Liver Macrophages: Old Dogmas and New Insights

Liver Macrophages: Multiple Players on the Same Team

Not so long ago, the term Kupffer cells (KCs) was synonymously used for hepatic macrophages, and relatively approximate methods (e.g., immunohistochemistry for F4/80 or cluster of differentiation [CD]68) were used to identify these cells in liver sections (Fig. 1). It is now well recognized that macrophages observed in the liver following injury are heterogeneous and may derive from different origins: liver‐resident macrophages or KCs and two patrolling populations of bone marrow monocyte‐derived macrophages (MoMFs) as well as peritoneal macrophages for subcapsular regions of the liver. Different cell origins have been linked to discrepancies in cell functionality as well as in responsiveness toward activating and recruiting signals, directly influencing the immune response outcome.

Kupffer Cells and Monocytes: The Sentinels and the Emergency Response Team

Due to their location in the sinusoids, their cytoplasmic expansions that allow them to sense both the blood circulation and the hepatocytes, and their high phagocytic abilities, KCs are among the primary cells exposed to and able to respond to liver insult.36 Key metabolic functions are undisputedly attributed to KCs. Most notably, they are pivotal in the phagocytosis of aging red blood cells and iron metabolism and have more recently been linked to lipid metabolism.75, 76 During homeostasis, KCs can uptake particulate peptide antigens to induce tolerogenic T‐cell responses.35 Following experimental liver injury, KC depletion dramatically prevents inflammation initiation and attraction of MoMFs, thus highlighting their role as sentinels in the liver.

An example of this sentinel role for KCs was demonstrated in a mouse model of steatohepatitis combining a high‐fat diet with alcohol feeding.77 In this model, sorted KCs did not show any tumor necrosis factor (Tnf)a, Il1b, or nitric oxide synthase 2, inducible (Nos2) gene expression induction, unlike MoMFs that were strongly activated toward a proinflammatory phenotype in a Notch homolog 1, translocation‐associated (NOTCH1)‐dependent manner.77 On the other hand, KCs are known to respond to gut‐derived LPS through TLRs following increased intestinal permeability caused by alcohol.78, 79 In this situation, KCs produce high amounts of IL‐6, monocyte chemoattractant protein 1/CCL2, and TNFα that drive inflammatory cell recruitment, as well as hepatocyte production of alarmins, acute phase proteins, and chemokines, thus increasing the inflammatory reaction.78, 79, 80, 81, 82 Interestingly, alcohol‐induced and gut‐derived LPS effects do not seem to be solely attributable to KCs. Inokuchi et al.83 demonstrated that wild‐type (WT) KC‐depleted mice receiving TLR4‐deficient bone marrow or TLR4‐deficient mice receiving WT bone marrow showed an intermediate liver injury and inflammation compared to that of WT and TLR4‐deficient mice exposed to ethanol challenge. The authors concluded that both MoMFs/KCs and hepatic stellate cells were implicated in the pathogenesis of alcoholic liver disease caused by LPS through TLR4 activation. Interestingly, it has also been nicely demonstrated that KCs potently produce IL‐10 and relatively low amounts of TNF following LPS challenge compared to peritoneal and splenic macrophages.84, 85 KCs were also shown in this study to have higher phagocytic capacities and to express higher levels of TLR inhibitory molecules, attributes that attest to their chronic exposure to endotoxins, their crucial roles in blood clearance, and their steady‐state immunotolerant phenotype.85

The “task diversification” between KCs and MoMFs is more difficult to assess in continuous chronic injury models because MoMFs may, at least in part, replace KCs in such conditions (see above). Nonetheless, studying the methionine/choline‐deficient dietary model of steatohepatitis revealed gene expression changes in sorted KCs that are mainly related to innate immune activation and metabolism. Contrastingly, MoMFs from steatotic livers had a higher expression level of fibrogenic and angiogenic genes than KCs from steatotic livers or cells sorted from control diet‐fed animals.60 Additionally, cell sorting from mice subjected to a combined fibrosis–cancer model induced by diethylnitrosamine followed by repetitive CCl₄ injections revealed distinct coexisting MoMF populations with specific gene expression patterns, suggesting inflammatory capacity (immature myeloid infiltrate), angiogenic and fibrogenic activity (classical tumor‐associated macrophages), or immune‐suppressive functions (myeloid‐derived suppressor cells).86

While KCs are stationary and crucial for initiating inflammation, drastic changes occur in the MoMF compartment following injury. MoMFs have been shown to be recruited to the injury site just a few hours after acute injury40, 53, 87 and account for the majority of hepatic macrophages in models of chronic liver injury.57, 60, 88 Their initial phenotype following recruitment in mice is characterized by high expression levels of pattern recognition receptors, multiple “polarization” markers, and flags of immaturity (e.g., Ly6C, CCR2, Csf‐receptor), supporting their ability to respond to signals and further mature at the site of injury.40, 53

