Bench-to-bedside review: Sepsis - from the redox point of view

The pathogenesis of sepsis and its progression to multiple organ dysfunction syndrome and septic shock have been the subject of investigations for nearly half a century. Controversies still exist with regard to understanding the molecular pathophysiology of sepsis in relation to the complex roles played by reactive oxygen species, nitric oxide, complements and cytokines. In the present review we categorise the key turning points in sepsis development and outline the most probable sequence of events leading to cellular dysfunction and organ failure under septic conditions. We have applied an integrative approach in order to fuse current state-of-the-art knowledge about redox processes involving hydrogen peroxide, nitric oxide, superoxide, peroxynitrite and hydroxyl radical, which lead to mitochondrial respiratory dysfunction. Finally, from this point of view, the potential of redox therapy targeting sepsis is discussed.


Introduction
Sepsis is a leading cause of death in critically ill patients; the mortality in patients admitted to intensive care units and given both antibiotics and supportive care varies from 20% up to more than 60% [1,2]. Th e problem is becoming even more prominent because the frequency of sepsis is increasing due to the prevalence of antibioticresistant pathogens [3]. Historically the widely accepted hypothesis was that sepsis was in fact an uncontrolled infl am matory response. Accordingly, the organism is killing itself while the pathogens are considered as bystanders [4,5]. However, this theory was based on animal studies that do not refl ect the clinical picture in humans [6]. Studies on animal models usually involve the application of large doses of endotoxin or pathogens, leading to much higher levels of circulating cytokines in com pari son to humans with sepsis and ending in death by 'cytokine storm' [6]. Recent data indicate that the immune system during sepsis actually shows two phases: an initial hyper-infl ammatory stage followed by a pro longed hypoinfl ammatory phase [7]. In addition, numerous studies and trials have shown that the application of antiinfl ammatory agents is futile, or can even worsen chances of survival [7,8]. Th ere are some exceptions, however. For example, Annane and co-workers [9] have shown that the application of selected corticosteroids at moderate doses may decrease the mortality of sepsis. In contrast to prevalently disappointing results for anti-infl ammatory therapies, a pro-infl ammatory approach seems to be more promising. For example, interferon-γ, a potent macrophage activator, improved survival of patients with sepsis [10]. IL-12, an immune stimulant, reduced mortality from subsequent sepsis when administered after burn injury [11]. Pertinent to these data, Hotchkiss and Karl [7] have recently proposed a coherent immunological approach in sepsis treatment, consisting of moderate suppression of the immune system during the hyperinfl ammatory mode in early sepsis and intensive promotion of activity of the immune system during the hypoinfl ammatory phase.
Immunohistochemical analysis has shown that in the majority of patients with sepsis, only lymphocytes and gastrointestinal epithelial cells die [12]. Th ese two types of cells normally undergo rapid turnover through apoptosis, and sepsis most likely accelerates this process. Cell death in the heart, lungs and kidneys has been observed to be minor, resulting in insignifi cant damage to these organs [13]. Studies in patients with acute renal failure have illustrated discordance between the level of renal dysfunction and the degree of tubular necrosis [14]. It is very interesting that most sepsis survivors who showed renal failure recover baseline renal function [15], suggesting that changes that may cause death are in fact reversible. It can be concluded that organ failure in sepsis is not provoked by tissue damage, but instead by the dysfunction of a substantial number of cells. A central paradigm in organ dysfunction is the concept of tissue hypoxia; the failure to adequately supply organs with oxygen [16]. However, non-survivors of sepsis exhibit an

Abstract
