From oxidative stress to redox homeostasis and redox signalling

Molecular oxygen (O₂) is essential for the survival of human beings and, more generally, of all aerobic organisms. Aerobic energy metabolism relies on oxidative phosphorylation, a crucial process by which the oxido-reduction energy of mitochondrial electron transport is eventually converted to the high-energy phosphate bond of ATP. Aerobic organisms use O₂ as the final electron acceptor for mitochondrial cytochrome c oxidase, which, in turn, represents the terminal functional element of the mitochondrial multicomponent NADH dehydrogenase enzymatic complex, which is able to catalyze the four-electron reduction of O₂, leading then also to H₂O formation (Figure 1). During mitochondrial oxidative phosphorylation and other electron transfer reactions, however, partially reduced and highly reactive O₂ metabolites, including superoxide anion (O₂^•-), hydrogen peroxide (H₂O₂) and hydroxyl radical (^•OH), can be formed within cells. These reactive O₂ metabolites are usually collectively referred to as 'reactive oxygen species' (ROS) and their generation in a biological environment exposes most living organisms to the so-called 'oxygen paradox': oxygen is necessary for life but it is also potentially hazardous since ROS may easily become a source of cell and tissue injury, as was recognized by early pioneers of free radical research [1–9]. However, as a natural consequence of this paradox, aerobic organisms have developed evolutionarily conserved mechanisms and strategies to carefully control the generation of ROS and other oxidative stress-related radical or non-radical reactive intermediates (that is, to maintain redox homeostasis), as well as to 'make use' of these molecules under physiological conditions as tools to modulate signal transduction, gene expression and cellular functional responses (the concept of redox signalling) [10–23].

At present, redox research is at the forefront of biomedical research in view of the expanding knowledge on the roles that increased and/or sustained levels of oxidative stress and related mediators have been described to play in major human diseases, including atherosclerosis, diabetes and cardiovascular diseases [14, 24–29], cancer [30, 31], neurodegenerative disorders [32–34], chronic liver [35–38] and lung diseases [39–41], to name just a few. Most of the conditions in which the role of oxidative stress and related mediators has been characterized belong to what one may define as chronic inflammatory/fibrogenic diseases, often involving chronic activation of wound healing.

About this review

This review has been designed as an attempt to offer a comprehensive, but not hyper-specialized, critical introduction to current knowledge in the field in order to introduce readers to the fascinating complexity of redox signalling and redox homeostasis regulation, with a major focus on chronic wound healing and liver fibrogenesis. This review will then offer a sequence of ready-to-use key information and concepts on major types of ROS, free radicals and oxidative stress-related reactive intermediates operating in living organisms, their sources in mammalian cells and tissues, and antioxidant defences. Along these lines, this review will not intentionally deal with all the details but, whenever possible, the interested reader will find indications for highly recommended and more detailed and specialized reviews and articles on specific topics.

Reactive oxygen species

ROS is a generic collective term indicating a number of active and reactive partially reduced O₂ metabolites. Some of them, such as O₂^•- and ^•OH, can be defined as true free radicals, which are reactive molecular species with an impaired electron in their outer orbital. Free radicals are paramagnetic and reactive chemical entities that can undergo redox reactions by interacting with surrounding molecules in order to regain the more stable non-radical condition. Other ROS, such as H₂O₂, are more properly pro-oxidant non-radical agents. Indeed, O₂^•-, ^•OH and H₂O₂ are by far the most relevant in physiological or pathophysiological conditions. Crucial information on major ROS, their sources and reactions are briefly summarized below and in Figures 1 and 2.

The process of lipid peroxidation and the generation of non-radical intermediates

Lipid peroxidation (Figure 3) is a term commonly used to indicate oxidative decomposition of the ω-3 (22:6) and ω-6 (18:2, 20:4) polyunsaturated fatty acids of membrane phospholipids [19, 42]. This process, which is very common in pathological conditions [21–23, 42], is usually initiated by the interaction of a ROS or other free radical with polyunsaturated fatty acids and exacerbated by the presence of divalent metal ions. This reaction leads to the formation of lipid radicals (L•) that, in turn, can react with available O₂ to generate lipid peroxyl radicals (LOO•). From this point the propagation phase of this chain reaction occurs, whereby LOO• interacts with other lipid molecules, resulting in the generation of lipid hydroperoxides (LOOH). These in turn undergo a degradative breakdown, leading to the generation of other radical species (LO• and LOO•) that further propagate lipid peroxidation, and to several aldehydic end-products, such as malonyldialdehyde (MDA) and 4-hydroxy-2,3-alkenals (HAKs) of different chain lengths, [19, 42] as well as to F₂-isoprostanes [53]. 4-Hydroxy-2,3-nonenal (HNE), the most active biological and pathophysiological HAK [23–25], and F₂-isoprostanes (so defined because of their PGF₂-like structure) are relatively stable and lipid soluble compounds that can diffuse from the site of generation and easily cross biological membranes. Moreover, as proposed more than 25 years ago for HNE [23–25, 35, 54] and more recently for F₂-isoprostanes [55, 56], these non-radical species can also act as mediators that are able to affect redox state, signal transduction and cell responses. Detection of HNE or F₂-isoprostanes in biological fluids or tissues is today considered one of the best ways to evaluate in vivo on-going oxidative stress [57].

Nitric oxide and reactive nitrogen species

NO is a small hydrophobic molecule that crosses cell membranes without needing channels or receptors [58]. It is generated by NO synthase (NOS) isoforms through the conversion of L-arginine to citrullin. Three types of NOS have been identified: endothelial NO synthase (eNOS), which is bound to plasma membranes and known to be strongly activated by the entry of calcium through membrane-bound receptors [59]; inducible NO synthase (iNOS), which was first identified in macrophages and then in other cells, including hepatocytes, is known to be up-regulated by pro-inflammatory cytokines and/or lipopolysaccharide (LPS), and is able to generate low levels of NO compared with the other NOS isoforms; and neuronal NO synthase (nNOS) (Figure 4).

NO exerts physiological effects by controlling vascular tone, cell adhesion, vascular permeability and platelet adhesion [12, 13, 15, 60]. It also exerts several potentially toxic effects, although many of these are more likely mediated by oxidation products included in the definition 'reactive nitrogen species' (RNS). In particular, NO is able to rapidly react with O₂^•- to form the much more powerful oxidant peroxynitrite (ONOO^-). Indeed, neither O₂^•- nor NO are particularly toxic in vivo because of efficient systems able to minimize their accumulation [61, 62]: O₂^•- is removed by SOD isoforms whereas NO is removed as a consequence of its rapid diffusion through tissues [63]. Under pro-inflammatory conditions, simultaneous production of O₂^•- and NO can be strongly activated and significant amounts of ONOO^- are generated, which may cause significant injury to different cellular structures.

Peroxynitrite

ONOO^- (see [58] and references therein) is a strong oxidant able to react directly with thiol groups, iron-sulphur centres and the active site -SH groups in tyrosine phosphatases. In physiological conditions, the production of ONOO^- is quite low and oxidative injury is minimized by endogenous antioxidant defences. When increased in pathological conditions, ONOO^- can act either as a direct oxidising species or indirectly by decomposing into highly reactive radicals. When ONOO^- acts as an oxidant, it produces nitrite and a hydroxide ion rather than isomerising to nitrate (Figure 4) and can react with proteins (tyrosine nitration or direct reactions with specific amino acids), lipids (lipid peroxidation) and nucleic acids (oxidative modifications in nucleobases). ONOO^- can also interact with mitochondria, reaching them from extra-mithocondrial compartments or being locally produced through the interaction of NO (generated by the mitochondrial NOS) and O₂^•-. Mitochondrial toxicity of ONOO^- results from direct oxidative reactions of principal components of the respiratory chain or from free radical-mediated damage. Persistent generation of significant levels of ONOO^- can lead to the induction of cell death, either apoptosis or necrosis (Figure 5; see also 'Mitochondria, nitric oxide, RNS and cell death' below).

How ROS and other oxidative stress-related reactive intermediates interact with biological macromolecules

ROS, NO, HAKS and other free-radical or non-radical reactive intermediates may interact with relevant biological macromolecules [7–9, 14, 15, 19, 21, 58], events that can easily lead to cytotoxic/damaging consequences or contribute to redox regulation and signalling.