Multifaceted Roles of Macrophages During Liver Injury and Regeneration

Despite the undisputable association between inflammation and tissue damage, it would be misleading to consider macrophages as foes during liver injury. In fact, macrophage depletion had detrimental consequences on liver disease resolution.89, 90 Several reviews pondered beneficial versus detrimental roles of macrophages during liver injury.91, 92, 93, 94 A generally accepted view is that proinflammatory macrophages, derived from MoMFs, are required for activating regenerative mechanisms but may increase tissue injury through uncontrolled inflammation while restorative macrophages promote inflammation resolution and tissue repair but are also implicated in aberrant tissue repair mechanisms, namely fibrosis and cancer.95, 96

In opposition to the old dogma that macrophage‐mediated inflammation is generally detrimental, a recent paper demonstrated that blocking Ly6C⁺ monocyte recruitment by using a chemokine receptor CCR2 antagonist delayed hepatitis B virus clearance in mice while it was enhanced by KC depletion using clodronate‐loaded liposomes.58 KC depletion led to increased Ly6C⁺ monocyte recruitment, thus accelerating virus clearance.58

Another situation in which liver macrophages were shown to be beneficial is following liver regeneration. Clodronate‐mediated KC depletion was reported to delay liver regeneration and to increase liver damage following partial hepatectomy in mice and rats.97, 98, 99, 100, 101 These effects seem to be partly due to the absence of IL‐6 production by KCs and are reversed after IL‐6 administration.101 Similarly, clodronate‐mediated macrophage depletion reduced the alternative liver regeneration pathway through liver progenitor cell (LPC) activation by reducing LPC accumulation in rats in a 2‐acetylaminofluorene/partial hepatectomy model.102 Low‐dose clodronate was used to deplete KCs and not MoMFs; this treatment reduced MoMF infiltration and tissue injury concomitantly with LPC accumulation in an LPC‐driven liver regeneration model (choline‐deficient ethionine‐supplemented diet),103 although another study reported that KC depletion reduced parenchymal invasion but not LPC proliferation in the same model.104 Alternatively, intravenous bone marrow‐derived macrophage injection has been shown to be sufficient to initiate regenerative mechanisms by inducing a ductular reaction in the healthy mouse liver through TNF‐related weak inducer of apoptosis (TWEAK).105 Macrophages have also been designated as the source of Wnt3a, thus favoring liver progenitor cell differentiation toward the hepatocyte fate during tissue regeneration.106 Human autologous bone marrow cell infusion also seems to be beneficial for chronic liver disease treatment.107 These effects may be attributed to immunomodulatory actions of transplanted cells and their ability to phagocytose cellular debris,108 arguing once more for the high relevance of the immune system in tissue repair. However, data on the clinical efficacy of macrophage cell therapy in patients are still awaited. Collectively, it is well documented that MoMFs and KCs play major roles in hepatocyte‐ and LPC‐mediated liver regeneration. Therefore, the use of macrophage‐suppressive drugs should be considered cautiously.

In contrast, KC inhibition using gadolinium chloride hexahydrate prevented liver cancer stem cell occurrence and tumor development in rats exposed to diethylnitrosamine, through inhibition of liver progenitor cell activation and proliferation.109 In a nonalcoholic steatohepatitis (NASH) model using melanocortin‐4 receptor‐deficient mice fed a Western diet followed by a low‐dose CCl₄ injection, it was reported that CD11c⁺ liver macrophage depletion prevented fibrogenesis.110 Similarly, the use of a CCR2/CCR5 antagonist that prevents MoMF recruitment to the liver in patients with NASH and murine models shows promising results notably by preventing fibrosis.111 Another good representation of these apparently conflicting macrophage functions is time‐controlled macrophage depletion. As such, selective CD11b‐expressing macrophage depletion using a diphtheria toxin receptor system during tissue scarring reduced collagen deposition, whereas macrophage depletion during the reparative phase led to reduced extracellular matrix degradation in the CCl₄ model.89