The pathogenesis of sepsis and its progression to multiple organ dysfunction syndrome and septic shock have been the subject of investigations for nearly half a century. Controversies still exist with regard to understanding the molecular pathophysiology of sepsis in relation to the complex roles played by reactive oxygen species, nitric oxide, complements and cytokines. In the present review we categorise the key turning points in sepsis development and outline the most probable sequence of events leading to cellular dysfunction and organ failure under septic conditions. We have applied an integrative approach in order to fuse current state-of-the-art knowledge about redox processes involving hydrogen peroxide, nitric oxide, superoxide, peroxynitrite and hydroxyl radical, which lead to mitochondrial respiratory dysfunction. Finally, from this point of view, the potential of redox therapy targeting sepsis is discussed. impaired capacity to increase tissue oxygen consumption in response to the augmentation of oxygen delivery [17]. Th is is reinforced by the fact that well-perfused endotoxaemic model animals still present acidosis [18]. In line with these facts, Fink [19] has postulated that impaired cellular O 2 utilisation, rather than inadequate oxygen delivery, may play an important role in sepsis development. As mitochondria represent the main consumers of O 2 , signifi cant attention has been focused on the role of mitochondrial dysfunction in sepsis [20]. Brealey and coworkers [21] demonstrated that there is an association between mortality in sepsis and mitochondrial dysfunction. In addition, a decrease in the number of mitochondria has been observed in patients with prolonged septic conditions [22], implying that sepsis-related dysfunction of mitochondria may result in mitoptosis. Th e direct consequence of mitochondrial dysfunction is a signifi cant fall in the level of ATP, as observed in the tissues of model animals and septic patients [21]. An activated immune system and dysfunctional mitochondria represent the two most potent sources of reactive molecular species in organisms [23]. Pertinent to this, it has been documented that sepsis is characterised by excessive production of reactive oxygen species (ROS) and reactive nitrogen species (RNS), both in the circulation (by activated immune system cells and endothelial cells) and in the aff ected organs (by dysfunctional mitochondria and due to a modifi ed anti-oxidative status) [16,24]. Many authors agree that ROS and RNS (Table 1) play a crucial role in the pathophysiology of sepsis [24][25][26]. Some of these species interfere with signalling cascades, while others provoke deleterious eff ects on various biological molecules and structures. It is clear that the increased production of signalling species (for example, nitric oxide ( • NO) and hydrogen peroxide (H 2 O 2 )) and strong oxidants (the hydroxyl radical ( • OH) and hypochlorous acid (HOCl)) act in synergy with collapse in energy metabolism to provoke cell dysfunction, which may result in organ failure and death.
When considering possible treatment strategies for sepsis patients, it is crucial to identify the main events in the development of sepsis and the schedule of changes of pathogenic factors. In the present review we focus on redox mechanisms of sepsis without under-estimating any other possible contributors to this pathology. We outline the most probable sequence of redox events leading to cellular dysfunction in sepsis. In addition, the current potential of and prospects for the application of redox therapy in sepsis are discussed.

Oxidative burst
A plasma oxidative burst takes place early in sepsis orchestrated by neutrophils, macrophages and endothe lial cells [27]. Th ese cells represent the innate immune system, our fi rst line of defence against invading micro organisms. Accordingly, they are equipped with diff erent ROS and RNS [23]. When activated by pathogens they generate superoxide ( • O 2 -), H 2 O 2 , • NO and HOCl, while • OH, peroxynitrite (ONOO -), the nitrogen dioxide radical ( • NO 2 ), the hydroperoxyl radical ( • OOH), and the peroxyl radical ( • ROO) are produced in secondary reactions ( Figure 1). Th e lack of this intrinsic oxidant system can be deleterious to the host as demonstrated in subjects lacking • O 2 generation. Mice defi cient in NADPH oxidase subunits and humans with chronic granulo ma tous disease who also lack NADPH oxidase activity are incapable of adequately performing bactericidal func tions [28,29]. However, the oxidant system may also signifi cantly modify redox settings in the organism.
Septic plasma shows a number of positive markers of oxidative stress (Table 2). Anti-oxidants scavenge reactive species, which leads to a decrease in the concentration of anti-oxidants. Th e level of free thiols decreases due to oxidation or nitrosylation. As a result, the total anti-oxidant capacity of plasma is reduced. In addition, uncharged radicals attack membranes, initiating a lipid peroxidation chain reaction ( Figure 1, reactions 6 to 8). Interestingly, two research groups have reported that the plasma total anti-oxidant capacity in sepsis is initially decreased but increases over time as the condition progresses [30,31]. Th is indicates that the redox status of plasma during sepsis follows a two-phase pattern characteristic of the immune system.
