The Power of Plasticity—Metabolic Regulation of Hepatic Stellate Cells

SUMMARY

Hepatic stellate cells (HSCs) are resident non-parenchymal liver pericytes whose plasticity enables them to regulate a remarkable range of physiologic and pathologic responses. To support their functions in health and disease, HSCs engage pathways regulating carbohydrate, mitochondrial, lipid, and retinoid homeostasis. In chronic liver injury, HSCs drive hepatic fibrosis and are implicated in inflammation and cancer. To do so, the cells activate, or transdifferentiate, from a quiescent state into proliferative, motile myofibroblasts that secrete extracellular matrix, which demands rapid adaptation to meet a heightened energy need. Adaptations include reprogramming of central carbon metabolism, enhanced mitochondrial number and activity, endoplasmic reticulum stress, and liberation of free fatty acids through autophagy-dependent hydrolysis of retinyl esters that are stored in cytoplasmic droplets. As an archetype for pericytes in other tissues, recognition of the HSCs’ metabolic drivers and vulnerabilities offer the potential to target these pathways therapeutically to enhance parenchymal growth and modulate repair.

INTRODUCTION

Hepatic stellate cells (HSCs) are a resident non-parenchymal liver cell population that first gained attention because they are the primary fibrogenic cell type in liver injury (Friedman, 2008). Yet, stellate cells (previously termed lipocytes, fat-storing, perisinusoidal, or Ito cells) are now appreciated as a remarkably plastic cell type that regulates hepatic growth, immunity, and inflammation, as well as energy and nutrient homeostasis. These functions rely on metabolic versatility and tight regulation of energy expenditure. Recent studies have brought to light the pathways underlying the cell’s plasticity, and their contributions to both liver homeostasis and the organ’s response to injury. Given the role of the liver in regulating whole-body energy metabolism, HSCs are also sensitive to systemic metabolic regulation and disruption. Moreover, since HSCs are liver-specific pericytes akin to similar cells with fibrogenic potential in other organs, they are a paradigm for metabolic regulation of tissue homeostasis outside the liver. Importantly, HSCs are also an overlooked determinant of immunometabolism in supporting liver function and the response to injury.

Stellate cells are derived from the septum transversum mesenchyme during development. Mesothelial and submesothelial cells of mesenchymal origin migrate inward from the embryonic liver surface, giving rise to HSCs and other perivascular mesenchymal cells (Asahina et al., 2009, 2011). In normal liver, HSCs have best been recognized for their storage of vitamin A as retinyl esters within perinuclear droplets; the cells are the primary depot for the body’s retinoid stores (Blomhoff and Wake, 1991) (see below).

The expanding application of single-cell transcriptomic methods to characterize HSCs has yielded a nuanced and deeper appreciation of their heterogeneity (Dobie et al., 2019; Ramachandran et al., 2019, 2020; Xiong et al., 2019). Single-cell RNA sequencing has thereby uncovered zonally distributed subpopulations of HSCs in the healthy liver—portal-vein-associated HSCs and central-vein associated HSCs (Dobie et al., 2019). During liver injury HSCs transdifferentiate, or “activate” into proliferative, fibrogenic myofibroblasts (MFBs) that acquire a range of feature that collectively perpetuate injury and fibrosis (Figure 1) (Tsuchida and Friedman, 2017); as liver injury subsides the number of HSC-derived MFBs decline through either apoptosis or reversion to an “inactivated” phenotype. These inactivated cells harbor unique epigenetic signatures, and these HSCs re-activate more readily upon re-injury (Kisseleva et al., 2012; Liu et al., 2020; Troeger et al., 2012).

Figure 1. Features of HSC Activation and Resolution.

Normally quiescent vitamin-A-rich pericytic cells, upon liver injury the cells initiate activation, rendering them responsive to many cytokines and soluble signals that trigger a full cellular transdifferentiation to myofibroblasts, whose features include enhanced proliferation, contractility, fibrogenesis, as well as altered matrix degradation properties, chemotaxis (directed migration), and the capacity to modify the inflammatory and immune milieu. Once injury resolves, the fate of these cells includes apoptosis, progressive senescence and/or deactivation to a more quiescent, intermediate cell phenotype. The relative distribution among these three fates during fibrosis resolution has not been determined, however.

Genes related to zonation are preserved following liver injury, and lineage tracing reveals that the dominant source of pathogenic MFBs are pericentral HSCs when the injury begins in that region (Dobie et al., 2019), underscoring the concept that patterns of activation overlap with the areas of injury. Moreover, in pericentral injury, periportal HSCs respond with increased proliferative capacity, but they may not transdifferentiate into collagen-producing MFBs (Dobie et al., 2019). Apart from their zonal distribution activated HSCs can also be subdivided into other subpopulations that are pro-regenerative (high growth factor expression), anti-regenerative (profibrogenic), and mixed (Krenkel et al., 2019). While the paradigm of “quiescent,” “activated,” and “inactivated” cellular states provides a valuable conceptual construct, it may not fully reflect intermediate or hybrid states. HSC heterogeneity might be manifested as different subtypes with variable capacity for activation, or with divergent contributions to regeneration, fibrosis, cancer, and immunomodulation, among others. For example, during regeneration, HSCs contribute to the initiation and termination of liver regeneration and are a major source of hepatocyte growth factor (HGF) in the liver (Bansal, 2016; Maher, 1993). They also support sinusoidal endothelial cell integrity through paracrine interactions, which in turn are critical for angiogenesis during repair and regeneration (DeLeve, 2013, 2015; Desroches-Castan et al., 2019).

Chronic liver injury promotes ongoing activation of HSCs, leading to accumulation of extracellular matrix proteins that disrupt the liver’s architecture and undermine its function (Henderson et al., 2013; Mederacke et al., 2013). HSC activation is triggered by a number of signals that serve as markers of cellular injury, including proinflammatory cytokines produced by infiltrating immune cells, hepatocyte apoptotic bodies, endothelial-cell-mediated growth factor activation, and increased reactive oxygen species (ROS) burden. The cells’ response to injury is potentiated by an array of paracrine and autocrine loops (Higashi et al., 2017), including fibrogenic signals such as transforming growth factor beta 1 (TGF-β1) and connective tissue growth factor (CTGF).

Activation of HSCs requires energy to support the cell’s ability to proliferate; secrete extracellular matrix, proteases, and cytokines; and migrate toward regions of cellular injury. To meet these energy demands, the cells harness a number of metabolic pathways similar to cancer cells, which also have high energy demands (Lane et al., 2020), although the latter have unregulated growth driven by major genomic and epigenomic disruptions not described in HSCs.