Protection from ROS and oxidants

The following categories of naturally occurring components may be defined.

Protection from lipid peroxidation: natural and synthetic antioxidants

According to Halliwell and Gutteridge [64] "an antioxidant is any substance that, when present at low concentrations compared to those of an oxidizable substrate, is able to significantly delay or inhibit oxidation of that substrate." Of course, this generic definition also includes primary antioxidants (free radical scavengers able to interact directly with and/or to block the initiating free-radical, such as mannitol) and synthetic molecules able to bind metal ions (for example, desferrioxamine). However, several authors, when using the word 'antioxidant', have in mind the so-called 'chain breaking' or 'secondary antioxidants', with α-tocopherol (vitamin E) being the naturally occurring prototype. These natural or synthetic molecules have a chemical structure (Figure 8) able to intercept radical intermediates produced during on-going lipid peroxidation, such as peroxyl or alkoxyl radicals, thus preventing (that is, 'breaking') the perpetuation of hydrogen abstraction in the chain reaction. Figure 6 offers an overview of the reactions involving α-tocopherol, including its re-cycling based on the involvement of GSH and ascorbate.

A number of additional and useful concepts regarding antioxidants (whether enzymic or not) should be considered: antioxidants binding metal ions are not usually consumed during the course of reaction; antioxidants able to decompose peroxides may be consumed or not, depending on their nature (for example, as enzymes GPXs are not consumed); chain breaking – as well as primary – antioxidants are usually consumed during on-going oxidative stress; many antioxidants have multiple mechanisms of action; some antioxidants (tocopherols, ubiquinol, carotenoids and flavonoids) will exert their effects in a lipid phase (that is, at the level of biological membranes) whereas others will do so in an aqueous phase (ascorbate, urate, GSH and other thiols).

Principles of redox homeostasis

To introduce the concept of redox homeostasis one can refer to the scenario depicted in Figure 9 and to the intuitive concept of oxidant/antioxidant balance, which is still the simplest way to begin understanding the complexity of redox mechanisms.

Redox sensors and the basis of redox-dependent transcriptional regulation

At this point one should move beyond the simple concept of oxidant/antioxidant balance by introducing the more refined notion of redox sensors as well as the principles of redox-dependent regulation of transcription. The key messages in this area (for more details see [18, 71]) can be summarized as follows.

Redox sensors in prokaryotic cells and yeast

Redox sensors were first described in bacteria, including the OxyR and SoxR redox sensitive transcription factors, the chaperon molecule Hsp33, the oxygen sensor FNR and others. All these 'redox receptors' have a structure designed to sense specific ROS, oxidants or other reactive intermediates. These ancestral redox sensors can essentially contribute to fast mechanisms designed to deal with ROS and to make adjustments allowing the survival of the bacteria (that is, to reset redox homeostasis). During evolutionary development these simple bacterial sensors have been replaced with more specifically designed proteins, such as yeast thiol peroxidases (enzymes belonging to the family of peroxyredoxins or GPXs), which contribute to H₂O₂ signalling (see Figure 10 for more details).

Redox sensors in higher eukaryotes

In higher eukaryotes redox regulation of transcription, as well as of signalling elements like protein phosphatases, relies on properties and strategies similar to those described for bacteria or yeast (cysteine-based oxidation/reduction cycles), which have been evolutionarily conserved. Here we still see thiol peroxidases affecting H₂O₂-dependent signalling, with some crucial differences since PRXs and GPXs have been reported to be involved in the modulation of signal pathways. Figure 11 illustrates established examples of three different mechanisms by which an increase in intracellular ROS may trigger transcription of redox sensitive genes: redox reactions directly involving either signalling components or transcription factors; nuclear translocation of transcriptional regulators that are maintained in an inactive form in another cellular compartment; and modulation of transcription by alterations in the so-called 'redox buffers'. More details can be found in the legend to Figure 11 and in [18, 71–73].

Patterns of fibrosis progression in CLDs

Fibrosclerotic progression follows distinct patterns that are intrinsically related to the aetiological cause of the CLD and the topographic site of tissue injury, as well as to the predominant pro-fibrogenic mechanism and the involvement of populations of pro-fibrogenic myofibroblast-like cells of different origin (MFs). Four main patterns of fibrosis have been identified and are described in detail in Figure 14.

Myofibroblast-like cells as pro-fibrogenic effectors in CLDs

MFs are pro-fibrogenic cells found in chronically injured liver in either experimental or clinical conditions. They are characterized by a positive stain for α-smooth muscle actin (α-SMA). The origin of liver MFs has been a matter of controversy for more than a decade but now there is substantial agreement on the following major concepts (summarized in Figure 15). First, three different phenotypes of MFs have been identified in fibrotic/cirrhotic livers of human patients or in animal models, including hepatic stellate cells (HSCs) that become activated or HSC/MFs (in capillarised sinusoids), interface myofibroblasts (MFs located at the interface between fibrotic septa and the surrounding parenchyma) and portal/septal myofibroblasts (MFs in the expanded portal areas or within fibrotic septa) [85, 86]. Second, liver MFs have multiple origins, with most originating from HSCs; indeed, most of our present knowledge comes from studies investigating this peculiar liver cell population, as recently reviewed by Friedman [81]. HSCs are likely to give rise to HSC/MFs and to most interface MFs; MFs can also originate from portal fibroblasts, which are also able to give rise to the phenotypically identical septal MFs. A significant number of MFs can also originate, in chronically injured human [87] and murine livers [88, 89], from bone marrow-derived mesenchymal stem cells (MSCs), suggesting caution when considering therapeutic procedures involving autologous transplant of bone marrow-derived stem cells [90]. Third, whatever the origin, all the mentioned 'precursors' of ECM-producing cells in CLDs are likely to undergo a similar process of activation and trans-differentiation that leads to the peculiar MF-like phenotype [82, 90]; thus, the actual 'feeling' is that HSC/MFs, and likely all activated MF-like cells, may share, from a functional point of view, the ability to exhibit a number of phenotypic responses [80, 81, 91] (summarized in Figure 15), including proliferation [76, 81, 83, 92, 93], synthesis and remodelling of ECM and of mediators, migration, contractility and the potential to undergo apoptosis [80, 81, 91, 94]. More details are given in the legend of Figure 15.

Major events, cells and mechanisms regulating liver fibrogenesis in CLDs

Several relevant events for fibrosclerotic progression of CLDs may be considered relatively independent of the specific aetiology, as detailed in Figure 13; in these conditions of persisting tissue injury, ROS and other related mediators, released by damaged cells or activated inflammatory cells, are likely to play a relevant role. Along these lines, it should be noted that, in chronic diseases, both necrotic and apoptotic, as well as apoptosis-like, forms of cell death have been reported to occur and have been detected in the same tissue section [80, 95, 96] and that hepatocyte apoptosis represents an effective pro-fibrogenic stimulus [97, 98].

To complete the oversimplified scenario offered in Figure 13 (more details in [74, 91, 95]), the following concepts should be recalled. First, in such a complex scenario several other cellular 'actors' are likely to be involved; Figure 16 summarizes all the cells involved in CLDs, indicating how they can contribute to fibrogenesis progression and, for some of them, to the generation of ROS or related reactive intermediates. Second, a major pathogenic role has to be attributed to several soluble factors (produced by different kind of cells; Figure 16) that can regulate both the state of activation of MFs and their phenotypic responses as well as the responses of other cells involved; these factors include platelet-derived growth factor (PDGF), transforming growth factor (TGF)β, connective tissue growth factor (CTGF), endothelin-1 (ET-1), monocyte chemotactic protein (MCP)-1, and tumour necrosis factor (TNF), to name just a few. Third, several signalling pathways, transcription factors and related transcriptional gene regulation have been dissected and identified as involved in the process of activation of HSCs or in mediating phenotypic responses of MFs, and most of these (see next section) are known to be redox-sensitive. Fourth, angiogenesis, pro-angiogenic cytokines and expression of related receptors are emerging as crucial factors potentially able to contribute actively to liver fibrogenesis [83, 92, 93]. Finally, oxidative stress as well as increased generation of ROS, HNE, NO and RNS have been unequivocally detected in all clinical conditions and animal models of fibrogenic CLD; moreover, administration of antioxidant agents in animal models usually offers prevention ([35, 36] and references therein).