It is important to note that redox conditions in the plasma of neonates with sepsis seem to be diff erent in comparison to other patient groups. Cherian and colleagues [32] reported that there were no signifi cant changes in the levels of ascorbate, glutathione and oxidative stress biomarkers in the blood of neonates and babies with sepsis when compared to age-matched controls. Th is may be attributed to the immaturity of the immune system. Neonates have a low neutrophil pool and reduced release of pro-infl ammatory cytokines (for Table 1. The main reactive species implicated in the pathogenesis of sepsis example, TNF-α), and their neutrophils have an impaired capacity to migrate to sites of infection, which may account for the less prominent oxidative burst in response to infection [33]. Fortunately, neonatal sepsis is very uncommon and shows a relatively low mortality rate of 5 to 10% (in developed countries) [34]. Th e pre sented facts regarding neonatal sepsis strongly imply that pronounced immune system-mediated redox changes may play an important role in sepsis initiation in adults.
Modifi ed redox status of septic plasma aff ects the intracellular milieu of surrounding tissues due to: (i) the capability of some reactive species, such as H 2 O 2 , • NO, and HOCl, to cross cellular membrane [23,35]; (ii) oxidative changes in the membrane (Figure 1, reactions 6 to 8) [36]; and (iii) the activation of specifi c receptors. After entering the cell, H 2 O 2 and • NO participate in secondary reactions ( Figure 1, reactions 1 to 4) to pro duce more reactive species, which provoke intracellular oxidative stress. Due to its high reactivity with bio molecules, HOCl is predicted to have a very limited diff usion radius in biological systems [37,38]. Th erefore, only a small fraction of HOCl manages to enter the cell, where it may provoke protein, DNA, lipid and mito chondria damage [38]. On the other hand, H 2 O 2 and • NO represent signalling molecules crucial for normal cell function [39]. Th ey activate specifi c sets of genes via a number of transcription factors at the nuclear level, of which NF-κB appears to have a central role in sepsis [40]. Other factors, such as nuclear factor E2-related factor 2 (Nrf2), may also be   The use of total anti-oxidant capacity assays is associated with some pitfalls; therefore, the results obtained with such methods should be taken with some caution. For example, urate interferes with both the ABTS and the TRAP assays, potentially leading to falsely high values [23]. Urate increases in many pathophysiological conditions, for example, in renal failure, which may develop in sepsis. Another problem is that plasma contains a complex mixture of anti-oxidants that react at diff erent rates, so some of them may not be detected by rapid assays [23]. ABTS, 2,2'-azino-bis(3-ethylbenzthiazoline-6-sulphonic acid); FOX, ferrous oxidation-xylenol orange; HPLC, high-performance liquid chromatography; TBARS, thiobarbituric acid reactive substances; TRAP, total radical-trapping antioxidant potential.
involved. Nrf2 is a transcription factor that regulates antioxidant response element (ARE)-mediated transcription of a plethora of anti-oxidant and protective genes to counteract the harmful eff ects of ROS or environmental carcinogens. Non-carcinogenic ARE inducers from a variety of chemical classes have been identifi ed to enhance the transcriptional activity of Nrf2 through S-alkylation of reactive cysteines within the cellular redox sensor protein Keap1 (Kelch-like ECH associated protein 1) [41]. Following dissociation from the cytosolic protein Keap1, a scaff olding protein that binds Nrf2 and Cul3 ubiquitin ligase for proteasome degradation, Nrf2 rapidly accumulates in the nucleus and transactivates the ARE in the promoter region of many anti-oxidant genes [42]. Nrf2 seems to represent an important regulator of oxidative stress in sepsis [43]. For example, it has been reported that lipopolysaccharide-induced shock and polymicrobial sepsis induces early and greater mortality in Nrf2-defi cent mice, which is related to the lethal innate immune response [44]. In contrast, transgenic mice with a decreased rate of Nrf2 degradation showed lower mortality in polymicrobial sepsis [45]. Consequences of the down-regulation of the Nrf2 host defence mechanism still remain to be fully explored.