The rapidly evolving breadth of the HSC’s functions and the highly contextual nature of its responses make understanding its metabolic regulation a new priority, with broader implications for pericyte metabolism in other tissues. In this review, we integrate emerging insights into HSC energetics, its crosstalk with cellular stress responses, and non-cell autonomous effects on liver homeostasis and immunometabolism to demonstrate how metabolic regulation of HSCs underlies their functions in health and disease, revealing a rich and wide-ranging biology.

HSC BIOENERGETICS

Carbohydrate Metabolism

Glycolysis and Gluconeogenesis

The reprogramming of HSC central carbon metabolism plays a major role in the cell’s activation, as they upregulate glycolysis to meet the energy demands of transdifferentiation into a myofibroblast (MFB) phenotype (Figure 2). Importantly, changes in carbon metabolism are not only a hallmark of the MFB phenotype, but they also contribute to activation. Studies interrogating glucose homeostasis regulation in HSCs rely almost entirely on cultured cells, and thus future studies will need to validate these findings in disease models in vivo.

Figure 2. Glucose and Mitochondrial Metabolism in Activated HSCs.

Exosomes from aHSCs, which contain GLUT1 and pyruvate kinase muscle isozyme 2 (PKM2), induce activation of qHSCs. Overall glycolytic flux is increased in activated HSCs relative to quiescent HSCs. Glucose influx capacity increases through upregulation of the glucose transporters GLUT1, GLUT2, and GLUT4. The pyruvate byproduct of glycolysis is shunted primarily toward lactate production, which accumulates in the cells despite increased expression of the lactate transporter MCT4 and increased lactate efflux. Lactate accumulation is a contributor to HSC activation. Despite a preferential shunting toward lactate production, overall mitochondrial activity is increased when HSCs activate. Additionally, there is an increase in mitochondrial ROS production. Major pro-fibrotic signals that contribute to the reprogramming of glucose metabolism include Hh signaling through Gli and leptin, both of which converge upon the transcription factor HIF-1α, an inducer of glucose transporters and glycolytic enzymes. DNA methyltransferase and histone methyltransferases induce epigenetic modifications that shift HSC metabolism toward increased glycolysis. TGF-β signaling increases HSC expression of RAGE, thereby making activated HSCs sensitive to global glucose metabolism, as AGEs are a result of hyperglycemia.

Compared with their quiescent counterparts, immortalized human HSCs (aHSCs) and culture-activated rat HSCs have increased glucose utilization rates (Gajendiran et al., 2018). Primary culture-activated cells have enhanced glucose transport capacity and glycolytic activity (Chen et al., 2012). Glucose transporter proteins, including GLUT1 (Chen et al., 2012). GLUT2 (Lin and Chen, 2011), and GLUT4 (Chandrashekaran et al., 2016), are expressed in primary culture-activated or immortalized rat HSCs and can be regulated under these conditions by high extracellular glucose (Chen et al., 2012) or purinergic signaling (Chandrashekaran et al., 2016). Interestingly, GLUT1 also is highly overexpressed in cancer cells, which primarily rely on anaerobic metabolism for ATP production; in cancer cells, glucose transport constitutes a rate-limiting step for ATP production (Amann and Hellerbrand, 2009).

In either immortalized human or primary murine HSCs, there is also increased mRNA expression of genes related to intracellular processing of glucose, hexokinase 2 (HK2), fructose-2,6-bi-sphosphatase-3 (PFKFB3), and pyruvate kinase (PK) (Mejias et al., 2020). Upregulation of PFKFB3 is mediated by stabilization of its mRNA through binding its 3-untranslated region by the cytoplasmic polyadenylation-element-binding protein 4 (CEBP) (Mejias et al., 2020). Enhanced glycolysis attributable to these enzyme inductions is accompanied by downregulation of genes involved in gluconeogenesis, including phosphoenolpyruvate carboxykinase 1 (PCK1) and fructose bisphosphatase 1 (FBP1) (Chen et al., 2012).

Increased glycolysis in cultured HSCs is accompanied by shunting of central carbon metabolic products away from the citric acid cycle. Activated HSCs have increased expression of pyruvate dehydrogenase kinase 3 (PDK3), which inhibits the conversion of pyruvate to acetyl coenzyme A, thereby channeling pyruvate toward lactate production. As noted above, a similar phenomenon, known as the Warburg effect, is observed in cancer cells (Liberti and Locasale, 2016; Vander Heiden et al., 2009). Like in cancer cells, this response may confer an evolutionary advantage by allowing them to prioritize ATP production rate over efficiency (i.e., ATP yield per glucose molecule) (Liberti and Locasale, 2016), or by diverting resources away from ATP production and toward proliferation (Vander Heiden et al., 2009). By analogy, both possibilities support the enhanced energy demands of aHSCs to secrete extracellular matrix (ECM) protein, migrate, and proliferate. These features are likely to be relevant to myofibroblasts in other tissues, but it is unclear whether the Warburg-like effect is more dominant in myofibroblasts than other proliferating cell types during injury. However, the enzyme pyruvate kinase M2 (PKM2) may represent a unique, shared link between aHSCs and cancer cells (Christofk et al., 2008), since it is induced in both cell types to promote aerobic glycolysis (Zheng et al., 2020). A PKM2 antagonist is antifibrotic in an animal model by promoting tetramerization of the enzyme, which attenuates its activity; in contrast, the activating effect of PKM2 is mediated by histone modifications, albeit in cultured HSCs but not yet established in vivo (Zheng et al., 2020).

Interestingly, lactate does not play a bystander role in HSC metabolism; rather, it is directly implicated in HSC activation and the perpetuation of the MFB phenotype. Despite upregulation of expression of the lactate export pump monocarboxylate transporter 4 (MCT4), intracellular levels of lactate remain elevated in aHSCs. Importantly, blocking the intracellular accumulation of lactate inhibits proliferation, suppresses genes associated with the MFB signature, and induces lipid accumulation and lipogenic genes (Chen et al., 2012).

Exosomes provide a mechanism for the rapid induction of glycolysis to support the metabolic reprogramming from quiescent HSCs (qHSCs) to activated HSCs (aHSCs), allowing for a synchronization of stromal cell injury response. Exosomes from aHSCs, which contain GLUT1 and PKM2, induce activation of qHSCs (Wan et al., 2019). The production of exosomes by aHSCs is potentiated by hypoxia-inducible factor 1-alpha (HIF-1α) signaling, which stimulates HSC activation in hypoxic and inflammatory conditions (Wang et al., 2013).