ROS and receptor-mediated signalling pathways

ROS have been shown to mediate a positive feedback on signal transduction elicited by, for example, PDGF, epidermal growth factor or nerve growth factor; this usually reinforces the receptor tyrosine kinases (RTKs) (Figure 17) and involves p21Ras and Rac, leading to activation of a subunit of non-phagocytic NOX, likely a gp91^phox analogue. TGFβ1, which operates by binding receptor serine/threonine kinase and involves Smads and Src kinases, as well as other relevant ligands in a scenario of chronic wound healing (as in CLDs), including interleukin (IL)-1, TNF, angiotensin II (Ang II), thrombin and insulin, have also been described to lead to activation of a NOX in non-phagocytic cells to generate O₂^•-, which will then spontaneously or enzymatically dismutate into H₂O₂. The reader may then envisage a scenario in which non-phagocytic cells involved in chronic wound healing receive signals from the extracellular milieu, such as those derived from peptide factors binding their receptors or extracellularly generated ROS or other oxidants, and then face an increase in intracellular ROS that affects signalling pathways by one or, more likely, two or more of the following mechanisms.

ROS and oxidative stress can activate defined transcription factors

Several transcription factors can be considered as redox sensitive but the two best characterized examples are NF-κB and AP-1. NF-κB is a transcription factor shown to respond to oxidative stress [106] and it is known to be involved in inflammatory reactions, in the control of cell growth and the balance between survival and apoptosis [107] and, possibly, necrotic cell death [108] (see below). NF-κB, a definition that, in mammalian cells, includes c-Rel, RelA (p65), RelB, NF-κB1/p50 and NF-κB2/p52 proteins, which all recognise DNA sequences called κB sites [107, 109, 110], is also involved in maintaining mitochondrial integrity and in regulating antioxidant activity [107, 110]. The redox-dependence of NF-κB relies on different mechanisms of activation that, depending on the specific target cell, may involve either an atypical phosphorylation of the Tyr42 residue of IκBα by the kinase Syk (thus, independently of IκB kinase (IKK)) or a more conventional H₂O₂-dependent activation through the classic IKK-dependent pathway [110], the latter being activated also by HOCl, singlet oxygen and peroxynitrite.

Pertinent to this review, a general model is emerging suggesting that all cytokines leading to NF-κB activation are likely to cause intracellular generation of ROS that are then responsible for IKK activation and IκBα degradation, with IL-1, TNF and LPS being the. best characterized examples (Figure 18). The concept here is again simple: ROS, produced intracellularly as a part of the response induced by inflammatory cytokines, contribute to reinforce the signal. Figure 19 oversimplifies the concepts described above (in the 'Principles of redox homeostasis' and 'Redox sensors and the basis of redox-dependent transcriptional regulation' sections) by proposing an intuitive model that relates the levels of ROS and oxidative stress to the overall response and even fate of the target cells.

Within the same scenario one can also easily include AP-1, a dimeric (homo- or heterodimer) transcription factor typically formed from c-Jun and c-Fos and involved in several physiological and pathophysiological processes. Activation of AP-1 occurs in the presence of low levels of ROS (mainly H₂O₂), IL-1, UV light, and γ-irradiation. Two mechanisms may lead to redox-dependent activation of AP-1: oxidative activation of JNKs that, in turn, phosphorylate Ser63 and Ser73 of the amino-terminal transactivation domain of c-Jun, a domain that is essential for functional activation [104]; and a mild shift in the redox state by different oxidants or ROS [14, 15].

A cautionary note has to be added: one should keep in mind that the DNA-binding activity of most transcription factors is redox sensitive in the opposite way. It has been shown that the binding of transcription factors to a DNA regulatory sequence requires reducing conditions because transcription factors must expose positively charged amino acid residues in their binding sites in order to be able to bind target DNA sequences (usually highly acidic and negatively charged). This introduces an apparent paradox: the binding site of a transcription factor presents redox-sensitive amino acids (cysteine, arginine) and, as is the case for NF-κB, oxidation of these critical residues may prevent its DNA-binding activity. This note is to underline that even in physiological conditions, the final response to redox changes relies on a delicate balance between pro-oxidant conditions needed to reinforce the signal and reducing conditions needed for the same signal (that is, the transcription factor) to be efficiently delivered in order to obtain the response. Pathophysiological conditions (Figures 9 and 19) can easily interfere with such a delicate balance by shifting redox homeostasis to a 'quasi-stable' but deregulated redox state.

Iron and its role in fibrogenic CLDs

Iron in its ferrous (Fe²⁺) or ferric (Fe³⁺) forms is critical for the life of all aerobic organisms, being included in haemoproteins like haemoglobin and myoglobin, metalloproteins, cytochromes and redox-dependent enzymes involved in oxygen and electron transport as well as in several other reactions, including oxygen sensing, NO sensing, DNA synthesis and transcriptional regulation. Major achievements in the understanding of hereditary hemocromatosis (HH) have disclosed in a detailed way [111, 112] crucial molecular and cellular mechanisms responsible for the control of iron homeostasis. Iron levels are carefully controlled in terms of adsorption, stores, plasma levels and transferrin saturation, whereas major physiological pathways for its excretion are lacking in higher organisms. The literature suggests that increased levels of hepatic iron (the liver being the most relevant site of storage) can significantly contribute to fibrogenic progression of a CLD.

Whatever the reason for increased hepatic iron levels, Figure 20 offers a simplified and 'iron-centric' view of how the metal may exacerbate oxidative stress and its consequences, which may range from redox-mediated cytotoxicity to ROS, radical and non-radical intermediate-related pro-inflammatory and pro-fibrogenic action (for more details, see [113]). Figure 20 (more details in [112–119]) also offers additional information detailing the role of excess iron in relation to HH, chronic HCV infection, NAFLD/NASH or alcoholic liver disease (ALD).

If one comes back to the central point (that is, how iron may contribute to fibrogenesis and CLD progression), several hypothesis have been proposed in which the pro-oxidant role of the metal ion remains prominent. The following concepts may be relevant: first, increased oxidative stress and lipid peroxidation have been detected in association with all major conditions of hepatic iron overload [35, 113]; second, in animal models evidence suggests that antioxidant supplementation is able to significantly prevent both iron-dependent chronic liver injury and excess ECM deposition [35, 120, 121]; and third, iron overload may induce cytotoxicity that primarily depends on the ability to generate oxidative stress and operates by inducing mainly mitochondrial damage (including damage to mitochondrial DNA) and destabilization of lysosomal membranes [122–124].

Another concept should be recalled: intracellular levels of iron are usually carefully controlled (as in hepatocytes or macrophages) by means of iron binding to cytoplasmic Iron regulatory proteins (IRP-1 and IRP-2) and the expression of genes involved in iron homeostasis through the binding of IRPs to Iron responsive element (IRE) sequences, including those for transferrin receptor (TfR) and ferritin (Ft). An uncontrolled rise in intracellular iron may lead to the formation of a low molecular weight pool of iron potentially able to convert O₂^•- and H₂O₂ into highly reactive •OH radicals or ferryl ions. IRPs may indeed represent a target for ROS and RNS, possibly as part of a more general scheme designed to protect the intracellular environment from oxidative stress [125]. IRP inactivation should, by down-regulating TfR expression and up-regulating Ft, decrease the intracellular labile iron pool, thus preventing amplification of iron-mediated oxidative damage. We do not know whether these regulatory mechanisms may be altered in CLDs, although it has been proposed that inactivation of IRP-2 (which is usually highly expressed in macrophages) by RNS may help to explain iron sequestration patterns of macrophages detected in tissues undergoing inflammation [125].

Copper is another transition metal acting as an excellent pro-oxidant catalyst that may contribute to enhance oxidative stress in Wilson's disease (WD), the human disease in which hepatic copper overload can occur. WD is an autosomal recessive disease caused by mutations in the gene encoding the copper-transporting P-type ATPase, ATP7B, required for copper biliary excretion [126, 127]. As for iron overload, copper overload has also been described to cause hepatic oxidative stress, leading to hepatocyte injury and subcellular damage to several structures [128], although other mechanisms, either oxidative or non-oxidative, may offer significant contributions [35, 36, 129].