Intracellular • NO generation
Pro-oxidative and pro-infl ammatory events in the plasma have a strong impact on the intracellular milieu. H 2 O 2 causes a cascade of intracellular events resulting in the liberation of NF-κB from its inhibitor IκB [46]. NF-κB can be activated via several other pathways, as reviewed by MacDonald and co-workers [40]. For example, TNF-α, one of the primary mediators of the innate immune system, binds to its specifi c receptor (TNFR), which leads to the activation of multi-component protein kinase (IκB kinase), IκB phosphorylation, and NF-κB release and translocation into the nucleus [40] Clearly NF-κB has a special place in sepsis development. In an animal study lipopolysaccharide administration resulted in NF-κB activation in several organs [47]. Two diff erent research groups have reported that NF-κB activity is signifi cantly higher in septic non-survivors in comparison to controls and survivors and that NF-κB activity predicts sepsisrelated mortality [48,49]. Finally, encouraging results of clinical trials with activated protein C could be at least partially attributed to its ability to block the expression of NF-κB-regulated genes [50]. NF-κB binds to DNA to enable the initiation of the transcription process of a number of genes, two of them encoding redox-active enzymes: inducible • NO synthase (iNOS) [51] and cyclooxygenase (COX)-2, which generates • O 2 as a by-product [52]. It is noteworthy that iNOS expression can be stimulated by hypoxia inducible factor-1, which is also activated by ROS [53]. In contrast to consti tutively expressed endothelial NOS, which produces local • NO concentrations in the nanomolar range, iNOS has been estimated to produce • NO in the micromolar range [54,55]. Th e role of iNOS expression in non-septic infec tions is to induce • NO-mediated vasodila tation, thus enabling adequate circulation in the aff ected areas and normal cardiac function [56]. However, over-expression of iNOS in sepsis aff ects normal cellular functions [57]. Pertinent to such dual (patho)physiological roles of iNOS in sepsis, it has been demonstrated in two separate studies that iNOS-defi cient animals exposed to endo toxin show lower cardiac output [58] but have improved survival [59] in comparison to control subjects.
Over-expression of iNOS occurs early in sepsis [54,60], taking place in nearly all the vital organs [61]. Patients with sepsis show increased • NO production and elevated nitrite and nitrate levels, which have been correlated with the severity of illness [62]. Th e negative role of iNOS in sepsis is also indicated by the fact that the administration of the iNOS substrate arginine results in a poorer outcome in sepsis [63]. Th e mechanism that makes • NO production deleterious to cells seems to be related to protein nitrosylation and nitration, which can modulate the function of various enzymes and ion channels as well as diff erent signalling pathways [23]. In addition, • NO effl ux from cells and a substantial increase in the • NO level in plasma may occur. • NO converts haemoglobin to methaemoglobin (MetHb), which leads to increased rigidity of red blood cells (RBCs) and RBC lysis [64]. Another potential cause of RBC lysis in sepsis is the ability of HOCl to induce membrane-pore formation in RBCs [65]. RBC lysis results in the release of MetHb into the plasma [64], refl ecting the severity of sepsis [66]. An increased level of free MetHb in the plasma may have several propagating eff ects on sepsis development: (i) it leads to increased availability of iron to pathogens, which may fuel their proliferation [67]; (ii) it has pro-infl ammatory eff ects by stimulating interleukins (IL-6 and IL-8) and E-selectin [68]; and (iii) MetHb is an activator of NF-κB [64], which may lead to further iNOS expression, thus closing a • NO-generating loop (Figure 2).