Epigenetic regulation also contributes to shifts in glucose metabolism in HSCs. DNA methyltransferase DNMT1 and histone methyltransferase G9a induce epigenetic modifications that shift HSC metabolism toward increased glycolysis (Barcena-Varela et al., 2020). The profibrotic cytokine TGF-β triggers recruitment of these enzymes to chromatin in cultured human HSCs. Furthermore, combined inhibition of DNMT1 and G9a returns glycolytic rates to those of quiescent HSCs in in vitro models of hypoxia-driven and TGF-β1-driven activation (Barcena-Varela et al., 2020).

Glycosylation

In addition to changes in carbohydrate catabolism, aHSCs display altered protein glycosylation profiles (Qin et al., 2012; Wu et al., 2017). The cell-surface glycome of aHSCs contains glycans that are favorable for galectin-1 (Gal-1) binding. Gal-1, which is upregulated in aHSCs relative to qHSCs, binds to neuropilin-1 (NRP-1), a receptor that induces angiogenesis, vascular permeability, and wound healing (Troeger and Schwabe, 2011). The interaction of Gal-1 and NRP-1 induces TGF-β- and PDGF-like signaling cascades (Wu et al., 2017) (TGF-β signaling is discussed further in the redox metabolism section, below).

Hyperinsulinemia and Hyperglycemia

Hyperinsulinemia and hyperglycemia, key features of the metabolic syndrome, can activate HSCs in culture but their specific contribution to HSC activation in the context of NAFLD and NASH is unknown. For example, HSCs grown in insulin-containing media undergo activation, as evidenced by a loss of lipid droplets (Lin et al., 2009), and have increased expression of CTGF, a key fibrogenic signaling molecule (Gao and Brigstock, 2004; Paradis et al., 2001). Primary murine HSCs cultured in high-glucose media have increased production of collagen type I, which is mediated by the SMAD3 transcription factor. Exposure to high glucose results in HSC upregulation of protein inhibitor of Stat4 (PIAS4), which represses transcription of SIRT1. Reduced SIRT1 leads to lysine hyperacetylation of SMAD3, thereby increasing affinity of SMAD3 for its profibrotic DNA-binding targets (Sun et al., 2016). Like many studies interrogating HSC metabolism; however, as noted, the links between hyperglycemia and hyperinsulinemia and HSC activation are based on culture studies, rather than in vivo studies. The only in vivo studies linking hyperinsulinemia to HSC responses describes enhanced mitochondrial volume but not typical features of HSC activation (Evert et al., 1998)

Advanced glycation end products (AGEs), which are generated by non-enzymatic reactions between proteins and sugars, also activate HSCs. In culture, murine HSC activation driven by TGF-β induces expression of the receptor for advanced glycation end product (RAGE), an effect that is abrogated through MAPK/ERK pathway blockade (Fehrenbach et al., 2001). Medium containing AGEs increases fibrogenic gene expression and ROS by cultured human HSCs (Iwamoto et al., 2008). In vivo, rats injected intraperitoneally with AGEs, develop more severe liver fibrosis following bile-duct ligation (BDL) compared with those not receiving AGEs. AGE-exposure also increased a-smooth muscle actin protein levels but does not induce fibrosis (Goodwin et al., 2013). This finding suggests that AGEs and hyperglycemia alone may not be sufficient to induce fibrosis; however, in the presence of tissue injury they may amplify fibrogenesis driven by other factors.

Mitochondrial Metabolism

Despite a preferential increase in glycolysis, the demands of maintaining a myofibroblast-like phenotype still require major energy contributions from oxidative phosphorylation, reflected in an increase in mitochondrial number (Chen et al., 2012) and activity (Gajendiran et al., 2018) during cellular activation. However, changes in mitochondrial metabolism from activation may not be limited to ATP production, as studies point to a role in ROS production as well (Jain et al., 2013) (the contributions of ROS to HSC activation are detailed in subsequent sections).

Compared with freshly isolated qHSCs in which mitochondria are sparse, culture-activated HSCs have abundant mitochondria (Chen et al., 2012; Friedman et al., 1985). This increased number is accompanied by heightened mitochondrial activity, as measured by elevated oxygen consumption rate, mitochondrial membrane potential, and expression of F₁-F_o ATPase (Gajendiran et al., 2018). Mitochondrial uncoupling inhibits HSC activation in vitro. Treatment with the chemical uncouplers FCCP and valinomycin reduce HSC expression of activation markers and blunt the response to the profibrotic cytokine TGF-β (Guimarães et al., 2012). Uncoupling also reduces ATP production by 30% (Guimarães et al., 2012), suggesting that despite upregulation of glycolysis, a significant and necessary portion of ATP production is still generated through oxidative phosphorylation. In vivo, enhanced mitochondrial volume in HSCs has been described in hyperinsulinemic mice (Evert et al., 1998), similar to changes described in in-culture-activated HSCs (Friedman et al., 1992, 1989, 1985). Interestingly, there are no studies to date that document similar features of mitochondria in aHSCs in fibrotic liver in vivo.

Fibroblast populations from other tissues reinforce a broader context for these findings. For example, in kidney fibroblasts TGF-β activity increases ROS production through inhibition of the mitochondrial protein complex III (Jain et al., 2013), a member of the electron transport chain. In cardiac fibroblasts, TGF-β stimulation increases mitochondrial ROS through NLRP3, a pattern recognition receptor implicated in the pathogenesis of inflammatory diseases (Bracey et al., 2014).

Vitamin A and Lipid Metabolism

Retinol Storage and Lipid Droplets

Regulation of vitamin A homeostasis is a defining characteristic of HSCs in healthy and injured liver. HSCs store 50%–95% of the body’s vitamin A (Blomhoff and Blomhoff, 2006; Saeed et al., 2017), which is comprised retinol and its metabolites. Over 95% of these retinoids are stored as retinyl esters (REs), which account for 30%–50% of HSC cytoplasmic lipid droplet content (Senoo et al., 2010). While RE may be a precursor for retinoic acid (RA) and hence increase retinoid signaling, they may also function as buffer that limits the amount of active retinoids. Retinoids in HSCs are associated with several perilipins, proteins that associate with the surface of lipid droplets. These include adipose differentiation-related protein (ADRP) (Fukushima et al., 2005; Lee et al., 2010), adipophilin (Straub et al., 2013), perilipin 5 (Lin et al., 2018), ATGL, and CGI-58 (Eichmann et al., 2015), as well as liver fatty-acid-binding protein (L-FABP) (Chen et al., 2013). Knowledge about the functional roles of perilipins in this cell type is limited, but in general greater perilipin expression reduces HSC activation, possibly by stabilizing retinoid droplets to attenuate their catabolism to fatty acids that provide energy (Lee et al., 2010; Zhou et al., 2020).

