Humoral and cell-mediated immune responses represent a highly specific and sophisticated defence mechanism against invading pathogens associated with minimal injury to host tissue. Unfortunately a considerable time must generally elapse following infection before these specific responses are evoked (in the absence of previous exposure or vaccination). Antigen-specific immune responses are usually preceded by a non-specific acute inflammatory response of varying degree. Host responses to inflammation are often similar whether evoked by Gram-positive or Gram-negative bacteria, viruses, parasites, or even in ‘sterile’ circumstances such as in acute pancreatitis or severe burns. In certain clinical situations the inflammatory response may aid survival but, as shown in Fig. 1 63, a penalty must always be paid in terms of host autoinjury. Sometimes the inflammatory response is itself the lethal event, the invading micro-organisms otherwise being quite harmless.

The concept of host autoinjury as the principal cause of death in many infections was given strong support when James D. Watson (of DNA double helix fame) demonstrated that a single mutation in a certain strain of mouse (C3H HEJ) was associated with virtually complete protection against the effects of an otherwise lethal injection of live Gram-negative bacteria or their associated endotoxins. By transplanting cells between ‘sensitive’ and ‘resistant’ strains it could be shown that the lethality of endotoxin was due to proteins synthesized by reticuloendothelial cells. This, and other research, enhanced the growing belief that the key to enhancing survival following severe sepsis did not lie with better antibiotics but with therapies that prevented induction of the inflammatory response to a lethal degree. Against this was the concern that blunting the inflammatory response might itself render the host succeptible to lethal secondary infection.

A major breakthrough in the understanding of sepsis has occurred in the last decade with the revolution in molecular biology. Techniques have been developed to characterize and assay the mediators that are evoked following severe inflammation which cause host autoinjury and death. These mediators are collectively refered to as cytokines.

Although there is no universally accepted definition, to be considered a cytokine a molecule must show the following features.

1.It must be a protein or polypeptide.

2.It must be a mediator of a component of inflammation.

3.It must be evoked as part of the immune response to an antigen (although micro-organisms acting through different receptors may be far more potent inducers than antigen-generated stimuli).

4.It must have no intrinsic chemical or enzymatic activity.

5.It must bind to specific protein receptors on target cells.

6.It must alter the behaviour of the target cell.

Cytokines may be regarded as the hormones of disease and like true hormones the target cell may be the cell of origin (autocrine action), an adjacent cell (paracrine action), or a distant cell (endocrine action). Cytokines are biologically active in concentrations a thousandfold less than true hormones. They also exist in multiple forms. Finally their biological effects long outlive their appearance in the circulation. These considerations explain why it has been so difficult to monitor cytokine activity during human critical illness.

A large number of cytokines (>30) have now been characterized. The bulk of available evidence suggests that the following cytokines are most important in the mediation of lethal sepsis: (1) tumour necrosis factor; (2) interleukin-1; (3) interleukin-6; (4) interleukin-8. Although these cytokines were originally believed to be principally the products of monocytes/macrophages it is now clear that they are synthesized to a significant extent by other cells including neutrophils, lymphocytes, endothelial cells, and glial cells in the central nervous system. The following is a brief description of the key inflammatory cytokines.

Tumour necrosis factor is synthesized as a 25-kDa precursor and processed to produce a 17-kDa product which is actively secreted into the circulation. A large amount of evidence supports a key role for this cytokine in critical illness.

1.Administration of tumour necrosis factor to laboratory animals or cancer patients will induce the entire spectrum of host alterations associated with severe infection in a dose-dependent manner.

2.In human volunteers receiving intravenous Escherichia coli endotoxin, the appearance of tumour necrosis factor in the circulation is temporally and quantitatively related to the clinical response to endotoxaemia.

3.In patients with overwhelming acute infection, e.g. meningococcal septicaemia or cerebral malaria there is a striking association between circulating TNF levels and the likelihood of death.

4.Agents which sensitize human beings or animals to the adverse effects of endotoxin often act to increase the biosynthesis of tumour necrosis factor (e.g. Corynebacterium parvum, Bacillus Calmette-Guerin, or experimentally d-galactosamine).

