Cellular and Molecular Basis of Liver Regeneration

Abstract

Recent advances in genetics and genomics have reinvigorated the field of liver regeneration. It is now possible to combine lineage-tracing with genome-wide studies to genetically mark individual liver cells and their progenies and detect precise changes in their genome, transcriptome, and proteome under normal versus regenerative settings. The recent use of single-cell RNA sequencing methodologies in model organisms has, in some ways, transformed our understanding of the cellular and molecular biology of liver regeneration. Here, we review the latest strides in our knowledge of general principles that coordinate regeneration of the liver and reflect on some conflicting evidence and controversies surrounding this topic. We consider the prominent mechanisms that stimulate homeostasis-related vis-à-vis injury-driven regenerative responses, highlight the likely cellular sources/depots that reconstitute the liver following diverse types of injuries and discuss the extrinsic and intrinsic signals that direct liver cells to proliferate, de-differentiate, or trans-differentiate while the tissue recovers from acute or chronic damage.

1. INTRODUCTION

The process of regeneration is vital for the survival of multicellular organisms owing to frequent exposure to toxins, pathogens, and disease. Although some organisms exhibit excellent capabilities of regeneration—for instance, planaria can give rise to an entire individual from just a 1/300^th fraction of the whole, axolotls can regenerate complete arms within weeks after amputation, and fruit flies can reconstruct their damaged wings following fragmentation—the regenerative capacity is limited to only a few contexts in mammals ¹. Intriguingly, the regeneration ability of mammals is progressively lost during postnatal development such that most adult injuries are repaired rather than regenerated. One organ system within mammalian species that retains a remarkable potential for regeneration is the liver ². The liver’s capability to regenerate in response to diverse types of injuries has fascinated mankind for centuries dating back to the age-old myth of Prometheus, whose liver was fed on eternally by Zeus’s eagle only to regrow every day.

The liver is a multi-functional organ serving numerous physiological roles ranging from nutrient processing and storage to xenobiotic clearance and detoxification ³. It is also responsible for the synthesis and secretion of bile to emulsify dietary fat in the intestines, production of serum proteins to transport lipophilic molecules in the plasma, and generation of blood clotting factors. Hepatocytes are the major cell type within the liver parenchyma, making up to 80% of the hepatocellular mass and performing much of its energy metabolism and detoxification functions. Because most liver injuries lead to hepatocyte death, the active course of regeneration is dedicated mainly to replenishing the lost hepatocyte population. The regenerative process, however, is not just limited to hepatocytes and actually involves multiple parenchymal (cholangiocytes) and non-parenchymal (Kupffer cells, liver sinusoidal endothelial cells (LSEC) and stellate cells) cell types that are important for providing the multitude of signaling factors that initiate, propagate and terminate liver regeneration.

The exceptional capacity of the liver to regenerate is evident from the fact that within a short period after partial hepatectomy (PHx), which typically involves surgical removal of two-thirds of the liver, the tissue grows back essentially to 100% of its original size ². This ability of the liver to regain its original size—measured as liver-to-body weight ratio—is required for maintaining homeostasis and is termed as the “hepatostat” ⁴. Nonetheless, impaired regeneration in humans following certain chemical injuries, viral infections, chronic inflammation or excessive alcohol consumption can lead to fulminant hepatic failure leaving transplantation the only option for these individuals ⁵. Chronic liver disorders present an additional set of problems such that normal architecture is replaced by fibrotic tissue, inflammatory insults trigger hepatitis, biliary overload results in cholestasis, and uncontrolled cell proliferation leads to hepatocellular carcinoma. Whereas hepatitis owing to viral infections is a major cause for chronic liver damage, alcohol abuse and nonalcoholic fatty liver disease—due to metabolic dysfunctions or malnutrition—also play significant roles. It is estimated that more than a million people die from liver failure every year, representing a leading cause of global death ⁶.

Over the past several decades, using both PHx and toxin-induced injury in rodents, liver biologists have modeled the process of regeneration and studied the proliferation kinetics of various cell populations along with temporal dynamics of signaling pathways that are turned ON or OFF for precise initiation and termination of the regenerative response (reviewed elsewhere) ^2,4,7. Despite considerable progress in the identification of individual pathways and their roles, our understanding of how liver regeneration progresses and how a system-wide integration of these signals takes place is incomplete. However, recent advances in lineage-tracing techniques along with the development of genome-wide strategies that can survey exact changes in the transcriptome, translatome, and proteome at the tissue and single-cell levels have ushered in a new era for addressing these questions. In this article, we focus on the cellular and molecular mechanisms that modulate “hepatocellular plasticity” and thus stimulate productive and efficient repair process after liver damage. We also tackle broad questions relating to the process of liver regeneration: what mechanisms dictate homeostasis- versus injury-related regenerative responses? Which cell types act as a source to repopulate the liver following diverse types of injuries? What guiding signals—extrinsic and/or intrinsic—instruct the liver cells to proliferate, de-differentiate, or trans-differentiate as part of the organ’s innate response to injury? How do these signals integrate with the transcriptional, post-transcriptional, and metabolic activities of cells to produce a coherent regenerative response, i.e., maintain essential liver functions while the tissue recovers from injury? and Why is regeneration compromised in chronically diseased livers? The expectation is that understanding the cellular and molecular basis of liver regeneration will provide not only fresh insights into disorders where the liver fails to heal itself correctly but also have direct clinical implications in stimulating normal and healthy tissue repair in affected individuals.

2. MULTITUDE OF INJURIES AND THE TAILORED RESPONSE

While repopulating the liver with hepatocytes predominates the regenerative response after injury, the origin of new hepatocytes can vary according to the type and extent of injury. This is because not all hepatocytes are made equal. The variation in hepatocyte subtypes arises largely from their relative position along the porto-central axis ⁸. In brief, the hepatic parenchyma is made up of a repeating microstructure comprising a hexagonal plate of cells, wherein a central vein occupies the center of the hexagon, and at each of the vertices lies a portal triad composed of the portal vein, hepatic artery and bile duct ³ (Figure 1A). Owing to this unique vascular arrangement, hepatocytes closer to the portal triad are exposed to a vastly different microenvironment compared to those near the central vein. For instance, periportal hepatocytes receive a mixture of blood from the portal vein, which is high in nutrients and toxins, and the hepatic artery, which is high in oxygen and circulating hormones. On the contrary, pericentral hepatocytes encounter much lower oxygen, nutrient, toxin, and hormone levels. This periportal–pericentral axis is believed to generate discrete metabolic zones within the liver parenchyma wherein complementary pathways operate in a non-overlapping manner and maintain optimal metabolic homeostasis ⁹. For instance, periportal hepatocytes prefer to carry out gluconeogenesis, oxidative phosphorylation, fatty acid oxidation, ammonia detoxification, and bile formation whereas the pericentral hepatocytes are better suited for carrying out glycolysis, TCA cycle, lipogenesis, glutamine synthesis, and drug metabolism ^9–11. This is because nearly 50% of hepatic genes are differentially expressed based on the position of a hepatocyte within the porto-central axis ¹². Until lately, the mechanisms leading to metabolic zonation were poorly understood, but multiple studies have now established that Wnt/β-catenin, Ras/Mapk/Erk, and YAP/Hippo signaling pathways create morphogenic gradients in the extracellular environment that drive zone-specific expression pattern of metabolic enzymes between the portal and central regions ^13–16. Hnf4α—a major trans-activator of genes associated with liver differentiation and metabolism ^17,18—acts as a determinant for zonal expression of many genes ^19,20, functioning through a cross-talk with the Wnt/β-catenin pathway ²⁰. Additionally, Hippo/YAP/Tead axis has also been implicated in regulating hepatocyte zonation via crosstalk with Hnf4α/FoxA1/2 ¹⁶.

Figure 1. Regenerative response of the liver to diverse injuries.