Blood vitamin A levels are tightly regulated, but specific mechanisms of vitamin A uptake and release in the healthy liver are not fully clarified (Saeed et al., 2017). Retinol uptake by HSCs occurs both as free retinol and as holo-RBP, the combination of retinol-binding protein (RBP) and retinol, although the latter is preferentially esterified (Ottonello et al., 1987; Trøen et al., 1994). Retinol esterification is primarily under the enzymatic control of lecithin:- retinol acyltransferase (LRAT) (Ajat et al., 2017; Blaner et al., 1985, 1990; Blomhoff et al., 1985; Matsuura et al., 1997; Trøen et al., 1994), although acyl-CoA:retinol acyltransferase (ARAT) also plays a minor role (Ajat et al., 2017; Ross, 1982).

As with many features of HSCs, the cells are heterogenous with respect to retinoid content (D’Ambrosio et al., 2011; Krenkel et al., 2019). Differential FACS sorting of primary HSCs—sorting for retinol autofluorescence versus collagen-promoter driven GFP expression—reveals at least two major subpopulations within the healthy liver (D’Ambrosio et al., 2011). Sorting for autofluorescence reveals a subpopulation with a higher capacity for retinol storage, while sorting for collagen-promoter driven GFP expression uncovers a subpopulation with a more MFB-like phenotype, suggesting that these HSCs may be “primed” to respond to liver injury. The upregulation of genes related to retinol catabolism and smaller lipid droplet size in the “primed” HSCs reveals some early changes to HSC lipid metabolism that may occur in vivo in the setting of liver injury (D’Ambrosio et al., 2011).

Retinol is thought to contribute to maintenance of the qHSC phenotype by its interaction with the retinoic acid receptor β (RARβ) and retinoid X receptor α (RXRα) nuclear receptors (Wang et al., 2002). Vitamin A treatment of HSCs prevents culture activation and can partially maintain qHSC marker expression (GFAP and PPAR-γ) while suppressing the expression of MFB markers (αSMA, Col1a1, and HSP-47) (Yoneda et al., 2016). Additionally, retinol exposure causes a partial reversion of aHSCs to a quiescent state (El Taghdouini et al., 2015; Hong et al., 2018; Lee et al., 2010). RA activates the nuclear receptors RARβ and RXRα, allowing them to bind to RA responsive elements located in the Col1α2 promoter (Wang et al., 2002), thereby providing a mechanism for endogenous RA-mediated potentiation of a quiescence phenotype in physiologic conditions (Allenby et al., 1993). The exact relationship between retinyl ester catabolism and signaling through RARβ and RXRα is not defined, however.

The loss of lipid droplets occurs histologically in a biphasic manner (Testerink et al., 2012). The first phase involves the breakdown of lipid droplets into smaller units, which are then redistributed from the perinuclear region to newly formed cellular extensions. Within the aHSC population, retinoid content is relatively homogeneous, with similar enzymatic activity and lipid composition regardless of size or localization. In this initial stage, the retinyl ester content of the lipid droplet declines, while relative triglyceride levels increase. In the second phase the lipid droplets have reduced size, until none are present in the fully transdifferentiated MFB.

Retinyl ester hydrolases (REHs) liberate REs from lipid droplets during HSC activation (Saeed et al., 2017). A number of potential enzymes have also been implicated, including adipose triglyceride lipase (ATGL/PNPLA2) (Taschler et al., 2015), PNPLA3 (Pirazzi et al., 2014; Taschler et al., 2015), hormone-sensitive lipase (Mello et al., 2008; Pang et al., 2011), and carboxyl ester lipase (Weng et al., 1999). However, other uncharacterized REHs may contribute to hydrolysis as well (Schreiber et al., 2012). Culture studies suggest retinol release from HSCs occurs independently of RBP or binding partners, as inhibition of protein secretion does not modulate retinol secretion (Friedman et al., 1993; Sauvant et al., 2001). However, the primary modality of retinol secretion is controversial and could be characterized by either the direct release of free retinol into the blood by HSCs, or holo-RBP secretion by hepatocytes (Senoo et al., 2010).

HSC activation appears to require loss of lipid droplets through autophagy, which may be critical to fueling this metabolically demanding cellular response (Figure 3) (Hernández-Gea and Friedman, 2012). HSC activation increases intracellular REH activity and subsequently provokes release of retinol without an increase in the release of REs. The lack of extracellular REH activity in aHSCs further implicates intracellular hydrolysis in mediating retinol loss (Friedman et al., 1993). Bile salt-independent REHs (Friedman et al., 1993), PNPLA3 (Pirazzi et al., 2014), ATGL (Taschler et al., 2015), and lysosomal acid lipase (LAL) (Grumet et al., 2016; Tuohetahuntila et al., 2017) have been implicated as enzymatic mediators of RE release from lipid droplets. Interestingly, hormone-sensitive lipase, which is an important REH in liver adipocytes (Mello et al., 2008), is not highly expressed by HSCs and does not affect HSC lipid mobilization (Taschler et al., 2015). In parallel to increased hydrolase activity, aHSCs also have a diminished capacity for retinol esterification, as HSC activation results in a rapid loss of LRAT expression (Kluwe et al., 2011).

Figure 3. Lipid Metabolism during HSC Activation.

Lipid metabolism in activated HSCs is characterized by a loss of retinyl ester-containing cytoplasmic droplets. The transcription factors PPAR-γ and SREBP-1c, adipogenic master regulators, are markers of HSC quiescence that are downregulated during activation. Lipid droplets are trafficked to the autosome for degradation. Vitamin A de-esterification also occurs under the control of lysosomal acid lipase REH activity is also increased, liberating free retinol to the extracellular space. Lipid droplet contents from the autolysosome are further metabolized by the cell through increased fatty acid β-oxidation. Conversion of some intracellular retinol to retinoic acid promotes increased RARβ and RXRβ transcription.

Consistent with engagement of autophagy, retinyl ester processing occurs partly within lysosomes. REs are trafficked to the lysosome, likely through the autophagy pathway, where they are broken down by LAL into retinol (Tuohetahuntila et al., 2017). The inhibition of LAL, which also catabolizes cholesterol esters and triglycerides, decreases HSC activation. Notably, despite inhibition of lysosomal RE degradation, culture activation of HSCs still results in a 75% loss of REs, implying a significant role of other pathways in processing of lipid droplets (Tuohetahuntila et al., 2017).