Spectra and Fluidity of Macrophage Polarization States

Macrophages present at a given time at an injury site are heterogeneous by nature. Recent unbiased approaches, such as single‐cell RNA sequencing, revealed the coexistence of various and mixed activation phenotypes, even in healthy human liver.39 This heterogeneity may also reflect the versatility of macrophages (Fig. 1). For instance, exposure of human MoMFs in vitro with typical extracellular stimuli, like cytokines, fatty acids, or danger signals, does not induce a well‐defined M1 or M2 phenotype but a broad spectrum of activation states.30 Similarly, feeding mice a Western diet (rich in carbohydrates, cholesterol, and fat) induced the transcriptomic and epigenomic reprogramming of myeloid progenitor cells, resulting in increased proliferation and enhanced innate immune responses.112 The ability of macrophages to be re‐educated, reprogrammed, or repolarized is a fascinating new area, especially for diseases in which the liver is highly populated with macrophages. Indeed, it seems to be possible to re‐educate macrophages from a classical so‐called M1 to M2 phenotype and vice versa.113 New technologies (sequencing techniques, imaging methods, unbiased big data analyses) allow for an exciting and fresh view on macrophage heterogeneity, for instance, regarding origin, polarization, localization in tissue, and function in relation to the disease state.114 However, it is important to accurately describe the experimental setting from which conclusions are drawn to allow comparability between models and researchers.115 Understanding the activating signals and intracellular signaling pathways that determine functional responses of macrophages or macrophage subsets in the liver holds great potential for advancing therapeutic options in liver diseases.

Extracellular vesicles are emerging as a significant intercellular communication tool, leading, for instance, to the activation of liver macrophages (both KCs and MoMFs) toward a proinflammatory and pathologic phenotype in alcoholic liver disease, notably characterized by TNFα and IL‐12/23 expression.116 This macrophage activation was partly attributed to increased heat shock protein 90 (HSP90) levels in hepatocyte‐derived extracellular vesicles. Extracellular vesicles may contain proteins as well as DNA and microRNAs (miRNAs) influencing gene expression in the cells that uptake vesicles. On the other hand, macrophages were shown to differentiate toward a matrix‐degrading phenotype when phagocytosing cell debris in culture or liposomes in vivo, thus accelerating fibrosis resolution.90 It is hence of crucial importance for extracellular vesicle‐related studies to be well designed in order to prevent data falsely attributed to exosome content.117

Bile acid accumulation in the liver during cholestasis, a widespread consequence of bile duct damage occurring during chronic liver disease or cholangiopathies, has also been shown to serve as an activating signal for liver macrophages. Interestingly, manipulating bile acid composition seems to represent a promising area of research because in the multidrug resistance protein 2 (Mdr2)‐deficient model of cholangitis, pharmacologic inhibition of the ileal apical sodium‐dependent bile acid transporter (ASBT) reduced liver fibrosis progression, cholestasis, as well as alanine aminotransferase and bilirubin levels, presumably by reduced toxic hydrophobic bile acid concentration.118 Along the same line, ursodeoxycholic acid (UDCA) treatment is currently the U.S. Food and Drug Administration‐approved first‐line therapy to limit primary biliary cholangitis progression.119 UDCA is a hydrophilic bile acid that is thought to limit bile toxicity to liver cells. UDCA and 24‐nor‐UDCA have notably been shown to reduce liver inflammation in a Schistosomiasis mansoni cercariae infection model.120 UDCA was also shown to reduce TNFα‐induced IL‐8 production by macrophages associated with a reduction in TNFα receptor–associated factor 2 (TRAF2) phosphorylation121 and to reduce LPS‐induced macrophage activation.122, 123, 124

Although all cells in one organism possess the same genetic material, gene expression modulation through epigenetic modifications controlled by histone remodeling, transcription factors, miRNAs, and long noncoding RNAs results in a plethora of distinct cell types. Recent studies have highlighted that macrophages are not exempt from these epigenetic modifications because they directly affect the polarization state of macrophages.125, 126, 127, 128, 129 Aside from macrophage‐depleting agents used in mice, such as clodronate‐loaded liposomes or KC depletion by gadolinium chloride, several approaches have been proposed to influence the macrophage phenotype. Macrophage‐targeting delivery methods for immunomodulatory drugs or silencing RNAs (siRNAs) include the use of nanoparticles, liposomes, glucan shell microparticles, and oligopeptide complexes.22, 130, 131 For instance, glucan‐encapsulated siRNAs against mitogen‐activated protein kinase kinase kinase kinase 4 (Map4k4) prevented mortality in LPS‐induced sepsis by inhibiting TNFα and IL‐1β production specifically in macrophages.132 Thus, the selective targeting of macrophages and the specific interference with a detrimental activation pathway may shift the immune response toward being beneficial. As such, it is tempting to believe that in situ macrophage reprogramming will become a major area of intervention in the future. These adaptable and flexible alterations of gene expression offer tremendous possibilities in macrophage polarization modulation. It is thus conceivable to alter macrophage activation from a pathogenic to a regenerative phenotype and to direct immune cells to a response “à la carte,” by orientating epigenetic modifications specifically in targeted macrophage populations (Fig. 1).

Translational Research: From Mouse Models to Human Diseases

A vast majority of the data referred to in this review is derived from animal models. While animal models are necessary to dissect molecular mechanisms leading to pathology, it is important to evaluate the human relevance of recent findings.