(A) Schematic of the hepatic lobule within an adult liver. Hepatocyte subpopulations with regenerative roles are differentially colored based on population-specific marker genes. (B) Homeostatic regeneration takes place due to normal hepatocyte turnover with aging. Axin2⁺ cells drive regeneration by expanding towards PV, while the Mfsd2a⁺ population shrinks over time, and TERT^High and HybHPs remain constant. (C) In acute injury specifically targeting CV regions (e.g. CCl₄), TERT^High and Mfsd2a⁺ hepatocyte populations expand towards central vein. While Mfsd2a⁺ cells expand radially outward, TERT^High hepatocytes divide asymmetrically producing daughter cells with either TERT^High or TERT^Low signatures. (D) In chronic CV injury, Mfsd2a⁺ cells expand radially, with their progenies repopulating the CV regions, with significant contributions from TERT^High hepatocytes as well. (E) During portal vein specific chronic injuries, Sox9⁺ hybrid (HybHPs) and Mfsd2a⁺ hepatocyte populations are depleted, with TERT^High hepatocytes providing a major source for new hepatocytes. (F) Schematic of the liver lobule with relative placement of hepatocytes along porto-central axis based on ploidy states. Lower ploidy states (2n, 4n) dominate near PV and CV, whereas higher ploidy states are present primarily within the core of the lobule. Under regenerative conditions such as after 2/3^rd partial hepatectomy (PHx), diploid hepatocytes exhibit higher proliferation rates compared to polyploid hepatocytes. CV: Central vein, PV: portal vein & PA: portal artery.

Although the role of hepatic zonation in metabolic specialization is well appreciated, it also serves important roles in dictating the regeneration response after injury. Certainly, most hepatocytes can re-enter the cell cycle and divide upon injury, but their response is often context-specific and driven by spatial coordinates within the hepatic lobule. About three decades ago, the ‘streaming liver’ hypothesis was proposed, wherein it was postulated that new hepatocytes arise from regions around the portal triad ²¹. Based on ³H-thymidine labeling experiments following PHx, extensive signal near the portal regions was evident compared to almost negligible signal near the central vein regions ²¹. However, later cell-tracing studies showed that hepatocytes within the lobule could proliferate irrespective of their position along the porto-central axis although with a slight predominance in the portal region ^22,23. Therefore, the source and anatomical niches of new hepatocytes continue to be an area of interest. Herein we discuss some of the recently identified subpopulations of hepatocytes with their geographical location and context-specific roles in tissue homeostasis and regeneration.

2.1. Homeostatic regeneration

As in most tissues, there is a constant need for new hepatocytes to replace the old ones once they die naturally—the approximate lifespan of an uninjured hepatocyte in mice is around 200–400 days ²³. This homeostatic renewal process ensures that an uninjured liver maintains the required hepatocyte number to support normal metabolic, biosynthetic, and detoxification functions. Whereas some other mature tissues rely on a stem cell niche for homeostatic regeneration—such as the stem cell crypts in intestines—the presence of a liver stem cell has been debatable ^24,25. Maintaining appropriate liver size without a stem cell pool is all the more remarkable, as it implicates that a fully differentiated cell type must balance its needs between proliferating and sustaining metabolic homeostasis. Recent advances in lineage-tracing strategies in mice has aided the identification of cells, which in most cases are distinct hepatocyte subpopulations that reconstitute the liver under homeostatic and/or injury conditions.

During embryonic liver development, hepatoblasts serve as bipotential progenitor cells giving rise to cholangiocytes and hepatocytes and they are often characterized by the expression of the transcription factor, Sox9 (SRY-related HMG-box transcription factor 9) ²⁶. Previous studies have pointed towards the periportal region as an organizational center for the creation of new hepatocytes, since there exists a population of Sox9⁺ hepatocytes around the ductal plate. These cells have also been called hybrid hepatocytes (HybHP) because they co-express cholangiocyte-specific genes along with key hepatocyte markers such as Hnf4α, and they proliferate extensively in response to chronic hepatocyte-depleting injuries ²⁷ (Figure 1C). But, the HybHP population in an uninjured liver stays relatively stable (approximately 5%) over the lifetime of an organism, negating the hypothesis that they might also give rise to new hepatocytes for homeostatic renewal ^21,27 (Figure 1B). Interestingly, an additional population of cells was identified around the portal triad of the hepatic lobule, marked by the expression of Mfsd2a (Major facilitator super family domain containing 2a), a gene previously recognized for its role in the maintenance of blood-brain barrier ²⁸. Mfsd2a⁺ periportal hepatocytes account for nearly 40% of hepatocytes, making up 8–9 layers centered around the portal triad of a young liver (Figure 1A). But, compared to HybHP, the Mfsd2a⁺ hepatocyte numbers regress with age in a healthy liver, accounting for only 3–4 cells layers in a 9-month old mouse.

The pericentral region of a hepatic lobule is associated with high expression of glutamine synthetase (GS) and activation of Wnt pathway genes due to nearby signaling from hepatic endothelial cells ²⁹. Recently, a small population of Axin2 (a Wnt-responsive gene, which encodes for a scaffolding protein critical for β-catenin degradation) expressing cells was identified adjacent to the central vein that can expand and replace up to 40% of hepatocytes across the lobule ²⁹ (Figure 1A). In comparison to most other hepatocytes, which are polyploid, the Axin2⁺ hepatocytes are diploid and they express high levels of Tbx3, a transcription factor involved in early embryogenesis. Strikingly, these cells are reliant on their microenvironment for maintaining an actively proliferating state, especially the short-range cytokines—Wnt2 and Wnt9b—produced by endothelial cells surrounding the central vein. Because hepatocytes residing closer to a portal zone are exposed to more toxic insults than the ones near a central zone, one could imagine pericentral hepatocytes as the preferred source of cells for homeostatic self-renewal. However, another study has described that Wnt/β-Catenin signaling ligands are present throughout the hepatic lobule and that the pathway is potentiated by Leucine-rich repeat-containing GPCRs 4/5 (LGR4/5) and R-spondin ligands ³⁰. The same study also found no evidence for higher proliferation rates in pericentral hepatocytes relative to the rest of the hepatic parenchyma. Furthermore, Lgr4 expression was demonstrated as a driving force for homeostatic self-renewal, and the Lgr4⁺ hepatocytes were detected throughout the lobule without any zonal dominance ³⁰. Contradicting the supposed role of pericentral hepatocytes in regulating hepatic homeostasis, a recent study showed that Lgr5 expression is confined to hepatocytes most adjacent to the central vein ³¹. It was further revealed that these Lgr5⁺ pericentral hepatocytes maintain their own lineage, and after pulse-chase labelling over 18 months their lineage made up <5% of all hepatocytes ³¹.

More recently, an interesting new study identified a subset of hepatocytes expressing high levels of Telomerase Reverse Transcriptase (TERT^High) that contributes to the repopulation of the liver during homeostasis and after injury ³². Using lineage-tracing from the endogenous Tert locus in mice, the researchers confirmed that TERT^High hepatocytes are relatively rare (represent 3–5% of hepatocytes in a young mouse liver) and are randomly distributed throughout the liver lobule (Figure 1A). However, over time, the TERT^High cells undergo clonal expansion through a self-renewal and differentiation mechanism yielding almost 30% of hepatocytes in a one-year-old uninjured mouse liver ³². The authors proposed a model wherein, just like stem cells, the TERT^High hepatocytes self-renew by giving rise to daughter cells that can be both TERT^High and TERT^Low. Transcriptomic profiling indicated that compared to TERT^Low hepatocytes, the TERT^High hepatocytes exhibit a de-differentiated/immature cell state. Intriguingly, the fraction of TERT^High hepatocytes in the liver dropped substantially over time ³², which might in part explain why the liver loses its regenerative potential with age.

2.2. Regeneration in response to injury

Besides homeostatic regeneration, the liver is also adept in responding to acute or chronic injuries triggered by viral infections or excessive exposure to hepatotoxins, drugs (e.g., Acetaminophen) and alcohol. Normally these harmful substances are metabolized and cleared by the liver’s detoxification machinery, but when overwhelmed, it leads to widespread death of hepatocytes, either restricted to a zone or throughout the hepatic lobule. To avoid liver failure, dying hepatocytes release signals to activate mechanisms extrinsic and/or intrinsic to the tissue, which promote hepatocyte proliferation and differentiation to recover the functional parenchyma. Herein we discuss the different modes of liver regeneration based on the injury context and highlight recent studies that have identified specific cell populations involved in this process.