Adipogenic Phenotype and Fatty Acids

Regulators of fatty acid content and metabolism control HSC activation. Peroxisome proliferator-activated receptor gamma (PPAR-γ) (Shao et al., 2016) and sterol regulatory-element-binding protein-1 (SREBP-1c) (Eberlé et al., 2004), two adipogenic transcriptional master regulators, are markers of HSC quiescence that are lost upon activation (She et al., 2005; Tsukamoto, 2005). Furthermore, ectopic induction of PPAR-γ and SREBP-1c can reverse HSC activation, demonstrating that an adipogenic transcriptional program maintains HSC quiescence (She et al., 2005; Tsukamoto, 2005), along with a growing list of transcriptional activators (Liu et al., 2020; Nakano et al., 2020; Wang and Friedman, 2020). Fatty acid synthesis-related genes under transcriptional control of PPAR-γ and SREBP-1c include acyl-CoA oxidase, solute carrier family 27 (fatty acid transporter) member 4, acetyl-CoA carboxylase, and fatty acid synthase (She et al., 2005). Fatty acids provide an important substrate for the esterification of retinol in qHSCs (Tsukamoto et al., 2006).

During activation, contents of HSC lipid droplets are metabolized to provide fatty acid species for oxidation, generating energy for activation/transdifferentiation under the control of PPAR-β (She et al., 2005), a known activator of fatty acid oxidation (Wang et al., 2003) (see above). PPAR-γ and SREBP-1c downregulation occurs through activation of the MAPK/ERK signaling cascade (Yan et al., 2012; Zhou et al., 2010), which is a nexus for several profibrogenic cytokines.

In early stages of activation, the most abundant fatty acids—palmitic, oleic, palmitoleic, and stearic acid—reach peak concentrations. This is followed by a decline in the total amount of free fatty acids in the later stages of activation, with a relative enrichment of arachidonic (C20:4) and docosahexaenoic (C22:6) acid and an upregulation of fatty acid elongases (Elovl5, Elovl6) and desaturases (Scd1) (Shmarakov et al., 2019). The overall decline in HSC fatty acid content is consistent with loss of an adipogenic phenotype. Fatty acid β-oxidation is an important energy source for HSCs undergoing activation, as inhibition of mitochondrial fatty acid catabolism blocks HSC activation (Hernández-Gea and Friedman, 2012). Interestingly, β-oxidation inhibition has analogous effects to autophagy inhibition, reinforcing the importance of autophagy to HSC lipolysis (Hernández-Gea and Friedman, 2012).

Acyl-CoA carboxylase (ACC), a regulator of fatty acid β-oxidation and de novo lipogenesis (DNL), has been implicated in HSC metabolic reprogramming during activation. ACC inhibition reduces HSC activation as measured by reduced αSMA expression and collagen production. The antifibrotic activity of ACC inhibitors is through inhibition of DNL, which is necessary for the induction of glycolysis and oxidative phosphorylation during HSC activation. While the molecular mechanism through which DNL regulates HSC metabolism has yet to be elucidated, ACC inhibition is being evaluated as a treatment for NASH, as it both reduces lipotoxicity in hepatocytes and inhibits HSC fibrogenesis (Bates et al., 2020); however, effects of ACC inhibition in vivo have not been directly ascribed to its effects on HSCs.

Cholesterol

Cholesterol contributes to the pathogenesis of fibrogenesis in NASH, as HSC activation can be induced by free cholesterol (FC) accumulation (Teratani et al., 2012; Tomita et al., 2014). Moreover, murine diets high in cholesterol are more effective in inducing NASH (Tsuchida et al., 2018; Wang et al., 2020). In cultured HSCs, FC accumulation increases Toll-like receptor 4 (TLR4) levels, sensitizing the HSCs to TGF-β-induced activation (Seki et al., 2007; Teratani et al., 2012). FC accumulation in HSCs is under the control of low-density lipoprotein receptor (LDLR) and miR-33a, a microRNA regulator of cholesterol metabolism, both of which are upregulated during HSC activation (Tomita et al., 2014). LDLR and miR-33a-mediated signaling impairs endosomal-lysosomal degradation of TLR4, which, in turn, downregulates bone morphogenetic protein and activin membrane-bound inhibitor (BAMBI) levels, a pseudoreceptor for TGF-β, thereby increasing HSC sensitivity to TGF-β (Onichtchouk et al., 1999; Teratani et al., 2012).

Oxidized LDL (ox-LDL) can directly activate HSCs, at least in culture. The scavenger receptors CD36 and lectin-like oxidized LDL receptor-1 (ox-LOX receptor-1, or LOX-1) mediate HSC ox-LDL uptake, resulting in an increased expression of profibrogenic genes (Kang and Chen, 2009; Schneiderhan et al., 2001). LOX-1 expression is upregulated by canonical Wnt signaling (Kang and Chen, 2009), which has been implicated as an HSC activator (Cheng et al., 2008; Myung et al., 2007).

Glutaminolysis

The immense metabolic demands of ECM production are met not only through reprogramming of carbohydrate and lipid metabolism but also protein metabolism. A comparison of the differentially expressed metabolic genes between qHSCs and aHSCs reveals that only 6% of such genes involve carbohydrate metabolism, while 38% are involved in protein metabolism, signified by a shift toward glutaminolysis (Du et al., 2018). Glutaminolysis is the conversion of the amino acid glutamine into α-ketoglutarate, a TCA cycle intermediate, that is typically observed in cancer cells. Activated HSCs have upregulated expression levels of the glutaminase GLS1, the rate-limiting enzyme of glutamine utilization in cancer cells. Increased glutaminase tGLS1 is abrogated by blocking yes-associated protein (YAP) signaling, which is a downstream effector of the fibrogenic agonist Hh. Glutamine deprivation of HSCs provokes lipid accumulation, increased PPAR-γ expression, and decreased collagen I expression, underscoring the importance of glutaminolysis as an energy source for aHSC (Du et al., 2018). In vivo, glutaminolysis by HSCs marks active fibrogenesis and its cell-specific antagonism represents a potential therapeutic target by depriving the cells of glutamine (Du et al., 2020). In contrast to glutamine, glucose deprivation does not have an antifibrogenic effect, despite increased glycolytic gene expression and glucose consumption by aHSCs (Chen et al., 2012; Du et al., 2018).

LINKING HSC METABOLISM TO STRESS RESPONSES

Oxidative Stress

ROS

Production of ROS is an important characteristic of aHSCs, as knockdown or inhibition of ROS-producing enzymes attenuate HSC activation. Enhanced ROS production corresponds with increased mitochondrial activity (Gajendiran et al., 2018).