Intracellular superoxide generation
Septic insult signifi cantly increases intracellular • O 2 - [69]. Th is is substantiated by the high activity of superoxide dismutase (SOD) in sepsis [22,27]. Th ere are several intra cellular • O 2 generators implicated in sepsis: NADPH oxidase, COX-2, xanthine oxidase and the electron transfer chain (ETC) in mitochondria. Th e entry of ROS from the plasma into the cell can initiate the expression of NADPH oxidase and COX-2 [40,52,70]. Th e early rise in • O 2 production in endothelial cells in sepsis is abolished by NADPH oxidase inhibitors [69], while the inhibition of COX-2 leads to a decrease in the production of ONOO -(an • O 2 derivative; Figure 1, reaction 3) in sepsis models [52,71]. At the later stages of sepsis development these two enzymes are accompanied by increased xanthine oxidase activity [72,73] and pronounced production of under physiological conditions [74]. Th e production is pronounced if the ETC is inhibited or uncoupled [74] as is the case under septic conditions. It has been documented that • NO actively inhibits mitochondrial respiration under sepsis-modelling conditions [75] and that such eff ects can be diminished by applying an iNOS inhibitor [76]. Fredriksson and co-authors [22] reported a signifi cant decrease in the activity of complex I (the main site of • O 2 production in the ETC [74]) in the muscles of septic patients. Finally, a very recent study noted ETC un coupling in severe sepsis and septic shock, the response being most pronounced in non-survivors [77]. In mitochondria, • O 2 is dismutated by MnSOD to H 2 O 2 , which enters the cytosol to close the redox loop ( Figure 2). In the reaction between • NO and • O 2 -, ONOOis produced, which is then protonated to form peroxynitrous acid (ONOOH), which in turn spontaneously decomposes to two highly reactive species -• OH and • NO 2 . These species damage mitochondria and in cooperation with ETC inhibition provoke mitochondrial dysfunction resulting in a fall in ATP. Superoxide is also produced in the cytosol via increased activities of three enzymes: NADPH oxidase, cyclooxygenase (COX)-2 and xanthine oxidase (XO). Indirectly, via DNA damage, poly (ADP-ribose) polymerase (PARP) activation and NAD + consumption, ONOOpromotes the production of • O 2 on complex I in the ETC, which depends on the NADH/NAD + ratio. Superoxide is dismutated in mitochondria by manganese superoxide dismutase (MnSOD) to H 2 O 2 , which closes two positive feedback redox loops. Intracellular • NO overproduction leads to • NO leakage into the plasma. There • NO provokes red blood cell (RBC) lysis while HOCH provokes pore formation in RBC membranes, thus freeing MetHb and increasing iron availability, which fuels pathogen proliferation. MetHb provokes the activation of NF-κB, thus closing the • NO-generating loop. The plus (+) and minus (-) symbols represent positive and negative eff ects on concentration, gene expression or activity, respectively. TNFR, TNF receptor.

Pathogen
Neutrophil

Th e interaction between • NO and • O 2
plays a fundamental role in cell and organ failure [78]. Superoxide reacts with • NO at a slightly faster rate than with SOD so when the level of • NO is in the high nanomolar or micromolar range • NO may outcompete SOD and react with • O 2 to generate ONOO - [79] (Figure 1, reaction 3), which is capable of inhibiting enzyme and membrane functions [23]. In particular, ONOOis responsible for decreased activity of the H 2 O 2 -removing enzyme glutathione peroxidase in sepsis [80]. Th e relevance of this interaction in sepsis is demonstrated by the fact that selenium, which is crucial for glutathione peroxidase meta bolism, shows clinical benefi ts in sepsis treatment [25,80].
Th e most important eff ects of ONOOin sepsis are exerted indirectly on mitochondria. It causes DNA singlestrand breakage in the nucleus, which signals poly(ADPribose) polymerase (PARP) to be activated. PARP catalyses the cleavage of NAD + , which leads to a decrease in the NAD + level and to an increase in the NADH/NAD + ratio [54,81]. A high NADH/NAD + ratio is known to raise • O 2 production on the mitochondrial complex I [74]. It appears that the ONOO --provoked increase in • O 2 produc tion on complex I represents the third major event in sepsis following the immune system-generated oxidative burst and iNOS over-expression. In vitro studies have shown that PARP activation causes energy depletion and cellular injury, while pharmaco logical inhibition of PARP improves the response of mitochondria to ONOOchallenge [81]. In addition, PARP inhibitors have been shown to diminish sepsis-provoked mitochondrial impairment [56]. Th e impor tance of the ONOO -/PARP/NAD + / complex I sequence of events in sepsis is illustrated by the fact that PARP-defi cient mice with sepsis show improved survival, while the administration of PARP inhibitors is benefi cial in a porcine model of sepsis [82,83]. Finally, Khan and co-workers [84] have shown that the application of liposomal NAD + on immunostimulated cells (modelling sepsis), lead ing to a decreased NADH/NAD + ratio, prevents dimin ished consumption of O 2 provoked by ONOO -.
As in the case of • NO-promoted • O 2 production in mito chondria, ONOO --provoked generation of • O 2 results in increased mitochondrial H 2 O 2 , which leaks from mitochondria to further promote NF-κB activation, thus closing another redox loop (Figure 2). In line with this, application of the ONOO --decomposition catalyst FeTPPS has been reported to reduce NF-κB activation in endotoxaemic mice [85].

Mitochondrial dysfunction
Mitochondria are specifi cally targeted by highly reactive products of ONOOdecomposition -• OH and • NO 2 [23] ( Figure 1, reaction 4). Th e hydroxyl radical non-selec tively damages membranes, proteins and DNA [24] while • NO 2 binds to proteins leading to increased susceptibility to protease activity [23]. Pertinent to this, research regarding mitochondrial structure after lipopolysac charide insult revealed oxidative stress and pronounced mitochondrial DNA damage [16,20]. In combination with ETC inhibition, oxidative modifi cations lead to mito chondrial dysfunction, which is intimately linked to a depleted cellular ATP pool and an energy defi cit charac teristic of sepsis (Table 3).
Dysfunctional mitochondria show increased susceptibility to mitoptosis. Morphological changes characteristic of mitoptosis have been documented in a feline model of acute endotoxaemia [86], while a decrease in mito chondrial content has been observed in patients with prolonged septic conditions [22]. A lower number of mitochondria may further compromise energy balance. Together, a fall in ATP concentration and oxidative stress exerted on mitochondrial metabolites lead to mitochondrial permeability transition pore opening with eventual organelle swelling and cytochrome c release. Once in the cytoplasm, cytochrome c can bind to apoptotic protease activating factor 1 (APAF-1) and thus activate caspase-9, which in turn activates caspase-3 and -7, culminating in apoptosis in some cell types. Anti-oxidant treatment can prevent mitochondrial swelling and cytochrome c release, pinpointing the role of ROS in mitochondrial dysfunction [87]. Mitochondrial dysfunc tion represents the fourth and fi nal major redox event in the pathogenesis of sepsis.
Patients typically die of sepsis days to weeks after the onset of the initial infection and in some cases even after the original infection has been eradicated [1,2]. Interestingly 10 to 40% of sepsis patients show negative cultures [1,2]. At the later stages of sepsis the immune system is in a hypoactive mode. Th erefore, the plasma oxidative burst can no longer be directly responsible for cell/organ dysfunction [7]. Th ese facts imply that infection could, under specifi c conditions, activate a self-sustaining and self-promoting redox cycle. Instead of shutting down upon pathogen eradication or in the immune hypoactive stage the 'sepsis redox cycle' continues to drain energy and prevents normal function from being re-established due to several positive feedback loops (Figure 2), even when there is no further infl ow of reactive species from plasma. It is important to note that in the 'sepsis redox cycle' H 2 O 2 crosses both cellular and mitochondrial membranes. Th is transport is facilitated by aquaporin-8 [88]. Interestingly, aquaporin-8 is down-regulated in sepsis [89], which may be a mechanism of defence targeted at stopping or slowing down the cycle. Th e pathophysiological redox profi le of one cell can spill over to surrounding tissue, due to the ability of H 2 O 2 and • NO to cross membranes, thus leading to organ failure.

Polymorphisms of redox-relevant genes and sepsis
Th e question why the activation of the immune system in some patients leads to sepsis and in others does not remains unresolved. On the basis of studies in identical twins genetic factors have been implicated in the determination of susceptibility to death from infections [90]. It seems that specifi c gene polymorphisms of some redoxactive enzymes are more frequent in patients with sepsis and other related conditions in comparison to the general population (Table 4). However, the exact role of gene polymorphisms in sepsis susceptibility and sepsis development, progression and outcome remains to be further elucidated.

Redox therapy in sepsis
Sepsis is treated by controlling the source of infection, administering antimicrobial therapy, assuring hemodynamic support with fl uid resuscitation and vasopressor drugs, inducing sedation or analgesia as needed and by other supportive therapies [91]. Despite these interventions, about one-third of all septic patients die before leaving hospital [1,2]. Th erefore, the development of additional treatment approaches, such as anti-oxidative therapy, is a research topic of urgent priority.
Although many promising results have been obtained from animal models, believable benefi t from anti-oxidant application in sepsis has rarely been translated into success in human clinical trials [92]. Th is may be explained by several reasons. Animal models of sepsis do not completely recapitulate the human disease or involve identical care delivered to human sepsis patients [93][94][95]. Th e onset and progression of sepsis to multi-organ failure occurs over days to weeks whereas in most animal models these processes are much faster, occurring within hours to days. Animal models of sepsis almost exclusively use young animals, which mimic the physiology of a young human. Th e pitfall here is that epidemiologic studies show that the older human population is more susceptible to sepsis. Moreover, in contrast to animal models, patients with sepsis commonly have co-morbidities [95]. Finally, models and clinical sepsis show very diff erent cytokine dynamics [93,94]. It should be stressed that therapeutic studies on animals sometimes involve the application of high doses of anti-oxidants exceeding recom mended human allowances that clinical researchers are reluctant to use in trials. In addition, animal studies frequently employ early application of therapy (or pretreatment), which is often not the case with clinical sepsis. Some clinical trials do not take into account all the main principles of anti-oxidative therapy. Halliwell and Gutteridge [23] made a list of questions that generally should be answered preceding the use of an anti-oxidant in humans: (i) Is it non-toxic? (ii) Which biomolecules will the anti-oxidant protect, and are such molecules implicated in the pathogenesis of the disease? (iii) How does the anti-oxidant work -by scavenging reactive species, increasing endogenous defence or by repairing the damage? (iv) Can the anti-oxidant-derived products cause damage? (v) Does the anti-oxidant interfere with redox signalling? Kohen and Nyska [96] sublimed the principles of anti-oxidant therapy as follows: 'Taking into consideration the high reactivity of ROS, their short life span, their continuous production in close proximity to biological targets and their ability to be modifi ed into other more reactive species, one realises that, in order to cope with these deleterious metabolites, the anti-oxidant should be administered to the body continuously, in high concentrations and targeted to the biological site susceptible to oxidative damage. ' Pertinent to this and the mechanism proposed here, it is clear that a redox approach in sepsis treatment should be targeted at the specifi c steps in the pathogenesis. For example, very promising eff ects of therapies targeting dysfunc tional mitochondria have been reviewed recently by Dare and colleagues [16] and Galley [20].
Targeted delivery of anti-oxidants into intracellular space is limited by mechanisms regulating the level of reducing agents in order to maintain normal functioning of signalling pathways that involve reactive species [97]. Th erefore, even if long-term supplementation raises blood anti-oxidants, it will have a limited eff ect on intracellular levels or redox status [97]. Hence, the majority of anti-oxidants perform their actions predominantly in plasma, which may have benefi cial eff ects during the hyper-infl ammatory stage of sepsis but are ineff ective against the self-sustaining 'sepsis redox cycle' taking place inside the cell. Th is raises the question of how to overcome refractory mechanisms and help the organism to fi ght intracellular oxidative stress against its own 'will' . In other words, in conditions related to intracellular oxidative stress, such as sepsis, reducing agents targeted to intra cellular space and capable of accumulating inside the cell are needed.
Potential candidates for such an approach are fructose 1,6-(bis)phosphate and ethyl pyruvate, which act as both antioxidants and energy molecules [98,99], and organisms in times of crisis will not reject energy supplies. Th ese compounds have already shown positive eff ects in sepsis models [100][101][102][103], but are yet to be studied in clinical trials. Molecular hydrogen that reacts directly with ROS should also be capable of overcoming the refractory response. Recently, Xie and co-workers [104] reported that H 2 inhalation signifi cantly improved the survival rate of septic mice in a concentration-and timedependent manner.
An alternative anti-oxidative approach is to suppress ROS production in mitochondria. It has been shown that maintenance of glucose in blood within the range 4.4 to 6.1 mM, instead of the normally higher range of 10.0 to 11.1 mM, leads to a decrease in mortality by more than 10% [105]. We propose that decreased availability of glucose could exert positive eff ects in sepsis by preventing the production of • O 2 on mitochondrial complex I. A lower level of glucose leads to decreased generation of NADH and hence to a decrease in the NADH/NAD + ratio. Another way to decrease the NADH/NAD + ratio is to apply liposomal NAD + , which has been shown to prevent diminished O 2 consumption in mitochondria under septic conditions [84]. As an alternative to feeding complex I, which represents a point of electron leakage and • O 2 production, mitochondrial complex II could be supplemented with succinate in order to bypass • O 2 production on complex I and in parallel to maintain normal energy metabolism. Several studies have confi rmed positive eff ects of succinate in sepsis [106][107][108].
Pertinent to the important role of iNOS and • NO in sepsis-related cell/organ dysfunction, therapeutic strategies capable of decreasing the level of • NO could be benefi cial in sepsis. Th e supplementation of selective iNOS inhibitors, such as aminoguanide, L-NAME and L-N6-(1-imminoethyl)lysine, has had positive eff ects in diff erent sepsis models [109][110][111][112][113]. It is interesting that a clinical trial of the non-selective NOS inhibitor L-NMMA in patients with septic shock was terminated early because of increased mortality. Such an adverse eff ect of high-dose L-NMMA may have resulted from the inhibition of endothelial NOS, whose activity is essential for microvascular homeostasis [114]. In addition to iNOS inhibition, • NO excess can be tackled using • NO scavengers. For example, diethyldithiocarbamate (DDC) demonstrates positive eff ects in septic shock mice, partially due to the ability of the iron-DDC complex to sequester • NO [115].

Conclusions
Organ failure in sepsis is provoked by cellular dysfunc tion, a process related to pro-oxidative conditions and dysfunctional mitochondria. Th e oxidative burst performed by the innate immune system is 'transferred' from the plasma into the cell, where it activates the production of ROS and RNS, resulting in non-physiological redox conditions, disturbed redox signalling, oxidative damage of diverse molecules and cellular structures and energy depletion. Th ese events are interconnected in a selfsustaining and self-promoting pathological process named here as the 'sepsis redox cycle' , which may result in the total loss of cellular function. Redox therapies targeting specifi c steps of the 'sepsis redox cycle' have shown some very promising results for sepsis treatment, but require further clinical evaluation. Although it is obvious that sepsis represents a multi-factorial patho physiology, studies investigating the eff ects of multi-target therapies are still scarce. We believe that instead of looking for a single 'magic bullet' for sepsis treatment, a multi-component 'magic bullets' therapeutic approach combining diff erent redox agents capable of suppressing specifi c steps in the pathogenesis of sepsis is clearly the way forward.