Targeting the liver clock improves fibrosis by restoring TGF-β signaling

Abstract

Background & Aims

Liver fibrosis is the major driver for hepatocellular carcinoma and liver disease related death. Approved anti-fibrotic therapies are absent and compounds in development have limited efficacy. Increased TGF-β signaling drives collagen deposition by hepatic stellate cells (HSC)/myofibroblasts. Here, we aimed to dissect the role of the circadian clock (CC) in controlling TGF-β signaling and liver fibrosis.

Methods

Using CC-mutant mice, enriched HSCs and myofibroblasts obtained from healthy and fibrotic mice in different CC-phases and loss-of-function studies in human hepatocytes and myofibroblasts, we investigated the relationship between CC and TGF-β signaling. We explored hepatocyte-myofibroblast communication through bioinformatic analyses of single-nuclei transcriptomes and validation in cell-based models. Using mouse models for MASH fibrosis and spheroids from patients with liver disease, we performed proof-of-concept studies to validate pharmacological targetability and clinical translatability.

Results

We discovered that the CC-oscillator temporally gates TGF-β signaling and this regulation is broken in fibrosis. We demonstrate that HSCs and myofibroblasts contain a functional CC with rhythmic expression of numerous genes, including fibrogenic genes. Perturbation studies in hepatocytes and myofibroblasts revealed a reciprocal relationship between TGF-β-activation and CC perturbation, which was confirmed in patient-derived ex vivo and in vivo models. Pharmacological modulation of CC-TGF-β signaling inhibited fibrosis in mouse models in vivo as well as patient-derived liver spheroids.

Conclusion

The CC regulates TGF-β signaling, and the breakdown of this control is associated with liver fibrosis in patients. Pharmacological proof-of-concept studies across different models uncover the CC as a therapeutic candidate target for liver fibrosis – a rising global unmet medical need.

Graphical abstract. Regulation of TGFβ expression by the liver circadian clock in the healthy and fibrotic liver.

In the healthy liver, the CC-oscillator controls the diurnal expression of numerous genes including TGFβ-SMAD pathway members. In MASH-induced liver fibrosis, the perturbation of the CC-oscillator leads to dysregulation of the TGFβ-SMAD pathway and perturbed hepatocyte and myofibroblast communication. Pharmacological targeting of the CC by an REV-ERBα agonist reduces TGF-β activity and restores cell-cell communication with inhibition of fibrosis progression.

Introduction

Persistent tissue injury initiates a fibrotic response as a wound healing program. Liver fibrosis is a common step in the pathogenesis of chronic liver disease (CLD) caused by viral, alcoholic, and metabolic liver injury. Advanced fibrosis is the major cause of liver disease-related mortality and can lead to hepatocellular carcinoma (HCC), one of the leading causes of cancer-related death[1]. A major obstacle for the development of efficient and safe anti-fibrotic therapies is the lack of understanding of the mechanisms driving CLD progression[2,3].

Due to high prevalence and rising incidence, metabolic dysfunction-associated steatotic liver disease (MASLD) progressing to metabolic dysfunction-associated steatohepatitis (MASH) is a major cause of liver fibrosis world-wide. Advanced fibrosis (stage F3, F4) is associated with decompensation, increased HCC risk, and death[4].

The pathogenesis of liver fibrosis is multi-factorial and mediated an interplay of hepatocytes, macrophages, and myofibroblasts[2]. While the transforming growth factor-β (TGF-β) pathway has been shown to play an essential role in fibrosis[2], the molecular mechanisms dysregulating TGF-β-controlled gene expression during fibrosis are only partially understood.

The circadian clock (CC) regulates several functions in the liver. The ‘clock’ functioning largely depends on the CC-oscillator, a transcriptional-translational feedback system[5]. The CC-oscillator drives daily rhythmicity of its own components as well as that of numerous target genes, through transcriptional, and post-transcriptional mechanisms ensuring liver homeostasis with BMAL1 and REV-ERBα/β as central regulators[5]. While CC perturbation has been shown to be associated with liver disease, the functional role of CC as a driver and therapeutic target for liver fibrosis is unknown.