5.Monoclonal antibodies which neutralize tumour necrosis factor have been shown to prevent death following administration of a lethal (LD&sub1;&sub0;&sub0;) dose of live Gram-negative organisms to animals, including primates).

The last piece of evidence is of key importance as it emphasizes the point that it is host-generated signals and not bacterial poisons per se that cause death in lethal acute Gram-negative infection.

Interleukin-1 exists in &agr; and &bgr; forms of which the latter is considered the more important mediator of sepsis. It is a 33-kDa polypeptide that has little structural similarity to tumour necrosis factor and acts through an independent receptor. Unlike tumour necrosis factor it has no leader sequence that allows its secretion by cells so that it remains cell bound and therefore extremely difficult to detect during sespis. Detectable levels of interleukin-1&bgr; may reflect lysis of the cell of origin rather than active secretion.

Administration of interleukin-1&bgr; to animals and human beings will replicate virtually all the responses seen in acute infection. It appears to be about 100 times less potent than tumour necrosis factor and has never been shown to be intrinsically lethal to any animal. It magnifies greatly the biological effects of tumour necrosis factor (by a factor of about 100) and it is currently believed that much of the lethality of acute infection is attributable to tumour necrosis factor and interleukin-1&bgr; acting synergistically. Furthermore tumour necrosis factor stimulates interleukin-1&bgr; production and agents which block the interleukin-1&bgr; receptor also prevent death in primates receiving an LD&sub1;&sub0;&sub0; dose of Escherichia coli endotoxin.

Interleukin-6 is a 26-kDa polypeptide produced principally by monocytes or endothelial cells which have been activated by tumour necrosis factor or interleukin-1&bgr;. Monoclonal antibodies against interleukin-6 have also been shown to provide protection in animals against lethal Gram-negative infection. Interleukin-6, when administered as a sole agent appears to be of low toxicity, the only significant host response being a marked increase in hepatic acute-phase protein synthesis. Recent evidence suggests that tumour necrosis factor or interleukin-1&bgr; increase the expression of GP 130, a target cell-associated glycoprotein that binds to the interleukin-6/interleukin-6 receptor complex and greatly enhances the intrinsic toxicity of this cytokine.

Induction of this 6- to 10-kDa polypeptide appears to parallel that of interleukin-6 and it resembles the latter in that it appears to have low intrinsic toxicity. However, when it is present in association with tumour necrosis factor or interleukin-1&bgr; it appears to have potent neutrophil activating and chemotactic properties which may well magnify the toxicity of the other cytokines. No definite role has yet been established for this cytokine in human sepsis.

Synthesis of tumour necrosis factor appears to be strongly repressed during health and it is not detectable in the circulation. Cells triggered by endotoxin or various other stimuli become ‘superinduced’ and elaborate tumour necrosis factor as their principal secretory product. Low levels of endotoxaemia appear to trigger tumour necrosis factor without significant circulating interleukin-1&bgr; and in this cytokine milieu, fever and neurohormonal and acute-phase protein alterations occur but without lethal responses such as hypotension or coagulopathy. These lethal responses occur if tumour necrosis factor levels become high enough for significant secondary synthesis of interleukin-1&bgr; or the other synergizing cytokines to occur. These observations may explain why monoclonal antibodies against any of the key cytokines, when given prophylactically, will prevent death following lethal bacterial challenge in animals.

It is well recognized that individuals differ greatly in their systemic response to a infected focus. For example, 20 per cent of individuals with established acute appendicitis show no fever, tachycardia, anorexia, or neutrophil leucocytosis. The remainder exhibit these changes to a varying extent. Even established peritonitis may be associated with minimal signs of systemic toxicity in some individuals while others rapidly progress to septic shock and death. Recent evidence suggests that this individual variation may be related to differences between individuals in the magnitude of the cytokine response following a standard stimulus. It has been shown that a variation of over one hundredfold exists between individuals in terms of monocyte-secreted tumour necrosis factor following a standard dose of endotoxin both in vitro and in vivo. The endotoxin-resistant mouse (C3H HEJ) has a defect on chromosome 4 that prevents synthesis of tumour necrosis factor. It appears that there are comparable human beings who secrete very little tumour necrosis factor after endotoxin stimulation (‘low responders’). It has recently been shown that responder status is closely related to haplotype on the Class 2 histocompatibility complex (i.e. tissue type). Tissue typing may be a method which can be used in the future to identify, prospectively, individuals at high risk of death if sepsis ensues (e.g. following major gastrointestinal surgery).