Substantial differences were identified in terms of KC‐ and MoMF‐specific identifiers between humans and mice (reviewed in Heymann and Tacke41) as well as polarization markers.115 At present, comparative studies on human versus mouse liver macrophage subsets are limited, although many key aspects of resident KC versus recruited MoMF, activation and recruitment signals, and metabolic activities appear to be conserved.39, 60, 133, 134 The liver microenvironment is also different between mouse models and human disease, and several studies have attempted to tackle species differences by using humanized livers. However, such studies currently require the mice to be immune deficient (e.g., severe combined immunodeficient [SCID] mice), thus limiting the ability to study macrophages in the context of a normal immune environment.

Much comparative work from mouse models to human diseases remains to be done. On the other hand, initial findings from macrophage‐directed therapies in early phase clinical studies, such as the CCR2/CCR5 inhibitor cenicriviroc in patients with NASH and fibrosis, indicate that principal observations can be successfully translated from bench to bedside.135

Prognostic Value of Macrophage‐Related Markers in Human Liver Disease

One particularly exciting area of translational research uses liver macrophage heterogeneity as the starting point for exploring novel biomarkers reflecting characteristics of human disease. For instance, staining liver samples with macrophage‐related markers, assessing hepatic or circulating monocyte/macrophage populations (by flow cytometry), and measuring macrophage‐related surface proteins in circulating blood have been suggested as novel biomarkers in hepatology.17, 22

One prototypical situation in which macrophage marker expression has been studied is liver cancer. It is generally believed that tumor‐associated macrophages mainly promote cancer progression by maintaining an immunotolerant tumor microenvironment. As such, CCL2 expression has a prognostic value in patients with hepatocellular carcinoma, and CCR2 blockade has been shown to inhibit cancer growth by favoring an antitumoral immune response136 and inhibiting angiogenesis.86 Recently, CD163, a macrophage‐specific marker with anti‐inflammatory functions, has been used to evidence the presence of tumor‐associated macrophages in cholangiocarcinoma. The authors concluded that CD163⁺ macrophage number correlated with cholangiocarcinoma stage.137 The presence of soluble CD163, released by activated macrophages in the blood, has also been correlated to hepatocellular carcinoma progression.138 However, the correlation between CD163 and hepatocellular carcinoma prognosis remains debatable.139

Soluble CD163 has, however, been widely described as a macrophage activation marker in patients with different liver diseases. Despite its anti‐inflammatory function as a scavenger receptor on macrophages, circulating CD163 seems to be associated with advanced disease states, disturbed intestinal barrier, and adverse prognosis. This has been convincingly demonstrated for NASH,140 viral hepatitis,141 autoimmune hepatitis,142 and decompensated cirrhosis.143 Interestingly, this marker dynamically responds to efficient treatment, such as for autoimmune disorders142 or viral hepatitis,144, 145 suggesting that it could be helpful monitoring treatment responses.

Similarly, CD206⁺ macrophages are classically defined as harboring an anti‐inflammatory phenotype, and a retrospective study identified high CD206⁺ macrophage infiltration in the liver to be a poor prognostic indicator in patients with hepatocellular carcinoma.146 CD206⁺ macrophages have, however, also been described as proinflammatory in patients with viral hepatitis.147 Moreover, monocytes are regarded as major contributors to the initiation and perpetuation of tissue damage, providing the soil for chronic liver disease, including fibrosis and cancer.148, 149 In line with these, it has been reported that soluble CD163 can be used in combination with the Model for End‐Stage Liver Disease score and other clinical scores to obtain a better prognostic for acute‐on‐chronic liver failure and cirrhosis.143, 150 Similarly, the expansion of myeloid cells (mainly monocytes) with an immune‐suppressive phenotype in peripheral blood has been linked to impaired antimicrobial responses in patients with acute‐on‐chronic liver failure.151

While these markers indicate adverse prognosis, there might be markers capturing the beneficial repair functions of macrophages as well. In this regard, MER receptor tyrosine kinase (MERTK) is expressed by monocytes and macrophages that have features of restorative phagocytes.152 Although these MERTK + macrophages seem to be associated with disease severity in acute‐on‐chronic liver failure,153 they apparently indicate proper resolution from acute liver failure in patients.154

Conclusions

Expanding knowledge on liver macrophages has certainly modified (or replaced) old dogmas in the field. For instance, it has become clear over the past years that liver macrophages cannot be sufficiently described as M1 or M2 cells. As recently suggested, all studies should precisely describe macrophage populations based on their origins, the activation signals, and a relevant choice of markers.115 The ongoing thorough macrophage characterization in a variety of liver diseases as well as the rapidly increasing knowledge of the biomolecular mechanisms implicated in gene expression regulation or cell communication associated with technical advances in targeted drug delivery may one day allow clinicians to provide their patients with personalized treatments instructing liver macrophages on how to properly orchestrate the liver response to a defined type of injury.