Results

The liver CC regulates TGF-β-related gene networks

To investigate the functional role of the liver CC-oscillator (Fig. 1A) in the regulation of TGF-β-target gene expression in non-diseased conditions, we first analyzed published circadian transcriptomes from livers of control and Bmal1 KO mice[6] using gene set enrichment analysis (GSEA) (Fig. 1B)[7,8]. In this model, the transcriptional activator Bmal1 is knocked-out (Bmal1 KO) in every tissue including the liver. We found that several TGF-β-target genes and hepatic stellate cell (HSC) activation signatures[9] are rhythmically expressed in control mice (Fig. 1B, Table S1 and S2), and this regulation is perturbed by the Bmal1 KO (Fig. 1B). We found that profibrotic transcription factors (TFs) Smad3 and Atf3 display circadian expression, and observed ultradian transcription pattern for Col3a1, Mmp2, Mmp9 and Mmp13, which are altered in Bmal1 KO mouse livers (Fig. S1A). We also performed analyses from livers of hepatocyte-specific Bmal1 KO mice (Bmal1^hep-/-) and littermate controls[10] (Fig. 1C). In control liver both hepatocyte- and HSC-enriched TGF-β-targets are rhythmically expressed (Fig. 1C). Importantly, in Bmal1^hep-/- livers, the circadian expression of these TGF-β-targets and of the Bmal1-controlled CC gene Rev-Erbα was abrogated (Fig. 1C). Chromatin immunoprecipitation followed by quantitative polymerase chain reaction (ChIP-qPCR) to analyze DNA-binding of BMAL1 and its partner CLOCK (Fig. 1A) revealed their rhythmic recruitment to promoter-enhancers of TGF-β-targets (Fig. 1D), a pattern identical to its binding to Dbp and Rev-Erbα. In Bmal1^hep-/- livers, BMAL1/CLOCK-binding to these TGF-β-targets was lost (Fig. 1D). Analyses of BMAL1/CLOCK-genome-wide binding in mouse livers[9] also confirmed its binding to Smad7 and Smad3 (Fig. S1B).

Fig. 1. TGF-β signaling-related genes display circadian rhythmicity in the mouse liver.

(A) Model of the CC-oscillator. (B) Heatmaps showing expression of HALLMARK TGF-β gene sets (left panel) and murine orthologs of human HSC-specific genes[9] (right panel) in liver of control and Bmal1 KO mice (GSE135898)[6]. (C) Normalized transcripts levels of indicated genes in liver of control and Bmal1^hep-/- mice[10]. (D) ChIP-qPCR of indicated genes using control and Bmal1^hep-/- liver[10] and indicated antibodies. (E) Normalized expression of indicated genes in control and Rev-Erbα/β^hep-/- liver (GSE143528)[11]. Data are expressed as mean ±SD. Fig. 1C, D were performed from same control and Bmal1^hep-/- mice. 3 mice/ time-point/group. Fig. 1C, D: Unpaired student’s T-test; bold dots: p < 0.05. Fig. 1E: Wald test and Benjamini-Hochberg correction; bold dots: FDR<0.05. Refer to Tables S1 and S2.

We then evaluated the relationship between the CC and TGF-β targets by analyzing their expression in the liver of hepatocyte-specific Rev-Erbα/β KO (Rev-Erbα/β^hep-/-; LDKO) mice[11]. Transcript analyses showed altered circadian expression of known profibrotic genes (Fig. 1E and S1C). In addition, we found a circadian variation of plasma TGF-β levels, which likely originates from multiple tissues including the liver, correlating with a diurnal activation of SMAD 2 and 3 (pSMAD2/3), and increased levels of other TGF-β targets in mouse livers (Fig. S1D-S1E). The ChIP assays revealed that SMAD2/3, SMAD4 binding to their cognate SMAD binding elements (SBE) in Smad7, Acta2, and Tgfbr1 is circadian and paralleling recruitment of RNA polymerase II (PolII) was reduced in Bmal1^hep-/- liver (Fig. S2A-S2C). Additionally, our analyses of an HSC-specific transcriptional signature which includes several TGF-β-target genes[12] revealed that their murine orthologues have a rhythmic expression in control liver, and this rhythmicity was disrupted upon Bmal1 KO (Fig. 1B and Table S2). Finally, we also analyzed the circadian transcription of several genes regulating TGF/BMP/Activin pathways (SMAD-dependent and -independent manner)[13] from whole liver tissue of control and Bmal1 KO mice, revealing rhythmic 24-hours expression for key members of these pathways (Fig. S3-4). These data suggest that pathways beyond TGF are also at play.

Collectively, these results suggest that the liver CC-oscillator controls the duration of expression (circadian gating) of TGF-β-target genes in the healthy liver. Hepatocytes are most likely the main contributors of the modulated gene expression. Given the hepatocyte-Bmal1-dependent modulation of the circadian expression of HSC-enriched genes (Fig. 1), we also speculate that the hepatocytic CC might regulate indirectly gene expression in liver non-parenchymal cells such as HSCs.

Metabolic stress in primary human hepatocytes (PHH) impairs expression of CC genes

The pathogenesis of MASH initiates with metabolic alterations, e.g., free fatty acid (FFA) accumulation-induced injury of hepatocytes. To model the status of CC genes in human hepatocytes during MASH, we treated PHH with FFA resulting in lipid accumulation (Fig. 2A-C) and perturbation of the expression of several CC-components (Fig. 2D and S5A). In synchronized PHH, FFA treatment altered the rhythmic expression of BMAL1 and REV-ERBA (Fig. 2C). FFA also increased the expression of TGF-β targets (Fig. 2D-F and S5A). Importantly, pharmacological inhibition of TGF-β activity by the SB505124 (SB) significantly prevented FFA-altered expression of CC-components and induction of TGF-β targets (Fig. 2G). Next, we explored the relationship between activated REV-ERBs and TGF-β targets expression. In FFA-treated PHH, we found that treatment with REV-ERB agonist SR9009[14] prevented (i) lipid accumulation, (ii) alteration of CC-gene expression, and (iii) induction of TGF-β targets (Fig. 2H-K). A part of inhibitory effect of SR9009 persists in a gene-specific manner even with REV-ERBA silencing (Fig. S5B-C), possibly reflecting a ‘trans-repressive’ effect through HDAC3 or other targets[15,16].

