AMPK maintains the activation of hepatic stellate cells through mitophagy-induced metabolic reprogramming

Abstract

The activation of hepatic stellate cells (HSCs), characterized by transdifferentiation from a quiescent state to a fibrogenic phenotype, is a core process of liver fibrosis. The metabolic reprogramming of HSCs plays a major role in this process to meet the high energy demands of myofibroblastic HSCs with multiple functions, such as extracellular matrix synthesis, migration, and proliferation. AMP-activated protein kinase (AMPK) is a gatekeeper of intracellular energy homeostasis, but its role in the activation of HSCs and the progression of liver fibrosis remains unclear. Here, we found that the phosphorylation of AMPK in HSCs was upregulated in liver tissues from metabolic dysfunction-associated steatohepatitis patients and from mice treated with carbon tetrachloride (CCl4) or bile duct ligation (BDL). HSC-specific deletion of two catalytic α-subunits of AMPK attenuated liver fibrosis in the CCl4 or BDL mouse model. In vitro analysis demonstrated that AMPK promoted HSC activation upon various profibrogenic stimuli. The activation of AMPKα-deficient HSCs was impaired due to the decreased mitochondrial oxidative phosphorylation but restored after treatment with the mitophagy inducer rapamycin. Mechanistically, both the AMPK–ULK1 and AMPK–Raptor pathways contribute to the maintenance of the mitophagy pathway and mitochondrial quality. These findings provide direct evidence of the crucial role of AMPK–mitophagy signaling in ensuring mitochondrial health and sufficient energy supply during HSC activation. In this study, AMPK was modulated in HSCs prior to activation, which is distinguished from previous investigations and thus provides new insights into the role of AMPK during distinct phases of HSC activation.

Introduction

Liver fibrosis, characterized by the excessive extracellular matrix (ECM) deposition and fibrous scar formation, is a central driving force for the progression from a number of chronic liver diseases, such as viral hepatitis, metabolic dysfunction-associated fatty liver disease (MAFLD), and metabolic dysfunction-associated steatohepatitis (MASH), to end-stage liver diseases, such as cirrhosis and hepatocellular carcinoma (Hernandez-Gea and Friedman, 2011). Upon liver injury, quiescent hepatic stellate cells (HSCs) undergo activation and transdifferentiate into scar-forming myofibroblasts, which migrate to the site of injury to produce collagen and promote liver recovery from injury. However, prolonged and superfluous activation of HSCs will cause ECM deposition, resulting in progressive fibrosis, defective repair, and ultimately cirrhosis (Barbosa Júnior Ade et al., 1993; Puche et al., 2013). Therefore, understanding how the activation of HSCs can be regulated is necessary for correctly diagnosing patients and developing novel antifibrotic strategies.

AMP-activated protein kinase (AMPK) is a critical cellular sensor of energy homeostasis responding to various stimuli, such as physiological, hormonal, and nutritional cues, to balance adenosine triphosphate (ATP) production with demand, which requires the elevated phosphorylation of AMPK (p-AMPK) at Thr172 within the α catalytic subunit (Hardie et al., 2012). It has been reported that hepatic p-AMPK is downregulated in the MAFLD model, and activation of AMPK attenuates high-fat diet-induced MAFLD, but knockout of hepatic AMPK does not further worsen it (Boudaba et al., 2018). Furthermore, hepatocyte-specific AMPK knockout aggravates liver damage in mouse MASH models via the caspase-6 pathway (Zhao et al., 2020).

Although the role of AMPK in hepatocytes is well characterized, its role in other liver cell types that contribute to fibrosis, such as HSCs, is less well understood. The myofibroblastic phenotype of activated HSCs, similar to the phenotype of highly proliferative cancer cells, is characterized by biosynthetic and energy-intensive processes such as increased ECM production, proliferation, and migration (Du et al., 2018; Khomich et al., 2019). Thus, energy-producing pathways involving glycolysis and mitochondrial oxidative phosphorylation are augmented during the activation of HSCs. This change in metabolic status is termed ‘metabolic reprogramming’. Growing evidence has shown that targeting metabolic reprogramming could be a new attractive way to control the switch of HSCs from a quiescent state to an activated state (Guimarães et al., 2012; Hernández-Gea and Friedman, 2012; Du et al., 2018; Bates et al., 2020; Mejias et al., 2020). Given that the bioenergetic demands increase dramatically when quiescent HSCs transdifferentiate into myofibroblasts, we suppose that AMPK is involved in this energy-demanding process.

We report here that conditional double-knockout of AMPKα1α2 in HSCs inhibits its activation and alleviates liver fibrosis. AMPKα deficiency in HSCs prevents the metabolic switch necessary for the induction of oxidative phosphorylation via mitophagy. This kind of metabolic regulation of HSC activation mediated by AMPK provides new insight into HSC activation modulation and antifibrotic strategies.