Chronic ethanol consumption and metabolism: induction of oxidative stress and related events

Chronic ethanol consumption can lead to ALD, which encompasses a large spectrum of pathological liver changes, ranging from simple fatty liver with minimal injury to alcoholic steatohepatitis (ASH) and, in more advanced stages, fibrogenic progression to cirrhosis. Progression of ALD is now considered a multifactorial process involving nutritional, environmental and genetic factors [130], and ethanol consumption also represents one of the major host-related factors able to accelerate progression of fibrosis towards cirrhosis in chronic HCV patients [131, 132] and, possibly, in patients affected by CLDs with a different aetiology. The role of ROS and oxidative stress in the pathogenesis of ethanol-induced liver injury has been extensively investigated [133–136]. Here the following relevant concepts are recalled. First, experimental and clinical data indicate that oxidative stress and lipid peroxidation are involved, with antioxidants and free radical scavengers being able, at least in animal models, to afford prevention [135]. Second, ethanol-related ROS can be produced by the mitochondria respiratory chain, ethanol metabolizing (and ethanol-inducible) cytochrome P450 2E1 (CYP2E1) in hepatocytes, but not in HSCs [137], and NOXs of activated Kupffer cells or infiltrating neutrophils [133–135]; NO produced by NO-synthase of Kupffer cells and other RNS has also been shown to contribute to ethanol-dependent hepatic injury [135, 136]. The CYP2E1 isoform can also lead to generation of the ethanol-derived hydroxyethyl radical. Third, ethanol-induced oxidative stress is likely to contribute to liver steatosis found in alcoholics [138] by causing an impairment in either mitochondrial lipid oxidation [139] or lipoprotein secretions, the latter being related to enhanced degradation of ApoB100 [140] and/or oxidative alteration of lipoprotein glycosilation in Golgi apparatus [135]. Fourth, CYP-2E1 generated ROS and formation of protein adducts by lipid peroxidation products may affect proteasomal degradation leading to cytoplasmic aggregates of cytokeratins 8 and 18, leading to the formation of Mallory's bodies [141]. More features of ethanol-induced oxidative stress are presented below (see the sections 'Ethanol-related redox mechanisms leading to mitochondrial damage and hepatocyte apoptosis', 'Redox mechanisms and chronic inflammatory response in CLDs', 'Redox mechanisms in liver fibrogenesis: pro-fibrogenic cells as a functional target' and 'Redox mechanisms in immune reactions associated with CLDs: fuel for chronic inflammation and fibrogenic progression').

ROS and oxidative stress-related reactive intermediates in NAFLD and NASH: their generation and role in causing steatosis

The term NAFLD refers to a wide spectrum of disorders having in common hepatic steatosis as a hallmark but also encompassing NASH, advanced liver fibrosis and cirrhosis. NAFLD is actually recognized as a major cause of liver-related morbidity and mortality, with a very high prevalence in Europe, USA and, more generally, in western countries (reviewed in [142–145]).

The traditional 'two-hit theory' proposed by Day and James [146] around ten years ago, which is still widely accepted [142–145], is summarized in the scheme presented in Figure 21, which also provides more details on mechanisms leading to increased circulating and hepatic levels of free fatty acids (FFAs), the ultimate cause of steatosis. There is consensus concerning the fact that long-term injury from hepatocyte triglyceride storage is responsible for the development of oxidative stress and, thus, significant intracellular generation of related reactive intermediates, with oxidative stress (detected in all clinical and experimental conditions of NAFLD/NASH [147–151]) representing the 'second hit', potentially favouring NAFLD progression. Along these lines, impairment of mitochondrial β-oxidation can lead to accumulation of FFAs in hepatocytes and FFAs are substrates (ω-oxidation) and inducers of cytochrome P450 isoforms CYP2E1 and CYP4A. Both isoforms (overexpressed in human and animal NASH) can generate ROS that, in turn, can elicit lipid peroxidation, a common hallmark of NASH [152–154]. In addition, ROS may also be generated as a consequence of increased FFA oxidation in peroxisomes (β-oxidation) by Acyl-CoA oxidase.

Other mitochondrial dysfunctions occur in NASH patients, some possibly due to up-regulation of uncoupling protein-2 (UCP-2) or representing the consequence of an oxidative stress-dependent derangement of mitochondrial membranes and/or the respiratory chain that may result in an increased release of ROS from mitochondria. The role of UCP-2 is still controversial and has recently been critically reviewed [155].

Oxidative stress and related reactive intermediates may contribute, as proposed for ethanol, to the genesis of steatosis by negatively affecting secretion of lipoproteins (either by enhancing degradation of Apo-B100 or by affecting lipoprotein glycosylation in the Golgi apparatus) [135, 140] or even by interfering with the regulation of lipid synthesis by the sterol regulatory element binding protein 1 (SREBP-1) or the peroxisome proliferator-activated receptor α (PPAR- α) [156]. Moreover, ROS have been proposed to contribute to insulin resistance (IR) itself: activation of stress-activated protein kinases by ROS can lead to impairment of the correct transduction of insulin-mediated signals through the induction of serine and threonine phosphorylation of IRS-1 (insulin receptor substrate-1) and the concomitant down-regulation of IRS-1 tyrosine phosphorylation [157, 158]. This has been confirmed by studies performed in hepatocytes overexpressing CYP2E1 [159].

Finally, the pivotal role of ROS in NAFLD progression has been shown by Xu et al. [160], who, in an elegant study (commentary in [161]), described the key role of the Nrf-1 gene, which is known to be involved in mediating activation of the oxidative stress-response. When using a mouse model of selective hepatic deletion of the Nrf-1 gene, the liver of these animals developed progressively all the characteristic features found in the progression of human NAFLD, including steatosis, apoptosis, necrosis, hepatitis, fibrosis and even liver cancer.

Chronic HCV infection and oxidative stress

The occurrence of oxidative stress is a common finding in human patients affected by chronic HCV infection (reviewed in [162]) and is likely to rely mostly on the actual conditions of persisting liver injury and chronic inflammation. However, experimental studies have proposed that HCV core protein may induce mitochondrial injury in hepatocytes, leading to increased generation of ROS [163, 164]. Mice overexpressing HCV core protein showed alterations of redox homeostasis in the absence of significant inflammation and were more susceptible to the hepatotoxin CCl₄ [163]. Moreover, overexpression of viral proteins in these transgenic mice was found to be associated with the development of steatosis and hepatocellular carcinoma, two common features of chronic HCV infection in humans [165].

Other researchers have provided similar evidence for the HCV non-structural protein 5A (NS5A), which has been shown to associate with the membranes of endoplasmic reticulum (ER) and increase generation of ROS and activation of NF-κB and STAT-3 [166] as well as of JNKs and p38 MAPK [167]. NS5A-mediated activation of NF-κB seems to operate through tyrosine phosphorylation of IkBα and its degradation by calpain protease [168]. Thus, HCV by itself may contribute to elicit a state of oxidative and nitrosative stress in affected patients and the following major concepts, as recently reviewed [169], should be outlined. First, only some HCV-related proteins, such as NS5A, NS3 and HCV core protein, are likely to play roles in mediating redox perturbations, which can include increased generation of both ROS and RNS, iNOS and COX-2 (NS5A and C). Second, significant alterations of redox state may help to explain synergistic effects of HCV and alcohol on the progression of disease. Lastly, ROS and ethanol consumption may influence HCV replication and affect the outcome of interferon therapy.

Recently, another concept has been introduced by a study performed on transgenic mice expressing the HCV polyprotein: HCV-induced excess generation of ROS causes down-regulation of hepcidin synthesis through inhibition of the DNA binding activity of C/EBPα [170], an event able to increase iron transport from duodenum and iron release from macrophages, which lead potentially to iron overload and related consequences.

Chronic cholestatic diseases and oxidative stress

The pattern of fibrosis in diseases of the biliary tract is peculiar and involves either significant alterations in the interactions between cholangiocytes and mesenchymal cells as well as generation of ROS and occurrence of oxidative stress. In adults and children, chronic cholestasis is usually the consequence of cholangiopathies due to autoimmunity, infectious or toxic agents, ischemia or genetically transmitted defects. All these cholangiopathies share features such as cholestasis and cholangiocyte loss associated with cholangiocyte proliferation and a variable degree of portal and periportal inflammation and fibrosis [171, 172]; relevant is the 'cross-talk' between cholangiocytes, portal MFs and/or HSCs, first demonstrated for PDGF-BB and PDGF-β R [173], with cholangiocytes being able to secrete IL-6, TNF, IL-8 and MCP-1 as well as PDGF, ET-1, TGFβ2 and CTGF (reviewed in [171]).