Fig. 2. FFA-drives CC-perturbation and induces TGF-β pathway gene expression in primary human hepatocytes (PHH).

(A) PHH were isolated from non-diseased patient livers and treated with FFA. (B) BODIPY staining confirms lipid accumulation in FFA-treated PHH. Green = neutral lipids. Blue = DAPI. The pictures show representative image (magnification X10). (C-D) FFA treatment perturbs CC gene expression and rhythmicity in PHH. (C) The graphs show relative gene expression of BMAL1 and REV-ERBA from synchronized PHH (low glucose) post-FFA treatment. One representative experiment out of 2 is shown. Unpaired student’s T-test; bold dots represent p < 0.05. (D) The graphs show means ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n =9). *p<0.05, **p<0.01, ***p<0.001, Mann-Whitney U test. (E-F) FFA treatment dysregulated CC and activates TGF-β signaling in PHH. One representative experiment in duplicate out of two if shown (see Fig S.17). Quantification of Western blot analysis was performed using ImageLab. *** p < 0.001 (Unpaired student’s T test) (G) TGF-β inhibitor reverts FFA-induced dysregulations in PHH. The graphs show mean ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n = 9) *, # p<0.05, **, ## p<0.01, ***, ### p<0.001 Kruskal-Wallis followed by Dunn’s multiple comparison tests (FFA VS mock or FFA-SB VS FFA). (H-K) Perturbation studies in PHH using the REV-ERBα agonist SR9009. (H-I) One representative experiment in duplicate out of two is shown (see Fig S.18). Quantification of Western blot analysis was performed using ImageLab. # or * p < 0.05 (One way ANOVA followed by Tukey’s multiple comparisons tests). (J) BODIPY staining in PHH (magnification X10). (K) SR9009 restores FFA-induces dysregulation of CC gene expression. PHH were treated with FFA and SR9009 or DMSO as control. The graphs show mean ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n = 9) *, # p<0.05, **, ## p<0.01, ***, ### p<0.001 Kruskal-Wallis followed by Dunn’s multiple comparison tests (FFA VS mock or FFA SR9009 VS FFA).

Collectively, our results suggest REV-ERBα as a negative regulator of TGF-β signaling in PHH and show metabolic alterations can perturb the expression of CC genes.

Perturbation of circadian rhythmicity in HSC correlates with fibrotic gene expression in vivo

HSC/liver myofibroblasts (MFs) are the central effectors of fibrosis[2,3]. While the rhythmic CC-oscillator in mouse hepatocytes is well characterized[5,6,9,11], the existence and status of this CC-oscillator in these cells during liver disease remains unknown.

Analysis of a single-nuclei RNA-sequencing (snRNA-seq) dataset from normal mouse liver[11] identified cell-specific markers and CC-genes in HSC (Fig. S6A-B). Employing a choline-deficient, L-amino acid-defined, high-fat diet (CDA-HFD) mouse model recapitulating features of MASH-induced fibrosis[17], we evaluated the status of HSC/MF CC. We isolated and characterized HSC from mice liver during circadian rest and active phases[18] (Fig. 3A and Fig. S7A-B). Transcriptomic analyses revealed that in HSC, CC-genes display characteristic rhythmicity (Fig. 3B-C). Bmal1 expression peaked at ZT0, which is known to transcribe Dbp, Rev-Erba (Nr1d1), Per1, and Per2 during rest phase[11] (Fig. 3B-C). In contrast, MFs from fibrotic livers of CDA-HFD mice displayed a perturbed CC-oscillator (Fig. 3B-C). Transcriptomic analysis identified 1,994 diurnally expressed genes (~5% of total genes covered) in HSC, which lost their rhythmicity in CDA-HFD mice (Fig. 3D-E and Tables S3-S4). In control HSCs, the basal level expression of fibrosis-related genes displayed a diurnal or ultradian rhythmicity (Fig. 3F, Fig. S7C). In contrast, we observed an upregulation and/or a perturbed rhythmicity of pro-fibrotic genes (Smad3, Tgfbr1, Runx1 and Atf3) and downregulation of anti-fibrotic genes (Smad7 and Mmp9) (Fig. 3F).

Fig. 3. Diurnal rhythmicity of gene expression is lost in hepatic stellate cells (HSCs) during fibrosis in vivo.

(A) Experimental approach showing HSC and MF isolation from mouse livers at different time points from control (chow diet) or CDA-HFD animals (n = 3 animals per group/time-point). (B-H) Diurnal rhythmicity of gene expression is lost in HSC/MF during MASH in vivo. (B) Heatmap and (C) normalized expression levels of indicated CC-genes in control and CDA-HFD HSCs/MFs (D) Status of circadian pathways (FDR < 0.05) in control and CDA-HFD HSC/MF (E) Heatmap of circadian genes expression in control and CDA-HFD HSC/MF. (F) Pathways enriched (FDR < 0.05) for circadian genes in control HSC. (G) Circadian enrichments of indicated pathways (FDR < 0.05) in control and CDA-HFD HSC. Panels CG: Wald test and Benjamini-Hochberg correction; bold dots: FDR < 0.05. Refer to Tables S1 to S7.