Results

AMPK is highly phosphorylated in activated HSCs

We first characterized the phosphorylation and activity of the critical energy sensor AMPK during the activation of HSCs, the principal cells responsible for liver fibrosis. To this end, we utilized freshly isolated primary mouse HSCs undergoing culture-induced activation and transdifferentiation from a quiescent to a myofibroblastic phenotype. The purity of HSCs exceeded 90% (Supplementary Figure S1). The phosphorylation of AMPK increased during myofibroblastic differentiation within 3 days after plating, as demonstrated by immunoblotting with the typical myofibroblastic marker alpha-smooth muscle actin (α-SMA) (Figure 1A and B). The activity of AMPK also increased, as the phosphorylation at Ser798 of regulatory-associated protein of target of rapamycin (Raptor), a classical downstream target of AMPK, was highly upregulated (Figure 1A and C).

Figure 1.

The phosphorylation of AMPK is highly upregulated during the activation of HSCs. (A–C) Immunoblotting and quantification of p-AMPK^Thr172 and p-Raptor^Ser798 in HSCs freshly isolated or cultured for 3 days. (D and E) Co-immunostaining of desmin and p-AMPK in mouse liver samples from chronic CCl4 model (three repeated injections of CCl4 every week for 8 weeks) (n = 6 or 9). (F and G) Co-immunostaining of desmin and p-AMPK in mouse liver samples from the BDL model (BDL for 10 days) (n = 3). (H–J) Immunoblotting and quantification of p-AMPK^Thr172 and p-Raptor^Ser798 in HSCs isolated from short-term CCl4 model (three repeated injections of CCl4 in a week). (K–M) Immunoblotting and quantification of p-AMPK^Thr172 and p-Raptor^Ser798 in HSCs isolated from the BDL model. (N and O) Co-immunostaining of desmin and p-AMPK in human liver samples of over-BMI patients or healthy control (NC) (n = 5 in each group). White arrows in D, F, and N indicate the region with or without desmin/p-AMPK double-positive staining. Scale bar, 100 μm. For immunoblotting, HSCs isolated from 2 or 3 mice were mixed and the quantification results were from three independent experiments. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001; t-test.

Next, the carbon tetrachloride (CCl4) and bile duct ligation (BDL) mouse liver fibrosis models were established. Liver sections from the mice receiving three repeated injections of CCl4 at 0.2 ml/kg in a week (short-term CCl4 model) or treated with BDL exhibited higher α-SMA expression and more collagen deposition (Supplementary Figure S2). Immunofluorescence staining showed that the p-AMPK level was elevated and accompanied by the increased expression of desmin, a marker of HSCs, in mice with CCl4- and BDL-induced liver fibrosis, demonstrating that phosphorylation of AMPK was induced during HSC activation (Figure 1D–G). Consistently, in the primary HSCs isolated from these mice, a significant increase in AMPK phosphorylation at Thr172 and Raptor phosphorylation at Ser798 was observed (Figure 1H–M).

To study AMPK phosphorylation status in human HSCs, we examined liver samples collected from MASH patients with abnormal body mass index (BMI) (Supplementary Figure S3A). Histological analysis showed the increased α-SMA expression and ECM deposition in the liver tissues of over-BMI patients (Supplementary Figure S3B and C). Immunofluorescence staining showed that the desmin/p-AMPK double-positive area in the liver tissues of over-BMI patients was ∼2.5 times greater than that in normal control (NC) tissues (Figure 1N and O). Above all, we concluded that AMPK phosphorylation is highly upregulated in activated HSCs.

AMPK mediates the effects of various profibrogenic stimuli in HSCs

Treatment with 1 mM CCl4 was reported to cause injury in primary mouse hepatocytes (Berger et al., 1986). In addition, palmitic acid (PA) also causes hepatocyte lipid overload and injury. To determine the role of highly phosphorylated AMPK during the transdifferentiation of HSCs to myofibroblasts in vitro, the supernatant of CCl4- or PA-injured primary mouse hepatocytes was used as conditioned medium (CM-CCl4 or CM-PA) to activate primary mouse HSCs, in the presence of si-AMPKα or its scramble (Figure 2A). The phosphorylation of AMPK in primary HSCs increased quickly after 6 h of coculture with CM-CCl4 (Supplementary Figure S4A and B) or 24 h of coculture with CM-PA (Supplementary Figure S4C and D). After 72 h of coculture with CM-CCl4 (Figure 2B–D) or CM-PA (Figure 2E–G), primary mouse HSCs showed significantly higher expression of α-SMA, collagen type I alpha 1 chain (Col1α1), and Fibronectin, indicating an activated state, while knockdown of AMPKα with siRNA dramatically inhibited the activation of HSCs and phosphorylation of AMPK.

Figure 2.

AMPK responses to different myofibroblastic stimuli. (A) Scheme of the preparation of conditioned medium from primary hepatocytes treated with CCl4 or PA and the treatment of primary HSCs in the presence of si-AMPKα or its scramble (Scr). (B–D) mRNA expression of α-SMA, Col1α1, and Fibronectin (B) and protein expression of α-SMA and p-AMPK (C and D) in HSCs after 72 h of coculture with CM-NC-Scr, CM-CCl4-Scr, or CM-CCl4-si-AMPKα. (E–G) The indicated mRNA expression and protein expression in HSCs after 72 h of coculture with CM-NC-Scr, CM-PA-Scr, or CM-PA-si-AMPKα. (H–J) The indicated mRNA expression and protein expression in HSCs after 48 h of treatment with or without glutamine, in the presence of si-AMPKα or its scramble. For immunoblotting, HSCs isolated from 2 mice were mixed and the quantification results were from three independent experiments. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 vs. CM-CCl4-Scr, CM-PA-Scr, and Gln⁺-Scr; t-test.