Oxidative stress and lipid peroxidation have been detected in the most relevant clinical conditions, including primary biliary cirrhosis and the bile duct-ligation model in the rat [174–178], with antioxidant providing a significant degree of protection in animal models (reviewed in [35, 36]). Generation of ROS, induction of oxidative stress and lipid peroxidation may represent the consequences of activation of inflammatory cells, as in the bile duct ligation (BDL) model [174], but ROS generation may even follow a direct effect of hydrophobic and cytotoxic bile acids on hepatocyte's mitochondria, thus mediating hepatocyte necrotic cell death and/or apoptosis [128, 179–182]. Recently, it has been suggested that intracellular generation of ROS may also mediate cholangiocyte apoptosis found in primary biliary cirrhosis [178].

Oxidative stress and genetic polymorphisms

Genetic polymorphisms may have a significant role in fibrotic progression of CLDs, as suggested by the broad spectrum of manifestations or responses that individual patients offer to the same aetiological agent or condition. Bataller and coworkers [183] have delineated a list of candidate genes involved and, not surprisingly, most, if not all, of the 'suspected' polymorphisms concern factors or enzymes that are known to be directly or indirectly either redox-sensitive or involved in the generation of ROS or other relevant reactive intermediates. The actual list includes genes encoding cytochrome P450 isoform CYP2E1, alcohol-dehydrogenases (ADHs), and Mn-SOD, as well as a number of cytokines, including TGFβ1 and TNFα.

Oxidative stress as an event able to induce cell death

Persisting liver injury and hepatocyte loss are common in CLDs and oxidative stress is likely to play a relevant role in inducing both necrotic as well as apoptotic cell death [136, 184–187]. Severe oxidative stress, able to significantly damage any relevant biological macromolecule and cellular structure, is an obvious candidate to induce necrosis and may represent the outcome of severe inflammatory response following acute liver injury, with activated Kupffer cells, neutrophils or recruited mononuclear cells from peripheral blood being the 'effectors' of increased generation of reactive species. High levels of intracellular oxidative stress may also be reached within hepatocytes damaged by specific toxins (for example, CCl₄ and acetaminophen) or aetiologies (for example, high levels of transition metal ions) or because of individual differences, including induction of peculiar cytochrome P450 isoforms (such as CYP2E1 in ASH/ALD or NASH), antioxidant status or even genetic polymorphisms.

In the scenario of a chronic inflammatory and fibrogenic disease, the first point to be discussed is the following: what do we mean, in terms of concentrations, by 'high levels of oxidative stress'? The literature offers the following considerations: reliable analytical (that is, quantitative) data on steady state concentrations of ROS as well as of other critical reactive intermediates are lacking for human liver of patients affected by a CLD; and the best data come from analyses performed on livers of animals undergoing experimental models of acute liver injury. In the liver of mice treated with acetaminophen, a model of severe and oxidative stress-dependent acute liver injury and failure, sophisticated techniques have detected a total concentration of approx 0.25 μM ROS, with H₂O₂ detected at levels of 0.15 μM [188]; in the CCl₄ model of oxidative stress-related acute liver injury, intrahepatic levels of HNE reached maximal values around 10 μM [189].

Data on acute liver injury should imply that during CLDs tissue levels of ROS and other mediators are likely to be lower; this has been shown for HNE levels, which ranged from 1–3 μM in chronic CCl₄ treatment and BDL (reviewed in [35, 36]). Although several researchers have the feeling that in some conditions (that is, the environment of biological membranes within the cell, the site of active inflammation, the concomitance of more 'pro-oxidant events') ROS may locally reach levels higher then those reported in [188], these data suggest caution when interpreting results obtained in vitro by exposing either suspensions of freshly isolated cells or cultured cells to unrealistic concentrations of ROS or HNE, such as in the range 0.1–1.0 mM. The problem of the 'steady state' concentration of reactive intermediates is not academic: reactive intermediates of oxidative stress can induce necrotic cell death, caspase-dependent apoptosis [184–186] or other intermediate forms, including apoptosis-like or necrosis-like cell death [96], even in the presence of the same pro-oxidant agent or condition [179–182, 190, 191]. This is relevant in the parenchymal chronic 'battlefield': whatever the aetiology of the CLD, signs of both necrosis and apoptosis can be found in the same section, in association with the other events of the chronic scenario (inflammation, fibrogenesis, angiogenesis, and so on) [185, 186]; moreover, apoptosis can indeed act in a pro-inflammatory and pro-fibrogenic way, thus sustaining progression of the CLD [192].

Some years ago, Kaplowitz [184] proposed that necrotic cell death may occur in cells exposed to very high levels of oxidative stress and be able to irreversibly damage mitochondria or to inactivate executioner caspases. This indeed may happen, as shown by in vitro experiments, in hepatocytes [19] or human HSC/MFs [189, 191] exposed to very high concentrations of HNE (50 μM or more). However, it is not really easy sometimes to identify which mode of cell death may predominate in different conditions of liver disease. Recently, Mahli and coworkers [186] have proposed that controversies about this specific feature may be solved by recognizing that apoptosis and necrosis frequently represent alternative outcomes of the same pathways leading to cell death. For example, even in conditions able to induce caspase-dependent apoptosis (activation of either intrinsic pathways or of death receptor-related pathways), severe mitochondrial changes with membrane depolarization and uncoupling of oxidative phosphorylation may result in ATP depletion and, thus, in the blocking of caspase activation and classic apoptosis.

Another relevant question is, is it really necessary to reach high levels of oxidative stress to induce irreversible cell death, for example, in hepatocytes? In theory, in the scenario of a CLD, oxidative stress should be quantitatively mild to modest on a tissue basis, but it is conceivable that additive or synergic factors may overlap, leading to higher levels that may locally (for example, in the site of an inflammatory flare) or, even more likely, at the level of the single cell 'make' or 'mark' the difference for one or more cells to survive or die. Two concepts should be stressed (interested readers can find more details in [107, 136, 186, 187, 190, 192–195]. First, necrosis or caspase-independent cell death is no longer seen as an accidental and uncontrolled form of cell death but is beginning to be envisaged as the result of crosstalk between several biochemical and molecular events, including an interplay between crucial signalling pathways [193]. ROS and increased levels of intracellular calcium are believed to be the main players in this scenario and necrosis, in vivo and in vitro, may function as a sort of back-up program when caspase activation is impaired. Second, both caspase-independent cell death and apoptosis, particularly when related to the engagement of death receptors (DRs) or Toll-like receptors (TLR) by their respective ligands, critically involves the kinase RIP (Receptor interacting protein), which is currently seen as one of the crucial, and redox sensitive, cellular crossroads determining whether cells live or die (Figure 22).

In the next sections a number of concepts and mechanisms involved in ROS-dependent and 'aetiology-related' forms of hepatocyte death are outlined to possible serve as a paradigm to understand the extremely complex scenarios occurring in CLDs.

ROS and mitochondria in cell death

Alterations of mitochondria have a role in different types of either caspase-dependent or caspase-independent cell death [196] and are strictly associated with ROS. If mitochondrial integrity is deranged, this will be associated with the dissipation of the mitochondrial inner transmembrane potential (ΔΨ_m), leading to mitochondrial outer membrane permeabilization and release of pro-apoptotic proteins such as cytochrome c, Smac, Diablo and AIF, leading then to apoptosis. Increased permeabilization of the mitochondrial outer membrane can also lead to increased intracellular ROS generation as a consequence of damage to the mitochondrial electron transport chain. ROS may also increase permeabilization of mitochondrial outer membrane by altering thiol groups of ANT (Adenine nucleotide translocase) or VDAC (Voltage-dependent anion channel), the latter being required for superoxide anion efflux from mitochondria. The overall message then is quite simple: mitochondria can represent not only a source of ROS but also a target for their action in relation to cell death.

