenaga et al. 2015). 24-week-old Abcb4ko mice display F2–F3 fibrosis and increased liver enzyme ALT, AST and ALP levels as compared to 12-week-old mice. Therefore, we selected this age to estimate the potential antifibrotic effect of gal- unisertib. In agreement with ex vivo studies in human liver slices (Serova et al. 2015; Luangmonkong et al. 2017), gal- unisertib treatment decreases pSmad2-positive staining in hepatocytes and non-parenchymal cells, including HSCs. No “healing” effect of galunisertib was observed regard- ing ECM deposition, although some fibrosis-related genes, including Collagen1 and Timp1 are downregulated at the mRNA level. This apparent discrepancy can be explained, e.g., the investigated Collagen/ECM genes are responsible only for a limited amount of ECM components deposited, while GT and PSR staining evaluate the whole fibrotic tis- sue. Furthermore, mRNA expression changes are not in any case immediately translated into protein data. This fact needs further and more detailed investigation. On the other hand, the decrease in Ln-332 and Fn protein expression upon galunisertib treatment is consistent with TGF-β signaling inhibition in HSCs, as these cells have been identified as the main source of both proteins, although not necessarily in fibrosis settings, but in the context of tumor stroma of liver cancer (Santamato et al. 2011). A shift in ECM composition is usually associated with an altered tissue microarchitecture and elasticity (An et al. 2009; Carraher and Schwarzbauer 2013) and has major impact on tumor cell behavior. One such impact is altered expression and nuclear translocation of β-catenin as a consequence of TGF-β triggered EMT, and together with increased expression of Ln-332, contributes to a more favorable microenvironment for HCC spread (Rygiel
et al. 2008; Fransvea et al. 2009; Yoshida et al. 2016). In particular, we previously showed that Ln-332 expression lev- els are positively correlated with HCC growth (Bergamini et al. 2007). Based on the present results, we suggest that inhibition of the TGF-β pathway in Abcb4ko mice with liver disease that progressed for 26 weeks interferes with tumor facilitating ECM rearrangements rather than with decreasing the total amount of fibrosis. This unveils a possible scenario, where TGF-β signaling paves the way for liver disease pro- gression into precancerous stages by regulating the cross- talk between stressed hepatocytes and stroma fibroblasts.
Therefore, we think the observed effects of galunisertib on Ln-332, Fn and β-catenin suggest a potential use of the drug to avoid carcinogenesis instead of blunting fibrogen- esis, as originally hypothesized. This being consistent with data previously described in preclinical experimental mod- els Giannelli and co-workers (2005). In this study, TGF-β1 cooperates with Ln-5 to induce EMT in noninvasive HCC cell lines as indicated by downregulation of Ecadherin, upregulation of snail and slug as well as nuclear transloca- tion of beta-catenin.
In conclusion, TGF-β signaling inhibition by galunisertib interferes with chronic liver disease progression by shifting the deposited ECM constituents, likely affecting the fate/ mesenchymal transition of non-parenchymal cells. This seems to be consistent with the effect of TGF-β in orchestrat- ing the microenvironment organization and composition to facilitate generation, growth and progression of HCC cells. Further investigations are required to understand the mecha- nistic influence of ECM remodeling on disease progression and regression. Further studies, e.g., with Abcb4ko mice at an older age that present with HCC, will be necessary to prove the findings and the subsequent current conclusions.

a
Fibronectin Laminin 332-γ2 β-Catenin

#
b c
F ibr on ec tin + sta ine d ar e a [% ]
8 10

La m in in 3 32 - γ 2 +
sta ine d a re a [% ]
7
8
6

5 6
4
3 4

2
2
1

35 # #

sta ine d
d
30

ar ea [% ]
25

β -C aten in +
20

0 0 0

Acknowledgements We thank Christof Dormann and Friedrich Behne for their excellent technical support.

Author contributions SH, JW and AD conceived the study, performed data analyses and wrote the manuscript. EC, MLC and AI performed the pathological evaluation. MPE, SD and GG performed critical revision of the manuscript. SD and GG provided supervisory support and corrected the manuscript. All authors read the final version of the manuscript.