2.2.1. Zonally-restricted injuries

Owing to strong pericentral expression and activity of many P450 enzymes, metabolism of certain toxins or xenobiotics occurs exclusively within zone 3 hepatocytes of the liver lobule ³³. Consequently, such toxic insults lead to hepatocyte injury and death in a zonally-restricted fashion. For instance, repeated exposure to carbon tetrachloride (CCl₄) or high doses of acetaminophen specifically ablates the pericentral population of hepatocytes, including the self-renewing Axin2⁺ cells ^28,34. Therefore, in these cases, separate source(s) of hepatocytes are required to replenish the lost population of hepatocytes (Figure 1C, D). Interestingly, while the HybHPs residing around the ductal plate, contribute less than 10% of new hepatocytes following a single exposure to CCl₄, they generate over 60% of hepatocytes in response to chronic CCl₄ treatment ²⁷. It is particularly noteworthy that the HybHP-derived hepatocytes not only repopulate the area around the central vein but they also eventually gain pericentral-specific gene expression profile, thereby re-establishing the pericentral niche. Independently, the Mfsd2a⁺ cell population was shown to expand and replace the majority of dying hepatocytes following repeated CCl₄ dosing ²⁸ (Figure 1D). However, this may not be entirely surprising as HybHPs likely represent a subset of Mfsd2a⁺ hepatocytes. While most studies have implicated periportal hepatocytes to often be involved in regenerating pericentral zone, a recent report using lineage tracing indicates that post-CCl₄ peri-injury midlobular hepatocytes replenish pericentral Axin2+ hepatic pools³⁵. Likewise, there are other de-differentiated cell populations including the Hnf4α⁺, Hnf1β⁺, Ck19⁻ hepatocytes in human liver, as well as the Hnf4α⁺, Afp⁺, EpCAM⁺ hepatocytes in murine livers that reside near a portal triad and can give rise to new hepatocytes in response to chronic liver damage ^36,37. The relationship and co-identity of these cell types vis-à-vis Axin2⁺ or Mfsd2a⁺ hepatocytes in the periportal niche are unknown.

While most xenobiotics injure hepatocytes around the pericentral area, certain chemicals such as allyl alcohol act as potent periportal necrogens ³⁸. Similarly, 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) administration or bile duct ligation in mice induces cholestatic injury and death of periportal hepatocytes ^39,40, including the HybHPs and Mfsd2a⁺ cells ^27,28 (Figure 1E). In response to these periportal hepatocyte-depleting injuries, proliferating hepatocytes are found primarily within the mid-lobular or pericentral regions likely through clonal expansion of TERT^High cells in these areas. Thus, it appears that when liver damage is restricted to a particular zone, distinct pool(s) of hepatocytes from other regions can expand to replace the damaged area and thereby restore liver functionality.

2.2.2. Regeneration after partial hepatectomy

Due to a shortage of donor livers, split liver transplantations are now increasingly common; however, the procedure’s success relies on the ability of donor grafts to regenerate and reach an appropriate liver mass in the recipient. Accordingly, animal models in which parts of the liver are surgically removed have gained direct clinical relevance as they can provide a mechanistic understanding of the regenerative process required for a successful liver transplant. After PHx, the resected liver lobes cannot be reconstructed; consequently, hepatocytes in the remaining lobes undergo compensatory hypertrophy (enlargement of hepatocytes) followed by hyperplasia (proliferation of hepatocytes) ⁴¹. The stromal cells also proliferate, but after the regenerating hepatocytes produce growth factors and mitogens to stimulate their cell division.

Remarkably, the extent of surgical resection determines the liver’s growth response wherein hepatocytes respond to 1/3^rd PHx only through hypertrophy, whereas a 2/3^rd PHx leads to both hepatocyte hypertrophy and proliferation ⁴². A simplified view of liver regeneration postulates that all hepatocytes proliferate at least once after 2/3^rd PHx, but, lineage-tracing studies contradict this view ⁴². Intriguingly, although most hepatocytes enter cell cycle after 2/3^rd PHx, many do not undergo a cell division ⁴². Furthermore, the rate of hepatocyte proliferation differs along the pericentral-periportal axis. For instance, the Mfsd2a⁺ periportal hepatocyte population expands much rapidly compared to pericentral hepatocytes after PHx; however, this difference is not as robust in response to chronic CCl₄ treatment ²⁸. Also, it is noteworthy that while specific hepatocyte populations proliferate in response to zonally-restricted injuries, newly generated hepatocytes after PHx originate from all parts of the lobule. These findings suggest the high likelihood that majority of hepatocytes have the potential to proliferate, albeit at different rates based on the factors such as spatially defined morphogens, or the ploidy status of hepatocytes. A recent study examined the differential ability of diploid vs. polyploid cells in repopulating the liver during liver regeneration ⁴³. While both diploid and polyploid hepatocytes could repopulate the liver, diploid hepatocytes were found to proliferate faster, and this difference was independent of their response to mitogenic signals ⁴³.

The liver attains a high percentage of polyploid hepatocytes during postnatal period of development such that over 70% of murine and 40% of human hepatocytes acquire greater than two homologous sets of chromosomes by adulthood ^44,45. Hepatic ploidy content is based both on the number of nuclei per cell (cellular ploidy) and the DNA content per nuclei (molecular ploidy). The primary mechanism for an increase in hepatocyte ploidy after birth is cytokinesis failure that produces tetraploid and octoploid hepatocytes with one or two nuclei ^46,47. Ploidy in the adult liver is dynamic, and hepatocytes can even reduce their ploidy via reductive mitosis, a process called ploidy reversal ⁴⁸. Interestingly, there is a pronounced increase in mono-nucleate polyploid hepatocytes in regenerating livers after surgery ⁴² (Figure 1F). A recent report suggested that the polyploid state is responsible for giving rise to aneuploid hepatocytes and that this is relevant in the context of chronic liver injury ⁴⁹. Interestingly, a single molecule imaging study found that cells with higher ploidy (4n & 8n) are found more often within the mid-lobular region whereas the majority of diploid (2n) cells are found immediately around the periportal and pericentral regions ⁵⁰. This distribution might explain why periportal and pericentral regions are the likely sites of hepatocyte repopulation, as they harbor diploid cells that proliferate faster compared to more polyploid, mid-lobular hepatocytes ⁴³. As expected, the pericentral Axin2⁺ cells were found to be largely diploid (~60%) ²⁹, but surprisingly, the mid-lobular TERT^High cells were polyploid ⁵¹. Therefore, while the relative position of these cells in the hepatic lobule agrees with their ploidy content, the capability to proliferate seems discordant. Future studies exploring the interplay between ploidy, hepatic zonation and hepatocyte subtypes are necessary to address these conflicting observations.

2.2.3. A ductular source for hepatocytes

Until this point, we discussed the roles of hepatocytes in liver regeneration, but, in cases of severe and prolonged liver injury, hepatocyte proliferation can be irreversibly blocked due to replicative senescence. Under such circumstances, it has been postulated that a reserve of bipotential liver progenitors—also referred to as oval cells ⁵²—expand through ductular reaction (DR), and then differentiate into hepatocytes or cholangiocytes to repopulate the liver ^53,24,54. These hepatobiliary progenitors are thought to reside near the canals of Hering, a structure that establishes the transition between hepatocytes and bile ductules. Tracking the origin of liver progenitor cells has been difficult owing to a lack of reliable markers that distinguish progenitors from de-differentiated cells. Nonetheless, the progenitor-like cells can emerge through DR when rodents are fed DDC-containing or choline-deficient, ethionine-supplemented (CDE) diets, which are known to induce rapid toxicity and turnover of hepatocytes ^{55–57,58,59}. DR is crucial for regeneration following biliary injury, and if inhibited leads to an increase in hepatic necrosis ⁶⁰, but the question of whether DR is a principal source of generating new hepatocytes is still controversial.

Recent cell-fate and lineage-tracing efforts did not find much evidence supporting the presence of hepatobiliary progenitors in healthy mouse livers and determined that, in response to both acute and chronic injury, essentially all new hepatocytes are generated through self-duplication ^27,61–66. Additional studies using Sox9 or Hnf1-β as lineage-tracing markers also reported insignificant roles for bipotential ductal progenitors in producing hepatocytes in multiple oval cell injury models ^67,68. On the contrary, evidence for ductular cells differentiating into hepatocytes came from an in vitro study wherein after CCl₄-triggered injury of pericentral hepatocytes, the Wnt pathway genes were activated in ductal cells around the periportal area ⁶⁹. These cells co-expressed Lgr5, a Wnt target gene associated with pericentral hepatocytes, and Sox9, a ductal progenitor marker. The FACS sorted Lgr5⁺ cells were successfully cultured into organoids, and could be differentiated into hepatocytes upon Notch and TGF-β signaling inhibition ⁶⁹. Another study however found Lgr5⁺ cells in the centrilobular but not periportal regions of uninjured livers. Differences in observations were accounted to genetic backgrounds and methodology ³⁰. Surprisingly, a recent study found that expansion of biliary epithelial cells (BECs) in DR is independent of LGR4/5-signaling, and instead relies on YAP1 and mTORC1 ⁷⁰.

Interestingly, the forced blockade of hepatocyte proliferation—through β1-Integrin ablation or p21 overexpression within pre-existing hepatocytes—was shown to generate a substantial number of new hepatocytes from BECs even after short-term damage ⁷¹. But, in this case, inhibiting the proliferation of pre-existing hepatocytes could have exaggerated the emergence of hepatocytes from BECs. Recently, Deng et al. used mTmG dual-fluorescent CRE reporter mice ⁷² to trace the contribution of hepatocytes and non-hepatocytes in repopulating the liver under short-term and long-term injury conditions ⁷³. Using cytokeratin 19 (CK19) as a tracing marker, it was demonstrated that BECs are an important source of new hepatocytes when severe liver damage is induced for a prolonged period of time. By feeding thioacetamide (TAA) and/or DDC-containing diets to mTmG mice, the authors confirmed that liver regeneration after short-term injury (TAA:12 weeks & DDC:4 weeks) depends entirely on hepatocyte self-duplication. However, in response to severe injury for a long-term (TAA & DDC >24 weeks), a significant fraction (~25%) of new hepatocytes are derived from the non-parenchymal cells (NPCs). Among NPCs, the BECs contribute ~10% of new hepatocytes after chronic injury in the absence of any additional genetic perturbations ⁷³. This study thus highlights the potential of DR cells to differentiate into hepatocytes after prolonged liver injury, and in part, clarifies some of the conflicting evidence in the literature in tracing the origin of newly regenerated hepatocytes ⁶⁶. Overall, with new studies demonstrating the exceptional degree of hepatocellular plasticity and the varied sources of regenerating cells that are specific to the type and extent of injury, the cell-of-origin debate is mostly settled. Although, it remains to be seen which underlying pathways make these hepatocellular subtypes differ in their response and if rules of engagement differ or remain the same. In this regard, the recent development of single-cell sequencing methods has opened a path to the discovery of cellular heterogeneity and dynamics at unprecedented resolution (Box 1).

Box: Applications of sssingle-cell RNA sequencing in liver regeneration & homeostasis.

A major drawback of bulk transcriptomic measurements is the lack of information on cellular heterogeneity. However, the latest advances in single-cell RNA sequencing and related computational methods now allow the characterization of any given cell population within a complex tissue under normal, disease, and regenerating conditions ¹⁸⁸. Furthermore, single-cell studies can be employed on MACS/FACS sorted cell populations to characterize the heterogeneity among similar cell types (Panel A). It is therefore not surprising that these new methodologies have taken the field of liver regeneration by a storm providing answers to long-standing questions such as which cell types or even subtypes repopulate the liver in response to various injuries. For instance, it was recently reported that within the larger pool of EpCAM⁺ biliary epithelial cells (BECs), a subset of BECs expresses YAP-driven gene programs, and these subsets are the ones that respond to the periportal injury signals ^70,189. Studies employing single-cell methods on human liver cell suspensions have also identified distinct intrahepatic macrophage populations and their subtypes with inflammatory and non-inflammatory phenotypes ¹⁸⁶. Likewise, new emergent subpopulations from diseased and regenerating livers are now being detected, such as the recent identification of NASH-associated macrophages ¹⁹⁰.

In addition to the intrinsic transcriptome differences between individual cells, the spatially-restricted factors can also affect their gene expression, and this is particularly pertinent in the hepatic lobule where the morphogen, nutrient, and oxygen gradients produce discrete metabolic zones (Panel B). Although the zonally-restricted expression of some hepatic genes such as Glutamine Synthetase (GLUL) and Phospho-enol Pyruvate Carboxykinase (Pck1) was known for a long time, recent single-cell studies have revealed that over 50% of hepatocyte genes follow a zonated expression pattern, and intriguingly this extends to other cell types in the mouse liver such as endothelial and stellate cells ^12,191,192. The zonated expression patterns can become important in specific disease contexts — for example a new single-cell study discovered that among the various hepatic stellate cell populations, the central zone restricted stellate cells are the primary drivers of pathogenic collagen deposition during liver fibrosis ¹⁹².

A number of laboratories in the past decade identified the hepatocellular sources that are important for liver homeostasis and regeneration ^29,32,73,193. While these studies helped to settle the debate on the cell-of-origin, the information about gene expression dynamics and differences in behavior for most cell types has remained a mystery. Fortunately, single-cell analysis can now be employed for identifying the cellular trajectories and gene expression dynamics of individual cell populations (Panel C) ^194,195. Exciting new methods are being developed that can estimate the rate of change among cells along a given trajectory. For instance, RNA velocity can measure the time derivative of the gene expression state by directly distinguishing between the unspliced and spliced mRNAs, producing a high-dimensional vector that predicts the future state of individual cells on a timescale of hours ¹⁹⁶. This type of analysis was recently applied to study differences between young and aged muscle stem cells, wherein it was found that although the cellular trajectories of young and old stem cells are identical, younger stem cells progress through the trajectory more swiftly than the aged cells ¹⁹⁷. A similar analysis of hepatic progenitors may help to identify differences in cellular dynamics after liver injuries while uncovering the gene regulatory networks that drive such dynamics (Panel C) ¹⁹⁸. Lastly, the single-cell methods can also address the cell-cell communication questions during different phases of liver regeneration, as was recently reported after CCl₄-mediated liver injury, wherein endothelial cells were identified as major drivers of Wnt-β-Catenin signaling for hepatocytes ³⁵.

Box Figure.

(A) Simplified pipeline for identification of new cell types from whole liver suspension as well as analysis of cellular heterogeneity and subtypes after cell sorting using FACS/MACS methods. (B) Isolation of individual cell types in the liver, and ordering of cells using pericentral and periportal dominant genes, to identify zonation patterns at a genome-wide scale. (C) Isolation of specific liver cell subtypes based on marker genes, ploidy status, and pre- or post-liver injury state to identify cellular dynamics with RNA velocity and/or trajectory analysis. Cell-specific trajectories can further resolve the differential transcriptome dynamics between various hepatocyte subtypes.

3. GUIDING SIGNALS, AND NEIGHBORING WHISPERS

An important facet of liver regeneration research has been the persistent search of extracellular signals that initiate, propagate, and terminate the regenerative process, i.e., control the exact timing of cell cycle progression of hepatocytes. Whereas the early investigations aimed to identify one single defining signal that drives hepatocyte proliferation, it is now accepted that this hypothesis is unlikely. In fact, multiple studies have repeatedly discovered that loss of any given signaling event often leads to a delay in regeneration rather than a complete failure ⁷⁴. Regardless of the extensive redundancy in regenerative mechanisms, identification of individual factors remains important as they likely act in parallel to promote maximal liver fitness. The factors guiding liver regeneration typically fall into two categories—extrinsic or intrinsic—and are a mixture of growth factors, cytokines, and metabolites.

3.1. Circulating factors – extrinsic to the liver

PHx is usually the preferred model to study liver regeneration due to its ability to induce a swift regenerative response without the inflammatory complications associated with toxin-mediated injuries ². Immediately after PHx, the first physical perturbation to the remnant lobes is the hemodynamic change. The portal vein and the hepatic artery supply blood to the liver, but after PHx, the total blood supply to the remaining liver stays unchanged. Therefore, theoretically after 70% PHx, the amount of blood flowing through the residual lobes should increase by three-fold. This surge in blood flow to the liver can induce cell proliferation via both mechanical stimuli from the increased flow and enhanced exposure to the signaling factors carried by the blood (Figure 2A). Indeed, portal vein pressure doubles after 70% PHx in mice ⁷⁵ and a prospective study of patients undergoing donor hepatectomy showed significant improvement in liver regeneration when immediate post-operative portal vein pressure was higher ⁷⁶.

Figure 2. Extrinsic and intrinsic signals regulating the hepatic regenerative response.

(A) Schematic of circulating factors in the hepatic lobule before and after 2/3^rd PHx. The elongated blue cells are endothelial cells, while the brown cells are hepatocytes. Post-PHx there is a decrease in circulating glucose levels, with increase in levels of bile acids, norepinephrine and multiple growth factors. Blood insulin and EGF levels increase after PHx due to secretions by pancreas and Brunner’s glands respectively. Additionally, increase in the rate of blood flow leads to turbulent flow in the hepatic sinusoids. (B) Cell-cell interactions in the form of paracrine and autocrine signals released by the resident liver cells. Non-parenchymal cells are exposed to circulating factors as well as increased blood flow rates leading to release of cytokines after PHx. TNFα and IL-6 released by Kupffer cells primes hepatocytes from G0 to G1 phase. Other growth factors including TGFα, HGF, and EGF are secreted by other resident cells to initiate hepatocyte proliferation. These cytokines prime the hepatocyte which then releases cytokines that serve both paracrine (VEGF, TGFα, PDGF) and autocrine (TGFα, amphiregulin) functions.

After PHx, an instantaneous rise in pressure and volume of blood flowing through the hepatic vasculature triggers turbulence exerting mechanical stress onto endothelial cells. It is known that certain extracellular matrix (ECM) remodeling events like the release of urokinase plasminogen activator (uPA) are initiated within the first 15 minutes after PHx ^77,78, and that increased mechanical stress on endothelial cells can directly stimulate uPA activity ⁷⁹. Additionally, it has been proposed that dilation of the liver vasculature after PHx causes stretching of sinusoidal endothelial cells leading to the release of growth factors, needed for initiating and sustaining liver regeneration ⁸⁰. Recently, Lorenz et al. demonstrated that the rate of liver growth during development correlates with blood perfusion to the organ and that mechanosensing by vascular endothelial cells stimulates them to release angiocrine signals, which induces proliferation of hepatocytes ⁸¹. By altering blood flow through the liver, it was discovered that vessel perfusion evokes β1 integrin and vascular endothelial growth factor receptor 3 (Vegfr3) activities in endothelial cells both of which are crucial for the production and release of hepatocyte growth factor (HGF), a key signal for the growth and survival of liver cells.

In addition to increased mechanosensing, greater exposure to circulating factors after PHx is also crucial for liver regeneration (Figure 2A). This was initially recognized in parabiosis experiments, wherein if carotid-to-jugular cross circulation was established between partially hepatectomized and normal rats, significant hepatocellular proliferation could be induced even in non-hepatectomized rats ⁸². The results indicated that factors were circulating from the hepatectomized rat to the non-hepatectomized rat and thereby stimulating a regenerative response. Later, HGF, epidermal growth factor (EGF), thyroid hormone, norepinephrine, and various cytokines were identified as humoral factors that direct liver regeneration. It is now well recognized that increased cytokine and norepinephrine levels in peripheral blood exert ‘auxiliary’ mitogenic effects on hepatocytes after PHx ^83,84, and that the factors released by other organs can collaborate to further enhance these effects. Most notably, Brunner’s glands in the duodenum releases EGF that is sequestered in the liver, and after PHx, circulating norepinephrine augments this secretion ².

As expected, after PHx, there is an instant change in the concentration of metabolites processed by the liver due to a decrease in liver-to-body-weight ratio. Therefore, metabolic dysregulation driven by hepatic insufficiency is considered one of the earliest signal that initiates the regenerative response ⁸⁵. For instance, owing to reduced glycogenolysis and gluconeogenesis, hypoglycemia sets in within hours of PHx, whereas dextrose administration to restore glycemic levels inhibits liver regeneration ^86,87. PHx-associated hypoglycemia leads to additional catabolic changes such as increased lipolysis, and interestingly, postponing the hypoglycemic response to PHx via dextrose infusion suppresses the increase in serum-free fatty acids and delays regeneration ⁸⁷. Changes in nutrients or serum metabolites are known to activate proliferative signals in hepatocytes including the accumulation of cyclin D1 and degradation of p21 via E3 ubiquitin ligase pathway; however, these responses to PHx are disrupted by dextrose infusion ⁸⁷.

Besides glucose, changes in serum bile acid concentrations are also important for liver regeneration. Circulating bile acid levels are rapidly elevated in response to PHx, and liver regeneration is compromised in rats in the absence of intestinal bile ⁸⁸. Also, supplementing bile acids at low doses stimulates hepatocellular mitoses and hepatomegaly after PHx, whereas lowering bile acid levels with a bile acid-binding resin delays liver regeneration ^89,90. Notably, in the clinic, external biliary drainage in patients following hepatectomy not only reduces serum bile acid levels but is also associated with worse liver regeneration ⁹¹. Bile acids normally signal through farnesoid X-activated nuclear receptor (FXR), and genetic ablation of FXR in mice impairs regeneration following PHx or CCl₄-induced liver injury ^89,90,92. Interestingly, global loss of FXR affects post-PHx recovery more severely in comparison to a hepatocyte-specific loss of FXR ⁹³. This is attributed to the entero-hepatic-FXR axis because delayed regeneration due to loss of intestinal FXR can be rescued by ectopic supplementation of fibroblast growth factor 15 (Fgf15), a signaling molecule that serves pleiotropic roles in bile acid homeostasis while fostering mitogenic stimulation of liver parenchyma through the entero-hepatic-FXR axis ^94–96. The effects of bile acids on liver regeneration are also influenced by gut microbiota, which can alter the composition and conjugation status of primary and secondary bile acids ⁹⁷.

3.2. Cytokines and growth factors – intrinsic to the liver

To proliferate after injury, quiescent hepatocytes have to re-enter and progress through the cell cycle (G1 to M phase), and eventually exit into the G0 phase once regeneration is complete. Intrinsic factors supporting this sequence of events are typically categorized into three stages: (i) the initiation or priming stage—commits cells into a state of replicative competence, (ii) the progression stage—expands the target cell population, and (iii) the termination stage—suppresses any further cell proliferation ^98,99. Among the hepatocyte priming factors (Figure 2B), the pro-inflammatory cytokines—tumor necrosis factor-α (TNF-α) and interleukin-6 (IL-6)—have received the most attention as loss of either factor impairs liver regeneration ^84,100,101. Kupffer cells (resident macrophages of the liver), the major source of TNF-α & IL-6, are stimulated by both C3 and C5 components of the complement system as well as by toll-like receptor 4 (TLR4) signaling via activation of the NF-κB signaling pathway ¹⁰². Interestingly, in uninjured conditions, an exposure to direct mitogens (e.g., TCPOBOP) can induce hepatocyte proliferation independent of TNF-α & IL-6, however, for compensatory proliferation after injury these factors become important for the transition of hepatocytes from G0 to G1 phase ¹⁰³.

After priming, the progression of hepatocytes through G1 and their subsequent replicative cycling involves additional signaling factors that can be described as mitogens, either complete or auxiliary. Complete mitogens refer to cytokines that can induce hepatocyte proliferation in vivo as well as in vitro. Auxiliary mitogens, however, cannot induce proliferation by themselves, but, they can promote and accelerate cell cycle entry. HGF was one of the first complete mitogens identified to promote liver regeneration, and it acts as a paracrine factor to induce hepatocyte proliferation by activating a tyrosine kinase signaling cascade ^104,105. HGF in a healthy liver is bound to the ECM in its inactive form, but immediately after PHx, uPA releases HGF from the matrix. Additional HGF is produced by hepatic stellate and endothelial cells after PHx, and both hepatocytes and cholangiocytes respond to this newly released HGF via the cell surface receptor C-Met leading to activation of cyclins and cyclin-dependent kinases (CDKs) that initiate DNA synthesis. Another set of complete mitogens pivotal to liver regeneration are the family of EGF receptor (EGFR) ligands, including EGF, heparin-binding (HB)-EGF and transforming growth factor α (TGFα) (Figure 2B). As discussed earlier, EGF is primarily secreted from the Brunner’s gland whereas HB-EGF is secreted by Kupffer and endothelial cells in the liver. TGFα is mainly produced by hepatocytes and can act in autocrine or paracrine fashion, though interestingly, TGFα KO mice exhibit no defects in liver regeneration ¹⁰⁶. The significance of HGF and EGFR ligands in stimulating liver regeneration is underscored by studies wherein individual loss of C-Met or EGFR receptors delays repair after injury ^107,108, but remarkably, combinatorial loss of both receptors leads to complete regeneration block and mortality within 2–3 weeks ¹⁰⁹.

Once the liver lobule is sufficiently reconstituted, proliferation must cease. The decision to stop is coordinated by a restoration of normal levels of circulating factors, reorganization of the ECM and increased expression/activity of terminating cytokines ¹¹⁰. The most well studied terminating factor for liver regeneration is TGF-β, an auto- and paracrine-acting cytokine released by hepatocytes, stellate, endothelial and Kupffer cells, that inhibits hepatocyte proliferation in vitro and in vivo ^111–116. TGF-β impedes cell proliferation by activating a cyclin-dependent protein kinase inhibitor, which blocks the function or production of CDKs and cyclins ¹¹⁷. Although TGF-β mRNA levels increase initially after PHx, the hepatocytes develop a transient resistance against TGF-β by downregulating TGF-β receptors and upregulating TGF-β inhibitory proteins SnoN and Ski ^111,118,119. Constraining TGF-β pathway at the priming stage thus allows hepatocytes to re-enter the cell cycle, but once DNA synthesis is complete, TGF-β sensitivity is re-established to limit further hepatocyte proliferation ^110,120. Another member of the TGF-β superfamily, activin A, is similarly elevated following PHx, and just like TGF-β, it hinders hepatocyte proliferation to terminate liver regeneration ^121,122. Regardless of the major role for TGF-β in promoting cell cycle exit, it is still not essential for termination of liver regeneration ¹²³; however, TGF-β signaling was recently shown to drive trans-differentiation of hepatocytes into cholangiocytes and form functional bile ducts in a mouse model of human Alagille syndrome ¹²⁴.

Other signals that influence termination of liver regeneration include Integrin-linked kinase (ILK) ¹²⁵ and Glypican-3 ¹²⁶. ILK belongs to a subfamily of Raf-like kinases and associates with β1 integrin under the plasma membrane to transmit ECM signals that inhibit hepatocyte proliferation. Hepatocyte-specific deletion of ILK in mice triggers a termination defect after PHx, resulting in a liver that is 58% larger than the actual pre-PHx mass ¹²⁵. This failure to terminate regeneration after PHx is due to continued cell proliferation driven in part by increased signaling through HGF/Met, β-catenin, and Hippo-YAP pathways. Glypican 3, on the other hand, is anti-proliferative heparan sulfate proteoglycan that is bound to the cell surface through a glycosylphosphatidylinositol anchor. Overexpression of Glypican-3 in the mouse liver blocks hepatocyte proliferation resulting in a significantly smaller liver at the end of regeneration partly through downregulation of cell cycle-related proteins and upregulation of growth arrest genes ¹²⁶. Thus, a variety of ECM-related intrinsic factors positively and negatively impact hepatocyte proliferation, and dynamic remodeling of ECM instructs these factors to synchronize the timely initiation, progression, and termination of liver regeneration. Besides cytokines and ECM factors, a recent study reported the requirement for Hnf4α, for the termination of regeneration after PHx ¹²⁷. Hnf4α levels decline rapidly following PHx, and this initial decrease followed by re-expression is required to enter and exit the cell cycle in a timely manner and resume normal function following regeneration ¹²⁷.

Recently, Caldez et al. used a combination of transcriptomics, metabolomics, and intravital imaging approaches to study the metabolic adaptation of regenerating hepatocytes and investigated how cell division and metabolism intersect to support liver regeneration ¹²⁸. Interestingly, hepatocyte proliferation after PHx in mice was accompanied by a sustained increase in redox equivalents, particularly NADH levels, indicating an increase in mitochondrial oxidation capacity during liver regeneration. However, when hepatocyte proliferation after PHx was blocked via liver-specific deletion of Cdk1, the remnant liver tissue exhibited compensatory cellular hypertrophy, reduced NADH levels, and increased flux through alanine aminotransferase producing higher levels of alanine and α-ketoglutarate ¹²⁸. Thus, to restore appropriate liver mass and functionality, proliferation-deficient hepatocytes can alter their metabolism to support compensatory growth via cellular hypertrophy.

4. INNER DECISIONS

After hepatocytes sense the injury-induced extracellular signals, a series of intracellular events are triggered that culminate into changes in the expression and activity of transcription and RNA processing factors, leading to well-orchestrated gene expression responses. These transcriptional and post-transcriptional responses reprogram hepatocytes to switch from quiescent to actively proliferating state.

4.1. Transcriptional response to cell signals

As discussed earlier, TNFα and IL-6 are the priming cytokines released by the stromal cells, and their effects on hepatocytes begin immediately after PHx. Interaction of IL-6 with the cell surface IL-6 receptor (IL-6R)/gp80 on hepatocytes triggers intracellular dimerization of gp130 proteins, thereby activating receptor-bound Janus kinases (JAKs), which phosphorylate multiple tyrosines on gp130 ¹²⁹. These phosphorylated tyrosines are essential for nuclear localization of the STAT transcription factors as well as activation of the Ras/MAPK/Erk pathway, both of which induce hepatocyte cell cycle entry by stimulating gene expression or activity of cell-cycle regulators ^84,130 (Figure 3). TNFα binding to its receptor TNFR1, on the other hand, leads to the recruitment of IKK kinase to the TNFR1 complex, where it becomes activated and subsequently initiates phosphorylation, polyubiquitination and proteasomal degradation of I-κB, an inhibitory partner of NF-κB ^131–133. I-κB degradation results in the release and translocation of NF-κB to the nucleus, where it stimulates transcription of target genes, including G1-phase cyclins that directly promote cell proliferation ¹³⁴ (Figure 3).

Figure 3. Signaling pathways coordinating the hepatic regenerative response.

Multiple signaling pathways are stimulated via circulatory factors as well as intrinsic (paracrine and autocrine) cytokines during liver regeneration. After an injury, Kupffer cells are stimulated by circulating LPS as well as the complement system. Kupffer cells play a major role in initiating hepatocyte transition from G0 to G1 phase via induction and release of TNF-α and IL-6. These factors eventually initiate multiple pathways within hepatocytes, including NF-κB, JAK-STAT, PI₃K-Akt, and AP1 signaling leading to transcriptional activation. Additionally, Kupffer cells also release the activated Wnt ligand that initiates Wnt/β-Catenin signaling. Hepatocyte growth factor receptor signaling is initiated via both HGF released from ECM by urokinase plasminogen activator (uPA), as well as secreted HGF from hepatic stellate and endothelial cells. EGF receptor signaling is initiated by EGF primarily derived from the Brunner’s gland. EGFR and HGFR signaling feed into the NF-κB and mTOR pathways via PI₃K-Akt signaling which can be regulated by miR-382. Sonic hedgehog (Shh) signaling is initiated by Shh secreted from Stellate cells, which in turn inactivates the inhibition of Smo receptor by the Ptch1 receptor. Activation of Smo leads to increased accumulation and localization of Gli to the nucleus and transcriptional response. Also, during regeneration, YAP predominantly localizes to the nucleus interacting with TEAD and β-Catenin proteins to activate transcription of target genes. In the regenerating phase, ESRP2 inhibition leads to depletion of the adult and accumulation of fetal splice isoforms of core Hippo pathway components (e.g., Yap1, Tead1, Nf2, Csnk1d), which suppress Hippo signaling, while promoting nuclear YAP-TEAD interactions and transcriptional activities.

Additionally, cross-talk between Hippo-YAP, Sonic-Hedgehog, and Wnt/β-Catenin signaling pathways is critical for proper liver regeneration ^135–137. Immediately after PHx, β-catenin and YAP proteins translocate to the hepatocyte nucleus, where they collaborate with LEF/TCF and TEAD family of transcription factors respectively, and turn on the transcription of pro-proliferative genes ^138,139,140 (Figure 3). This is in part driven by increased Wnt binding to the Lrp5/6 and Frizzled receptors, which recruit β-catenin degradation complex to the membrane, decrease activation of Hippo core kinases MST1/2, and consequently, spare β-catenin and YAP degradation so they both translocate to the nucleus ^138,141. Intriguingly, during the progression stage of liver regeneration, HGF binding to the c-Met receptor also promotes nuclear translocation of β-catenin in hepatocytes, which increases EGFR transcription and thus extends the mitogenic effects of growth factors ¹⁴². Conversely, Ring finger 43 (RNF43) protein induces degradation of Wnt receptors Lrp/Frizzled; however, when R-spondin binds to LGR4/5/6, it sequesters RNF43, which potentiates the Wnt/β-catenin pathway ³⁰ (Figure 3). Fittingly, R-spondin 1 injection accelerates liver regeneration by promoting hepatocyte proliferation, whereas LGR4/5 deletion hinders hepatocyte proliferation, impairing liver growth and regeneration after PHx ³⁰.

The Hippo-YAP signaling pathway supports the differentiated state of hepatocytes and is central for cell fate determination in healthy and regenerating livers ¹⁴³. The mammalian homologs of Hippo—MST1/2 (serine/threonine kinases)—phosphorylate and activate LATS1/2 kinases, which in turn, phosphorylate YAP resulting in its cytoplasmic sequestration and proteasomal degradation. In healthy adult hepatocytes, YAP is almost undetectable, but after injury, there is a rapid decrease in YAP phosphorylation due to inactivation of MST1/2, which results in its increased YAP abundance and nuclear localization ¹⁴⁰. Once YAP accumulates in the nucleus, it partners with TEAD proteins to activate transcription of pro-proliferative genes, including many cyclins ¹⁴⁴ (Figure 3). Once the liver size is restored, MST1/2 kinases are reactivated, YAP gets re-phosphorylated, and its nuclear levels decline, which silences the pro-proliferative gene expression. Accordingly, absence of YAP activity after PHx leads to inefficient regeneration owing to reduced hepatocyte proliferation, whereas YAP overexpression promotes hepatocyte proliferation and causes liver overgrowth ¹⁴⁰. Interestingly, YAP/TAZ proteins were recently discovered as integral parts of the β-catenin degradation complex, which was proposed to act as a sink for these proteins ¹⁴⁵.

Liver regeneration is also directed by the sonic hedgehog (Shh) pathway ¹⁴⁶. After PHx, Shh is released by the hepatic stellate cells, and it binds to the Patched (Ptch) receptor on stellate cells as well as nearby hepatocytes, relieving its inhibitory role on smoothened (Smo) receptor. Liberated Smo then inhibits factors that normally phosphorylate and degrade glioma-associated oncogene (Gli) transcription factors in the cytoplasm (Figure 3). Accumulation and nuclear translocation of Gli factors, in turn, induces transcription of target genes that coordinate the survival and proliferation of hepatocytes ¹⁴⁷. Thus, both Shh and YAP activities increase within regenerating hepatocytes and blocking hedgehog signaling by hepatic stellate cells reduces nuclear YAP accumulation in neighboring hepatocytes and therefore inhibits their proliferation ¹⁴⁸. Further studies should resolve how Hippo-YAP, Sonic-Hedgehog, and Wnt/β-Catenin pathways cooperate among regenerating hepatocytes and stromal cells, especially during chronic liver injuries.

4.2. Post-transcriptional regulation of gene expression

The vast majority of liver regeneration studies have focused on signaling pathways and their influence on various transcriptional programs. However, in the past several years, there has been a surge in reports implicating post-transcriptional gene regulatory processes like alternative splicing, translational regulation and non-coding RNAs (ncRNAs) in liver regeneration. A recent study integrated in vivo polyribosome profiling ¹⁴⁹ with high-resolution transcriptome analyses, revealing global reprogramming of mRNA splicing and translation in regenerating hepatocytes ^39,150. The authors showed that in response to toxin injury, expression of several alternative splicing factors is translationally regulated to activate a neonatal splicing program, which enables hepatocyte proliferation and liver regeneration. Hundreds of exons were identified that are alternatively spliced in regenerating hepatocytes and a majority of them were noted to encode functionally relevant peptide segments containing subcellular targeting signals, protein-binding domains, and intrinsically disordered motifs that can rewire protein-protein interactions.

Importantly, a set of alternative exons were identified within components of Hippo-YAP pathway, especially in Nf2, Csnk1d, Yap1 and Tead1 mRNAs that are coordinately regulated by Epithelial splicing regulatory protein 2 (ESRP2) during liver regeneration ^39,150. ESRP2 normally promotes the generation of adult splice isoforms of Hippo pathway genes that favor quiescence ¹⁵¹. However, upon chronic liver injury, ESRP2 is translationally repressed to reactivate the neonatal splicing program that rewires the Hippo signaling pathway and allows quiescent hepatocytes to proliferate. Accordingly, Esrp2 knockout mice exhibit excessive hepatocyte proliferation after injury, whereas maintaining high expression of ESRP2 blocks hepatocyte proliferation by inhibiting the formation of neonatal splice isoforms of the Hippo-YAP pathway ^39,150. Thus programmed changes in alternative splicing seem to constitute a fundamental mechanism that imparts a higher level of adaptability following injury. Future studies are likely to identify additional splicing events that support liver repair and regeneration under specific injury contexts. However, it would be interesting to know whether (and how) cytokine or growth factor signaling pathways activate or repress RNA splicing factors, including ESRP2, to reprogram alternative splicing and fine-tune regenerative response of the liver.

An emerging body of evidence is revealing that, by directly modulating gene expression, ncRNAs provide an additional layer of regulation on top of transcriptional control mechanisms. One major class of ncRNAs, known as microRNAs, is now considered particularly crucial for liver regeneration ¹⁵². microRNAs bind complementary regions on mRNAs to trigger their degradation and translation repression ¹⁵³. It has been shown that microRNAs are important in the adult liver to maintain hepatocyte maturity since their loss in Dicer KO livers leads to many cellular and metabolic phenotypes ^{154, 155}. microRNAs act as promoters and inhibitors of liver regeneration by modulating the expression of cyclins, cyclin-dependent kinases and transcription factors that function as regulators of hepatocyte cell cycle and proliferation ¹⁵³. Recent microRNA profiling studies found dynamic changes in microRNA expression (e.g., miR-122, miR-192) both after PHx and in toxin-injured livers ^156,157. Subsets of microRNAs are upregulated at particular phases of liver regeneration and then rapidly downregulated before transitioning to the next phase. One study showed that miR-21 is upregulated within hours after PHx, and it accelerates liver regeneration by inhibiting NF-κB signaling via the ubiquitin ligase Pellino¹⁵⁸ as well as activation of FoxM1b, a key regulator of DNA synthesis ¹⁵⁹. miR-382, on the other hand, is more critical for the progression phase of regeneration wherein it regulates PTEN expression and Akt activity, ultimately promoting the G1/S phase transition of hepatocytes ¹⁶⁰.