Several classic activators of HSCs converge on ROS-mediated activation, especially through upregulation of NAPDH oxidase (NOX) enzymes (Figure 4). Generation of ROS is viewed as a “feed forward” mechanism of fibrogenic cell activation, particularly through TGF-β (Hellerbrand et al., 1999). Redox imbalances activate the latent form of TGF-β, while TGF-β signaling generates redox imbalances (Liu and Desai, 2015). TGF-β acts primarily through SMAD transcription factors (Breitkopf et al., 2006) and the MAPK/ERK pathway to activate a fibrogenic transcriptional program (Tsukada et al., 2005). TGF-β signaling upregulates several NAPDH oxidase enzymes—NOX1, NOX2, NOX4, and NOX5 (Andueza et al., 2018; Sancho et al., 2012) — which increase H₂O₂ generation leading to enhanced collagen production. Other inducers of NOX in HSCs include platelet-derived growth factor (PDGF) (Adachi et al., 2005), leptin (De Minicis et al., 2008), angiotensin II (Ang II) (Bataller et al., 2003), and advance glycation end products (Iwamoto et al., 2008). Like TGF-β signaling, PDGF (Adachi et al., 2005), leptin (De Minicis et al., 2008), Ang II (Bataller et al., 2003), and AGEs (Chang et al., 2004) act through the phosphorylation and activation of the MAPK/ERK signaling pathway. The knockdown or inhibition of NOX enzymes blunt the potency of stimuli to fibrogenesis, underscoring their importance to HSC activation in liver fibrosis (Adachi et al., 2005; Bataller et al., 2003; De Minicis et al., 2008; Hanafusa et al., 1999; Iwamoto et al., 2008; Tsukada et al., 2005). H₂O₂ facilitates the binding of the transcription factor C/EBPβ to its cognate element on the alpha-1 type I collagen (col1α1) promoter, thereby increasing its expression (García-Trevijano et al., 1999). The TGF-β promoter also contains C/EBPβ responsive elements (Abraham et al., 2009), supporting the “feed forward” element of TGF-β/ROS signaling. However, SMADs, which are downstream effectors of TGF-β, attenuate C/EBPβ-mediated TGF-β upregulation (Abraham et al., 2009), suggesting that the TGF-β/ROS circuit is tightly regulated in HSCs.

Figure 4. Redox Metabolism in HSC Activation.

Profibrotic stimuli (TGFb, leptin, AGEs, angiotensin II, and PDGF) converge upon a MAPK/ERK signaling pathway that leads to increased NADPH oxidase (NOX) expression. Increased NOX expression increases H₂O₂ levels, which facilitates the finding of the transcription factor C/EBPβ to its cognate element on the Col1α1 promoter, thereby increasing collagen expression. The increase in H₂O₂ levels is accompanied by an increased capacity of HSCs to handle oxidative stress, such as through increased glutathione and superoxide dismutase levels. Combined with responsiveness of collagen expression to H₂O₂, this points toward a signaling role for H₂O₂ in activated HSC, which is traditionally considered to be a marker of cell stress.

Redox metabolism in HSCs also acts a feedforward mechanism for activation; TGF-β signaling increased H₂O₂ levels through induction of NOX expression, while TGF-β itself converted from its latent from to its active form in the presence of H₂O₂

While the profibrogenic contributions of NOX-generated ROS are well characterized, a definitive role for mitochondrial ROS in HSC activation has not been established. Studies conducted in embryonic (Ranganathan et al., 2001), kidney (Jain et al., 2013), and heart fibroblasts (Bracey et al., 2014) demonstrate increased mitochondrial ROS production during cellular activation. Furthermore, mitochondrial ROS metabolized to H₂O₂ by manganese superoxide dismutase (Sod2) upregulate matrix metalloproteinase transcription in embryonic fibroblasts, which increases Sod2 activity resulting in increased MMP-1, MMP-2, MMP-3, and MMP-7 expression (Ranganathan et al., 2001). Conversely, an endocannabinoid-mediated increase in mitochondrial ROS induces HSC cell death (Siegmund et al., 2007), consistent with the classical role of mitochondrial ROS as a regulator of apoptosis (Fleury et al., 2002; Giorgio et al., 2005; Li et al., 2003). Differential regulation of these two major axes of ROS production may explain the more variable contributions of mitochondrial ROS, in contrast to the definitively profibrogenic nature of NOX enzymes (Brunati et al., 2010).

RNS

The functions of reactive nitrogen species (RNS) in HSC biology are not as well clarified as those of ROS. Nearly all liver cell types, including HSC, can produce nitric oxide (NO·) (Feder et al., 1993). NO· has been implicated as a promoter of HSC apoptosis (Dong et al., 2015; Jin et al., 2017; Langer et al., 2008) and a downstream effector of adiponectin (Langer et al., 2008), an adipokine with antifibrotic properties (Urtasun et al., 2008). Furthermore, NO· production by Kupffer cells (resident hepatic macrophages) and activated neutrophils can antagonize the stimulatory activity of ROS on collagen expression, potentially by acting as a free radical scavenger of ROS species (Casini et al., 1997; Friedman, 2003).

Antioxidant Responses

The endogenous antioxidant activity of HSCs is not well characterized. In vitro, TGF-β-activated HSCs have increased glutathione content (De Bleser et al., 1999), an observation that is consistent with studies in culture-activated HSCs (Proell et al., 2007). Additionally, varying degrees of HSC activation correspond to different redox states. In a study of late-stage cellular activation using the M1–4HSC cell line, which is considered an “activated HSC” based on loss of lipid droplets and expression of α-SMA, TGF-β promotes further activation, leading to a fully transdifferentiated myofibroblastoid M-HT cell (Proell et al., 2007). In culture, M1–4HSCs contain higher intracellular ROS content despite having lower levels of NOX activity compared with M-HTs (Proell et al., 2007). The fully transdifferentiated M-HTs have enhanced superoxide dismutase activity and glutathione levels (Proell et al., 2007), which may be necessary to curb ROS-induced cellular damage, while still allowing the cells to function as intracellular signaling molecules. Studies of HSCs isolated from animal models of fibrotic disease report conflicting data about the increase in HSC glutathione content with activation (De Bleser et al., 1999; Maher et al., 1997). Additionally, increased antioxidant availability attenuates the effects of profibrotic signals such as TGF-β and PDGF (Foo et al., 2011) and reduces fibrosis in animal models of disease (Jia et al., 2015; Serviddio et al., 2014).

ER Stress

Quiescent HSCs display moderately developed rough ER, which becomes enlarged and more abundant when they transdifferentiate to a MFB phenotype, consistent with increased capacity for ECM protein synthesis (Friedman, 2008; Friedman et al., 1985; Minato et al., 1983). The resultant accumulation of proteins in the ER initiates the unfolded protein response (UPR), which is an early marker of HSC activation (Mannaerts et al., 2019). Activated HSCs develop ER stress due to a combination of high-volume production of ECM proteins and exposure to ROS.

The UPR is a homeostatic response to protein folding stress that coordinates multiple functions to preserve ER function. The UPR-associated transcription factors initiate a number of ER stress-alleviating actions, including increased chaperone and foldase gene expression, activation of the autophagy pathway to direct damaged ER portions and protein aggregates for lysosomal degradation, and regulation of glucose metabolism. Patients with more severe fibrosis have higher levels of ER stress markers, such as GRP78, CHOP, and p-PKR-like endoplasmic reticulum kinase (PERK), compared with patients with milder fibrosis (Koo et al., 2016). Notably, the UPR alone does not activate qHSCs, as chemical induction of ER stress using tunicamycin is not sufficient to induce activation (Mannaerts et al., 2019). However, amelioration of ER stress through overexpression of Grp78 reduces the expression of profibrotic genes, including TGF-β, α-SMA, Col1α1, and Timp1, suggesting that the induction of ER stress through increased biosynthetic load is a driver of HSC activation (Koo et al., 2016). SMAD2, a member of the SMAD family of TGF-β signaling effectors (Breitkopf et al., 2006), provides a direct link between ER stress and HSC transcriptional reprogramming. ER stress activates PERK, which increases degradation of MIR18A, a microRNA inhibitor of SMAD2 expression (Koo et al., 2016).

ER stress in HSCs can also result from increased ROS burden, which is characteristic of aHSCs. HSC treatment with H₂O₂, as well as ethanol-feeding of rats, induce the UPR, as demonstrated by an upregulation of ER stress markers ATF4, ATF6, IRE1, and XBP1 (Hernández-Gea et al., 2013; Kim et al., 2016). The roles of IRE1 and XBP1 as mediators of autophagy are described in further detail below.

Autophagy

Macroautophagy, (“autophagy”), is a cellular stress response that engages auto-lysosomal digestion of macromolecules and organelles to generate intracellular nutrients and energy (Mehrpour et al., 2010; Mizushima et al., 2008). Several activation-related metabolic changes described above converge on autophagy, including increased lipolysis (Hernández-Gea et al., 2012; Tuohetahuntila et al., 2017), ROS production (Zhang et al., 2017), and ER stress (Hernández-Gea et al., 2013; Kim et al., 2016).

Lipid droplet mobilization, a key feature of HSC activation, is driven by autophagy, as HSCs from mice deficient in ATG7, a key autophagy mediator, fail to lose lipid droplets or activate following liver injury (Hernández-Gea et al., 2012). While early efforts examining the impact of lipid droplet mobilization focused on the release of free retinol outside the cell following intracellular retinyl ester hydrolysis (Friedman et al., 1993), it is now clear that a critical consequence of hydrolysis is the liberation of free fatty acids to “fuel” cellular activation. In support of this model, addition of endogenous oleic acid can rescue activation in autophagy-deficient HSCs, whereas inhibition of fatty acid oxidation has the opposite effect (Hernández-Gea et al., 2012). On the other hand, HSCs lacking lipid droplets in mice due to deficiency of lecithin-retinol acyltransferase (LRAT) still can fully activate but do not undergo spontaneous activation (Kluwe et al., 2011). This paradox—that lack of retinoid content due to LRAT deficiency does not alter activation, yet retinoid loss in other models accompanies HSC activation—is not easily reconciled, but one possibility is that these LRAT-deficient cells utilize an alternative energy source other than fatty acids.

Other studies have further defined cellular interactions that support autophagy. For example, PI3K-CIII, a kinase regulator of autophagy engages Rab25, a small GTPase, to direct the recognition of lipid droplets by the autophagy pathway (Zhang et al., 2017). Interestingly, Rab25 activity in HSCs is partially regulated by ROS, which activates Rab25 to increase autophagic flux (Zhang et al., 2017).

ER stress is functionally upstream of autophagy in the sequence of HSC activation (Kim et al., 2016). IRE1, which promotes HSC activation and fibrogenesis, signals through the p38-MAPK pathway to activate autophagy, as evidenced by IRE1 mediated accumulation of the autophagy marker LC3II (Hernández-Gea et al., 2013). Furthermore, the profibrogenic effects of XBP1, the downstream effector of IRE1, are abrogated through suppression of ATG7 expression, a critical autophagy mediator (Hernández-Gea et al., 2013).

NON-CELL-AUTONOMOUS EFFECTS OF HSC METABOLISM

Hepatic Stellate Cell-Hepatocyte Metabolic Crosstalk

Hepatocytes, which makes up >80% of liver cells and represent the main metabolic workhorse of the body, interact with HSC metabolism through the exchange of retinoid metabolites and signaling modalities. During HSC activation, free retinol released by HSCs (Friedman et al., 1993; Sauvant et al., 2001) is rapidly taken up in a contact-dependent fashion by hepatocytes (Ikeda et al., 1997; Sauvant et al., 2001), which release it into circulation as holo-RBP (Bellovino et al., 1999, 1996; Sauvant et al., 2001). Apart from loss of retinoid during HSC activation, it is not known if hepatic metabolism of vitamin A directly contributes to liver injury and fibrosis. The total liver content of retinoid declines with progression of human liver disease (Leo et al., 1993), and in two murine models of NASH there is depletion of retinol-RBP4 with redistribution of retinoid to hepatocytes rather than HSCs, the significance of which is not yet clear (Saeed et al., 2020).

Signaling pathways downstream of Hedgehog (Hh) and leptin receptor binding regulate glucose metabolism in HSCs (Chandrashekaran et al., 2016; Chen et al., 2012). Hh signaling has been broadly implicated in wound healing, and Hh ligands are produced by hepatocytes during liver injury (Jung et al., 2010; Wang et al., 2016). These ligands act upon neighboring HSCs to induce activation through the HIF-1α transcription factor (Chen et al., 2012), a master regulator of energy metabolism that activates a repertoire of target genes, including glucose transporters and glycolytic enzymes (Lee et al., 2004). Since hypoxia upregulates oxygen-independent ATP production pathways, HIF-1α mediated regulation of glucose metabolism upon activation is concordant with the Warburg-effect-like phenomenon observed in aHSCs. In addition, interactions between hepatocytes and HSCs contribute to steatohepatitis through neuropilin1-mediated regulation of lgfbp3 and SerpinA12 (Arab et al., 2020), as well as signaling by glutamate (Choi et al., 2019), Taz/Hedgehog (Wang et al., 2016), and Notch (Schwabe et al., 2020; Zhu et al., 2018).

