Cytokines and Escherichia coli Sepsis

CHLOÉ KAECH,¹ PIERRE-YVES BOCHUD,² AND THIERRY CALANDRA¹*

[SECTION EDITORS: FERRIC FANG AND MARTIN KAGNOFF]

Posted September 1, 2006

Infectious Diseases Service, Department of Medicine, Centre Hospitalier Universitaire Vaudois, CH-1011 Lausanne, Switzerland,¹ and Institute for Systems Biology, 1441 North 34th Street, Seattle, WA 98103²

*Corresponding author. Phone: 41-21-314-1037, Fax: 41-21-314-1033, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

In the twenty-first century, sepsis remains an important cause of infection-related morbidity and mortality. Encouragingly, an epidemiological survey performed between 1979 and 2000 in acute care hospitals in the United States indicated that the in-hospital mortality of sepsis has decreased from 27.8% for the period 1979 to 1984 to 17.9% for the period 1995 to 2000, suggesting that significant advances in the management of these patients have been made during the past 20 years (104). However, there is still an important margin for improvement, as severe sepsis and septic shock, the two feared life-threatening complications of sepsis, remain associated with a mortality in the range of 30% to 70% (12).

The epidemiology of sepsis has changed during the past decades. Several reports have revealed a major shift in the proportion of bloodstream infections caused by gram-negative and gram-positive bacteria (29,104,139). While gram-negative bacteria were predominant in the 1960s and early 1970s, the frequency of gram-positive bacteremia has markedly increased over the past 30 years. We have recently compiled microbiological data from 30 large clinical trials or epidemiological studies (30). As shown in Table 1, Escherichia coli and staphylococci are the most frequent bacterial pathogens isolated from the blood of patients with severe sepsis and septic shock (23,30,104). Nowadays, gram-positive bacteria are a very common cause of severe sepsis and septic shock; in some institutions, staphylococci have surpassed E. coli.

Table 1Microorganisms most frequently isolated from patients with severe sepsis and septic shock

Considerable progress has been made in understanding the pathogenesis of sepsis over the past 20 years. One of the most substantial achievements in the field of sepsis has been the recognition of the essential role played by the innate immune system in the natural host defense against microbes (77, 85, 106). Sensing of microbial pathogens, either in tissues in contact with the host’s environment or in the systemic circulation after invasion of the bloodstream, is carried out by sentinel cells of the innate immune system (i.e., monocytes/macrophages, dendritic cells, natural killer cells, and granulocytes). Microbial products such as endotoxin, peptidoglycan, and other bacterial cell wall components are powerful activators of innate immune cells. These microbe-associated molecular motifs bind to a family of microbial recognition molecules expressed by immune and nonimmune cells, including CD14, Toll-like receptors (Chapter Toll-Like Receptors), NLR/NOD proteins ( Chapter NLRs: Nucleotide-Binding Domain and Leucine-Rich-Repeat-Containing Proteins), and peptidoglycan-recognition proteins. Ligand-activated receptors turn on signal transduction pathways and the transcription of immune genes, resulting in the expression of costimulatory molecules at the cell surface and in the release of immunoregulatory effector molecules in the extracellular compartment.

Cytokines are an important family of mediators that orchestrate the innate and adaptive host defense responses. They are small proteins (usually less than 30 kDa), whose expression is, with few exceptions, induced rather than constitutive. Pleiotropism (i.e., the capacity for a given cytokine to stimulate several cell types) and redundancy (i.e., the ability of different cytokines to exert similar effects) are typical characteristics of these effector molecules. Cytokines frequently stimulate each other’s expression, giving rise to a broad and dense network of interacting molecules that exert autocrine, paracrine, and endocrine activities through interactions with specific receptors expressed on target cells. Tumor necrosis factor, the interleukins, the chemokines, the interferons, and the colony-stimulating factors are all members of the cytokine family of messenger molecules that are released by both immune and nonimmune cells. Cytokines have been implicated in the regulation of both proinflammatory and anti-inflammatory host responses that are key players in the pathogenesis of sepsis. Proinflammatory cytokines recruit and activate cells of the innate immune system and induce the production of proinflammatory mediators such as cytokines themselves, eicosanoids, platelet-activating factor, and free radicals (including nitric oxide). Proinflammatory cytokines also enhance the expression of the major histocompatibility complex I (MHC I) and II molecules and participate in the activation and proliferation of B and T lymphocytes. Anti-inflammatory cytokines, on the other hand, mitigate the inflammatory process by either inhibiting or counteracting these effects. Today we know that the clinical manifestations of severe sepsis and septic shock result from an overwhelming proinflammatory cytokine response. If the host survives the early stage of disease, the proinflammatory phase is followed by a compensatory anti-inflammatory phase that leads to immunosuppression and increased susceptibility to bacterial or fungal superinfection (see Fig. 1).

Fig. 1Schematic representation of the role of cytokines in innate immune responses against microbial pathogens. +, beneficial effects; −, detrimental effects; MAMP, microbe-associated molecular patterns; MRR, microbial recognition receptors.