Funding This work was supported by the BMBF program LiSyM (SD: Grant PTJ-FKZ: 031 L0043), e:Bio-Modull-II: MS_DILI and by the Italian Ministry of Health, Ricerca Corrente 2018.
Compliance with ethical standards

Conflict of interest The authors declare no conflict of interest.

References
Alberts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P (2007) Molecular biology of the cell. Garland Science, London
An SS, Kim J, Ahn K, Trepat X, Drake KJ, Kumar S et al (2009) Cell stiffness, contractile stress and the role of extracellular matrix. Biochem Biophys Res Commun 382(4):697–703. https://doi. org/10.1016/j.bbrc.2009.03.118
Bedossa P, Poynard T (1996) An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology 24(2):289–293. https://doi.org/10.1002/hep.51024
0201
Bergamini C, Sgarra C, Trerotoli P, Lupo L, Azzariti A, Antonaci S et al (2007) Laminin-5 stimulates hepatocellular carcinoma growth through a different function of alpha6beta4 and alpha- 3beta1 integrins. Hepatology 46(6):1801–1809. https://doi. org/10.1002/hep.21936
Bissell DM (2001) Exp Mol Med 33(4):179–190. (Review)
Cao Y, Agarwal R, Dituri F, Lupo L, Trerotoli P, Mancarella S et al (2017) NGS-based transcriptome profiling reveals biomarkers for companion diagnostics of the TGF-β receptor blocker galunisertib in HCC. Cell Death Dis 8(2):e2634. https://doi.org/10.1038/cddis
.2017.44
Carraher CL, Schwarzbauer JE (2013) Regulation of matrix assembly through rigidity-dependent fibronectin conformational changes. J Biol Chem 288(21):14805–148014. https://doi.org/10.1074/jbc. M112.435271
Cho JJ, Hocher B, Herbst H, Jia JD, Ruehl M, Hahn EG et al (2000) An oral endothelin-A receptor antagonist blocks collagen synthesis and deposition in advanced rat liver fibrosis. Gastroenterology 118(6):1169–1178
de Gouville AC, Boullay V, Krysa G, Pilot J, Brusq JM, Loriolle F et al (2005) Inhibition of TGF-beta signaling by an ALK5 inhibitor protects rats from dimethylnitrosamine-induced liver fibrosis. Br J Pharmacol 145(2):166–177. https://doi.org/10.1038/sj.bjp.0706172 Dituri F, Mazzocca A, Fernando J, Papappicco P, Fabregat I, De Santis F et al (2013) Differential inhibition of the TGF-β sign- aling pathway in HCC cells using the small molecule inhibitor LY2157299 and the D10 monoclonal antibody against TGF-β receptor type II. PLoS One 8(6):e67109. https://doi.org/10.1371/
journal.pone.0067109
Dooley S, Ten Dijke P (2012) TGF-β in progression of liver disease. Cell Tissue Res 347(1):245–256. https://doi.org/10.1007/s0044 1-011-1246-y (Review)
Fickert P, Fuchsbichler A, Wagner M, Zollner G, Kaser A, Tilg H et al (2004) Regurgitation of bile acids from leaky bile ducts causes sclerosing cholangitis in Mdr2 (Abcb4) knockout mice. Gastroenterology 127(1):261–274. https://doi.org/10.1053/j.gastr o.2004.04.009
Franco DL, Mainez J, Vega S, Sancho P, Murillo MM, de Frutos CA et al (2010) Snail1 suppresses TGF-beta-induced apoptosis and is sufficient to trigger EMT in hepatocytes. J Cell Sci 123(Pt 20):3467–3477. https://doi.org/10.1242/jcs.068692
Fransvea E, Mazzocca A, Antonaci S, Giannelli G (2009) Targeting transforming growth factor (TGF)-βRI inhibits activation of beta1 integrin and blocks vascular invasion in hepatocellular carcinoma. Hepatology 49(3):839–850. https://doi.org/10.1002/hep.22731