It has been reported that the activation of HSCs is highly dependent on glutamine for energy metabolism (Du et al., 2018). Therefore, we also investigated whether AMPK could be phosphorylated by glutamine treatment to promote its myofibrogenic effects on HSCs. Similarly, the phosphorylation of AMPK was rapidly switched on 1 h after the addition of glutamine (Supplementary Figure S4E and F). The activation of HSCs was significantly promoted with increased expression of α-SMA, Col1α1, and Fibronectin after 48 h in glutamine-containing medium, while knockdown of AMPKα inhibited the activation of HSCs and phosphorylation of AMPK (Figure 2H–J).

Overall, these results suggest that under conditions of chronic liver fibrosis, various profibrogenic stimuli, such as danger-associated molecular patterns (DAMPs) released from damaged hepatocytes, and metabolites, such as glutamine, can activate AMPK in HSCs. Furthermore, these profibrogenic stimuli activate HSCs in an AMPK-dependent manner. AMPK is probably a crucial common downstream mediator of a variety of profibrogenic stimuli that promote HSC activation in liver fibrosis.

Deletion of AMPKα attenuates liver fibrosis in the CCl4 mouse models

To elucidate the role of AMPK in the pathogenesis of liver fibrosis, homozygous AMPKα1/α2-floxed mice (Control) were crossed with mice expressing an HSC-specific Cre recombinase (Lrat-Cre) to generate AMPKα1α2 conditional knockout mice, referred to as α1α2HKO mice hereafter (Supplementary Figure S5A). The protein expression of AMPKα in HSCs was notably lower in α1α2HKO mice than in Control mice (Supplementary Figure S5B) but did not change in hepatocytes or Kupffer cells (Supplementary Figure S5C and D). Compared with Control mice, α1α2HKO mice did not show any difference in phenotypes in terms of liver fibrosis under normal housing conditions (Supplementary Figure S5E and F).

Both α1α2HKO mice and Control mice were treated with CCl4 for 8 weeks to induce liver fibrosis (Supplementary Figure S6A). Sirius red staining showed significantly less collagen deposition in the liver tissue of CCl4-treated α1α2HKO mice (Figure 3A and B), indicating the less severity and extent of liver fibrosis. Consistently, hepatic hydroxyproline content was significantly lower in CCl4-treated α1α2HKO mice (Figure 3C). Furthermore, hepatic mRNA levels of fibrogenic genes, such as α-SMA, Col1α1, Col3α1, Col4α1, and Vimentin (Figure 3D), and α-SMA expression (Figure 3E–G) were found significantly decreased in CCl4-treated α1α2HKO mice, indicating that the activation of HSCs was inhibited. Meanwhile, hepatic matrix metalloproteinase-2 (MMP2) and MMP9 activities were higher and serum tissue inhibitor of metalloproteinases 1 (TIMP-1) concentration was lower in CCl4-treated α1α2HKO mice than in the Control mice (Supplementary Figure S7C). The serum levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) significantly decreased in CCl4-treated α1α2HKO mice, indicating the improved liver function (Figure 3H and I).

Figure 3.

AMPKα knockout attenuates liver fibrosis in the CCl4 model. (A and B) Sirius red staining on mouse liver tissues was photoed and the average percentage of Sirius red-positive area in the liver was quantified by Vectra and Inform, counting in 8–10 different fields for each sample (n = 13). (C) Concentration of hydroxyproline in the liver (n = 13). (D) Hepatic mRNA expression of Col1α1, Col3α1, Col4α1, and Vimentin (n = 9). (E) Western blot analysis of protein expression of α-SMA in liver lysates. (F and G) α-SMA immunofluorescence was quantified by ImageJ, counted in 6–8 different fields for each sample, and presented as fold of the Control (n = 5). (H and I) Concentrations of ALT and AST in the serum (n = 13). Scale bar, 100 μm. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 vs. Control; t-test.

In another acute liver injury model induced by APAP (Supplementary Figure S6C), hepatic mRNA levels of α-SMA, Col1α1, and Fibronectin and α-SMA expression were significantly reduced in APAP-injured α1α2HKO mice compared with the Control mice (Supplementary Figure S8A–C), supporting that the loss of AMPKα prohibits HSC activation. Moreover, serum ALT and AST levels were significantly lower in APAP-injured α1α2HKO mice than in the Control mice, indicating that liver injury was attenuated by knocking out AMPKα in HSCs (Supplementary Figure S8D and E). Meanwhile, Sirius red staining and hematoxylin and eosin (HE) staining showed significantly less collagen deposition in the liver tissue of APAP-treated α1α2HKO mice than in the Control mice (Supplementary Figure S8F–H).