Death receptor activation, ROS, NF-κB and sustained activation of JNK: to die or not to die, that is the question

Probably the best detailed example of ROS involvement in cell death is related to the activation of death receptors, the mitochondria-related generation of ROS and the subsequent sustained activation of JNK. Activation of pathways leading to increased mitochondrial ROS generation and related to cell death (apoptotic or not) may be elicited by TNF, mainly by acting on the TNF receptor TNFRI, but a significant role for ROS has also been described in the activation of Fas as well as other death receptors, such as DR-4 and DR-5 [107]. The reader can then easily understand the intrinsic relevance of what we are going to describe by thinking about the complex scenario of CLDs of different aetiology, where chronic inflammation and ligand-mediated activation of death receptors is a very common event (reviewed in [186]). TNF interaction with its type I receptor will generate a complex sequence of events (more detail is given in Figure 23) that are designed to lead either to survival or cell death, with ROS forcing them towards the latter [197–202]. Indeed, TNF-TNFRI interaction may serve as a paradigm for describing the role of ROS (here generated mostly by mitochondria) since (see below) several conditions leading to the increased generation of ROS will converge on sustained JNK activation and its consequences. Interestingly, TNF (Figure 23) may even lead to ROS-dependent and JNK-mediated necrotic cell death through the activation of Nox1: however, this has been described in fibroblasts and, at present, we do not know whether it may apply to hepatocytes or other liver cell populations [195, 202].

ROS-dependent sustained activation of JNK: a common step in oxidative stress-dependent cell death

Mitochondria-derived ROS are implicated in TNF-dependent apoptotic and non-apoptotic cell death by inducing a ROS-dependent sustained activation of JNK (Figure 24), the latter being an event commonly seen in other conditions resulting in increased generation of ROS [107, 186, 187, 193, 199, 203, 204]. Several mechanisms have been described to explain ROS-mediated activation of JNK but three should be considered. The group of Karin has described for TNF-mediated cell death that ROS are able to inhibit JNK phosphatases, which are the JNK-inactivating enzymes [203, 204]. Alternatively, it has been described that ROS may act by removing a thioredoxin-dependent inhibitory action on the JNK upstream kinase ASK1 (MAP3K) [104, 205, 206], although others [203] were not able to confirm ASK1 involvement in TNF-mediated cell death. ROS may even operate by inducing a PARP-1 over-activation resulting in depletion of NAD+ and ATP as well as in a downstream involvement of TRAF2/RIP1 mediating sustained JNK activation and cell death, but it is still unclear how TRAF2 and RIP1 may be involved following PARP-1 activation [207, 208].

Whatever the mechanism, there is no doubt that ROS-dependent sustained JNK activation can promote cell death, being effective in inducing mitochondrial outer membrane permeabilization; although there is still some controversy on the molecular 'targets' of ROS-activated JNK, a number of hypotheses [187, 194, 199, 209] have been proposed to explain this JNK-mediated event (summarized in Figure 24). The reader should also remember that NF-κB will act to prevent cell death by a number of mechanisms, some of them potentially affecting ROS-mediated sustained JNK activation, such as specifically the reported up-regulation of SOD-2 (that is, the mitochondrial SOD isoform), in TNF-dependent cell death [210]. According to the protective role of NF-κB, it has been shown that NF-κB inhibition can indeed sensitize hepatocytes to TNF-induced apoptosis by means of JNK sustained activation [211].

Free fatty acids, endoplasmic reticulum stress, oxidative stress and cell death: hepatocyte injury in NAFLD and other CLDs

As elegantly reviewed by Parekh and Anania [143], liver injury in NAFLD can be considered essentially as the consequence of increased hepatocyte stores of FFAs. Overall, the most relevant mechanisms leading to hepatocyte injury in these metabolically altered conditions involve an excess of FFAs: directly inducing hepatocyte apoptosis and stimulating production of TNF, which is considered in this context as an adipocytokine; increasing Fas ligand binding to Fas (CD-95) receptor in steatotic hepatocytes, leading to apoptosis; leading to impaired mitochondrial or peroxisomal β-oxidation of FFAs accumulating in hepatocytes – as previously described (see the section 'ROS and oxidative stress-related reactive intermediates in NAFLD and NASH: their generation and the role in causing steatosis'), this eventually leads to generation of ROS and lipid peroxidation products, mainly HNE, which in turn may cause cell injury and death; inducing ER stress and the so-called 'unfolded protein response' (UPR), which is potentially able to result in the induction of a form of caspase-dependent cell death involving mitochondria.

The first three mechanisms are all intrinsically related to ROS generation, as already described. If the interplay between FFAs, ER stress and oxidative stress is considered, the scenario can be summarized as follows (more details in [212, 213]). In normal conditions the ER is the site devoted to protein entry into the secretory pathway. The ER environment involves a protein-folding machinery that is based on protein chaperones, proteins designed to catalyze folding and proteins and systems able to sense and detect unfolded or misfolded proteins. The latter systems prevent secretion of misfolded proteins and direct them to be degraded; UPR signalling pathways, specifically designed to avoid accumulation of unfolded proteins in the ER lumen, are essential for adaptation to altered homeostatic conditions, including redox changes. In this context, two concepts should be underlined: first, the ER is a unique oxygen folding environment in which disulfide bonds can be formed – unfortunately, even in normal conditions protein folding can lead to ROS generation, likely as by-products of protein oxidation occurring in the ER; and second, oxidative stress, whatever the source, can affect ER and activate UPR, the latter initially operating as an adaptative mechanism to preserve cell function and favour cell survival. Several environmental or metabolic insults may then result in an alteration of protein folding within the ER [212, 213], including redox changes, depletion of calcium, energy deprivation, elevated protein traffic within the ER compartment, altered post-translational modifications and impairment of glycosylation and Golgi processing and, as also described in the case of NAFLD, excess storage of FFAs (reviewed in [143]). Indeed, signs of ER stress have been detected in hepatocytes of mice fed high fat diets or ob/ob mice (mice deficient in leptin that are obese and diabetics), including phosphorylation of pancreatic ER kinase (PERK), translation initiation factor 2 (TRAF2) and JNK. Moreover, ER stress has been shown to exacerbate hepatocyte insulin resistance [214]. Figure 25 summarizes the most relevant features of ER stress, following accumulation of FFAs and also involving oxidative stress and ROS, the latter once again significantly contributing to multiple pathways leading to hepatocyte death (for more details, see [212, 213]).

ER stress-induced apoptosis has also been implicated in chronic hepatitis C and ALD [186]. With respect to HCV, several studies have shown that HCV-related proteins, including either core proteins as well as E1 and E2 proteins of the envelope, are able to induce overexpression of C/EBP homologous protein (CHOP) via the UPR response, ER depletion of calcium, and apoptosis in liver cells (HCV replicon cells) [215, 216]. Moreover, ER stress and apoptosis were also found in the livers of HCV core transgenic mice [215]. More recently, it has been reported that activation of the Gadd153 gene by HCV in HCV replicon cells is able to sensitizes these cells to oxidant injury [217] and that both HCV and, interestingly, HBV proteins are able to induce ER stress and to up-regulate protein phosphatase 2A (PP2A) through activation of CREB [218]. In the case of ALD, ER stress involvement has been attributed to hyper-homocysteinemia, since betaine treatment in a mouse model of ALD can promote removal of homocysteine and prevent not only ER stress but also liver steatosis as well as induction of apoptosis. Excellent reviews dedicated to ER stress and HCV- or alcohol-related liver injury published recently by Kaplowitz and coworkers [219–221] also take into consideration the role of ROS.

