Uncover Root Causes of Liver Fibrosis

MASLD and MASH are usually associated with systemic metabolic dysfunction, including obesity, diabetes, hypertension, and dyslipidemia. Excessive energy substrate is associated with de novo lipogenesis in the liver, while dysfunctional adipose tissue results in the release of excessive free fatty acids (35). Eventually, these adaptations overwhelm the liver’s buffering capacity, causing hepatic mitochondrial dysfunction and aberrant adipose tissue–liver crosstalk, leading to accumulation of toxic lipids and reactive oxygen species, and ultimately resulting in mitochondrial ER stress (35–37). This metabolic injury in hepatocytes then triggers cellular activation, death, or senescence, causing stimulation of inflammatory and fibrogenic signalling cascades that propagate activation of nonparenchymal cells (e.g., hepatic stellate cells [HSCs] and macrophages) and ultimately lead to the chronic inflammation and fibrosis characteristic of more advanced disease (38, 39). Hepatocyte lipid metabolism’s role as a central driver of MASLD pathogenesis is emphasized by large-scale GWAS, where the majority of genetic polymorphisms associated with the development and progression of MASLD and liver fibrosis (e.g., PNPLA3, TM6SF2, MBOAT7, and HSD17B13) (40) are genes predominantly expressed by hepatocytes in the liver that encode proteins responsible for nutrient processing, lipid handling, and the resultant hepatic mitochondrial redox state (41). Hence, much therapeutic focus in MASLD has been placed on targeting metabolic pathways in hepatocytes, for example inhibiting de novo lipogenesis (e.g., Aramchol), reducing energy availability (e.g., GLP-1 and/or glucagon agonists), or enhancing lipid handling (e.g., thyroid hormone receptor [THR] β analog [resmetirom], FXR agonist [obeticholic acid], PPAR agonist [lanifibrinor]) (35). Unfortunately, many of these interventions remain unproven in human MASH. What has been less clear in the field are the specific mechanisms and signalling pathways by which injured/dying hepatocytes in human liver result in nonparenchymal cell activation at different disease stages and spatial locations in the liver, how they lead to the establishment of fibrosis and contribute to disease progression, and which specific pathogenic mediators or cell subpopulations can be targeted therapeutically. Single-cell and spatial technologies are now yielding new insights into these key unanswered questions.

Epithelial cell plasticity. Hepatocyte injury is the key trigger of fibroinflammatory responses in MASH, driving a focus on applying single-cell methodologies to better dissect hepatocyte heterogeneity and transcriptional responses in regulating disease pathogenesis (Figure 1). As discussed in Table 1, the implementation of snRNA-seq has circumvented the difficulties of isolating viable hepatocytes from diseased human tissue (14, 42) and provided more clarity on human hepatocyte heterogeneity and transcriptional responses to disease. The most comprehensive study currently available included snRNA-seq data on approximately 70,000 hepatocytes from 47 patients across the full MASLD/MASH disease spectrum with a range of fibrosis stages (43). Hepatocytes showed most transcriptional changes according to disease severity of any cell type, which was most apparent in patients with advanced MASLD cirrhosis. Notably, markers of hepatocyte zonation such as GLUL and ASS1, which distinguish pericentral and periportal hepatocytes in healthy liver, respectively, are progressively more coexpressed in the same hepatocytes as MASLD progresses, as demonstrated by snRNA-seq and immunofluorescent staining (43). This observation mirrors spatial mass spectrometry data, where zonation patterns of lipids in the liver are lost in more advanced human MASLD (44).

This transcriptional reprogramming of hepatocytes in MASLD also resulted in accumulation of a subpopulation coexpressing hepatocyte and biliary epithelial (cholangiocyte) markers (e.g., KRT7, CFTR, EPCAM) that progressively expanded with MASLD severity and potentially derive from hepatocytes (43). However, significant plasticity was also observed in the cholangiocyte compartment, with expansion of cholangiocytes coexpressing hepatocyte markers (e.g., ALB, ASGR1, TTR, ASS1, PCK1, ABCC2, GPC5, HNF4α) noted across the MASLD spectrum (43). These biphenotypic cholangiocytes likely represent expanded biliary epithelial cells, key to the “ductular reaction” that has been demonstrated to be functionally important in hepatocellular regeneration following chronic injury in mice (45–48). In human MASLD, the association between ductular reaction and increased fibrosis is well recognized (49–51), while in rodent models of CLD, these biphenotypic ductular cells have been shown to promote myofibroblast activation, ECM deposition, and inflammatory cell infiltration (52–56) via secretion of key mediators such as PDGF (57), osteopontin (58), and chemokines (56, 59–61). Due to their transcriptional similarities and coexpression of both hepatocyte and cholangiocyte markers, it remains unclear whether hepatocyte-derived and cholangiocyte-derived biphenotypic epithelial cells exert functional differences in regulating fibrosis. These populations could feasibly have a distinct spatial location and local cellular niche regulating their functions. Application of high-resolution ST in MASLD tissue samples will hopefully shed further light on this.