Deletion of AMPKα attenuates liver fibrosis in the BDL mouse model

In addition, we performed BDL operations to induce progressive liver fibrosis in mice (Supplementary Figure S6B). BDL-treated α1α2HKO mice exhibited less fibrosis (Figure 4A and B) and lower hydroxyproline content (Figure 4C, showing a decreasing tendency, P-value = 0.09) in the liver compared with BDL-treated Control mice. Knockout of AMPKα in HSCs attenuated the increase in hepatice mRNA expression of α-SMA, Col1α1, Col3α1, Col4α1, and Vimentin induced by biliary fibrosis (Figure 4D) and suppressed the expression of α-SMA in the BDL model, as shown by protein expression and α-SMA immunofluorescence (Figure 4E–G). The serum levels of total bilirubin (T-bil), direct bilirubin (D-bil), and ALT, as indices of biliary cell and hepatic cell damage, were all significantly lower in BDL-treated α1α2HKO mice than in the Control mice (Figure 4H–J).

Figure 4.

AMPKα knockout attenuates liver fibrosis in the BDL model. (A and B) Representative images of Sirius red staining and quantification of Sirius red-positive area in the liver (n = 6 or 7). (C) Concentration of hydroxyproline in the liver (n = 6 or 7). (D) Hepatic mRNA expression of Col1α1, Col3α1, Col4α1, and Vimentin (n = 6 or 7). (E) Protein expression of α-SMA in liver lysates. (F and G) α-SMA immunofluorescence and quantification, presented as fold change of the Control (n = 5). (H–J) Concentrations of T-bil, D-bil, and ALT in the serum. Scale bar, 100 μm. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01 vs. Control; t-test.

AMPK is needed for the activation of HSCs

To investigate the mechanism underlying the antifibrotic effects of HSC-specific knockout of AMPKα in vivo, we isolated HSCs from Control and α1α2HKO mice (hereafter referred to as Control HSCs and α1α2HKO HSCs, respectively). There was no difference in α-SMA, Col1α1, and Fibronectin protein expression between the Control and α1α2HKO HSCs at the time of isolation (Supplementary Figure S9). However, AMPKα deficiency inhibited the spontaneous transdifferentiation of primary mouse HSCs to myofibroblast-like cells during in vitro culture for 7 days, as demonstrated by the decreased mRNA expression of α-SMA, Col1α1, and Fibronectin (Figure 5A) and α-SMA protein expression (Figure 5B–D) in α1α2HKO HSCs.

Figure 5.

AMPK is important for the activation of HSCs. (A–D) HSCs isolated from Control or α1α2HKO mice were cultured for 7 days. (A) mRNA expression of α-SMA, Col1α1, and Fibronectin. (B and C) Western blot analysis of α-SMA protein expression. (D) α-SMA expression indicated by Immunofluorescence. (E–K) HSCs were isolated from short-term CCl4-injured Control or α1α2HKO mice. (E–H) Morphology and droplets. Droplets were stained with LipidTox and analyzed by ImageJ. (I) mRNA expression of α-SMA, Col1α1, and Fibronectin. (J and K) Protein expression of α-SMA, Col1α1, Fibronectin. Scale bar, 50 μm. The quantification results were from three independent experiments and presented as fold change of Control. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 vs. Control; t-test.

Then, HSCs were isolated from short-term CCl4-injured Control and α1α2HKO mice. HSCs lack of AMPKα showed fewer myofibroblast-like features but more and larger lipid droplets compared with the Control HSCs (Figure 5E–H), suggesting a less activated state after injury. Furthermore, mRNA and protein expression levels of α-SMA, Col1α1, and Fibronectin were significantly lower in HSCs lack of AMPKα (Figure 5I–K).

Quiescent HSCs reside in the space of Disse between sinusoidal endothelial cells and hepatocytes, whereas activated HSCs involved in tissue repair can migrate to injury sites. Therefore, a wound healing assay was performed to evaluate the migration ability of Control and α1α2HKO HSCs. At 12 h and 24 h after scratching, the 3-day cultured primary Control HSCs quickly migrated into the wound, exhibiting wound closure rates of ∼15% and ∼30%, respectively, while the 3-day cultured α1α2HKO HSCs showed the rates of ∼7% and ∼20%, respectively (Supplementary Figure S10A and B). After 7 days of culture in vitro, the migration of primary α1α2HKO HSCs was much slower than that of the Control HSCs (Supplementary Figure S10C and D).

Knockout of AMPKα disrupts mitophagy-induced metabolic reprogramming of activated HSCs by impairing mitochondrial function

Because HSCs and other fibroblasts require metabolic reprogramming to differentiate into myofibroblast-like cells (Chen et al., 2012), we then investigated whether inhibiting AMPK reduces HSC activation by changing the metabolic state of HSCs. Following 3 days of culture in vitro, the ATP concentration in α1α2HKO HSCs was only approximately one-third of that in Control HSCs (Figure 6A).

Figure 6.