Ethanol-related redox mechanisms leading to mitochondrial damage and hepatocyte apoptosis

This section and Figure 26 offer a number of additional concepts (see also the section 'Chronic ethanol consumption and metabolism: induction of oxidative stress and related events' above) that are mostly focussed on the role of ethanol and ROS-dependent injury of mitochondria in determining hepatocyte cell death [133–136, 222]. Chronic ethanol consumption can promote intramitochondrial formation of ROS and decrease/deplete the mitochondrial content of GSH, thus making mitochondria even more susceptible to oxidative injury [223, 224]. Lipid peroxidation has a major role in the impairment of mitochondrial oxidative phosphorylation [223]. Mitochondrial DNA is oxidatively altered in animals treated with ethanol [225] and this may account for the high levels of mitochondrial DNA deletions found in the liver of alcoholic patients [226] and the ethanol-related impairment of hepatic respiratory activity (presumably by affecting the synthesis of subunits of the electron transport chain encoded by mitochondrial DNA). Intramitochondrial generation of ROS and oxidative stress can induce the collapse of mitochondrial membrane potential and promote mitochondria permeability transition (MPT), possibly by favouring Bax translocation to mitochondria [225, 227]; induction of MPT can, alternatively, lead to mitochondrial swelling and then to necrotic cell death as well as to cytochrome c release and induction of caspase-dependent apoptosis [196, 228–230]. The hepatic inflammatory reaction may have a role in ethanol hepatotoxicity, with a major role attributed to activated Kupffer cells (reviewed in [133, 135, 231]), in which ethanol may 'switch on' a ROS- and NF-κB-related cycle leading to an amplified release of TNF; indeed, hepatocytes undergoing ethanol-induced oxidative stress may be more sensitive to the pro-apoptotic action of TNF [232], including increased sensitivity to TNF-related MTP, which may also depend on ROS- or HNE-mediated changes to ERK1/2- or PI3K-related survival signals [233, 234].

Mitochondria, nitric oxide, RNS and cell death

NO and related RNS are additional key players in the chronic inflammatory settings that characterize chronic wound healing and, more specifically, fibrogenic CLDs. NO and RNS theoretically can promote or prevent apoptotic cell death by interfering with either mitochondria-dependent or mitochondria-independent signalling pathways [235–238]. The most relevant feature is represented by the cellular redox state, with major determinants being the steady state concentration of ROS (mainly O₂^•-) and the rate between ROS and NO generation; as a rule, NO can act in a preventive way when ROS levels are low and vice versa [235–238].

Cytoprotective action is likely to rely on NO-dependent inhibition of ^•OH generation in the presence of an iron-catalysed Fenton reaction or on inhibition of propagation of lipid peroxidation (NO being able to react with lipid alkoxyl or lipid hydroperoxyl radicals). In the presence of higher levels of ROS, the right NO/O₂^•- ratio or the right levels of O₂, NO may lead to the generation of highly reactive RNS, such as N₂O₃ or ONOO^-, at levels that are able to induce more aggressive oxidation, nitrosation/S-nitrosation and nitration of different biological macromolecules, eventually leading to either necrotic or apoptotic cell death. Figure 5 summarizes these two alternative scenarios (that is, preventive versus injurious) that can be elicited by NO and RNS, with emphasis on the major molecular mechanisms that may be involved (more details in [235–243]). Here one should just recall that if liver injury is concerned, NO levels (and then those of RNS) are usually increased in CLDs as a consequence of up-regulation of iNOS in hepatocytes, endothelial cells, Kupffer cells and, possibly, HSC/MFs [235–238, 241, 242]. This is likely to be significantly related to levels of TNF in the chronic inflammatory environment as well as to increased translocation of gut-derived endotoxins to the portal circulation, as clearly described in the case of chronic ethanol ingestion and the subsequent interaction of endotoxins with CD14 and TLR-4 in Kupffer cells [243].

Oxidative stress and induction of cell death in HSC/MFs

In the past decade an emerging issue in hepatology has been that liver fibrosis and even cirrhosis may be potentially reversible. Recovery from either acute or chronic liver injury in animal models is characterized by apoptosis of HSC/MFs, reduction of TIMP (tissue inhibitor of metalloproteinases) levels and progressive degradation of fibrillar fibrotic ECM. In this scenario, the sensitivity of HSCs and HSC/MFs to pro-apoptotic stimuli has been investigated to gain basic knowledge for a putative cell targeted antifibrotic therapy [79–82, 91, 94, 95, 244]. Pertinent to this review, HSC/MFs exposed to oxidative stress can undergo caspase-independent cell death but this usually requires very high concentrations of either H₂O₂, O₂^•- or HNE, which are hardly comparable with levels observed in vivo [94, 189, 246]. This has been substantiated mainly for human cells, which, perhaps, may be more resistant than activated rat or murine cells, possibily because human HSC/MFs when activated overexpress Bcl-2 and other survival pathway proteins [94].

The overall message is that HSC/MFs are likely to survive the conditions of oxidative stress usually operating in CLDs, and rather (see below) are more likely to sustain their pro-inflammatory and pro-fibrogenic responses.

Functional responses of inflammatory cells to oxidative stress-related intermediates

Apart from the NF-κB-related increased expression of inflammatory cytokines and chemokines [15, 106, 107], ROS and RNS may act as signalling intermediates by activating tyrosine kinases and inhibiting tyrosine phosphatases, resulting in an enhancement of tyrosine phosphorylation events known to regulate anti-microbial and host defence functions in leukocytes [251–253]. The following concepts should be underlined. First, ROS are likely to be involved in the process of phagocytosis, possibly by leading to amplification of the stimulating signal that follows engagement of Fc receptors on the surface of phagocytic cells, as reported for neutrophils, where ROS seem able to increase either the cross-linking of the FcãRIIIb [254], as well as by contributing, in neutrophils and macrophages, to amplification of the signal by modulating the activity of the tyrosine kinase Syk, a FcãR downstream signalling element [255]. Second, ROS-mediated inhibitory modulation of the activity of PTPs (for example, CD45, SHP-1, HePTP) has been shown to regulate signalling events involved in the activation of T lymphocytes [256, 257]. In addition, CD45 has long been shown to be relevant for LTB4- and C5a-induced chemotaxis and low affinity FcãR signalling in neutrophils [258, 259]: since ROS have been shown to be potent inhibitors of CD45 [260], they may interfere with physiological responses to these chemoattractants. Third, ROS may have a role in the apoptosis-related removal of leukocytes during inflammatory responses, as clearly shown for neutrophils and other cells, again by involving tyrosine phosphorylation, CD45 and SHP-1 [261–264], but also through the NOX-dependent, Lyn-mediated activation of the inositol phosphatase SHIP [265]. Similarly, peroxynitrite has also been shown to enhance apoptosis in leukocytes [266, 267].

An additional concept is that intermediates able to cross the leukocyte plasma membrane may affect, in a paracrine way, the behaviour of surrounding cells, as shown originally for endothelial cells [268, 269] and now accepted for many other cells with roles in chronic pro-inflammatory, angiogenic and fibrogenic environments. Accordingly, the lipid soluble aldehyde HNE, at levels compatible with those described in CLDs, up-regulates the expression of the pro-fibrogenic cytokine TGFβ1 in both rat Kupffer cells and human monocyte/macrophage cells [270]. These data may help to explain the scenario observed when using the in vivo experimental CCl₄-dependent chronic model: administration of α-tocopherol to rats undergoing this protocol resulted not only in a reduction of oxidative stress, lipid peroxidation and HNE, but also in decreased synthesis of both collagen type I and TGF β1 [271, 272]. Moreover, HNE and other HAKs have been reported to stimulate leukocyte chemotaxis at very low concentrations (0.1 μM; reviewed in [21, 35, 55]), suggesting that α-tocopherol and other chain-breaking antioxidants may prevent experimental liver fibrosis (reviewed in [21, 35, 36, 55]) by either preventing or inhibiting selected phenotypic responses of activated MF-like cells (see the section 'Redox mechanisms in liver fibrogenesis: pro-fibrogenic cells as a functional target' below) or, at least in part, by inhibiting leukocyte recruitment due to HNE or HAKs. Indeed, both ROS and HNE have been shown to up-regulate MCP-1 expression in vivo and in vitro, and then to sustain recruitment/activation of monocytes/macrophages and Kupffer cells as well as to attract HSC/MFs [76, 273, 274].

Pro-inflammatory response of activated HSC/MFs to ROS and HNE: the strange case of MCP-1

A peculiar example of interplay between the generation of reactive intermediates from oxidative stress, inflammatory response and fibrogenesis is represented by the case of MCP-1 (CCL2), which can recruit and activate monocytes and T lymphocytes and plays a major role in the formation and maintenance of the inflammatory infiltrate in different pathological conditions [275–281]. MCP-1, which is overexpressd in human CLDs and experimental models of liver injury [273, 282, 283], can be synthesised by activated macrophages and Kupffer cells as well as by HSC/MFs and biliary epithelial cells [283, 284]. With regard to HSC/MFs, MCP-1 expression is stimulated by pro-inflammatory cytokines, such as IL-1 and TNF [282, 283], thrombin [285], engagement of integrin receptors [285] and both ROS and HNE [273, 286]. Indeed, as for TNF or other chemokines such as IL-8 and RANTES, MCP-1 depends on the activation of the redox sensitive transcription factors AP-1 and NF-κB [287]. In human HSC/MFs, ROS can stimulate MCP-1 expression through involvement of NF-κB, whereas HNE does not involve NF-κB and more likely operates through an AP-1-related mechanism [288], possibly involving activation of PKC [282], as shown for the PKCβ isoform in monocyte/macrophage cell lines [289].