(A) Perturbation of CC in LX2 stellate cells induces an increase in fibrotic gene expression.

REV-ERBA KO LX2 stellate cells were generated and synchronized. The graphs show mean ± SD of relative mRNA expression normalized to GAPDH mRNA. One representative experiment out of 3 performed in triplicate is shown (see Figure S12-13). (B) HLMF were isolated from patient liver tissues and exposed to FFA. (C-D) FFA treatment perturbs CC gene expression in HLMF and induces fibrotic gene expression. The graphs show means ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n =9). *p<0.05, **p<0.01, ***p<0.001, Mann-Whitney U test. (E) Perturbation of CC and fibrotic gene expression are reversed by the TGFβ inhibitor SB505124. The graphs show mean ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n = 9) *, # p<0.05, **, ## p<0.01,***, ### p<0.001 Kruskal-Wallis followed by Dunn’s multiple comparison tests (TGFβ VS mock or TGFβ +SB VS TGFβ). (G) SR9009 inhibits TGF-β signaling in HLMFs. One representative experiment out of two if shown (see Fig S.19). Quantification of Western blot analysis was performed using ImageLab.

Further analyzing the significant cycling genes, we performed gene set variation analysis (GSVA) to identify the pathways regulated by CC in HSCs. HSCs displayed diurnal rhythmicity of metabolic, stress response/cytokine signaling, cellular biochemical processes, collagen, and ECM synthesis/remodeling pathways (Fig. S8 and Tables S5-S6). Analysis of the individual genes leading to the enrichment of these rhythmic pathways identified several key TFs, enzymes, and signaling intermediates (Table S7). Importantly, these enriched pathways were significantly (FDR < 0.05) perturbed in MFs of CDA-HFD mice, along with upregulation and altered rhythmicity of pathways driving liver fibrosis, including the TGFβ signaling pathway (Fig. 3G). Activation of the TGFβ signaling pathway is mediated through the canonical SMAD2/3 dependent pathway and non-canonical SMAD2/3 independent pathways such as JNK, MAPK or PI3K/AKT[19]. While Smad3 and Smad7 display significant circadian variation over time in HSC, our GSVA analysis showed that SMAD-independent pathways were not rhythmic (Table S6), indicating that regulation of fibrotic genes are mediated predominantly by the TGFβ-SMAD pathway. Additional analyses of the circadian transcription of genes regulating TGF/BMP/Activin pathways, revealed rhythmic 24-hours expression for multiple key members of these pathways and support that pathways beyond TGF may also be at play (Fig. S9-10).

Collectively, these data indicate that HSCs contain a functional CC-oscillator with numerous diurnally expressed genes controlling physiological functions, and that metabolic injury disrupts the HSC CC-oscillator promoting the expression of pro-fibrotic genes.

REV-ERBα activation impairs TGF-β signaling in human HSC and myofibroblasts

The functionality of the human HSC CC-oscillator and its impact on expression of TGF-β targets was explored by generating Crispr-Cas9-guided knockout of BMAL1 and REV-ERBA in LX2 stellate cells (Fig. S11A). Circadian synchronization demonstrated perturbed expression of CC-genes and TGF-β targets in REV-ERBA KO and BMAL1 KO LX2 cells (Fig. 4A and Fig. S12-S13). Moreover, REV-ERBA KO increases the expression of profibrotic genes (Fig. 4A). Then, HSCs were isolated from patient-derived liver tissues (Fig. 4B). Of note, isolated HSCs quickly acquire characteristics of human liver myofibroblasts (HLMFs) in cell culture. FFA treatment affected expression of CC-components in a gene-specific manner (Fig. 4C) and induces upregulation of TGF-β target expression (Fig. 4D). Inhibition of TGFβ signaling using SB431542 restored CC gene expression (Fig. 4E). Furthermore, REV-ERBα activator SR9009 reduced the phosphorylation of SMAD2/3 in TGF-β-treated HLMFs (Fig. 4F). As observed for PHH (Fig. 2), a part of the inhibitory effect of SR9009 persisted in gene-specific manner with REV-ERBA silencing (Fig. S11B).

Fig. 4. SR9009 inhibits fibrotic gene expression in human liver myofibroblasts (HLMF).

(A) Perturbation of CC in LX2 stellate cells induces an increase in fibrotic gene expression. REV-ERBA KO LX2 stellate cells were generated and synchronized. The graphs show mean ± SD of relative mRNA expression normalized to GAPDH mRNA. One representative experiment out of 3 performed in triplicate is shown (see Figure S12-13). (B) HLMF were isolated from patient liver tissues and exposed to FFA. (C-D) FFA treatment perturbs CC gene expression in HLMF and induces fibrotic gene expression. The graphs show means ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n =9). *p<0.05, **p<0.01, ***p<0.001, Mann-Whitney U test. (E) Perturbation of CC and fibrotic gene expression are reversed by the TGFβ inhibitor SB505124. The graphs show mean ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 3 experiments performed in triplicate (n = 9) *, # p<0.05, **, ## p<0.01,***, ### p<0.001 Kruskal-Wallis followed by Dunn’s multiple comparison tests (TGFβ VS mock or TGFβ +SB VS TGFβ). (G) SR9009 inhibits TGF-β signaling in HLMFs. One representative experiment out of two if shown (see Fig S.19). Quantification of Western blot analysis was performed using ImageLab.

Collectively, these results indicate a reciprocal relationship between REV-ERBα function and expression of TGF-β targets in HLMFs. This also suggests that activating REV-ERB could dampen TGF-β-signaling in liver disease and might attenuate fibrosis progression.

Crosstalk and communication of the hepatocyte CC with HSCs

Since our transcriptomic analyses highlighted a perturbed expression of HSC-enriched genes in hepatocyte-specific CC KO mice (Fig. 1), we aimed to study the impact of CC perturbation in hepatocytes on HSC gene expression. Analysis of snRNA-seq data from livers of hepatocyte-specific KO of Rev-Erbα/β livers (LDKO)[11] revealed ~300 perturbed HSC genes including CC-genes (Bmal1, Npas2, Rora, and E4bp4), TFs (Mlxipl, Srebf1, Pparβ), metabolic regulators (Fasn, Insig2, Acly), and ligands/receptors (Nrg4, Pi3kr1, Pdgfrβ, Il1r) (Fig. 5A-B and Table S8).

Fig. 5. Crosstalk and communication of the hepatocyte circadian clock (CC) with HLMF.

(A) Volcano plot showing differential expression of genes in HSC of control and hepatocyte-specific Rev-Erbα/β knock-out (Rev-Erbα/β^hep-/-; LDKO)[11] mice. (B) UMAPs showing expression of Mlxipl in different cell types of liver in control and LDKO mice. Black arrow = Mlxipl expression in HSC. (C-E) PHH-HLMF co-culture model showing that the perturbation of the hepatocyte CC modulates the HLMF phenotype. (C) Experimental approach (see method) (D-E) The graphs show means ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 2 experiments performed in triplicate (n = 6). *p<0.05, **p<0.01, Mann-Whitney U test. Refer to Tables S8.

To functionally validate a crosstalk between human hepatocyte CC and HLMF-enriched profibrotic gene expression, we knocked-down BMAL1 (siBMAL1) in PHH and subsequently co-cultured them with HLMFs, treated with TGF-β (Fig. 5C). Specific silencing of hepatocyte BMAL1 (Fig. 5D) led to a significantly increased expression of HLMF-enriched genes ACTA2, COL27A1 and TGFB (Fig. 5E). The magnitude of these changes was donor-dependent. Collectively, these results suggest that the hepatocyte CC can influence HSC gene expression, possibly by regulating the levels and activities of different TFs, secretory proteins and receptor-ligand pairs.

Pharmacological targeting of REV-ERBs inhibits liver fibrosis progression in vivo

To explore the CC as a therapeutic target for MASH-induced fibrosis, we studied whether SR9009, a pharmacological activator of REV-ERBα[14], modulates fibrosis in CDA-HFD fed C57Bl/6J mice, a state-of-the-art model for MASH fibrosis (Fig. 6A). We confirmed REV-ERBs target engagement in whole liver tissues by an SR9009-induced decrease of REV-ERBs transcriptional targets Bmal1 and Clock (Fig. 6B). SR9009 treatment resulted in significant inhibition of liver fibrosis as shown by a decrease of collagen deposition (CPA, collagen proportionate area) (Fig. 6C-D), hydroxyproline levels (Fig. 6D) and fibrotic gene expression and content (Fig. 6E, G). Liver steatosis and inflammation remained unchanged (Fig. 6F, Fig. S14) in line with transcriptomic analyses revealing that SR9009 treatment only a had modest effect was observed on lipid metabolism (Fig. S15, Table S9). In contrast, CDA-HFD-driven perturbations of xenobiotic, steroid, and amino acid metabolism were partially restored (Fig. S15, Table S9). Consistent with histopathological analyses, RNA-Seq and CIBERSORTx cell fraction modeling[20] showed similar abundance of the major liver cell populations in treatment and vehicle-treated animals (Fig. S14). Moreover, RNA-Seq/CIBERSORTx analyses revealed reduced levels of Rev-Erbα (Nr1d1) in hepatocytes of mice subjected to CDA-HFD diet (Fig. 6H-I, Table S9-S10). Upon SR9009 treatment Rev-Erbα expression was restored resulting in decreased expression of its down-stream targets Bmal1, Clock, and Cry1 (Fig. 6I).

Fig. 6. REV-ERBs agonist SR9009 inhibits liver fibrosis progression in a MASH fibrosis model in vivo.

(A) Study design. Male C57Bl/6 mice fed with standard chow diet or choline-deficient, L-amino acid-defined, high-fat diet (CDA-HFD) for 12 weeks. Animals received vehicle control or SR9009 for 6 weeks. Chow diet n=10; CDA-HFD + vehicle n=15; CDA-HFD + SR9009 n=14. (B) Target engagement assay. SR9009 treatment efficacy was confirmed by the decrease in Bmal1 and Clock expression, the downstream target of Rev-Erba. (C-E) SR9009 treatment improves liver fibrosis. (C) Representative stainings for hematoxylin & eosin (H&E), Sirius red and Oil Red O are shown (scale bar = 100 µm). Fibrosis levels were evaluated by (D) quantitative digital analysis (collagen proportional area, CPA), hydroxyproline quantification and (E) fibrotic gene expression measurement. (F) Lipid accumulation by Oil Red O staining quantification. (G) SR9009 treatment reduced expression of Il1 and Tnfa in mouse liver. The graphs show mean ±SD of % area, µg/mg of tissues or relative mRNA expression normalized to GAPDH mRNA. *, # p<0.05, **, ## p<0.01,***, ### p<0.001, ****, #### <0.0001 (vehicle vs. chow or SR9009 vs. vehicle) Mann-Whitney U test. (H-I) CIBERSORTx -pathway analyses on mouse liver RNA-Seq was used to predict (H) overall liver cell type abundances and (I) hepatocyte CC gene expression in the indicated study groups. Refer to Tables S9 and S10.

To validate the therapeutic effect of SR9009 in a mouse model closer to the patient, we applied a human liver chimeric mouse model (HLCM) (Fig. 7A). HLCM are generated using Fah−/−/Rag2−/−/Il2rg−/− (FRG)-NOD mice with livers repopulated by PHH expressing human REV-ERBα in the chimeric livers[21]. Feeding HLCM with CDA-HFD recapitulates key features of liver fibrosis. While these mice lack T- and B-cells, the mouse livers contain macrophages, myofibroblasts and endothelial cells[21]. Immunohistochemistry confirmed the degree of humanization (FAH, Fig. 7B), a lack of T-cells (CD3; Fig. S16A), and the presence of macrophages (CD68; Fig. S16A). Treatment of CDA-HFD HLCM mice with SR9009 inhibited liver fibrosis as shown by a decrease of collagen deposition (Sirius red), HSC activation (αSMA), and reduction of fibrotic gene expression (Fig. 7B-E). Lipid accumulation (Oil-Red-O staining) remained unchanged after SR9009 treatment, as well as blood levels of triglycerides and FFA (Fig. S16 B-E). SR9009 treatment had no detectable effect on apoptosis (Fig. S16 F-G).

Fig. 7. SR9009 inhibits fibrosis in human liver chimeric mice (HLCM).

(A) Study design. FRG-NOD mice were engrafted with PHH to generate HLCM (see method) and subjected to CDA-HFD for 12 weeks. Animals received vehicle control or SR9009 for 4 weeks. CDA-HFD + Vehicle n = 3; CDA-HFD + SR9009 n = 4. (B-D) SR9009 treatment improves liver fibrosis in HLCM. (B) Representative staining images for hematoxylin & eosin (H&E), Sirius red, Fumarylacetoacetate Hydrolase (FAH) and alpha smooth muscle actin (aSMA) are shown (scale bar = 500 µm). Fibrosis levels were evaluated by (C) quantitative digital analysis (collagen proportional area, CPA) and aSMA quantification (% of area) and (D) fibrotic gene expression measurement. (E) SR9009 treatment reduces liver inflammation. The graphs show mean ± SD of % area or relative mRNA expression normalized to GAPDH mRNA. *p<0.05, **p<0.01, ***p<0.001, Mann-Whitney U test.

Deregulated expression of CC genes is associated with liver fibrosis progression in patients and SR9009 treatment reduces fibrosis gene expression in patient liver spheroids

Transcript analyses of CC-components in patients with metabolic liver disease (GSE126848)[22] showed decreased PER1, PER2, REV-ERBs, and CRY2 expression, while CLOCK expression was increased (Fig. 8A). In a second cohort of MASLD patients with varying degrees of fibrosis (F0 to F4; GSE135251)[23], we also observed that the expression of PER1, PER2 and REV-ERBs was reduced in the liver (Fig. 8B). Reduced PER1, PER2, and REV-ERBA expression was observed in early (F2) and more pronounced in the advanced stages (F3-F4), suggesting a possible ‘burn-out’[23] (Fig. 8B).

Fig. 8. Expression of CC-genes in MASH patients and pharmacological targeting of the CC in patient liver spheroids.

(A) Transcript levels of indicated CC-genes in patient livers (GSE126848)[22]. Wilcoxon test; ^*p<0.05, ^**p<0.01 *** p < 0.001 and **** p < 0.0001. (B) Transcript levels of indicated CC-genes in the liver of patients with varying levels of fibrosis, as indicated (GSE135251)[23]. Wilcoxon test; ^*p<0.05, ^**p<0.01. (B-C) SR9009 treatment restores TGF-β induced-CC and pro-fibrotic gene perturbations in patient derived spheroids. (B) Multicellular liver spheroids were generated from patient liver tissue and treated with TGF-β and SR9009 or DMSO treatment The graphs show mean ± SD of % of mRNA normalized to GAPDH (mock = 100%) from 2 experiments performed in triplicate (n = 6). *p<0.05, **p<0.01, ***p<0.001 1 Kruskal-Wallis followed by Dunn’s multiple comparison tests. (C) Multicellular liver spheroids were generated from cirrhotic liver tissue and treated with SR9009 or DMSO. The graphs show mean ± SD of mRNA normalized to GAPDH from 1 experiment performed in triplicate (n = 3).

Finally, we targeted the CC in patient-derived liver spheroids which include epithelial cells, fibroblasts, and immune cells modeling key features of the liver microenvironment[24]. TGF-β and SR9009 treatment of non-diseased liver tissue-derived spheroids revealed a SR9009-induced rescue of TGF-β-reduced REV-ERBA expression (Fig. 8C), and an inhibition of TGF-β target gene expression by SR9009 (Fig. 8C). Remarkably, in fibrotic patient-derived spheroids SR9009 effectively reduced TGF-β-regulated fibrotic gene expression (Fig. 8D). These results support a clinical impact of CC as a therapeutic candidate target for liver fibrosis.

Discussion

Our study has the following key findings: (1) the liver CC-oscillator ‘gates’ the expression of TGF-β target genes (Fig. 1 and Fig. S1) in the healthy liver as shown by analyses of Bmal1KO and Rev-Erbα/β^hep-/- CC-mutant mice. Although some of our data are on a correlative level, our functional data from multiple experiments suggest that the CC-controlled pulsatile TGF-β target genes expression in non-diseased liver may act as a break from driving its constant activity (Figs. 1 and 3), which could otherwise drive fibrosis. (2) MASH metabolic injury perturbs the liver CC which is associated with increased expression of TGF-β target genes (Figs. 2-4). (3) Targeting the liver CC clock provides an opportunity for a previously undiscovered anti-fibrotic therapy.

Mechanistically, our results suggest that CC-oscillators are present in both hepatocytes and HSC – the key driver cells of liver fibrosis. The alterations of HSC-enriched genes in hepatocyte-specific CC mutants raised the possibility of hepatocyte-HSC crosstalk in which the hepatocyte CC possibly influences HSC gene expression. This hypothesis is further supported by loss-of-function studies in PHH-HLMF co-culture studies although the detailed mechanism remains to be determined (Fig. 5C-E).

We found that TGF-β targets are rhythmically expressed not only in human hepatocytes but also in HSC. We validated the impact of the human HSC CC-oscillator on TGF-β targets by Crispr-guided KO of CC genes and synchronization experiments (Fig. 4), as a surrogate for in vivo HSC-specific CC mutational studies. TGF-β signaling has different biological roles in hepatocytes and HSC, but includes some common key receptors and effectors e.g., Smad7, Tgfbr1. The expression of these common genes indicates ‘functional’ TGF-β signaling, although whether activation of some of these genes in hepatocytes influence on HSC activation during disease remains to be explored. Note that the phase of peak expression for some of these TGF-β target genes are different between mouse and human cells, likely reflecting a gene- and species-specific differential enhancer usage. Interestingly, CC-regulation of Tgfb1 and Smad3 expression has also been observed in kidney, adipose tissue and heart[25,26]. GWAS studies have implicated different CC genes in various pathologies, including metabolic disease affecting liver and HCC risk[27]. However, an association of CC genes polymorphisms with fibrosis has not been found.

Through multiple approaches we uncovered a “crosstalk” between the CC-oscillator and TGF-β signaling. In this crosstalk the CC-component Rev-Erbs play an important role in repressing TGF-β signaling. Moreover, targeting REV-ERBα by a small molecule SR9009 inhibited liver fibrosis progression in in vivo and ex vivo models (Figs. 6-8). We chose REV-ERBα as: (i) its level was reduced during fibrosis and elevated TGF-β and FFA-signaling in mice, patient tissues, HSC/HLMF, and PHH, (ii) its activity can be modulated by agonists (SR9009) and (iv) such agonists can be safely administered in mice. While the liver CC is known to regulate metabolism, we only observed a modest effect on lipid metabolism upon SR9009 treatment (Figs. 6 and 7) which did not result in a major reduction of liver steatosis in vivo (Fig. S14). This is consistent with the well described effect of SR9009 on fatty acid and glucose metabolism predominantly in muscle and adipose tissue[28]. Alternatively, the CDA-HFD model which is designed to study predominantly liver fibrosis [29] may not model entirely the CC metabolic effects in the liver.

A strength of our study is the validation of results in patient-derived model systems including animal models, spheroids and primary cells, the consistency of results across complementary model systems and patient cohorts supporting its clinical validity and future therapeutic translatability.