Loss of AMPKα leads to a less myofibroblastic state of HSCs with decreased metabolic reprogramming and impaired mitophagy. (A) Concentration of ATP in 3-day cultured Control and α1α2HKO HSCs. (B–D) Representative transmission electron micrographs show mitochondrial morphology in HSCs isolated from short-term CCl4-injured Control or α1α2HKO mice. White arrows indicate the normal mitochondrial ridge in the Control group, and yellow arrows indicate the injured mitochondrial ridge in the α1α2HKO group. The number of mitochondria and the percentage of injured mitochondria were quantified (n = 22 pictures). (G–I) HSCs were isolated from Control and a1α2HKO mice. (E and F) ΔΨm of HSCs was measured by TMRE (1 μM) staining. (G–I) Western blot analysis of p-ULK1^S555, p-Raptor^S798, total ULK1, and total Raptor levels. (J and K) Representative micrographs of HSCs transfected with Cox-EGFP-mCherry on Day 5 for 36 h and quantification of mCherry (red)/EGFP (green) ratio by Image J (n = 8–10 fields per group). (L–N) α-SMA, LC3, and P62 protein expression in HSCs isolated from short-term CCl4-injured Control or α1α2HKO mice. (O) Western blot analysis of the indicated protein expression in Control or α1α2HKO HSCs treated with or without rapamycin (1 μM) for 24 h. Scale bar, 20 μm. The quantification analyses were from three independent experiments and presented as fold change of Control. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 vs. Control; t-test.

Recent studies have indicated that bioenergetic processes, including glycolysis and oxidative phosphorylation, are affected by both quality and quantity of mitochondria (Hill et al., 2012). Transmission electron microscopy of activated HSCs isolated from short-term CCl4-injured Control and α1α2HKO mice revealed that the percentage of mitochondria with disrupted cristae significantly increased in α1α2HKO HSCs, while the number of mitochondria remained the same (Figure 6B–D), suggesting that AMPK is essential in maintaining mitochondrial health and integrity during HSC activation in vivo. Detection of mitochondrial transmembrane potential (ΔΨm) by tetramethylrhodamine methyl ester (TMRE) revealed markedly decreased ΔΨm, indicating impaired mitochondrial function, in α1α2HKO HSCs (Figure 6E and F), in line with the irregular shape of mitochondria ridge in α1α2HKO HSCs (Figure 6B).

Mitophagy is a critical mechanism by which cells ensure mitochondrial quality; however, the relationship between HSC activation and mitophagy has not been investigated systemically. We found that the mitochondrial un-coupler carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone and the mitophagy inhibitor mitochondrial division inhibitor 1 could diminish the transforming growth factor β1-stimulated α-SMA protein expression in LX-2 cells, suggesting a crucial role of mitophagy in HSC activation (Supplementary Figure S11). AMPK, a master regulator of autophagy/mitophagy through mammalian target of rapamycin 1 (mTOR1) and UNC-51-like kinase 1 (ULK1), has been reported to regulate mitophagy in other cell types. To determine whether AMPK regulates mitochondrial quality through mitophagy in HSC activation, we examined the phosphorylation of Raptor at Ser798 and ULK1 at Ser555, the sites phosphorylated by AMPK to promote mitophagy. As expected, both phosphorylation levels were markedly reduced in α1α2HKO HSCs (Figure 6G–I), implying that AMPKα knockout in HSCs may attenuate mitophagy through both pathways. To directly monitor mitophagy in HSCs, we transfected Control and α1α2HKO HSCs with Cox-EGFP-mCherry, a mitochondria-targeted protein that normally exhibits yellow fluorescence and turns red in acidic lysosomes after mitophagy occurs. α1α2HKO HSCs showed a significant reduction in red puncta, indicating impaired mitophagy (Figure 6J and K). Consistently, the level of lipidated microtubule-associated protein light chain 3 (LC3), another core component of mitophagy, decreased, while the expression of Sequestosome 1 (P62) increased in HSCs isolated from short-term CCl4-injured Control and α1α2HKO mice (Figure 6L–N).

Moreover, we treated HSCs with rapamycin, an inhibitor of mTOR1, to promote mitophagy after AMPKα knockout. As shown in Figure 6O, α1α2HKO HSCs treated with rapamycin (1 μM) exhibited obviously increased expression of α-SMA protein. In addition, HSCs treated with the mitophagy inducer MTK458 demonstrated analogous experimental outcomes (Supplementary Figure S12), confirming that AMPK–mTOR1-mediated autophagy/mitophagy plays a key role in HSC activation.

Collectively, these results support the assumption that AMPK-mediated mitophagy is crucial for HSC activation as it supports the maintenance of mitochondrial health.

Discussion

HSC activation is a central driving force of liver fibrosis originated from different etiologies (Tacke and Weiskirchen, 2012). During HSC transdifferentiation, metabolic reprogramming occurs to ensure heightened energy demand of transdifferentiated myofibroblastic HSCs. The metabolic reprogramming of HSCs involves increased glycolysis, glutaminolysis, and mitochondrial oxidative phosphorylation to ensure sufficient ATP production (Trivedi et al., 2021). However, elevated glutaminolysis and mitochondrial oxidative phosphorylation produce excessive reactive oxygen species (ROS), which can promote the classic apoptotic pathway or pyroptosis through NLRP3 activation (Chung et al., 2020). Therefore, a specific machinery is necessary to ensure mitochondrial health and function during HSC activation.