Hepatic expression of claudin 1 (CLDN1), a member of the tight junction family of proteins, was increased in patients with MASLD (and other etiologies of CLD) and correlated with more advanced fibrosis (62). scRNA-seq and snRNA-seq data localized CLDN1 expression to hepatocytes, cholangiocytes, and biphenotypic epithelial cells as well as HSCs (62). Notably, inhibition of CLDN1 in a range of in vivo and in vitro models abrogated fibrosis and HCC formation, potentially due to reduced cellular plasticity, inhibition of ductular reaction, as well as more direct effects on myofibroblast activation and ECM production (62). Inhibition of CLDN1 via monoclonal antibody was noted to be safe in nonhuman primates (62), with an active phase II clinical trial evaluating CLDN1 inhibition in patients with head and neck cancer (Clinicaltrials.gov NCT06054477). Hence, CLDN1 inhibition is potentially an appealing target for modulation of fibrosis in MASLD.

Notch signalling was shown to increase in hepatocytes from patients with MASH and fibrosis, while in a longitudinal analysis, patients who responded to the treatment in the PIVENS trial (pioglitazone versus vitamin E versus placebo) (63) demonstrated reduced hepatocyte Notch activation (64). In mouse models of MASLD, inhibition of hepatocyte Notch reduced fibrosis despite no change in hepatocyte injury or steatosis, while overexpression of Notch exacerbated fibrosis (64). snRNA-seq on human and mouse MASLD liver tissue identified expansion of a MASH-associated hepatocyte subpopulation expressing high levels of the activation receptor tyrosine kinase ephrin type B receptor 2 (EphB2) (65). EphB2 was shown to be a downstream transcriptional target of the Notch pathway and promoted inflammatory cytokine and chemokine secretion from hepatocytes; accordingly, inhibition of hepatocyte EphB2 expression in a mouse MASH model reduced inflammatory cell recruitment and attenuated fibrosis (65). Hence, pathological Notch signalling may connect hepatocyte injury, inflammation, and fibrosis in MASH.

Beyond transcriptional changes, lipotoxicity in MASLD can drive hepatocyte death, which regulates local inflammatory and fibrogenic responses (66). Specifically, hepatocyte apoptosis was associated with more advanced MASH and fibrosis (67) and suggested to promote disease progression (68). Caspase inhibitors, which inhibit apoptosis and attenuate liver fibrosis in rodent MASH models (69, 70), were tested in clinical trials for MASLD, albeit with disappointing results so far (71). Alternative forms of programmed cell death may also be relevant; for example, necroptosis has been suggested as a predominant driver of cell death in MASLD (66). Interestingly, necroptotic (but not apoptotic) hepatocytes in MASH livers upregulate the “don’t eat me” molecule CD47, while hepatic macrophages show increased expression of the CD47 ligand SIRPα (72). Inhibiting either CD47 or SIRPα improved necroptotic hepatocyte clearance and attenuated fibrosis, highlighting this axis as a possible therapeutic target (72). Dead or dying hepatocytes may also signal directly to HSCs to promote a profibrogenic phenotype, e.g., via release of mitochondria-derived damage-associated molecular patterns (DAMPs) (73), secretion of high-mobility group box-1 (HMGB1) (74), or activation of the purinergic receptor P2Y14 on HSCs through the production of UDP-glucose and UDP-galactose (75). Targeting downstream fibroinflammatory responses to hepatocyte death might prove a more specific and tractable antifibrotic therapeutic option than global inhibition of cell death pathways, with lower potential for off-target effects or inducing the persistence of premalignant epithelial cells.

A fraction of hepatocytes develop a senescent phenotype, a state of permanent cell cycle arrest. Hepatocyte senescence, likely induced by DNA damage and telomere shortening, has been shown to correlate with fibrosis stage and predict adverse clinical outcomes in patients with MASLD (76). Senescent hepatocytes secrete a range of autocrine and paracrine factors (called the senescence-associated secretory phenotype [SASP]) that can regulate responses of adjacent epithelial and nonparenchymal cells and control local inflammation and fibrosis (77). However, before such concepts can be effectively translated, further data are needed to define the transcriptome, spatial niche, and cellular interaction partners of senescent hepatocytes in human MASH, to dissect the pathological versus protective aspects of this process. High-resolution spatial approaches will likely address these questions.