In this chapter, we review the critical role played by cytokines in the pathogenesis of E. coli sepsis. We focus our attention on prototypic proinflammatory and anti-inflammatory cytokines and their influence on mortality in experimental animal models of E. coli endotoxemia and of live E. coli sepsis. For the sake of clinical relevance, we also review the results of clinical trials on anticytokine therapy in patients with severe sepsis or septic shock. Readers who would like to acquire additional or more in-depth information on specific cytokines should refer to the comprehensive and systematic reviews of cytokines published elsewhere (69, 125).

PROINFLAMMATORY CYTOKINES

TNF

Tumor necrosis factor (TNF) is a secreted 17-kDa cytokine produced by a broad range of cells, including myeloid cells (monocytes, macrophages, dendritic cells, and neutrophils). Together with lymphotoxin-α, lymphotoxin-β, and Fas ligand, TNF belongs to a family of ligands that bind to a group of structurally related receptors comprising the two TNF receptors, the lymphotoxin beta receptor, and the TNF/nerve growth factor family (reviewed in reference 155). TNF itself binds to two receptors coexpressed on a majority of cells and tissues: the TNF type I receptor (TNFRI, also designated as p55-TNFR) and the TNF type II receptor (TNFRII, also designated as p75-TNFR) (reviewed in reference 143). TNF activates myeloid cells and triggers the synthesis of proinflammatory mediators such as cytokines (including TNF itself), eicosanoids, platelet-activating factor, and free radicals (including nitric oxide). Cells of the immune system are recruited to sites of inflammation by TNF via the induction of endothelial adhesion molecules and chemokines. TNF also plays an important role in the synthesis of acute-phase proteins by hepatocytes. TNF is a potent inducer of apoptosis in many cell types, including inflammatory cells, fibroblasts, and myocytes. TNF is a pyrogenic cytokine that causes anorexia and can induce shock by decreasing vascular resistance, causing capillary leak, and depressing myocardial function.

Both TNF receptors are shed from the cell membrane after cell activation and circulate as soluble molecules in the bloodstream. Soluble p55-TNFR and p75-TNFR function as naturally occurring inhibitors of TNF activity (140, 150). Sensu stricto, soluble TNF receptors are anti-inflammatory molecules, but they are discussed in this section together with TNF. Shedding of the TNF receptors renders cells hyporesponsive to the actions of TNF and thereby protects them from overstimulation. Moreover, soluble TNF receptors bind TNF, acting as decoy receptors to prevent cell activation. However, binding of TNF to its soluble receptors may prolong its half-life, and TNF may be released from the receptor at a later stage of disease, as has been shown for p75-TNFR (8, 61, 110, 111).

Experimental Data.

The recognition of the critical role played by TNF in endotoxic and gram-negative shock has been a major step forward in our understanding of the pathogenesis of sepsis (21). TNF was found to be a proximal mediator of the inflammatory response induced by endotoxin (107). Infusion of TNF in animals reproduced the symptoms and signs of sepsis (145), while anti-TNF antibodies protected mice or baboons from shock induced by endotoxin (22, 146). Several groups of investigators subsequently confirmed that anti-TNF antibodies conferred protection against E. coli endotoxic shock (Table 2). Furthermore, treatment with anti-TNF antibodies reduced mortality in animals injected intravenously with live E. coli. Note, however, that the protective effects of anti-TNF antibodies were generally lost when treatment was administered after endotoxin or bacteria. Mice carrying a knockout mutation in p55-TNFR were also protected against E. coli endotoxemia. In E. coli peritonitis, blocking TNF activity with anti-TNF antibodies administered before or at the time of infection, or by employing p55-TNFR knockout mice, did not improve mortality in lethal peritonitis models and reduced survival in sublethal infection models (Table 2).

Table 2Evaluation of the role of TNF, IL-1, and IL-6 in animal models of Escherichia coli endotoxemia and sepsis

Taking advantage of the fact that soluble TNF receptors act as decoy receptors, chimeric proteins were engineered by fusing the extracellular portion of p55-TNFR or p75-TNFR with the Fc portion of human immunoglobulin G (IgG). Four chimeric constructs were produced, using either IgG1 or IgG3 as the fusion partner (soluble p55-TNFR-IgG1, p55-TNFR-IgG3, p75-TNFR-IgG1, and p75-TNFR-IgG3). The TNF-neutralizing activity of these constructs was tested in mouse and baboon models of E. coli endotoxemia and sepsis. When used prophylactically, soluble p55-TNFR-IgG constructs were observed to reduce mortality of endotoxic shock or systemic E. coli sepsis (Table 2). In some instances, administration of soluble p55-TNFR-IgG several hours after the induction of shock was also protective. In contrast, no protection was observed in models in which soluble p75-TNFR-IgG constructs were used (Table 2). This difference is probably due to the fact that, as mentioned above, p75-TNFR-IgG seems to have a less stable interaction with TNF, which might lead to a persistent release of TNF from the immunoglobulin fusion construct during the course of infection (61).