Finally, data on human and rat fibrotic livers indicate a direct correlation between oxidative stress, hepatic levels of MCP-1 and the number of monocytes infiltrating the injured liver [273, 284]. Moreover, MCP-1 can significantly stimulate chemotaxis of human HSC/MFs, another putative redox-sensitive pro-fibrogenic feature [290].

Pro-fibrogenic action of oxidative stress revealed by antifibrotic action of antioxidant molecules

The hypothesis of a causative involvement of oxidative stress in fibrogenesis relies on an impressive number of experimental studies leading to the same final scenario: antioxidant supplementation is able to significantly prevent fibrotic progression in animal models of CLDs by reducing the extent of oxidative stress and/or lipid peroxidation (reviewed in [35, 36], with comments in the section 'Antioxidants as antifibrotic therapeutics for CLDs?' below). The first antioxidants used were silymarin [291, 292], α-tocopherol [271, 272], silybin [174] and S-adenosylmethionine [293], and most laboratories were able to describe a temporal sequence of events, suggesting overall that oxidative stress and lipid peroxidation precede or are concomitant with HSC activation and collagen deposition [294–297]. Similarly, experimental studies have also reported that NO was able to prevent both lipid peroxidation and collagen deposition [298, 299].

HNE and ROS can up-regulate pro-collagen type I expression in HSC/MFs: a puzzle that in the end makes sense

ROS and HNE affect other phenotypic responses of HSC/MFs: the specific mediator can make the difference

ROS, but not HNE, mediate proliferation of HSC/MFs

Activation of HSCs as well as proliferation of HSC/MFs have been suggested to rely on the activation of NF-κB as well as on the expression of the c-myb proto-oncogene (reviewed in [323]). Indeed, when rat HSC/MFs are co-cultured with HepG2 cells overexpressing CYP2E1 as a source of ROS, they start to express α smooth muscle actin (α-SMA; a marker of MF-differentiation) and to proliferate [324]. These effects have been prevented by using antioxidants or inhibitors of the Na⁺/K⁺ exchanger [323, 325–328]. It has been suggested that the mitogenic action of ROS may rely on a crucial cysteine residue in Raf-1, MEK and Erk signalling elements; indeed, treatment of HSC/MFs with N-acetylcysteine is followed by activation of Erk, phosphorylation of the transcription factor Sp1 and up-regulation of p21Cip1/WAF1 expression, eventually leading to cell cycle arrest in G1 phase [329]. More recently, Adachi et al. [330] have proposed that the mitogenic action of PDGF-BB (see also Figure 28), and possibly also its chemotactic activity, may rely on ROS generation through involvement of NOX, leading to activation of MAPKs, particularly p38MAPK. In these experiments, performed on a human immortalised line of HSCs and on murine HSC/MFs, DPI was able to prevent the PDGF-BB-dependent proliferative response, which was restored by adding H₂O₂ together with PDGF-BB. Ang II has also been shown to up-regulate proliferation of HSC/MFs through involvement of NOX and ROS [318].

Results obtained with HNE as well as with other HAKs of different chain length substantially differ from those obtained with ROS: HNE and HAKs do not elicit proliferation of human HSC/MFs when employed at pro-fibrogenic doses (1–5 μM) [245, 288, 300, 331] but rather, when used at a pro-fibrogenic dose (1 μM), they induce a block in DNA synthesis by selectively inhibiting PDGF-β receptor intrinsic tyrosine kinase activity and downstream signalling pathways [288, 331]. This peculiar effect of HNE and HAKs is transient, with sensitivity to PDGF-BB being recovered within 48 hours and associated with increased expression of PDGF-β receptor subunits. Interestingly, a similar block in PDGF-dependent signalling and proliferation has been described in human cells when exposed to very high levels of ROS, either H₂O₂ or superoxide anion [246, 331].

Such a discrepancy between the effects reported for ROS and HNE is likely to depend on the different mechanisms of action: although Uchida and coworkers [23, 332] have proposed that, in some cells, HNE may result in mitochondrial damage and subsequent release of ROS (at high concentrations, however, such as 20 μM or more) or by trapping thiols, HNE is more likely to act as a non-oxidant agent by forming adducts to proteins by means of nucleophilic Michael type reactions [9, 19, 21], as shown in the case of JNK activation [288]. Moreover, HNE does not activate NF-κB [21, 23] in human HSC/MFs [288] as ROS do, and can even inhibit c-myb [21], which has been involved in ROS-mediated activation of proliferation [323]. HNE exerts its pro-fibrogenic action only on fully activated human HSC/MFs [21, 189] and, even more relevant, does not apparently act as an activating factor for rat HSCs early in primary culture [333]. This differs from what was suggested for ROS [75, 77, 80, 81] and is likely to occur because quiescent HSCs can remove H₂O₂ less efficiently than fully activated cells [310, 312], whereas HSC/MFs are more sensitive to HNE because they lack isoforms of GSH-S-transferase and aldehyde dehydrogenase able to remove or inactivate HNE [288, 334]. Finally, HNE seems much more selective in its activity towards HSC/MFs, since it causes the up-regulation of only a limited number of genes, including those encoding collagen type I, TGFβ1, and TIMP-1 [21, 189].

Oxidative stress-mediated immune responses as fuel for inflammation and fibrogenesis

In CLDs, with ALD patients being a reference for the amount of evidence available, oxidized or modified epitopes may induce humoral as well as cell-mediated immune responses able to significantly contribute to the maintenance of hepatic inflammation in the natural history of ALD. This is crucial in the specific ethanol-related scenario that is often dominated by increased translocation of LPS and endotoxins from the gut to the portal circulation (and the consequent effect on Kupffer cells) and with regard to the fact that ethanol-derived oxidative stress can increase the hepatotoxic action of TNF: indeed, in either chronically ethanol fed rats or alcoholics the presence of high titres of IgG against antigens modified by lipid peroxidation or oxidative stress correlate well with increased TNF production and, quite reasonably, the severity of inflammatory infiltrate [249].

One can envisage a scenario in which, in the presence of continuous antigen stimulation, cytokines released by lymphocytes can actively sustain Kupffer cell-mediated release of cytokines and chemokines as well as of ROS and NO, which, in turn, may contribute to the perpetuation of oxidative injury, the inflammatory response and fibrogenesis. Interestingly, as many as 60–80% of ALD patients with an advanced stage of the disease have significantly increased levels of anti-phospholipid autoantibodies (aPL-Ab) that recognize oxidized cardiolipin and phosphatidylserine (reviewed in [249]). aPL-Ab, by recognizing and binding to specifically oxidized epitopes in apoptotic hepatocytes (not in living cells, with phosphatidylserine being oxidized during apoptosis and before exposure on plasma membranes), may affect the ability of Kupffer cells to recognize and phagocytose apoptotic cells. Since phagocytosis of apoptotic cells usually involves increased release of TGFβ1 and IL-10 and down-regulation of TNF and IL-12 production, an impaired disposal of apoptotic hepatocytes by Kupffer cells will negatively affect the anti-inflammatory response of Kupffer cells. Moreover, Kupffer cells as well as other phagocytes may be further activated (pro-inflammatory) by increased recognition of aPL-Ab bound to the surface of apoptotic cells through the IgG Fc receptors. Finally, impaired disposal of apoptotic cells and/or bodies may, at the same time: further increase the inflammatory response because of post-apoptotic cell lysis; maintain conditions that further sustain development of aPL-Ab; and act as a pro-fibrogenic stimulus, as shown by already cited studies in which phagocytosis of apoptotic bodies by HSC/MFs resulted in the NOX- and ROS-dependent increased stimulation of pro-collagen type I synthesis [97, 322].