George J, Roulot D, Koteliansky VE, Bissell DM (1999) In vivo inhibi- tion of rat stellate cell activation by soluble transforming growth factor beta type II receptor: a potential new therapy for hepatic fibrosis. Proc Natl Acad Sci USA 96(22):12719–12724. https:// doi.org/10.1073/pnas.96.22.12719
Giannelli G, Bergamini C, Fransvea E, Sgarra C, Antonaci S (2005) Laminin-5 with transforming growth factor-beta1 induces epi- thelial to mesenchymal transition in hepatocellular carcinoma. Gastroenterology 129(5):1375–1383. https://doi.org/10.1053/j. gastro.2005.09.055
Giannelli G, Villa E, Lahn M (2014) Transforming growth factor-β as a therapeutic target in hepatocellular carcinoma. Cancer Res 74(7):1890–1894. https://doi.org/10.1158/0008-5472.CAN-14-
0243 (Review)
Giannelli G, Mikulits W, Dooley S, Fabregat I, Moustakas A, ten Dijke P et al (2016) The rationale for targeting TGF-β in chronic liver diseases. Eur J Clin Invest 46(4):349–361. https://doi.org/10.1111/ eci.12596 (Review)
Gomori G (1950) A rapid one-step trichrome stain. Am J Clin Pathol 20(7):661–664
Hammad S, Hoehme S, Friebel A, von Recklinghausen I, Othman A, Begher-Tibbe B et al (2014) Protocols for staining of bile cana- licular and sinusoidal networks of human, mouse and pig livers, three-dimensional reconstruction and quantification of tissue microarchitecture by image processing and analysis. Arch Toxi- col 88(5):1161–1183. https://doi.org/10.1007/s00204-014-1243-5
Hiramatsu K, Aoyama H, Zen Y, Aishima S, Kitagawa S, Nakanuma Y (2006) Proposal of a new staging and grading system of the liver for primary biliary cirrhosis. Histopathology 49(5):466–478. https
://doi.org/10.1111/j.1365-2559.2006.02537.x
Ikenaga N, Liu SB, Sverdlov DY, Yoshida S, Nasser I, Ke Q et al (2015) A new Mdr2(-/-) mouse model of sclerosing cholangi- tis with rapid fibrosis progression, early-onset portal hyperten- sion, and liver cancer. Am J Pathol 185(2):325–334. https://doi. org/10.1016/j.ajpath.2014.10.013
Kim KH, Kim HC, Hwang MY, Oh HK, Lee TS, Chang YC et al (2006) The antifibrotic effect of TGF-beta1 siRNAs in murine model of liver cirrhosis. Biochem Biophys Res Commun 343(4):1072– 1078. https://doi.org/10.1016/j.bbrc.2006.03.087
Kocabayoglu P, Friedman SL (2013) Cellular basis of hepatic fibrosis and its role in inflammation and cancer. Front Biosci (Schol Ed) 5:217–230 (Review)
Lammert F, Wang DQ, Hillebrandt S, Geier A, Fickert P, Trauner M et al (2004) Spontaneous cholecysto- and hepatolithiasis in Mdr2-
/- mice: a model for low phospholipid-associated cholelithiasis. Hepatology 39(1):117–128. https://doi.org/10.1002/hep.20022
Luangmonkong T, Suriguga S, Bigaeva E, Boersema M, Oosterhuis D, de Jong KP et al (2017) Evaluating the antifibrotic potency of galunisertib in a human ex vivo model of liver fibrosis. Br J Phar- macol 174(18):3107–3117. https://doi.org/10.1111/bph.13945
Mauad TH, van Nieuwkerk CM, Dingemans KP, Smit JJ, Schinkel AH, Notenboom RG et al (1994) Mice with homozygous disruption of the mdr2 P-glycoprotein gene. A novel animal model for studies