Mesenchymal cell activation. As with other fibrotic disorders, myofibroblasts expand in MASLD liver tissue and adopt ECM-producing, migratory, immunomodulatory, and contractile properties that orchestrate disease progression (78). HSCs become activated following hepatic injury (79) and have been shown to be the main source of myofibroblasts in different mouse models of liver fibrosis, including MASLD (80). Indeed, scRNA-seq analysis from patients with cirrhosis of different etiologies identified a population of PDGFRA+ ECM-expressing mesenchymal cells populating the fibrotic niche and predicted to derive from HSCs based on RNA velocity analysis (14). However, transcriptionally distinct populations of vascular smooth muscle cells and portal fibroblasts demonstrated in scRNA-seq studies (14, 81) highlight substantial heterogeneity in the hepatic mesenchymal compartment. HSCs themselves are heterogeneous, with clear patterns of zonation observed across the liver lobule (81, 82). Human HSCs in fibrotic liver can be partitioned into myofibroblastic HSCs (myHSCs), enriched in ECM-related molecules, and cytokine- and growth factor–enriched HSCs (cyHSCs), which express high levels of factors such as HGF (83). In advanced liver disease, cyHSCs, which normally exert protective functions, differentiate into myHSCs to promote disease progression, increased liver stiffness, and the development of HCC (83). The concept of myofibroblast heterogeneity and early activated HSCs/myofibroblasts being as a hub of cytokine and growth factor production before transitioning into a more ECM-producing myofibroblast subpopulation was also identified in rodent MASH- and CCl4-induced fibrosis (82, 84). However, while the balance of cyHSCs and myHSCs may influence MASLD pathogenesis, the specific signals regulating this transition between cyHSC and myHSC (and potentially back again) need further study.

Nonetheless, abundant data exist describing the mediators that promote transdifferentiation of quiescent HSCs into ECM-producing myofibroblasts, with TGF-β signalling being the key driver (79, 85). However, off-target effects complicate therapeutic targeting of ubiquitous pathways such as TGF-β. Single-cell approaches and modeling of cell-cell communication potentially enable identification more specific molecules and pathways regulating ECM-producing myofibroblasts (86, 87), for example, PDGF/PDGFRA, TNFSF12/TNFRSF12A, IL-1β/IL-1R1, and AREG/EGFR between scar-associated macrophages (SAMacs) and myofibroblasts or Notch signalling between scar-associated endothelial cells and myofibroblasts in advanced cirrhosis (14). An snRNA-seq study of 9 MASH patients demonstrated a MASH-associated HSC phenotype enriched for autocrine signalling (88). These findings were recapitulated in a mouse MASH model that identified the neurotrophin-3–neuronal receptor tyrosine kinase (NTF3/NTRK3) ligand-receptor pair as an autocrine pathway that promotes fibrogenic activity in HSCs and can be therapeutically inhibited in vivo using LOXO-195, a highly specific NTRK3 kinase domain inhibitor (88). Further recent snRNA-seq and single-cell ATAC-seq data identified transcriptional regulators of HSC activation in MASH, highlighting HSC SERPINE1 as a cell-autonomous driver of fibrogenic activity (89). Bulk profiling has also informed the identification of novel molecules that promote HSC activation; for example, proteomics revealed elevated soluble folate receptor γ (FOLR3) as a driver of HSC activation in MASH, via modulation of TGF-β signalling (90). The cellular source of FOLR3 in the MASH liver remains uncertain but should become clear in more detailed analyses of scRNA-seq and ST data from human samples. In addition to activating signals, HSCs also demonstrate loss of quiescence signals in MASH. scRNA-seq and ATAC-seq analyses in murine MASH identified NR1H4/FXR activity as a key feature of quiescent HSCs that is lost during activation (91). FXR agonists such as obeticholic acid are being actively tested in patients with MASLD (92) and may provide a therapeutic approach for maintaining HSC quiescence.

snRNA-seq analysis has also identified a senescent HSC subpopulation (93). These senescent HSCs expanded in MASH livers and demonstrated an inflammatory and fibrogenic gene expression profile in both human disease and mouse models (93). Senescent HSCs appeared to derive from activated HSCs and upregulated a series of markers, including urokinase plasminogen activator receptor (uPAR), MRC1/CD206, SLC9A9, PTPRB, and STAB2 (93). Notably, targeting senescent cells using chimeric antigen receptor (CAR) T cells directed at uPAR was shown to attenuate fibrosis in a mouse MASH model (94). However, uPAR expression is not specific to senescent HSCs, so it remains uncertain whether selective targeting of senescent HSCs will attenuate or potentially exacerbate fibrosis by promoting the persistence of ECM-producing myofibroblasts (95).