The induction of the cytokine cascade represents an inappropriately severe host response to an inflammatory stimulus and leads to death through the following mechanisms.

1.Hypothermia;

2.Coagulopathy;

3.Visceral hypoperfusion and reperfusion injury;

4.Lactate acidosis and acidaemia;

5.Myocardial depression;

6.Peripheral vasodilation;

7.The capillary leak syndrome which is believed to induce profound hypotension and multiple organ failure.

All of these alterations may be induced by tumour necrosis factor or interleukin-1&bgr; but they do not act directly. Figure 2 64 illustrates the principal efferent mechanisms by which cytokines exert their lethal effects.

This field is still in its infancy and only a small fraction of the potential agents that may be of value have been tested in infected human beings. In severe surgical sepsis (e.g. pancreatic abscess or acute gastrointestinal anastomosis disruption) mortality is still ‘sticking’ at 50 to 70 per cent and there is little prospect that better antibiotics or haemodynamic support will alter this. There is good reason, however, to believe that modification of the cytokine cascade might reduce mortality. Figure 2 64 illustrates the theoretical ways in which this could be done. The advantages and disadvantages of each approach will be considered briefly.

Blockade of endotoxin is an attractive concept because the toxic moiety of the molecule (lipid A) is common to all Gram-negative strains. Furthermore endotoxin is not a fundamental component of the immune system and therefore inhibiting endotoxin should not cause immunosuppression in any way. A human monoclonal antibody against lipid A has recently been evaluated in Phase 3 clinical trials and has been shown to reduce mortality following Gram-negative bacteraemia by 39 per cent. This represents a major advance in the therapy of Gram-negative infection but the use of this agent is complicated by the following considerations: (1) it is expensive; (2) there is no ideal way to predict whether shock is caused by a Gram-negative organism while there is still a therapeutic window and probably only 40 per cent of individuals receiving this agent will, in retrospect be candidates to benefit from it; (3) it is ineffective against septic shock induced by non-Gram-negative pathogens. Other methods which have been proposed for modifying cytokine production following endotoxaemia include:

1.Blocking the effects of endotoxin on its target cell. The mechanism whereby endotoxin triggers the target cell to secrete cytokines has recently been clarified. Firstly the lipid A component of endotoxin binds to the LPS binding protein (LBP). The endotoxin/LBP complex then binds to the CD14 receptor of target cells which acts as the signal triggering cytokine synthesis. It is possible that anti-CD14 antibodies may render target cells resistant to endotoxin but this has not been investigated therapeutically. Additionally, the azurophilic granules of neutrophils contain an endogenous inhibitor of the LPS binding protein called bactericidal permeability-increasing protein (BPI). Levels of bactericidal permeability-increasing protein appear to be deficient in severe sepsis and it can be shown that exogenously administered bactericidal permeability-increasing protein significantly protect animals from the lethal effects of endotoxin. The clinical evaluation of bactericidal permeability-increasing protein is about to commence.

2.Blocking the intracellular signal generated by the interaction between endotoxin, its binding protein, and the CD14 receptor. Unfortunately this secondary signal has not yet been identified but its characterization might allow therapies that rendered the individual temporarily resistant to the effects of endotoxin.

3.Blockade of cytokine synthesis. Glucocorticoids are potent inhibitors of tumour necrosis factor synthesis but only if given prophylactically. It appears that these agents stimulate inhibitors that impede the signals stimulating cytokine synthesis but are powerless to arrest activation once it is in progress. Current clinical evidence does not favour a role for these agents in most instances of severe sepsis.

This is an extremely promising approach as unlike antiendotoxin antibodies these agents might be more generally applicable in severe inflammatory states rather than just those initiated by endotoxin. Clinical trials have now commenced with antitumour necrosis factor and anti-interleukin-1 antibodies and results should soon be available. Experimentally they are most effective when given prophylactically and concerns have been expressed as to their potential role in established sepsis. Against this animal models of lethal endotoxaemia are not really representative of typical human sepsis and septic shock and there is a small body of evidence suggesting that antitumour necrosis factor antibodies significantly attenuate hypotension associated with septic shock, although it has not yet been shown that they diminish mortality.

Concerns have been raised about the inherent immunogenicity of anticytokine antibodies and the use of natural or synthetic inhibitors of cytokine function may circumvent this problem. Approaches undergoing or about to undergo clinical evaluation include:

1.Use of a naturally occuring interleukin-1 receptor antagonist. Animals or human beings exposed to endotoxin elaborate a 23- to 25-kDa peptide that competes with interleukin-1 for occupancy of its receptor but has no agonist properties. This moiety is termed the interleukin-1 receptor antagonist (IL-1 ra). It prevents death in otherwise lethal endotoxaemia. It has also been tested in Phase II studies in patients with various types of sepsis (Gram-negative, Gram-positive, and fungal) and was found to reduce mortality significantly (from 45 to 16 per cent). If this finding is confirmed in double-blind clinical trials it will represent a major advance in the therapy for severe sepsis.

2.Use of soluble tumour necrosis factor receptors. Soluble tumour necrosis factor receptors are detectable in the bloodstream during health and become elevated during severe sepsis. Again the shedding of receptors into the bloodstream may represent a beneficial host response to limit cytokine activity and this protective mechanism may become overwhelmed in severe sepsis. Gram-negative septic shock in the baboon may be attenuated by use of large amounts of soluble tumour necrosis factor receptors administered exogenously. Attempts are being made to conjugate soluble tumour necrosis factor receptor fragments to an antibody to increase their half-life in the circulation.

Although anticytokine therapy is in its infancy it is clear that following severe inflammatory or infective episodes the host not only releases large amounts of potentially lethal cytokines but also releases potential antidotes to these cytokines such as soluble receptors and endogenous antagonists. Lethal effects of cytokines are probably seen when these antidotes are exhausted, leading to unopposed synergistic cytokine activity. The current interpretation is that both interleukin-1 and tumour necrosis factor orchestrate the deleterious effects of infection and that blocking the activity of either one of these cytokines prevents the full consequences of the disease.

As stated earlier septic shock and multiple organ failure have a multifactorial aetiology but surgical patients are particularly at risk.

The major thrust of cytokine research has been involved with attenuating the host response to acute infection. It is becoming apparent that cytokines such as tumour necrosis factor and interleukin-1 may also mediate chronic events in sepsis including immunosuppression, accelerated proteolysis, and generalized cachexia and anorexia. At present the data are confusing and controversial and different but similar models reveal conflicting results. If cytokines are of importance as mediators of cachexia it is probable that they act in concert with other mediators such as stress hormones and by altering neurohormonal set-points within the hypothalamus and elsewhere within the central nervous system.

Beutler B. Cytokines in shock 1992. In: Lamy M, Thijs LG, eds. Update in Intensive Care and Emergency Medicine. Berlin: Springer-Verlag, 1993: 51–67.

Dinarello CA. Blocking cytokines in infectious disease. In: Lamy M, Thijs LG, eds. Update in Intensive Care and Emergency Medicine. Berlin: Springer-Verlag, 1993: 362–76.

Dinarello CA, Wolff SM. Mechanisms of disease—the role of interleukin-1 in disease. N Engl J Med 1993; 328: 106–13.

Lowry SF, van Zee KJ, Moldawer LL. Strategies for modulation of tissue cytokine responses to sepsis. In: Lamy M, Thijs LG, eds. Update in Intensive Care and Emergency Medicine. Berlin: Springer-Verlag, 1993: 345–61.

Michie HR, et al. Detection of circulating tumor necrosis factor following endotoxin administration. N Engl J Med 1988; 318: 1481–6.

Ohlsson K, et al. Interleukin-1 receptor antagonist reduces mortality from endotoxin shock. Nature 1990; 348: 550–2.

Ziegler EJ, et al. Treatment of Gram-negative bacteremia and shock with human antiserum to a mutant Escherichia coli. N Engl J Med 1982; 307: 1225–30.

Ziegler EJ, et al. Treatment of Gram-negative bacteremia and septic shock with HA–1A human monoclonal antibody against endotoxin. N Engl J Med 1991; 324: 429–36.