lncRNAs are another class of ncRNAs that serve essential roles in cell proliferation and differentiation. Unlike microRNAs, lncRNAs resemble an mRNA transcript, and they are often 5’-capped, spliced, and 3’-polyadenylated ¹⁶¹. lncRNAs typically regulate proteins and other RNAs in cis or trans by recruiting factors to particular sites and can impact gene expression through a number of mechanisms, including changes in chromatin architecture or by influencing mRNA transcription, processing and translation ¹⁶². lncRNA associated with liver regeneration 1 (LALR1) was one of the first lncRNA identified in relation to liver regeneration. It is upregulated within the first 24 hours after PHx, and it boosts hepatocyte proliferation by promoting cell cycle progression both in vivo and in vitro ¹⁶³. Interestingly, HGF induces the expression of LALR1 and LALR1 accelerates hepatocyte proliferation via suppression of Axin1 gene expression in trans. This is in part due to LALR1-driven recruitment of CTCF to the Axin1 promoter region, which activates of Wnt/β-catenin signaling, and induces cyclin D1 mRNA transcription. Likewise, lncRNA MALAT1 is also upregulated within 24h of PHx, it accelerates hepatocyte proliferation by promoting cell cycle progression, and both HGF and p53 antagonistically regulate its activity during liver regeneration ¹⁶⁴. Moreover, just like LALR1, MALAT1 also activates Wnt/β-catenin signaling by suppressing the expression of Axin1, which stimulates the transcription of cyclin D1.

Through genome-wide expression profiling, several attempts have been made to identify additional lncRNAs important for regeneration, and while these identified hundreds of candidates, only a few were characterized extensively ^165,166. Amongst them, lncHand2 was recently implicated in promoting liver regeneration by triggering HGF-c-Met signaling ¹⁶⁶. Notably, the authors discovered that lncHand2, a conserved lncRNA transcribed in the opposite direction to its nearby protein-coding gene Hand2, is selectively expressed in the nuclei of pericentral hepatocytes, its expression peaks around 36 hours PHx, and that CRISPR/Cas9 deletion of lncHand2 in mice hinders hepatocyte proliferation and repopulation capacity. Mechanistically, it was found that lncHand2 recruits the Ino80 chromatin-remodeling complex onto the Nkx1–2 promoter to initiate its transcription, which induces c-Met expression to support HGF-c-Met signaling ¹⁶⁶. Thus, it appears that several ncRNAs collaborate with existing signaling pathways to support feed-forward and/or feedback loops of gene transcription, expanding liver’s regeneration capacity in response to physical or chemical injuries. Further characterization of lncRNAs that are evolutionarily conserved and exhibit interesting expression patterns after PHx will undoubtedly reveal additional candidates that either oppose or reinforce the activities of core regenerative pathways.

5. REGENERATION FAILURE IN ACUTE AND CHRONIC LIVER DISEASE

Despite being able to withstand a multitude of injuries, the liver can fail to regenerate both in response to acute or chronic damage. Acute liver failure in humans occurs because of severe viral infections or exposure to toxic doses of certain pharmaceuticals such as acetaminophen. Normally, after acute liver injury, hepatocytes proliferate proportionately until homeostasis is restored ⁴. However, in cases of acute liver failure, damage-induced cell death exceeds the liver’s ability to replace the sheer number of dying hepatocytes. This can happen due to two main reasons: (i) an injury threshold is reached after which the surviving cells are inhibited from further proliferation ¹⁶⁷, or (ii) the liver gets repopulated with functionally immature, fetal-like cells that are incapable of carrying out vital hepatocyte-related functions ¹⁶⁸. Chronic liver disorders, instead, present a different set of problems as repeated injuries impede regeneration owing to replicative exhaustion, which provokes hepatocyte senescence. The liver might still be able to recover from this over time if no further damage accrues, but in many cases, additional insults erode the remaining mature hepatic epithelia, and the liver turns fibrotic ¹⁶⁹.

Hepatic inflammation due to viral infections is a significant cause of liver damage worldwide; but, excessive alcohol consumption and nonalcoholic fatty liver disease are also major risk factors for both severe hepatitis and eventual cirrhosis. Compromised regeneration of fatty livers is well recognized clinically ^170–172, as moderate-to-severe steatosis increases the danger of graft failure upon liver transplantation ^173,174. Experimentally, high-fat diet-induced hepatic steatosis worsens the hepatocellular injury and impairs liver regeneration after PHx ^175,176, although, GADD34-mediated inhibition of the integrated stress response can improve regeneration of fatty livers in mice ¹⁷⁷. In alcohol-fed animals, hepatic steatosis is accompanied by increased production of reactive oxygen species, progenitor cell activation, and diminished replication of mature hepatocytes ^178–180. The accumulation of progenitor cells in alcohol-injured liver seems to reconstruct a more fetal-like tissue at the expense of mature hepatic parenchyma explaining why livers of some patients with severe alcoholic hepatitis fail to regenerate properly ^181,182. Usually, the dying cells are recognized by liver-resident macrophages, which release pro-inflammatory cytokines like TNF-α and IL-6 to amplify the injury signal and initiate repair process. But, in alcoholic and nonalcoholic fatty liver disease, enhanced hepatic translocation of bacteria and their endotoxic products—owing to increased leakiness of the gut—stimulate stromal and infiltrating immune cells to produce reactive oxygen species, TNF-α, and other damage-associated molecular patterns (DAMPs), triggering incessant inflammation and an abnormal regenerative response ^179,183,184. These elevated inflammatory signals activate quiescent hepatic stellate cells into scar-forming myofibroblasts, which deposit excessive ECM further inhibiting hepatocyte proliferation and liver regeneration ^169,185.

6. CONCLUSIONS AND OUTLOOK

The use of advanced genetics and genomics methods in model organisms has unraveled many of the cellular and molecular mechanisms of liver regeneration. These methodologies have revealed strong evolutionary conservation and robustness of the process, and “hepatocellular plasticity” seems like an underlying feature of the liver in responding to various injuries. However, the challenge remains not only to understand the normal mechanisms of regeneration, but also to address the aberrant regenerative responses observed in the clinic and how they differ from normal healthy regeneration. Also, the vast majority of studies focus on hepatocytes or use whole liver tissue for analyses. But, individual cell types within the liver serve unique and irreplaceable roles both during normal and regenerative conditions. It is, therefore, obligatory to study the molecular events manifested by each of the individual cell types under these circumstances.

With the advent of super-resolution microscopy, CRISPR-mediated genetic engineering, and single-cell genomics, the field is on the verge of addressing fundamental questions that have remained unanswered in the past. For instance, Halpern et al. provided the first single-cell approach in mouse livers to show that over 50% of hepatic genes exhibit distinctly zonated expression pattern along the porto-central axis ¹². Another study reported unbiased isolation and sequencing of single cells from human autopsied livers, identifying several previously unknown hepatocyte and non-parenchymal cell subpopulations ¹⁸⁶. Remarkably, while the original study in the mouse detected zonated gene expression pattern in hepatocytes ¹², recent single-cell analysis of human livers uncovered zonation of gene expression in hepatocytes and surprisingly in endothelial cells ¹⁸⁷. The authors also identified a population of EpCAM⁺ progenitor cells that co-express both hepatocyte- and cholangiocyte-specific markers ¹⁸⁷. Given the rapid advances in above-mentioned technologies, it may soon be possible to map individual cell trajectories with phenotypic data from whole tissues making for an exciting future for liver regeneration studies.

Our current understanding of cellular transitions that bring a mature, quiescent hepatocyte to a de-differentiated, proliferative state is incomplete. With multiple groups discovering distinct hepatocyte subpopulations in regenerating livers and their roles being context-specific, one may ask, if there are different starting states and transition paths for reaching the same destination. Likewise, differential proliferation rates of diploid and polyploid hepatocytes under identical contexts raise the question if starting states make it difficult/easier to transition from quiescence to proliferation. Presently most studies focus on the transcriptome, as it serves as a reliable proxy for the cell state. However, recent experimental and computational advances are enabling quantitative and scalable assessment of the genomic architecture, chromatin states, and transcription factor occupancies at both bulk and single-cell levels. The eventual hope is to use this understanding of hepatocellular plasticity and renewal for developing targeted therapeutics for various human liver diseases.