The study has the following limitations: (1) While the CDA-HFD is accepted as state-of-the-art model and widely used to study compounds targeting liver fibrosis, it is limited by only partially modeling MASH metabolism and steatosis[29], (2) the 2D in vitro experiments have limitations including the fact that PHH de-differentiate during culture and HSC quickly activate. Furthermore, additional studies using combined transcriptome and secretome analyses will be needed to unravel the detailed mechanisms of crosstalk. (3) The study is limited by a lack of HSC-specific knock-out of CC-genes in vivo, which could further illuminate the role of the HSC ‘clock’ disruption in liver disease progression. Future studies, e.g., in the PDGFRβ CreER mouse model recently developed by Hamberger et al.[30] may help to understand whether CC-oscillators in hepatocytes and HSCs may be required to de-regulate TGF-β signaling and the regulation of fibrosis-associated genes.

Targeting the CC provides novel opportunities to treat liver fibrosis. A key differentiator of the CC-targeting approach is its anti-fibrotic efficacy as demonstrated across all models (Figs. 6-8), which addresses a key unmet medical need in the treatment of liver fibrosis. While we did not observe any major adverse effects on liver health, further studies are needed to investigate the safety profile of CC-targeting compounds in patients. Moreover, as our preclinical models are focused on MASH as a liver injury for fibrosis, additional studies are needed to show efficacy and impact also in fibrosis of other etiologies such as chronic viral hepatitis and alcoholic liver disease. Collectively, our pharmacological proof-of-concept studies across several models including patient-derived models uncover the CC as a previously undiscovered therapeutic candidate target for liver fibrosis – an important and rising global unmet medical need.

Materials and methods

Mouse liver HSC isolation

HSC/MFs were isolated from 8 weeks old mice fed with chow diet or CDA-HFD for 12 weeks. Animals were sacrificed at ZT0 (7 am), ZT6 (1 pm), ZT12 (7 pm), and ZT18 (1 am). Once the animals were sacrificed, livers were collected and directly processed for MF isolation (see supplemental methods)[18]. The experiment (Inserm U1110) was approved by the local ethic committee and authorized by the French ministry of higher education and research (APAFIS#29396-2021012915164103 v2).

Statistical analyses

Individual experiments were reproduced at least 2-3 times, independently (except otherwise stated). PHHs and HLMFs for ex vivo experiments were obtained from tissues of 2-3 independent donors and each condition/treatment was conducted in triplicates. Due to the rarity of patient-fibrotic liver tissues, experiments involving patient-derived fibrotic spheroids were performed one time in triplicate. Individual values are presented in bar charts. GraphPad Prism 8 software was used for statistical analyzes. The data are presented as the mean ± SD and were analyzed by the unpaired Student’s t-test or the two-tailed Mann-Whitney U test as indicated in figure legends, after determination of distribution by the Shapiro-Wilk normality test. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***) were considered significant. Comparisons of 3 or more groups were analyzed by Ordinary one-way Anova followed by Dunnett’s or Tukey’s multiple comparison tests or Kruskal-Wallis test followed by Dunn’s multiple comparison tests as indicated in the legends, after determination of distribution by the Shapiro-Wilk normality test. p < 0.05 (* or #), p < 0.01 (** or ##), p < 0.001 (*** or ###) were considered significant.

Highlights.

• -The circadian clock (CC) is a regulator of TGF-β signaling in the healthy liver.
• -The breakdown of CC control in MASH is associated with liver fibrosis progression.
• -Proof-of-concept studies in several models unravel targetability by a small molecule.
• -Pharmacological targeting the CC provides opportunities to improve liver fibrosis.

Impact and implications

Liver fibrosis due to metabolic diseases is a global health challenge. Many liver functions are rhythmic throughout the day being controlled by the circadian clock (CC). Here we demonstrate that the regulation of the CC is perturbed upon chronic liver injury and this perturbation contributes to fibrotic disease. By showing that a compound targeting the CC improves liver fibrosis in patient-derived models, this study provides a novel therapeutic candidate strategy to treat fibrosis in patients. Additional studies will be needed for clinical translation. Since the findings uncovers a previously undiscovered profibrotic mechanism and therapeutic target, the study is of interest for scientists investigating liver disease, clinical hepatologists and drug developers.

Abbreviations

CDA-HFD

choline-deficient, L-amino acid-defined, high-fat diet

ChIP-qPCR

chromatin immunoprecipitation followed by quantitative polymerase chain reaction

CLD

chronic liver disease

CC

circadian clock

CPA

collagen proportionate area

FDR

false discovery rate

GSEA

gene set enrichment analysis

GSVA

gene set variation analysis

HSC

hepatic stellate cell

HCC

hepatocellular carcinoma (HCC)

LDKO

hepatocyte-specific KO of Rev-Erbα/βlivers

HLCM

human liver chimeric mouse model

HLMF

human liver myofibroblasts

MASH

metabolic dysfunction-associated steatohepatitis

MASLD

metabolic dysfunction-associated steatotic liver disease

MF

myofibroblasts

PH

primary human hepatocytes

SBE

SMAD binding elements

TF

transcription factors

TGF-β

transforming growth factor-β

Data availability statement

The RNA-sequencing data has been deposited at GEO (GSE264250 for mouse HSC and MF and GSE272293 for MASH/fibrosis model). Data referring to heatmaps and full-length western blots are available in supplementary information. All data are available from the corresponding author upon reasonable request.