Autophagy is a vital mechanism used by cells to respond to extracellular or intracellular signals to maintain cell health or adapt to ever-changing demands (Levine and Kroemer, 2019). Autophagy, specifically lipophagy, has been reported to play an essential role in HSC activation by releasing fatty acids from lipid droplets to provide energy (Thoen et al., 2011). Mitophagy, one type of selective autophagy, is important to ensure mitochondrial integrity and function in many cell types (Montava-Garriga and Ganley, 2020). However, a link between mitophagy and HSC activation has not yet been established. Here, we reported for the first time that the AMPK–mitophagy pathway is an essential mechanism ensuring mitochondrial health and function during HSC activation.

We demonstrated that the phosphorylation of AMPKα was upregulated during HSC activation in vitro and in vivo. Culture-activated primary HSCs exhibited the elevated AMPK phosphorylation. Immunofluorescence staining showed the increased p-AMPK level in the liver tissues from over-BMI patients and from mice with CCl4- or BDL-induced liver fibrosis. These results suggest that AMPK can rapidly respond to cell-intrinsic signals or extracellular cues. Various stimuli in fibrotic livers, such as inflammatory cytokines and profibrogenic cytokines, can trigger the activation of HSCs. In particular, DAMPs, such as lipid peroxide and ROS, released by injured hepatocytes play a vital role in HSC activation (Wobser et al., 2009; Das et al., 2017; van Grunsven, 2017). We found that conditioned medium from CCl4- or PA-treated hepatocytes promoted the activation of HSCs in an AMPK-dependent manner, since AMPKα knockdown by siRNA attenuated the expression of a set of fibrogenic genes. These data are of importance, as they provide the first evidence that AMPK is a common downstream sensor of various stimuli in fibrotic livers and plays a crucial role in HSC activation.

We also provided data showing that the loss of AMPKα in HSCs in vivo markedly attenuates liver fibrosis in three different mouse models, suggesting that AMPK activation in HSCs is a shared mechanism that drives liver fibrosis or injury with different etiologies. Mechanistically, the loss of AMPKα in HSCs impairs the health and function of mitochondria. AMPKα-deficient HSCs exhibited a reduction in ATP production and the distorted mitochondrial structure, implying that AMPK is a vital factor in maintaining mitochondrial integrity and capacity. By using Cox-EGFP-mCherry to monitor mitophagy, we found that the loss of AMPKα blocked mitophagy. Consistently, LC3 expression decreased, while P62 expression increased. AMPK promotes mitophagy through two major ways. AMPK can inhibit the activity of mTOR1, thus releasing ULK1 to initiate autophagy/mitophagy; it can also directly phosphorylate and activate ULK1, followed by translocation of ULK1 to mitochondria to facilitate mitophagy (Herzig and Shaw, 2018). We found that the loss of AMPKα in HSCs impaired both pathways, as evidenced by the reduced p-Raptor^Ser798 and p-ULK1^Ser555 levels. Rapamycin, an inhibitor of mTOR, has been demonstrated to attenuate mitochondrial dysfunction via the activation of mitophagy (Li et al., 2014). We found that rapamycin treatment attenuated the impairment of mitochondrial oxidative phosphorylation in HSCs isolated from α1α2HKO mice, suggesting that AMPK inhibits the activity of mTORC1 to promote mitophagy during HSC activation.

Previous studies have shown that AMPK phosphorylation reduces fibrosis in mouse models and activated primary HSCs. Here, we identified several key distinctions that may explain discrepancies between our findings and prior work. (i) Most studies linked AMPK activation to fibrosis reduction at the whole-animal or hepatocyte level, without directly manipulating AMPK activity specifically in HSCs. Our study uniquely targets AMPK in HSCs, isolating its role in HSC biology. (ii) While some studies connected AMPK to HSC inactivation, most relied on pharmacological agents with unclear AMPK-targeting specificity. Furthermore, these studies often lacked dependency validation (e.g. AMPK knockdown) to confirm causality, limiting mechanistic resolution. (iii) Critically, nearly all prior studies modulated AMPK after HSC activation or fibrosis initiation. Our work focuses on preemptive AMPK manipulation in quiescent HSCs, revealing a context-dependent role that differs from post-activation therapeutic effects.

In summary, this study elucidates the essential role of the energy sensor AMPK in the activation of HSCs. Our findings further underscore the critical contribution of energy metabolism to the activation of HSCs. An AMPK–mTOR1/ULK1–mitophagy pathway is essential for maintaining mitochondrial health and function during HSC activation in vitro and in vivo, implying that therapies inhibiting AMPK activity in HSCs may confer beneficial effects in fibrosis treatment. Since hepatocyte AMPK activation is reported to attenuate MASH and fibrosis, ideal AMPK-based antifibrotic therapy should take the targeted delivery to HSCs into consideration to avoid unwanted side effects. Moreover, these findings afford us a valuable indication that one protein can assume diverse roles in different effector cells, potentially leading to disparate influences on the progression of diseases. Consequently, it is essential to delineate the specific function of the protein by targeting investigations to discrete cellular contexts. Overall, our novel findings may provide insights leading to potential therapeutic strategies for liver fibrosis.

Materials and methods

Human liver samples

Human samples of obesity-associated liver (n = 5) were obtained from patients undergoing laparoscopic sleeve gastrectomy. Paraffin-embedded human normal liver tissues were obtained from those adjacent to the benign tumor of the liver (n = 5). The study protocol was approved by the the Clinical Review Board and Ethics Committee at Huashan Hospital of Fudan University, and all samples collected were from subjects who provided informed consent for their tissues to be used for research purposes (No. 2016-230).

Animals

All animal experiments were approved by the Animal Care and Use Committee of Shanghai Institute of Materia Medica, Chinese Academy of Sciences, where the experiments were conducted. All animals were housed in a temperature-controlled (22°C ± 2°C) room with a light/dark cycle of 12 h/12 h.

To induce chronic liver fibrosis with CCl4, mice were injected intraperitoneally with corn oil or CCl4 (0.2 ml/kg) three times a week for 8 weeks. Mice were sacrificed to collect serum and liver samples at 48 h after the last injection of CCl4. To induce chronic liver fibrosis with BDL, mice were anesthesitized by isoflurane. After a midline abdominal incision was made, the common bile duct was isolated and doubly ligated with 6–0 silk. Then, the portion of the bile duct between the two ligatures was resected. In sham-operated animals, the bile duct was similarly manipulated but not ligated. Mice with BDL or sham operation were sacrificed after 10 days. To induce acute liver injury, mice were injected intraperitoneally with APAP (20 mg/ml) and sacrificed 24 h after the injection. To isolate activated HSCs, mice were injected intraperitoneally with corn oil or CCl4 (v/v, 10%) for three times in a week (short-term CCl4 model). Primary mouse HSCs were isolated at 24 h after the last injection of CCl4.

Primary HSC isolation and culture

Mouse HSCs were isolated from adult BALB/c or C57BL/6J mice by pronase/collagenase liver perfusion and density centrifugation with Nycodenze (Weiskirchen et al., 2017). Perfusion through inferior vena cava was initiated with 30 ml EGTA solution and then continued with 30 ml Pronase E and Collagenase P solution. The liver was removed, minced, and incubated (10–20 min, 37°C, under agitation) in pronase/collagenase solution containing 2 mg/ml DNase. HSCs were separated by Nycodenze density gradient centrifugation (Accurate Chemicals) and cultured in Dulbecco’s modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS) and penicillin/streptomycin.

Preparation of conditioned medium

Primary mouse hepatocytes were isolated using Selgen’s two-step perfusion method and maintained in low-glucose DMEM supplemented with 10% FBS (Hommes et al., 1970). Hepatocytes treated with 1 mM CCl4 for 24 h (Berger et al., 1986) were switched to fresh complete cell medium (low-glucose DMEM) for another 24 h. Then, the supernatant was centrifuged and collected as CM-CCl4.

Similarly, primary hepatocytes treated with 250 μM PA for 24 h were switched to fresh complete cell medium for another 24 h. Then, the supernatant was centrifuged and collected as CM-PA.

For glutamine treatment assay, primary HSCs were cultured with complete cell medium for 4 days and then switched to DMEM without glutamine (Gln⁻) or with 4 mM glutamine (Gln⁺) for 48 h.

Generation of conditional AMPK knockout mice

To obtain HSC-specific AMPKα1/α2 double-knockout (α1α2HKO) mice, AMPKα1/α2-floxed mice were first generated by mating homozygous AMPKα1-floxed mice (stock No. 014141, Prkaa1fl, Jackson Laboratory) with AMPKα2-floxed mice (stock No. 014142, Prkaa2fl, Jackson Laboratory). Lrat-Cre mice were produced by Shanghai Biomodel Organism Science & Technology Development Co. Ltd according to Mederacke et al. (2013). Next, AMPKα1/α2-floxed mice were crossed with Lrat-Cre mice to generate HSC-specific HKO mice.

Biochemical analysis

Serum ALT, AST, T-bil, and D-bil levels were measured using Sysmex JCA-BM6010C. Liver hydroxyproline content after hydrolysis of ∼100 mg liver samples in 6 N HCl at 120°C for 22 h was measured as previously described (Reddy and Enwemeka, 1996).

MMP activity determination by gelatin zymography

MMP activity of liver tissue was assessed with gelatin zymography. Liver tissues were homogenized in Tris–HCl buffer (pH 7.6) to prepare protein samples with 5× loading buffer without sodium dodecyl sulfate (SDS). The proteins were separated on a 7.5% acrylamide gel containing gelatin, followed by gel washing and staining as previously described (Bencsik et al., 2017). Areas of enzyme activity appear as white bands against a dark blue background.

Histological and immunofluorescence analysis

Liver sections were fixed in 4% paraformaldehyde for 48 h and embedded in paraffin. Liver damage and collagen deposition were assessed by HE staining and Sirius red staining, respectively. Image analysis was performed with a positive pixel count algorithm using PerkinElmer software.

For immunofluorescence, slides were blocked for 60 min in 1× phosphate-buffered saline (PBS) containing 5% bovine serum albumin and 0.3% Triton X-100 (blocking buffer) after rehydration antigen retrieval. Blocking buffer was aspirated, and sections were incubated overnight at 4°C with an anti-rabbit α-SMA antibody. Afterward, slides were washed three times in 1× PBS and incubated for 1 h at room temperature with the respective secondary antibody (AF555). Slides were then washed and mounted with anti-fluorescence quench mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI). Relative quantification of immunofluorescent staining was performed using ImageJ software (National Institutes of Health).

Detection of mitochondrial transmembrane potential

ΔΨm was detected by TMRE, which emits a red fluorescence that can be detected by fluorescence microscopy. The level of TMRE fluorescence in stained cells can be used to determine whether mitochondria in a cell have high or low ΔΨm (Crowley et al., 2016). Briefly, primary HSCs and LX-2 cells were treated with TMRE (1 μM) in phenol red-free medium for 20 min and then washed with PBS. Photos of all samples were taken with Opera Phenix (PerkinElmer) and calculated with Columbus.

Quantitative PCR

Total RNA was isolated from cells or livers using the RNAiso Plus reagent (TaKaRa). One microgram of total RNA was reverse-transcribed using the PrimeScript RT reagent kit (TaKaRa). The resulting cDNAs were amplified using a 2× SYBR Green qPCR Master Mix (Vazyme or ABclonal) and an Applied Biosystems QuantStudio Q5 instrument (Thermo Fisher Scientific). The expression was normalized to that of the Gapdh gene.

Transmission electron microscopy

Activated HSCs were isolated from short-term CCl4-injured Control or α1α2HKO mice (n = 3 or 5). The samples were fixed in 2.5% glutaraldehyde (pH 7.4) for 2 h and embedded in agarose with low melting point. After washed three times with 0.1 M phosphate buffer (pH 7.2) and fixed in 1% osmic acid at 4°C for 2 h, the samples were gradient-dehydrated in a graded series of ethanol. Subsequently, the samples were embedded in Epon-Araldite resin for penetration and placed in a mould for polymerization. Then, ultrathin sections were counterstained with 3% uranyl acetate and 2.7% lead citrate and collected for microstructure analysis with a HT7800 transmission electron microscope. For each sample, total mitochondria and the mitochondria bearing cristae disruption were quantified, and the percentage of mitochondria with disrupted cristae was calculated. Criteria for disrupted cristae included any observable disorganization, vacuolization, or dissolution of cristae within mitochondria (Mottillo et al., 2016; Bal et al., 2017).

ATP detection

ATP concentration was determined by the commercially available CellTiterGlo^® Luminescent Cell Viability Assay kit (Promega). Primary mouse HSCs plated for 3 days were lysed in 100 μl CellTiter-Glo reagent, and luminescent signal was recorded according to the manufacturer’s recommendations. ATP disodium salt was used to generate an ATP standard curve with different concentrations of ATP (0–16 nM).

Multiplex immunohistochemistry

Multiplex staining was performed using TSA 5-color kit (D110071-50T, Yuanxibio), according to manufacturer’s instruction. Primary antibodies against desmin (Cat# ab32362) and p-AMPK (Cat# 2535L) were sequentially applied, followed by horseradish peroxidase-conjugated secondary antibody incubation (1:1, Cat# DS9800, Lecia Biosystems; 1:1 Cat#A10011-6/A10012-6, WiSee Biotechnology) and tyramide signal amplification (MD110051, WiSee Biotechnology). The slides were microwave heat-treated after each TSA operation. Nuclei were stained with DAPI (D1306, ThermoFisher) after all the antigens above were labeled. The stained slides were scanned to obtain multispectral images using Vectra (PerkinElmer) and analyzed with PerkinElmer’s Inform to calculate the ratio of co-stained area.

Western blotting

The cells were lysed with SDS loading buffer. Equal amounts of denaturated total cell lysates were separated on 10% SDS–PAGE gel. After electrophoresis, proteins were transferred to a polyvinylidene difluoride membrane, blocked with 5% fatty acid-free milk for 1 h at room temperature, and incubated with the primary antibodies as described previously overnight at 4°C. After incubation with secondary antibodies (1:5000, Jackson ImmunoResearch), the blots were visualized using the ECL detection kit (GE Healthcare). Western blotting data were quantified with ImageJ software.

Data statistics

Data are presented as mean ± standard error of the mean (SEM). Statistical analysis was performed using GraphPAD Prism (GraphPAD Software Inc.). Statistical significance between groups was evaluated using unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) test or two-way ANOVA with Fisher’s LSD test. P < 0.05 was considered statistically significant.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82130099 to J.L.), the Key New Drug Creation and Manufacturing Program of China (2019ZX09201001-003-010 to Y.Z.), Shanghai Institute of Materia Medica (CASIMM0120225006-2 to J.L.), Taishan Scholar Foundation of Shandong Province (tstp0648 to J.L.), and the Natural Science Foundation of Shandong Province (ZR2024QC378 to Hanmin Wang).

Conflict of interest: none declared.

Author contributions: Hanmin Wang, G.W., T.Y., H.L., Hanlin Wang, and Y.S.: experiments and procedures; Hanmin Wang, G.W., T.Y., Y.L., J.L., and Y.Z.: concept and design; Y.L., R.H., J.L., and Y.Z.: supervision; Hanmin Wang, G.W., and T.Y.: writing—original draft; Y.L., J.L., and Y.Z.: writing—review & editing.