A subpopulation of portal fibroblasts with mesenchymal stem cell features (PMSCs) was identified in mice using scRNA-seq (96). PMSCs and PMSC-derived myofibroblasts expressed a gene signature (Col1a2, Col15a1, Igfbp6, Loxl1, Mgp, Thy1, Slit2) that facilitated distinction from HSCs. Slit2 in particular was specific to PMSCs, and SLIT2+ myofibroblasts were identified in the fibrotic niche of cirrhotic human liver of varying etiologies including MASLD, suggesting that PMSC-derived myofibroblasts may contribute to scar deposition in human MASLD (96). Spatially, SLIT2+ PMSC-derived myofibroblasts were found adjacent to vessels and in close proximity to SLIT2– myofibroblasts (presumed to be HSC derived) in fibrotic human liver, while SLIT2 itself has been shown to promote HSC activation (96, 97). This suggests that interactions between different mesenchymal cell types may regulate fibrogenesis in CLD. The precise role of this phenomenon in human MASLD pathogenesis remains to be determined.

Chronic inflammation and SAMac accumulation. Chronic inflammation is a key feature of MASLD and its fibrotic microenvironment (Figure 2). The innate immune system has been a major focus of scRNA-seq studies, particularly cells of the monocyte-macrophage lineage that strongly regulate fibrosis in preclinical models (98–101). Initial studies identified a distinct population of TREM2+CD9+SPP1+GPNMB+ macrophages that expand in cirrhotic liver and accumulate in the fibrotic niche (14). These SAMacs are derived from the recruitment and differentiation of monocytes rather than resident liver macrophages (Kupffer cells, KCs) and have been shown to promote HSC activation and proliferation in vitro (14, 102), suggesting a potential target population for antiinflammatory and antifibrotic therapies. Notably, transcriptionally similar SAMac populations were also described in fibrosis in other organs, suggesting conserved pathophysiological mechanisms between different fibrotic diseases (103). To confirm that accumulation of SAMacs in the fibrotic niche is not simply a feature of end-stage cirrhosis, deconvolution of bulk liver RNA-seq data across the full MASLD disease spectrum using annotated reference scRNA-seq data demonstrated that SAMac expansion correlates with fibrosis in earlier-stage disease (14), and that accumulation of SAMacs was associated with adverse clinical outcomes in patients with MASLD (8). Additionally, circulating levels of TREM2, a characteristic SAMac marker, shows promise as a serum biomarker of fibrosis in MASLD (104). Overall, these data highlight the potential role of SAMacs in the evolution of fibrosis in MASLD and other causes of CLD.

A population of TREM2+CD9+SPP1+GPNMB+ macrophages known as lipid-associated macrophages (LAMs) that are transcriptionally similar to SAMacs was also reported in various mouse models of MASLD (102, 104–107). Spatial analysis using high-plex in situ hybridization, antibody staining, and unbiased ST localized LAMs adjacent to bile ducts in healthy liver and in areas of steatosis in the MASLD liver (29), suggesting that monocytes recruited into areas of tissue injury may differentiate into LAMs/SAMacs within this niche.

To interrogate the mechanisms by which SAMacs regulate fibrosis, ligand-receptor interaction analyses from scRNA-seq data have been used to dissect candidate ligands expressed by SAMacs that are predicted to signal to HSCs/myofibroblasts to promote activation and/or proliferation (86, 87). A combination of soluble mediators, including GM-CSF, IL-17A, and TGF-β1 induced SAMac differentiation from circulating monocytes in vitro, while in vivo blockade of these mediators in the mouse carbon tetrachloride (CCl4) CLD model attenuated SAMac differentiation (102). The effect was most striking for TGF-β1 inhibition, where HSC activation and SAMac number were reduced in models of CLD and lung injury (102), indicating that both are at least partially dependent on TGF-β signalling.