of nonsuppurative inflammatory cholangitis and hepatocarcino- genesis. Am J Pathol 145(5):1237–1245
Nieto MA, Huang RY, Jackson RA, Thiery JP (2016) EMT: 2016. Cell 166(1):21–45. https://doi.org/10.1016/j.cell.2016.06.028 (Review)
Okuno M, Akita K, Moriwaki H, Kawada N, Ikeda K, Kaneda K et al (2001) Prevention of rat hepatic fibrosis by the protease inhibitor, camostat mesilate, via reduced generation of active TGF-beta. Gastroenterology 120(7):1784–17800. https://doi.org/10.1053/ gast.2001.24832
Popov Y, Patsenker E, Fickert P, Trauner M, Schuppan D (2005) Mdr2 (Abcb4)-/- mice spontaneously develop severe biliary fibrosis via massive dysregulation of pro- and antifibrogenic genes. J Hepatol 43(6):1045–1054. https://doi.org/10.1016/j.jhep.2005.06.025
Qi Z, Atsuchi N, Ooshima A, Takeshita A, Ueno H (1999) Block- ade of type beta transforming growth factor signaling prevents liver fibrosis and dysfunction in the rat. Proc Natl Acad Sci 96(5):2345–2349. https://doi.org/10.1073/pnas.96.5.2345
Rodon J, Carducci MA, Sepulveda-Sánchez JM, Azaro A, Calvo E, Seoane J et al (2015) First-in-human dose study of the novel trans- forming growth factor-β receptor I kinase inhibitor LY2157299 monohydrate in patients with advanced cancer and glioma. Clin Cancer Res 21(3):553–560. https://doi.org/10.1158/1078-0432. CCR-14-1380
Rygiel KA, Robertson H, Marshall HL, Pekalski M, Zhao L, Booth TA et al (2008) Epithelial-mesenchymal transition contributes to portal tract fibrogenesis during human chronic liver disease. Lab Invest 88(2):112–123. https://doi.org/10.1038/labinvest.3700704
Santamato A, Fransvea E, Dituri F, Caligiuri A, Quaranta M, Niimi T et al (2011) Hepatic stellate cells stimulate HCC cell migration via laminin-5 production. Clin Sci (Lond) 121(4):159–168. https
://doi.org/10.1042/CS20110002
Serova M, Tijeras-Raballand A, Dos Santos C, Albuquerque M, Paradis V, Neuzillet C et al (2015) Effects of TGF-beta signalling inhibi- tion with galunisertib (LY2157299) in hepatocellular carcinoma models and in ex vivo whole tumor tissue samples from patients. Oncotarget 6(25):21614–21627. https://doi.org/10.18632/oncot arget.4308
Vartak N, Damle-Vartak A, Richter B, Dirsch O, Dahmen U, Ham- mad S et al (2016) Cholestasis-induced adaptive remodeling of interlobular bile ducts. Hepatology 63(3):951–964. https://doi. org/10.1002/hep.28373
Williams MJ, Clouston AD, Forbes SJ (2014) Links between hepatic fibrosis, ductular reaction, and progenitor cell expansion. Gas- troenterology 146(2):349–356. https://doi.org/10.1053/j.gastr o.2013.11.034
Yoshida K, Murata M, Yamaguchi T, Matsuzaki K, Okazaki K (2016) Reversible human TGF-β signal shifting between tumor suppression and fibro-carcinogenesis: implications of smad phospho-isoforms for hepatic epithelial-mesenchymal transi- tions. J Clin Med 5(1):E7. https://doi.org/10.3390/jcm5010007 (Review)

Affiliations
Seddik Hammad1,2 · Elisabetta Cavalcanti3 · Julia Werle1 · Maria Lucia Caruso3 · Anne Dropmann1 · Antonia Ignazzi3 · Matthias Philip Ebert4 · Steven Dooley1 · Gianluigi Giannelli3

1 Molecular Hepatology Section, Department of Medicine II, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany
2 Department of Forensic and Toxicology, Faculty
of Veterinary Medicine, South Valley University, Qena, Egypt
3 National Institute of Gastroenterology, “S. de Bellis” Research Hospital, Castellana Grotte, Bari, Italy
4 Department of Medicine II, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany