As a consequence of cohabiting with divergent microbes (e.g. virus, bacteria and fungi), animals have to deal with various microbial infections. Epithelial barriers provide the first layer of defence by limiting the access of potential pathogens. If they are breached, the host's innate immune system mounts an immediate but nonspecific biological response – termed inflammation – at the infection site, to confine and remove invading pathogens. If the invading pathogens are effectively eliminated, inflammation resolves normally to restore immunological homeostasis (Ref. Reference Serhan and Savill1); however, if not, invading pathogens or pro-inflammatory mediators such as tumour necrosis factor (TNF) or other cytokines can leak into the bloodstream, triggering a systemic inflammatory response that may lead to sepsis (Fig. 1).
Figure 1. A microbial infection can trigger a local or systemic inflammatory response. The disruption of an epithelial barrier allows invasion of microbial pathogens, which elicit an innate immune response at the site of infection. If invading pathogens are effectively eliminated by phagocytes, local inflammation resolves normally to regain immunological homeostasis. If invading pathogens are not effectively eliminated, they can leak into the bloodstream, and trigger a potentially injurious systemic inflammatory response (such as sepsis).
Sepsis refers to a systemic inflammatory response syndrome resulting from a microbial infection. As a continuum of increasing clinical severity, ‘severe sepsis’ is defined as sepsis associated with one or more acute organ dysfunctions (Ref. Reference Dellinger2). Septic shock is severe sepsis with organ hypoperfusion and hypotension (defined as systolic blood pressure less than 90 mmHg) that are poorly responsive to fluid resuscitation. Despite recent advances in antibiotic therapy and intensive care, sepsis is still the most common cause of death in intensive care units (Ref. Reference Dellinger2). Here, we briefly review the prevailing theories of sepsis as an uncontrolled systemic inflammatory response, and discuss potential therapeutic agents that target clinically more feasible, late-acting mediators of experimental sepsis, such as HMGB1.
Local innate immune response to mild infection
The innate immune system comprises phagocytes (such as macrophages, monocytes and neutrophils), mast cells, eosinophils, basophils and natural killer cells. It constitutes a front line of defence against most microbial infection by eliminating invading pathogens and initiating an inflammatory response.
Elimination of invading pathogens
Neutrophils and monocytes continuously patrol the body to search for invading pathogens, and infiltrate into infected/injured tissues upon detecting microbial products (Ref. Reference Luster, Alon and von Andrian3). Neutrophils arrive at the infection site early and in high numbers, and thus usually kill more invading bacteria than other phagocytes (Ref. Reference Scapini4). However, neutrophils are short-lived, with an average lifespan of 1–2 days: after engulfing and killing several bacteria, neutrophils exhaust intracellular enzymes and subsequently undergo apoptotic cell death. Upon reaching extravascular tissues, monocytes can differentiate into tissue-specific macrophages. Macrophages can ingest and eliminate larger pathogens that are not handled by the neutrophils; in addition, they remove the cell debris of apoptotic neutrophils in order to resolve an inflammatory response (Ref. Reference Liles5).
The recognition of pathogens by phagocytes is mediated by host bridging proteins called opsonins (such as complement or antibodies) (Ref. Reference Mosser6). The specific recognition of apoptotic cells is achieved through cell-surface receptors for phosphatidylserine or opsonins (such as MFG-E8) (Ref. Reference Wu, Tibrewal and Birge7). After binding to these opsonins, phagocytes engulf pathogens or damaged cells, and eliminate them through the generation of reactive oxygen species and hydrolytic enzymes.
Initiation of the innate inflammatory response
Upon recognition of molecules shared by groups of related microbes (called pathogen-associated molecular patterns; PAMPs) by pattern-recognition receptors (such as the Toll-like receptors; TLRs), innate immune cells can initiate an inflammatory response. Well-known PAMPs include bacterial endotoxin (lipopolysacharides; LPSs), peptidoglycan, and microbial unmethylated CpG-DNA (Refs Reference Hemmi8, Reference Krieg9). Although there is a structural similarity among various TLRs, each TLR can recognise a specific type of PAMP. For instance, TLR2 is essential for the recognition of lipoproteins, peptidoglycan and lipoteichoic acids of most Gram-positive bacteria (Ref. Reference Brightbill10); TLR4 recognises endotoxin of Gram-negative bacteria (Ref. Reference Poltorak11); and TLR9 recognises microbial unmethylated CpG-DNA (Ref. Reference Hemmi8).
Engagement of various TLRs by specific PAMPs leads to production and release of cytokines (such as TNF and the interleukins IL- 1 and IL-6) and chemokines (such as IL-8, and the ‘macrophage inflammatory proteins’ CCL3 and CCL4) (Ref. Reference Akira and Takeda12). Chemokines are responsible for recruiting more innate immune cells to the site of infection or injury (Ref. Reference Baggiolini and Loetscher13), whereas cytokines can activate these immune cells to produce more pro-inflammatory mediators (Ref. Reference Balkwill14). Although an appropriate inflammatory response is required for host defence against infection, an uncontrolled systemic inflammatory response may contribute to the pathogenesis of lethal inflammation diseases such as sepsis.
Systemic innate immune response to severe infection
The prevailing theories of sepsis as an uncontrolled systemic inflammatory response are supported by extensive studies employing various animal models of sepsis.
Animal models of experimental sepsis
Experimental sepsis is induced in animals by three common strategies: infusion of exogenous bacterial toxin (endotoxaemia); infusion of exogenous bacteria (bacteraemia); and faecal contamination of the peritoneal cavity induced by caecal ligation and puncture (CLP). Each of these models has particular strengths and weaknesses with respect to its ability to mimic the clinical progression of human sepsis (Ref. Reference Wichterman, Baue and Chaudry15).
Endotoxaemia
Endotoxaemia is induced by intraperitoneal or intravenous injection of known amounts of bacterial endotoxin to animals. It provides a model to investigate pathogenic roles of pro-inflammatory mediators in lethal systemic inflammation. Depending on the doses, endotoxin can induce transient/nonlethal or persistent/lethal haemodynamic cardiovascular responses. Thus, endotoxemia is considered as a model of septic shock rather than sepsis (Ref. Reference Wichterman, Baue and Chaudry15). Other bacterial products (such as CpG-DNA) can also be used to induce septic shock in animals.
Bacteraemia
Bacteraemia is induced by intravenous or intraperitoneal infusion of exogenous viable bacteria into the host. Because many exogenous bacteria may not colonise or replicate well in the host, the doses of bacteria required to induce lethality do not mimic those inducing a typical host response to infection in the clinical setting (Ref. Reference Wichterman, Baue and Chaudry15). Since various bacteria strains may induce different cytokine responses, the bacteraemia model is useful to study the host response to a particular pathogen.
Caecal ligation and puncture
Sepsis can be induced by surgical perforation of the caecum, a technique known as CLP (Ref. Reference Wichterman, Baue and Chaudry15). This procedure allows bacteria spillage and faecal contamination of the peritoneal cavity, mimicking the human clinical disease of perforated appendicitis or diverticulitis. The severity of sepsis, as reflected by the eventual mortality rates, can be controlled surgically by varying the size of the needle used for caecal puncture. CLP in animals induces similar, biphasic haemodynamic cardiovascular, metabolic and immunological responses to those observed during the clinical course of human sepsis. Thus, the CLP model is considered as the most clinically relevant model for experimental sepsis.
Inflammatory mediators of experimental sepsis
In response to systemic infection, innate immune cells release large amounts of pro- and anti-inflammatory mediators that collectively determine the outcome of systemic inflammation.
Pro-inflammatory mediators
Various microbial PAMPs (e.g. LPS and CpG-DNA) stimulate innate immune cells to release a wide array of pro-inflammatory mediators, including nitric oxide (Ref. Reference Lowenstein, Dinerman and Snyder16), TNF (Ref. Reference Tracey17), IL-1 (Ref. Reference Dinarello and Thompson18), leukaemia inhibitory factor (LIF) (Ref. Reference Block19), interferon (IFN)-γ (Ref. Reference Heinzel20), and macrophage migration inhibitory factor (MIF) (Refs Reference Bernhagen21, Reference Calandra22, Reference Bozza23). Extensive studies employing animal models of sepsis suggest that various pro-inflammatory mediators, individually or in combination, contribute to the pathogenesis of lethal systemic inflammation. For instance, neutralising antibodies against bacterial products (such as endotoxin) (Ref. Reference Davis24) or an early pro-inflammatory cytokine (TNF; Ref. Reference Tracey17) reduce lethality in an animal model of endotoxaemic/bacteraemic shock.
Anti-inflammatory mediators
Microbial products also stimulate innate immune cells to produce anti-inflammatory mediators, such as prostaglandin E2 (PGE2) (Ref. Reference Knudsen25), IL-10 (Refs Reference Oswald26, Reference Bogdan, Vodovotz and Nathan27) and transforming growth factor (TGF)-β (Refs Reference Tsunawaki28, Reference Miller-Graziano29), which counter-regulate or suppress potentially injurious pro-inflammatory mediators. Another local feedback mechanism is through spermine, a ubiquitous biogenic molecule that accumulates at sites of infection or injury and post-transcriptionally inhibits endotoxin-induced release of multiple pro-inflammatory cytokines (e.g. TNF, IL-1β, CCL3 and CCL4) from macrophages and monocytes (Refs Reference Zhang30, Reference Zhang31, Reference Wang32, Reference Zhang, Wang and Tracey33). Recent evidence suggests that the central nervous system can also attenuate the peripheral innate immune response through efferent vagus nerve signals to tissue-resident macrophages (Ref. Reference Borovikova34). This effect is mediated by the principal neurotransmitter of the vagus nerve, acetylcholine, which inactivates macrophages via nicotinic cholinergic receptors (Ref. Reference Borovikova34).
Current available therapies for human sepsis
Currently, available therapies for sepsis are still limited to several simple clinical interventions: (1) appropriate broad-spectrum antibiotics; (2) physiological doses of steroidal anti-inflammatory drugs (e.g. hydrocortisone); (3) adjunctive therapy with an anticoagulant agent (e.g. activated protein C); (4) early goal-directed therapies (EGDTs) to restore tissue oxygen delivery; and (5) intensive insulin therapy to regain normoglycaemia (Table 1).
Table 1. Current available therapies for sepsis
Antibiotics
Once the infecting agents are identified, appropriate broad-spectrum antibiotics are immediately administered to patients to facilitate elimination of bacterial pathogens (Ref. Reference Dellinger2). However, administration of antibiotics may also trigger release of bacterial products (such as endotoxin or CpG-DNA) that may further stimulate innate immune cells to release pro-inflammatory cytokines. Thus, anti-inflammatory agents may be useful to pharmacologically modulate a potentially injurious inflammatory response.
Steroidal anti-inflammatory drugs
Steroidal anti-inflammatory drugs refer to a group of steroid-like molecules that can reduce an inflammatory response. Although high-dose steroid (e.g. methylprednisolone at 30 mg/kg or dexamethasone at 0.5 mg/kg body weight/day) was harmful to septic patients (Ref. Reference Lefering and Neugebauer35), one study showed low-dose steroid therapy (50 mg hydrocortisone every 6 h, plus 50 µg oral fludrocortisone for 7 days) was beneficial for septic patients with adrenal insufficiency (i.e. poor endogenous responses to steroid-inducing hormones) (Ref. Reference Annane36). However, a more recent multicentre clinical trial indicated that intravenous hydrocortisone (50 mg every 6 h for 5 days) did not improve 28-day survival of patients with septic shock, regardless of whether patients were responsive or nonresponsive to steroid-inducing hormones (corticotrophin) (Ref. Reference Sprung37). It raises the question of whether the dose and the timing of an anti-inflammatory agent are critical for the successful management of human sepsis (Ref. Reference Dellinger2).
Activated protein C
The systemic inflammatory response is integrally related to intravascular coagulation and endothelial activation. As a major regulator of haemostasis, thrombin has both pro- and anti-coagulant properties. The procoagulant activities of thrombin include proteolytic activation of blood-clotting factors (Va, VIIIa and XI), cleavage of fibrinogen to form a fibrin clot, and stimulation of platelet aggregation. The anticoagulant effect of thrombin is regulated by thrombomodulin, a cofactor that is expressed on the luminal surface of vascular endothelium. Following engagement of thrombin to thrombomodulin, the ability of thrombin to catalyse procoagulant reactions is inhibited, but its ability to activate a plasma anticoagulant, activated protein C, is enhanced >1000-fold, which results in inactivation of blood clotting factors. In a large clinical trial, human recombinant activated protein C (drotrecogin alfa) reduced 28-day mortality (from 30.8% in the placebo group to 24.7% in the experimental group) (Ref. Reference Bernard38), but was accompanied by a 1.5% increase in haemorrhagic complication risk (Ref. Reference Bernard38). Thus, activated protein C has been approved by the US Food and Drug Administration only for patients with severe sepsis, who are more likely to die if otherwise not treated.
In addition to activated protein C, other anticoagulation agents, such as tissue factor pathway inhibitor and antithrombin III, have also been tested in clinical sepsis trials. Although both these agents were beneficial in preclinical studies or Phase I or II clinical trials, they failed to reduce 28-day mortality rates in Phase III clinical trials (Refs Reference Warren39, Reference Abraham40).
Early goal-directed therapies
As a supportive treatment, EGDT employs extremely tight control of several physiological parameters (such as central venous pressure, mean arterial blood pressure, central venous oxygen saturation, and haematocrit) with discrete interventions of crystalloid fluid, vasopressors and blood transfusions. In a controlled, randomised, prospective clinical trial, EGDT – combining volume resuscitation, catecholamine therapy and transfusion – effectively reduced mortality rates of patients with septic shock (from 46.5% in the placebo group to 30.5% in the experimental group) (Ref. Reference Rivers41). While achieving a significant reduction in mortality through relatively simple interventions, this approach is labour intensive, requiring intensive and continuous staff commitment (Ref. Reference Sama42). Furthermore, a recent multicentre clinical trial indicated that low-dose vasopressin (0.01–0.03 U/min) alone did not significantly reduce the 28-day mortality rate of patients with septic shock (35.4% in the vasopressin group versus 39.3% in the noradenaline control group; P = 0.26) (Ref. Reference Russell43).
Insulin
In critically ill patients, hyperglycaemia frequently occurs, and this has long been perceived as a beneficial metabolic response to stress that ensures glucose supply to insulin-insensitive organs (such as the brain and the immune system). However, this notion has recently been challenged by two landmark publications revealing that tight blood glucose control with intensive insulin therapy significantly improved morbidity and mortality in critical ill septic patients (Refs Reference van den Berghe44, Reference van den Berghe45). Nevertheless, the zeal for infusing insulin has been tempered by the announcement of unsuccessful multicentred clinical trials (Ref. Reference Brunkhorst46).
HMGB1 as a novel therapeutic target for experimental sepsis
The early kinetics of systemic TNF accumulation in sepsis makes it a difficult therapeutic target in clinical settings (Ref. Reference Tracey17), prompting a search for other, late pro-inflammatory mediators that may offer a wider therapeutic window. Here we briefly review evidence that supports extracellular HMGB1 as a potential novel therapeutic target.
Intracellular HMGB1 as a DNA-binding protein
HMGB1 is constitutively expressed in many cell types, and a large ‘pool’ of preformed HMGB1 is stored in the nucleus as a result of the presence of two lysine-rich nuclear localisation sequences (Refs Reference Chen47, Reference Bonaldi48). It contains two internal repeats of positively charged domains (‘HMG boxes’ known as the ‘A box’ and ‘B box’) in the N-terminus, and a continuous stretch of negatively charged (aspartic and glutamic acid) residues in the C-terminus. The HMG boxes enable HMGB1 to bind chromosomal DNA and fulfil its nuclear functions, including determination of nucleosomal structure and stability, and regulation of gene expression (Ref. Reference Bustin49).
Extracellular HMGB1 as an alarmin signal
Recently, a number of structurally diverse, multifunctional, ubiquitous host proteins – such as HMGB1 and heat shock protein 72 (HSP72) – have been categorised as ‘alarmins’ based on the following shared properties (Ref. Reference Oppenheim and Yang50) (Fig. 2).
Figure 2. Extracellular HMGB1 functions as an alarmin signal. HMGB1 is actively secreted by innate immune cells in response to exogenous bacterial products (e.g. endotoxin or CpG-DNA) or endogenous inflammatory stimuli (e.g. TNF, IFN-γ or hydrogen peroxide), and passively released by damaged or virus-infected cells. Extracellular HMGB1 sustains an inflammatory response by stimulating migration of innate immune cells, facilitating innate recognition of bacterial products (e.g. CpG by TLR9 and endotoxin by TLR4), activating various innate immune cells, and inhibiting phagocytosis of apoptotic neutrophils. Thus, HMGB1 can function as an alarmin signal, which recruits, alerts and activates various innate immune cells, thereby sustaining a potentially injurious inflammatory response. Abbreviations: HMGB1, high-mobility group box 1 protein; IFN-γ interferon, γ; TLR, Toll-like receptor; TNF, tumour necrosis factor.
Active release and passive leakage
In response to exogenous bacterial products (such as endotoxin or CpG-DNA) (Refs Reference Wang51, Reference Ivanov52) or endogenous inflammatory stimuli (e.g. TNF, IFN-γ or hydrogen peroxide) (Refs Reference Wang51, Reference Rendon-Mitchell53, Reference Tang54), innate immune cells actively release HMGB1 in a dose- and time-dependent manner (Fig. 2). Lacking a leader signal sequence, HMGB1 cannot be actively secreted via the classical secretory pathway from the endoplasmic reticulum through the Golgi complex (Ref. Reference Wang51). Instead, activated macrophages/monocytes acetylate HMGB1 at its nuclear localisation sequences, leading to sequestration of HMGB1 within cytoplasmic vesicles and subsequent extracellular release (Refs Reference Bonaldi48, Reference Rendon-Mitchell53, Reference Gardella55). In addition, HMGB1 can be released passively from damaged cells (Ref. Reference Scaffidi, Misteli and Bianchi56) or cells infected by viruses (e.g. West Nile virus, Salmon anaemia virus, and Dengue virus) (Refs Reference Chu and Ng57, Reference Joseph58, Reference Chen59), and such HMGB1 similarly triggers an inflammatory response (Ref. Reference Wang60) (Fig. 2).
Stimulation of cell migration
Accumulating evidence indicates that HMGB1 can stimulating migration of neurites (Ref. Reference Fages61), smooth muscle cells (Ref. Reference Degryse62), tumour cells (Ref. Reference Huttunen63), mesoangioblast stem cells (Refs Reference Palumbo64, Reference Palumbo65), monocytes (Ref. Reference Rouhiainen66), dendritic cells (Refs Reference Yang67, Reference Dumitriu68) and neutrophils (Ref. Reference Orlova69) (Fig. 2). It raises a possibility that extracellular HMGB1 may recruit cells to sites of infection or injury (Ref. Reference Degryse62), thereby functioning as a potential chemokine (Ref. Reference Degryse and de Virgilio70).
Facilitation of innate recognition of microbial products
Recent studies suggested that HMGB1 can facilitate recognition of bacterial products (e.g. CpG-DNA or LPS) by innate immune cells (such as macrophages and dendritic cells) (Refs Reference Ivanov52, Reference Tian71, Reference Silva72) (Fig. 2). For instance, extracellular HMGB1 can bind to biologically active microbial CpG-DNA, and facilitate its innate recognition by the intracellular TLR9 receptor, thereby augmenting CpG-DNA-induced inflammatory responses (Refs Reference Ivanov52, Reference Tian71).
Activation of innate immune cells
Extracellular HMGB1 binds to several cell-surface receptors, including the receptor for advanced glycation end products (RAGE), and pattern-recognition receptors such as TLR2 and TLR4 (Refs Reference Park73, Reference Yu74). Consequently, HMGB1 activates innate immune cells (Refs Reference Park73, Reference Yu74, Reference Park75, Reference Kokkola76, Reference Pedrazzi77) or endothelial cells (Refs Reference Fiuza78, Reference Treutiger79) to produce pro-inflammatory cytokines, chemokines and adhesion molecules (Fig. 2). Notably, the ‘A box’ of HMGB1 functions as an antagonist of HMGB1 (Refs Reference Yang80, Reference Kokkola81), whereas the ‘B box’ recapitulates the cytokine activity of full-length HMGB1 (Refs Reference Li82, Reference Messmer83).
In vitro, exogenous HMGB1 appears to accumulate on the macrophage cell surface within 4–6 h of HMGB1 incubation (Ref. Reference Li84), which correlates with the kinetics of HMGB1-induced release of pro-inflammatory cytokines (Ref. Reference Andersson85). It is not yet known whether engagement of exogenous HMGB1 to cell-surface receptors (such as TLR2, TLR4 or RAGE) induces cell-surface clustering of ligand–receptor complexes (Ref. Reference Li84), thereby activating various innate immune cells.
In the brain, exogenous HMGB1 induces the release of pro-inflammatory cytokines (Ref. Reference Agnello86) and excitatory amino acids (such as glutamate) (Ref. Reference Pedrazzi87), induces fever (Ref. Reference O'Connor88), and exacerbates cerebral ischaemic injury (Ref. Reference Liu89). In the lung, HMGB1 induces neutrophil infiltration and acute injury (Refs Reference Abraham90, Reference Ueno91, Reference Lin92). Considered together, these studies indicate that extracellular HMGB1 can function as an alarmin signal to recruit, alert and activate innate immune cells, thereby sustaining a potentially injurious inflammatory response.
Inhibition of phagocytotic elimination of apoptotic neutrophils
As mentioned above, macrophages recognise apoptotic cells through cell-surface receptors for phosphatidylserine. Interestingly, a recent study indicated that HMGB1 could interact with phosphatidylserine on the cell surface of apoptotic neutrophils, and consequently inhibit phagocytotic elimination of apoptotic neutrophils by macrophages (Ref. Reference Liu93). Impaired clearance of apoptotic cells may allow excessive accumulation of late apoptotic and/or secondary necrotic cells, which may directly (Ref. Reference Bell94), or indirectly (by activating phagocytes), release pro-inflammatory mediators (such as HMGB1) (Ref. Reference Qin95). Thus, extracellular HMGB1 may sustain rigorous inflammatory responses by multiple mechanisms including interference with phagocytotic elimination of apoptotic neutrophils (Ref. Reference Liu93) (Fig. 2).
Pathogenic role of HMGB1 in diseases
Accumulating evidence has supported a pathogenic role for extracellular HMGB1 in infection- or injury-elicited inflammatory diseases.
Experimental sepsis
In murine models of endotoxaemia and sepsis (induced by CLP), HMGB1 is first detectable in the circulation 8 h after the onset of lethal endotoxaemia and sepsis, subsequently increasing to plateau levels from 16 to 32 h (Refs Reference Wang51, Reference Yang80). This late appearance of circulating HMGB1 precedes and parallels the onset of animal lethality from endotoxaemia or sepsis, and distinguishes HMGB1 from TNF and other early proinflammatory cytokines (Ref. Reference Wang96) (Fig. 3).
Figure 3. Early versus late mediators of septic lethality. Vertebrates subjected to septic insult succumb at latencies of up to several days, long after serum TNF has returned to basal levels. By contrast to early pro-inflammatory cytokines, systemic HMGB1 accumulation occurs in a delayed fashion, which precedes and parallels the onset of septic lethality (Refs Reference Wang96, Reference Wang, Yang and Tracey146). This delayed systemic accumulation makes HMGB1 a better therapeutic target with a wider therapeutic window for experimental sepsis. CLP, caecal ligation and puncture; HMGB1, high-mobility group box 1 protein; IFN-γ, interferon γ; TNF, tumour necrosis factor.
The pathogenic role of HMGB1 as a late mediator of lethal endotoxaemia was originally examined using HMGB1-specific neutralising antibodies, which conferred a dose-dependent protection against lethal endotoxaemia (Ref. Reference Wang51) and endotoxin-induced acute lung injury (Ref. Reference Abraham90). In a more clinically relevant animal model of sepsis (induced by CLP), delayed administration of HMGB1-specific neutralising antibodies beginning 24 h after the onset of sepsis, rescued mice from lethal sepsis in a dose-dependent manner (Refs Reference Yang80, Reference Qin95). Similarly, anti-HMGB1 antibodies conferred protection in a rat model of sepsis (induced by CLP) (Ref. Reference Suda97). In contrast, administration of exogenous HMGB1 to mice recapitulates many clinical signs of sepsis, including fever (Ref. Reference O'Connor88), derangement of intestinal barrier function (Ref. Reference Sappington98), and tissue injury (Refs Reference Abraham90, Reference Ueno91). Taken together, these experimental data establish extracellular HMGB1 as a critical late mediator of experimental sepsis, with a wider therapeutic window than early pro-inflammatory cytokines (Fig. 3).
Ischaemic tissue injury
By contrast to the delayed systemic HMGB1 accumulation in experimental sepsis, HMGB1 functions as an early mediator of ischaemia–reperfusion (I–R) injury (Refs Reference Tsung99, Reference Oozawa100, Reference Wu101, Reference Liu102). Prophylactic administration of HMGB1-specific neutralising antibody conferred significant protection against hepatic I–R injury in wild-type mice, but not in a TLR4-defective (C3H/HeJ) mutant, implicating TLR4 in HMGB1-mediated hepatic I–R injury (Ref. Reference Tsung99). Similarly, treatment with HMGB1 antagonist (such as HMGB1 box A) significantly reduced myocardial ischaemic injury in wild-type mice, but in this case not in RAGE-deficient mutants, indicating a potential role for RAGE in HMGB1-mediated ischaemic injury (Ref. Reference Andrassy103).
In addition, HMGB1-specific neutralising antibodies have been proven protective against ventilator-induced acute lung injury (Ref. Reference Ogawa104), severe acute pancreatitis (Ref. Reference Sawa105), and haemorrhagic shock (Ref. Reference Yang106), supporting a pathogenic role for extracellular HMGB1 in various inflammatory diseases. However, HMGB1 is capable of attracting stem cells (Ref. Reference Palumbo64), and may be important for tissue repair and regeneration. Therefore, like other cytokines, extracellular HMGB1 may have protective roles when released at low amounts (Ref. Reference Li, Sama and Wang107). It is thus important to pharmacologically modulate, rather than abrogate, systemic HMGB1 accumulation to facilitate resolution of a potentially injurious inflammatory response.
Other pro-inflammatory mediators of sepsis
In addition to HMGB1, other pro-inflammatory mediators (such as complement anaphylatoxin, C5a and MIF) also accumulate in the circulation in sepsis (Refs Reference Bernhagen21, Reference Calandra22, Reference Bozza23, Reference Czermak108), and contribute to the pathogenesis of sepsis. For instance, blockade of MIF with neutralising antibodies as late as 8 h after onset of experimental sepsis improved survival in mice (Ref. Reference Calandra22). Similarly, blockade of C5a or its cell-surface receptors (C5aR or C5L2) with specific neutralising antibodies protects animals against lethal sepsis (Refs Reference Czermak108, Reference Riedemann109, Reference Rittirsch110), supporting a role for C5a in the pathogenesis of sepsis. Intriguingly, C5L2 may play an important role in the regulation of HMGB1 release, because HMGB1 release was somewhat impaired in C5L2-deficient mice following septic insult, and C5L2-deficient peritoneal macrophages following LPS stimulation (Ref. Reference Rittirsch110). Thus, many known (such as HMGB1, C5a and MIF) or as yet unidentified pro-inflammatory mediators may synergistically interact with each other and collectively contribute to the pathogenesis of sepsis.
Novel HMGB1-targeting therapeutic agents
With a limited number of effective therapies available for patients with sepsis, it is important to search for other agents capable of inhibiting clinically accessible late mediators, such as HMGB1. As discussed below, several agents have been proven protective against experimental sepsis partly through attenuating systemic HMGB1 accumulation (Table 2).
Table 2. Potential therapeutic agents for experimental sepsis
Abbreviations: CLP, cecal ligation and puncture; EGCG, epigallocatechin 3-gallate; HMGB1, high-mobility group box 1; IVIG, Intravenous immunoglobulin; LPC, lysophosphatidylcholine; LPS, lipopolysaccharide; PACAP, pituitary adenylate cyclase-activating polypeptide; TNS IIA-SS, tanshinone IIA sodium sulphonate; VIP, vasoactive intestinal peptide.
Anticoagulant agents
Antithrombin III
Although antithrombin III failed to reduce mortality rate in a large sepsis clinical trial (Ref. Reference Abraham40), a recent study suggested that antithrombin III could attenuate endotoxin-induced systemic HMGB1 accumulation, and reduced endotoxaemic lethality (Ref. Reference Hagiwara111). The mechanisms by which antithrombin III, a liver-derived anticoagulant glycoprotein, inhibits HMGB1 release remain to be investigated.
Thrombomodulin
As mentioned above, another anticoagulant molecule, thrombomodulin, can interact with thrombin to activate protein C. Interestingly, human soluble thrombomodulin (ART-123) can physically bind to HMGB1 protein, thereby inhibiting an HMGB1-mediated inflammatory response. Indeed, ART-123 conferred significant protection against lethal endotoxaemia partly by attenuating HMGB1-mediated inflammatory response (Ref. Reference Abeyama112). It is not yet known, however, whether ART-123 confers similar protection in more clinically relevant animal models of sepsis.
Danaparoid sodium
A third anticoagulant, danaparoid sodium, also prevents blood clotting by inactivating thrombin. It is often used for individuals who cannot be given heparin because of heparin-induced thrombocytopaenia. Intriguingly, danaparoid sodium effectively protected rats against endotoxin-induced acute lung injury by attenuating systemic HMGB1 accumulation (Ref. Reference Hagiwara113).
Intravenous immunoglobulin
Intravenous immunoglobulin (IVIG) refers to IgG immunoglobulins (antibodies) pooled from the plasma of many healthy blood donors. It is usually given intravenously as a plasma protein replacement therapy to patients with various inflammatory diseases due to acute infections, autoimmune disease, or immune deficiencies. A recent study indicated that IVIG dose-dependently protected rats against sepsis-induced lung injury and lethality by attenuating systemic HMGB1 release (Ref. Reference Hagiwara114). The mechanisms by which IVIG suppresses systemic HMGB1 release remain poorly understood. Notably, it has recently been found that human IgGs can bind to HMGB1, and potentially interfere with ELISA detection of HMGB1 (Ref. Reference Urbonaviciute115). It is thus important to ask whether IVIG indeed attenuates systemic HMGB1 accumulation, or simply interferes with ELISA detection of HMGB1 in serum samples.
Endogenous hormones
Insulin
A recent study indicated that hyperglycaemia, induced by infusion of glucose immediately following endotoxaemia, aggravated endotoxin-induced HMGB1 release and lung injury (Ref. Reference Hagiwara116). By contrast, intensive blood glucose control by insulin conferred protection against endotoxin-induced acute lung injury, and endotoxaemic lethality (Ref. Reference Hagiwara116). It is currently unknown whether the observed protective effects are dependent on insulin's anti-inflammatory activities or its blood-glucose-modulating properties (Ref. Reference Wang117).
Neuropeptides
Vasoactive intestinal peptide (VIP) is a short-lived small peptide hormone that is produced by the gut, pancreas and brain. It can induce smooth muscle relaxation, and is involved in communication between brain neurons. In animal models of sepsis induced by CLP or bacteraemia, administration of VIP attenuated systemic HMGB1 accumulation, and consequently reduced animal lethality (Ref. Reference Chorny and Delgado118). Consistently, replenishing septic animals with recombinant HMGB1 completely reversed VIP-mediated protective effects (Ref. Reference Chorny and Delgado118), confirming a pathogenic role for HMGB1 in experimental sepsis.
Another member of the VIP family, the pituitary adenylate cyclase-activating polypeptide (PACAP), shares 68% amino acid sequence identity with VIP. It is abundantly expressed in the central and peripheral nervous systems, and functions as a parasympathetic and sensory neurotransmitter. Interestingly, administration of PACAP peptide also significantly attenuated circulating HMGB1 levels, and similarly protected mice against lethal endotoxaemia (Ref. Reference Tang119).
The neuropeptide urocortin, which belongs to the corticotropin-releasing factor family, is expressed in the brain and may be responsible for regulation of appetite. In animal models of sepsis induced by CLP or bacteraemia, administration of urocortin attenuated systemic HMGB1 accumulation and reduced animal lethality (Ref. Reference Chorny and Delgado118), supporting a therapeutic potential for neuropeptides in experimental sepsis.
Ghrelin
Ghrelin is a stomach-derived hormone that is responsible for regulating the appetite – increasing it before eating and decreasing it afterwards. Intriguingly, plasma ghrelin levels are significantly decreased in septic animals (Ref. Reference Wu120), and administration of ghrelin promoted a dose-dependent protection against sepsis-induced acute lung injury and lethality (Refs Reference Wu120, Reference Wu121, Reference Chorny122). Ghrelin may exert its protective effects through multiple mechanisms, such as by attenuating systemic HMGB1 release and by facilitating bacterial elimination (Ref. Reference Chorny122). Intriguingly, ghrelin may attenuate systemic accumulation of pro-inflammatory cytokines partly via the vagus nerve (Ref. Reference Wu121), suggesting that pharmacological stimulation of the vagus nerve may be an effective therapy for experimental sepsis.
Vagus nerve stimulation
The vagus nerve is the structural basis for the cholinergic anti-inflammatory pathway (Ref. Reference Tracey123), which inhibits the innate immune response via the release of acetylcholine. Acetylcholine binds to α7 nicotinic acetylcholine receptors of various innate immune cells (Ref. Reference Wang124), thereby counter-regulating potentially injurious innate immune responses. Indeed, stimulation of the vagus nerve by physical methods (e.g. electrical or mechanical) (Refs Reference Huston125, Reference Huston126) or chemical agents (such as the cholinergic agonists nicotine and GTS-21) (Refs Reference Wang127, Reference Pavlov128) conferred protection against lethal endotoxaemia and sepsis partly by attenuating systemic HMGB1 accumulation.
Stearoyl lysophosphatidylcholine
An endogenous phospholipid, stearoyl lysophosphatidylcholine (LPC), has recently been proven protective against experimental sepsis by stimulating neutrophils to destroy ingested bacteria in a mechanism dependent on hydrogen peroxide (Ref. Reference Yan129). However, stearoyl LPC also confers protection against lethal endotoxaemia (Ref. Reference Yan129), implying that it may exert protective effects through an additional, bactericidal-independent mechanism (Ref. Reference Wang, Czura and Tracey130). Indeed, administration of stearoyl LPC significantly attenuated circulating HMGB1 levels (Ref. Reference Chen47), indicating that stearoyl LPC protects against experimental sepsis partly by facilitating elimination of invading pathogens and partly by attenuating systemic HMGB1 accumulation (Ref. Reference Wang, Czura and Tracey130).
Ethyl pyruvate
Ethyl pyruvate is an aliphatic ester derived from pyruvic acid, which is a final product of glycolysis and the starting substrate for the tricarboxylic acid cycle (Ref. Reference Fink131). It dose-dependently inhibits LPS-induced release of early (e.g. TNF) and late (e.g. HMGB1) pro-inflammatory cytokines, and protected mice against experimental sepsis even when treatment was started as late as 12–24 h after the onset of disease (Ref. Reference Ulloa132).
Chinese medicinal herbs
Traditional herbal medicine has formed the basis of folk remedies for various inflammatory ailments. Out of several dozen commonly used Chinese herbs (Ref. Reference Zhu133), we found that aqueous extracts of danggui (Angelica sinensis), green tea (Camellia sinensis), and danshen (Salvia miltorrhiza) efficiently inhibited endotoxin-induced HMGB1 release, and protected animals against experimental sepsis (Refs Reference Li84, Reference Wang134, Reference Li135).
Danggui
Danggui has been traditionally used to treat gynaecological disorders (such as abnormal menstruation) (Ref. Reference Piersen136). Its aqueous extract dose-dependently inhibited LPS-induced HMGB1 release in macrophage and monocyte cultures, partly by interfering with HMGB1 cytoplasmic translocation (Ref. Reference Wang134). Furthermore, danggui extract rescued mice from lethal sepsis even when the first dose was given at 24 h after onset of disease (Ref. Reference Wang134). The active components responsible for these beneficial effects remain a subject of future investigation.
Green tea
Green tea brewed from the leaves of Camellia sinensis contains a class of biologically active polyphenols called catechins. Epigallocatechin 3-gallate (EGCG), which accounts for 50–80% of the total catechin, is effective in attenuating endotoxin-induced HMGB1 release by macrophage and monocytes (Ref. Reference Li84). In addition, EGCG dose-dependently inhibited HMGB1-induced release of TNF, IL-6 and nitric oxide in macrophage cultures (Ref. Reference Li84). Interestingly, EGCG completely abrogated accumulation/clustering of exogenous HMGB1 on the macrophage cell surface (Ref. Reference Li84), suggesting that EGCG inhibits HMGB1 cytokine activities by preventing its cell-surface accumulation/clustering.
In vivo, repeated administration of EGCG conferred a dose-dependent protection against lethal endotoxaemia, and rescued mice from lethal sepsis even when the first dose of EGCG was given at >24 h after onset of sepsis (Ref. Reference Li84). Consistently, delayed administration of EGCG significantly attenuated circulating levels of HMGB1, as well as surrogate markers of experimental sepsis (such as IL-6 and chemokine KC) (Refs Reference Li84, Reference Osuchowski137). Considered together, these experimental data indicate that EGCG protects mice against lethal sepsis partly by attenuating systemic HMGB1 accumulation, and partly by inhibiting HMGB1-mediated inflammatory response.
Danshen
Danshen has been widely used in China for patients with cardiovascular disorders (Refs Reference Ji, Tan and Zhu138, Reference Cheng139). Danshen contains abundant red pigments (termed tanshinone I, tanshinone IIA, and cryptotanshinone) (Fig. 4), which effectively attenuated LPS-induced HMGB1 release (Ref. Reference Li135). A water-soluble derivative (sodium sulphonate) of tanshinone IIA (TSN IIA-SS) at concentrations (100 µm) that completely abrogated LPS-induced HMGB1 release, only partially attenuated LPS-induced release of four out of 62 cytokines (IL-12p70, IL-1α, platelet factor 4 and CCL12) (Ref. Reference Li135), indicating a specificity for TSN IIA-SS in inhibiting LPS-induced HMGB1 release. Despite a structural resemblance (i.e. the presence of a fused four-ring structure) between tanshinones and steroidal anti-inflammatory drugs (such as dexamethasone and cortisone) (Fig. 4), tanshinones inhibit LPS-induced HMGB1-release in a glucocorticoid-receptor-independent mechanism (Ref. Reference Li135).
Figure 4. Steroid-like tanshinones and water-soluble derivatives. Several steroid-like pigments (tanshinone I, tanshinone IIA, and cryptotanshinone) of the medicinal Chinese herb danshen (Salvia miltiorrhiza) are structurally similar to steroidal anti-inflammatory drugs (such as dexamethasone and cortisone) – that is, they all have a fused four-ring structure (Ref. Reference Zhu133). Tanshinone IIA sodium sulphonate is water-soluble, and widely used in China as a cardiovascular medicine. MW, molecular weight.
More importantly, repeated administration of TSN IIA-SS, beginning at >24 h and followed by additional doses at >48, >72 and >96 h after the onset of sepsis, dose-dependently rescued mice from lethal sepsis (Ref. Reference Li135). Notably, administration of TNS IIA-SS dose-dependently attenuated circulating HMGB1 levels in septic mice (Ref. Reference Li135), suggesting that TSN IIA-SS confers protection against experimental sepsis partly by inhibiting systemic HMGB1 accumulation.
Clinical implications
For complex systemic inflammatory diseases such as sepsis, it appears difficult to translate successful animal studies into clinical applications. For instance, although neutralising antibodies against endotoxin (Ref. Reference Davis24) or cytokines (e.g. TNF) (Refs Reference Tracey17, Reference Beutler, Milsark and Cerami140) are protective in animal models of endotoxaemia or bacteraemia, these agents failed in sepsis clinical trials (Refs Reference Ziegler141, Reference Ziegler142, Reference Abraham143). This failure partly reflects the complexity of the underlying pathogenic mechanisms of sepsis and the heterogeneity of the patient population (Refs Reference Dellinger2, Reference Cohen144). It may also be attributable to pitfalls in the selection of (Reference Serhan and Savill1) feasible therapeutic targets or drugs, (Reference Dellinger2) optimal doses and timing of drugs, and (Reference Luster, Alon and von Andrian3) nonrealistic clinical outcome measures (such as mortality rates) (Refs Reference Dellinger2, Reference Cohen144).
Nevertheless, the investigation of pathogenic cytokines in animal models of diseases has led to the development of anti-TNF therapy for patients with debilitating chronic inflammatory diseases, such as rheumatoid arthritis (Ref. Reference Feldmann and Maini145). Consequently, a chimaeric anti-TNF monoclonal antibody (infliximab) and a soluble TNF-receptor–Fc fusion protein (sTNF-R–Fc; etanercept) have been approved by regulatory authorities in the USA and Europe for treating rheumatoid arthritis. Since pro-inflammatory cytokines are indeed pathogenic in human inflammatory diseases (such as rheumatoid arthritis), it is necessary to continue the search for clinically feasible therapeutic targets and drugs for other inflammatory diseases (such as sepsis).
Will HMGB1 ever become a clinically feasible therapeutic target for human sepsis? We cannot answer this question until HMGB1-neutralising antibodies have been tested for efficacy in large clinical trials. Although HMGB1 appears to be a feasible therapeutic target for experimental sepsis (Refs Reference Wang96, Reference Wang, Yang and Tracey146, Reference Lotze and Tracey147), its levels in unfractionated crude serum of septic patients did not correlated well with disease severity (Refs Reference Sunden-Cullberg148, Reference Angus149). Upon separation of serum proteins by ultrafiltration through membrane with a defined molecular weight cut-off (100 kDa), a 30 kDa HMGB1 band was detected (by western blotting analysis) in both low (<100 kDa) and high (>100 kDa) molecular weight serum fractions of many septic patients. Furthermore, HMGB1 levels in the low (<100 kDa) serum fraction were significantly higher in septic patients who died of sepsis than those who survived (Ref. Reference Wang51). This observation suggested a possibility that HMGB1 may interact with other serum components to form large (>100 kDa) complexes.
Indeed, many exogenous bacterial products (such as endotoxin or CpG-DNA) (Refs Reference Ivanov52, Reference Tian71, Reference Silva72) or endogenous proteins (such as human thrombomodulin, IgG1 or IL-1) (Refs Reference Urbonaviciute115, Reference Sha150) may physically interact with HMGB1 to form various complexes. It is not yet known how these and as yet unidentified HMGB1-binding molecules affect the biological activities, or immunodetection, of HMGB1 in septic patients (Refs Reference Urbonaviciute115, Reference Sunden-Cullberg148). In addition, chemical modifications may similarly affect the biological activities of HMGB1. For instance, a recent study indicated that reactive oxygen species (ROS) may oxidise HMGB1 to form an intramolecular disulphide bond between the thiol group of Cys106 and Cys23 or Cys45, and consequently abolish HMGB1-mediated immunostimulatory activities (Ref. Reference Kazama151). Because Cys106 is located within the 18-amino-acid cytokine domain of HMGB1 B box, it will be important to investigate whether oxidisation similarly affects biological activities of HMGB1 in future studies.
Will any specific HMGB1 inhibitor ever become a therapeutic agent for human sepsis? One of the most selective HMGB1 inhibitors, TSN IIA-SS, has already been used in China as a medicine for patients with cardiovascular disorders (Ref. Reference Ji, Tan and Zhu138). Even in septic animals, TSN IIA-SS reduced total peripheral vascular resistance, and yet increased cardiac stroke volume and cardiac output (Ref. Reference Li135). Because HMGB1 may function as a myocardial depressant factor by reducing contractility of cardiac myocytes (Ref. Reference Tzeng152), it is plausible that TSN IIA-SS improves cardiovascular function partly by attenuating HMGB1 release. The dual effects of TSN IIA-SS in attenuating late inflammatory response and improving cardiovascular function make it a promising therapeutic agent for sepsis.
Conclusions and perspectives
The ubiquitous nuclear protein HMGB1 is released by activated macrophages/monocytes, and functions as a late mediator of experimental sepsis. First, circulating HMGB1 levels are elevated in a delayed fashion in endotoxaemic and septic animals. Second, administration of exogenous HMGB1 to mice induces fever, derangement of intestinal barrier function, and tissue injury. Third, administration of anti-HMGB1 antibodies or inhibitors rescues mice from lethal experimental sepsis even when the first dose is given 24 h after onset of sepsis. Taken together, these data establish HMGB1 as a late mediator of experimental sepsis with a wider therapeutic window than early mediators such as TNF.
HMGB1-specific neutralising antibodies and small-molecule inhibitors (such as tanshinone IIA derivative) have been proven protective in animal models of experimental sepsis. Currently, the intricate mechanisms by which various agents attenuate systemic HMGB1 release and protect against experimental sepsis remain poorly understood. In addition, it is not yet known whether a better protection could be achieved by combinational therapy with several anti-HMGB1 agents. It is thus important to further explore the therapeutic potential of these HMGB1-inhibiting agents in future studies.
Acknowledgements and funding
We are grateful to the peer reviewers for their critical and constructive comments. Work in the authors' laboratories was supported by grants from the National Institutes of Health, National Institute of General Medical Science (R01GM063075 and R01GM070817 to H.W.).