While these examples of paracrine interactions between HSCs and hepatocytes are limited, they hint at a rich repertoire of bidirectional signaling yet to be uncovered, governing not only metabolism but also injury, inflammation, regeneration, and cancer. The maturation of technologies for single-cell analysis, coupled with powerful informatics to predict and validate cell-cell interactions, will yield accelerating progress.

HSCs as Effectors of Immunometabolism

While emerging concepts of immunometabolism focus primarily on immune cells (Hotamisligil, 2017; Lee et al., 2018), HSCs also contribute significantly to convergent pathways of metainflammation and tissue injury, especially in alcoholic and non-alcoholic steatohepatitis (NASH) (Schwabe et al., 2020). Through their control of retinoid metabolism, they regulate immune cells function in liver as well as in systemic immunity (Lee and Jeong, 2012; Wang et al., 2002; Weiskirchen and Tacke, 2014) (see section on Retinol Storage and Lipid Droplets, above). Stellate cells produce and respond to inflammatory mediators including cytokines and chemokines that can amplify the liver’s response to injury (Roh and Seki, 2018). Conversely, they are often overlooked determinants of the liver’s unique immune tolerance (Weiskirchen and Tacke, 2014) through induction of T regulatory cells (Heine et al., 2016; Jiang et al., 2008) and myeloid-derived suppressor cells (Heine et al., 2016). HSCs are also proposed as antigen-presenting cells (Winau et al., 2007), and they can respond to innate immune ligands through Toll-like receptor signaling and inflammasome activation (Guo et al., 2009; Miura et al., 2012; Seki et al., 2007; Watanabe et al., 2007). Interactions between HSCs and inflammatory cells can drive inflammation and injury through paracrine signaling. For example, HSC interactions with macrophages through the MER-TK receptor contribute to fibrosis in experimental NASH (Cai et al., 2020), as well as crosstalk with NK, NKT (Notas et al., 2009), and gamma delta T cells (Seo et al., 2016).

There is fragmentary knowledge about specific immunometabolites from HSCs that engage with immune cells, but several examples are notable: (1) release of all-trans RA by HSCs promotes a tolerogenic phenotype of dendritic cells through induction of arginase-1 and inducible nitric oxide synthase (Bhatt et al., 2014); (2) HSCs can expand FoxP3⁺ regulatory T cells in response to LPS by inducing the enzyme indoleamine 2,3-dioxygenase 1 (IDO1), an immunoregulatory molecule, which catabolizes tryptophan to generate kynurenine (Kumar et al., 2017). Interestingly, inhibition of IDO reduces proliferation in cultured HSCs (Oh et al., 2017); (3) asymmetric dimethylargine (ADMA) is catabolized by dimethylarginine dimethylaminohydrolase (DDAH), which reduces TGFβ1-mediated activation in primary rat HSCs (Liu et al., 2018); and (4) the bacterial product microcystin-leucine arginase (MC-LR) can activate HSCs through Hedgehog signaling, and administration of this molecule induces hepatic fibrosis in vivo, but its impact on others cell types might account in part for this effect (Gu et al., 2020)

THERAPEUTIC IMPLICATIONS AND FUTURE DIRECTIONS

Although therapies based on metabolic regulation of HSCs are limited to date, clarifying metabolic pathways provides a clear opportunity for new and potentially targeted interventions to treat both liver disease and systemic illnesses that engage the liver. While a comprehensive review of therapeutics is beyond the scope of this review, currently those drugs targeting metabolic pathways to attenuate fibrosis are being investigated in NASH, including: (1) bile acids, which have anti-apoptotic and ER stress relieving properties (Golan-Gerstl et al., 2017), (2) FXR agonists, which regulate glucose, lipid, and bile acid metabolism (Younossi et al., 2019), (3) thyroid hormone receptor-beta agonists, which moderate lipid and glucose homeostasis (Harrison et al., 2018), (4) PPAR agonists, which are adipogenic master regulators (Ratziu et al., 2016), and (5) vitamin A coupling for HSC targeting, among others (Lemoinne and Friedman, 2019; Sato et al., 2008).

A broad array of cell injury signals, including proinflammatory cytokines, apoptotic bodies, endothelial-cell-mediated growth factors, and ROS, conspire to induce HSC activation. As the key event that leads to hepatic fibrogenesis, HSC activation to amplify ECM production exacts a heavy bioenergetic toll on HSCs, demanding broad metabolic reprogramming that mirrors findings from cancer metabolism, including a Warburg-like effect, a reliance on glutaminolysis, and increased autophagic flux.

At the same time, HSCs also contribute to liver and immune homeostasis through functions that do not necessarily require their activation. This concept is reinforced by analyzing recently published single-cell RNA sequencing data to define expression of metabolism-associated genes in both murine and human HSCs from normal and diseased liver. Interestingly, expression of many of these genes is not altered with activation, suggesting that perhaps there is post-transcriptional regulation of some of the metabolic pathways they support, and/or their activity is unchanged with activation (Figure 5). In fact, there has been no systematic clarification of which metabolic functions of the cells are strictly dependent on their transdifferentiation/activation and which are associated with basal cellular homeostasis. As noted above, the original concept of HSCs as either “activated,” “quiescent,” or “inactivated” may be oversimplified, and the emergence of different subtypes of HSCs in different contexts is increasingly apparent. Single-cell-based approaches, coupled with advances in “-omic” technologies (metabolomics, proteomics, and epigenomics), could reveal greater HSC heterogeneity within specific homeostatic and disease contexts, yielding further insight into how metabolic regulation supports these varied cellular phenotypes. In leveraging these new technologies, novel paradigms of cellular energetics in fibrogenic cells, and new therapeutic vulnerabilities may emerge.

Figure 5. Metabolism-Related Gene Expression by HSCs from Healthy and Diseased Liver Based on Published Single-Cell RNA Sequencing Studies.

Human healthy and cirrhotic liver single-cell RNA sequencing datasets were extracted from Ramachandran et al. (2019) and from a study analyzing health and NASH mouse liver (Xiong et al., 2019). Data were processed using Seurat 3.0 (Butler et al., 2018; Stuart et al., 2019) using default parameters with cell clusters assigned using gene markers reported in the original publications. Mesenchymal cell clusters that likely represent HSCs are highlighted in yellow. “Percent Expression” refers to percentage of cells from each cell type that express the gene of interest. “NK,” natural killer cells; “MP,” mononuclear phagocytes; “ILC,” innate lymphoid cells; and “DC,” dendritic cells.