Spatially resolved high-plex immunostaining of human biopsies identified a IBA1+CD16loCD163lo subpopulation of disease-associated macrophages derived from monocytes and spatially located in portal areas in close proximity to the KRT19+ ductular cells in patients with advanced MASH fibrosis, as well as other causes of CLD, including PSC (33). This close spatial relationship suggests that these cells could have functional relevance in the ductular reaction, given that macrophages are known to regulate the ductular reaction in mice (108) via secretion of soluble mediators such as Wnts (109) or TWEAK (110). Furthermore, macrophage-hepatocyte crosstalk can directly control hepatocyte mitochondrial function, lipid accumulation (111), and clearance of senescent hepatocytes (112), all important factors in epithelial dysfunction observed in MASLD (see above). How these direct epithelial-macrophage interactions can be modulated to abrogate fibrosis should be a focus of future work.

Macrophages are producers of inflammatory mediators, including activation of the NLRP3 inflammasome (resulting in release of proinflammatory cytokines IL-1β and IL-18), an important driver of fibrosis in MASLD models (113, 114). The transmembrane molecule membrane-spanning 4-domains A7 (MS4A7) was identified in TREM2+ SAMacs from MASLD livers, and MS4A7 deletion in mouse MASH reduced SAMac expansion, liver inflammation, HSC activation, and fibrosis (115). Lipid droplets derived from steatotic hepatocytes were shown to promote SAMac differentiation, MS4A7 expression, and NLRP3 inflammasome activation, with inflammasome activation being at least partially dependent on MS4A7 expression in a cell-intrinsic manner (115). These data potentially provide a mechanistic link between hepatocellular injury, SAMac differentiation, inflammation, and fibrosis; MS4A7 therefore warrants further exploration as a therapeutic target in human MASLD. The transcription factor EGR2 (116) and Notch signalling (117) were both also recently implicated in SAMac differentiation and fibrogenesis in MASLD mouse models. However, current studies aimed at investigating molecular drivers of liver fibrosis are mainly based on mouse models and underestimate the complexity of interactions regulating fibroinflammatory processes in MASLD. Application of spatial omics technologies in human MASLD samples should help clarify these interactions.

Some molecules expressed by SAMacs appear to have antiinflammatory antifibrotic functions in MASLD. The efferocytosis receptor TREM2 is a prime example, as several groups have shown that TREM2 deficiency exacerbates liver inflammation and fibrosis in MASLD models (118, 119), suggesting that TREM2 agonism may be an effective therapeutic strategy. The complexity of TREM2 in the liver is further highlighted by the presence of TREM2+ macrophages in healthy human livers, albeit at a lower proportion than in MASH (120), while resident KCs were recently reported to upregulate TREM2 in certain inflammatory contexts (121). Hence, despite numerous candidate antifibrotic targets expressed by SAMacs, it remains unclear which candidates are adequately specific to pathogenic macrophages and selectively inhibit profibrotic functions without disrupting their role in physiological repair and fibrosis regression.

Given their role in lipid metabolism and pathogen clearance, tissue-resident KCs may also have a role in MASLD pathogenesis. In rodent models, embryologically derived KCs (EmKCs) are the main macrophage population in healthy livers but undergo transcriptional reprogramming and cell death in the context of MASLD (122, 123). scRNA-seq has identified two major subsets of EmKCs: CD206loESAM– KC1, characterized by the expression of immune signatures, and CD206hiESAM+ KC2, which are involved in metabolism (124). Notably, KC2 ablation or depletion of the fatty acid transporter CD36 in this subset prevented diet-induced obesity (124). The presence of similar KC subpopulations in human MASLD is yet to be confirmed. In MASLD, the EmKC niche is repopulated with monocyte-derived macrophages that acquire a KC-like phenotype, termed MoKCs (106, 122). Interestingly, there are some suggestions that MoKCs remain functionally distinct from EmKCs, with a more pronounced inflammatory profile and increased liver injury (122, 125). The transcription factor HIF-2α was shown to simultaneously promote KC death and inflammatory activation of monocyte-derived macrophages in MASH, while deletion of HIF-2α protected against inflammation and fibrosis both in vivo and in vitro (126). Whether it is feasible to rebalance the aberrant macrophage compartment in human MASH remains unknown but should be the focus of future studies.

Of course, the chronic inflammatory microenvironment in MASH livers includes numerous other innate and adaptive immune cell types, which have also been studied using single-cell approaches and have variously been associated with the propagation of fibrosis (summarized in Table 2 and Figure 2). More detailed evaluation of which cell populations are the most pertinent drivers of fibrosis at different stages of human MASLD will help rationalize which aspects of this complex inflammatory milieu represent tractable antifibrotic therapeutic targets.