The results obtained with blockade of TNF activity illustrate a pattern that also applies to other proinflammatory cytokines. Indeed, neutralization of proinflammatory cytokines through prophylactic administration of neutralizing antibodies, of soluble receptors or of natural antagonists, or through gene deletions improves survival in E. coli endotoxemia and in sepsis induced by intravenous injection of live E. coli. On the other hand, neutralization either has no effect or is harmful in E. coli peritonitis, as illustrated by an increased mortality in sublethal infection models. These findings, which might appear paradoxical at first glance, might be due to the fact that E. coli endotoxemia and sepsis caused by intravenous bolus injection of live E. coli are models of systemic, rapidly lethal disease. Overwhelming and disseminated stimulation of the innate immune system induces a devastating proinflammatory cytokine storm, and attenuation of the proinflammatory cytokine response thus improves outcome. In contrast, in subacute sepsis models such as E. coli peritonitis, infection is initially localized. The infectious focus triggers a local host response aiming at the eradication of the infectious focus prior to systemic dissemination. Inhibition of antimicrobial innate immune defenses leads to unchecked microbial growth and dissemination of the infection. Neutralization of proinflammatory cytokines impairs the host’s capacity to keep the infection localized and thus worsens prognosis. Once the infection has become generalized, blockade of the systemic proinflammatory cascade might be beneficial. At this point, however, the neutralizing antibodies may no longer be fully active, depending on their half-lives. Taking these reflections into account, one can imagine that neutralization of proinflammatory cytokines may be a delicate balancing act between causing harm through impairment of infection control and improving prognosis through attenuation of an overwhelming systemic proinflammatory cytokine storm. Depending on the cytokine or the experimental setting, the balance might be tipped in one direction or the other. For experimental sepsis caused by direct intravenous injection of live E. coli, the advantages of an attenuated proinflammatory response most likely outweigh the disadvantages of an impaired antimicrobial defensive response. The opposite is likely to be the case in primarily focal sepsis models, such as E. coli peritonitis.

Clinical Data.

Given the pivotal role played by TNF in experimental sepsis and the fact that elevated concentrations of TNF were detected in the circulation of patients with sepsis (38, 71, 154), anti-TNF treatment strategies were investigated as adjunctive therapy for severe sepsis and septic shock. More than 6,000 patients with severe sepsis or septic shock were enrolled in eight clinical trials of anti-TNF monoclonal antibodies. There was no significant difference in mortality between patients treated with placebo and patients treated with anti-TNF antibodies in any of these trials (Table 3). In a subgroup analysis of one trial that included only patients with elevated IL-6 serum levels (defined as IL-6 >1,000 pg/ml), anti-TNF therapy led to a significant reduction of mortality (126). Pooling the results of the eight studies, anti-TNF therapy was associated with a 2.9% absolute reduction of mortality (36.7% versus 39.6%) (41 ). However, a crude meta-analysis should be interpreted with great caution, because the two pooled treatment groups may differ in terms of patient characteristics and prognostic factors.

Table 3Clinical trials of anticytokine therapies for patients with severe sepsis or septic shock

The efficacy and safety of the soluble p55 and p75 TNFR-IgG fusion proteins were tested in 1,927 patients enrolled in three clinical trials (Table 3). In the first of these three trials, an unexpected but statistically significant, dose-dependent increase of mortality was observed in 141 patients with septic shock who had been treated with soluble p75-TNFR-IgG1 (65). Mortality rates were 30% in patients who received placebo or low-dose p75-TNFR-IgG1, and 48% and 53% in patients who received medium or high doses of p75-TNFR-IgG1, respectively. The increased mortality in the treatment group might have been due to persistent release of TNF from the Ig fusion construct or to prolonged neutralization of TNF associated with high doses. In the other two studies, patients with severe sepsis or septic shock were randomized to receive either a single dose of p55-IgG1 or placebo (4, 5). In a prospectively defined subgroup of patients with severe sepsis or early septic shock, treatment was associated with a reduction of mortality approaching significance, but this beneficial effect was not seen in patients with refractory septic shock.

The reasons for failure of anti-TNF therapy in clinical trials are many and include the enormous complexity of the sepsis syndrome, which is only partially mimicked by experimental animal models, inappropriate dosing and timing of cytokine-directed therapy, heterogeneity and underlying conditions of the target population, and limitations of outcome measures. In animal models, intervention occurs before or during the very early stages of sepsis, when proinflammatory cytokine levels are still rising and both organ damage and vascular leakage are minimal. In contrast, patients who are included in sepsis trials already have fully symptomatic disease by definition and are thus treated at a much later stage, when many patients are switching from the proinflammatory to the anti-inflammatory phase of sepsis, and organ damage is already apparent. A detailed analysis of the pros and cons of cytokine-directed therapy in patients with sepsis is beyond the scope of this review. Readers who would like to acquire additional or more in-depth information on this complex topic should refer to the comprehensive and systematic reviews published elsewhere (1, 60, 103, 115).

IL-1 and IL-1 Receptor Antagonist (IL-1ra)

Three of the seven members of the interleukin-1 (IL-1) gene family (IL-1α, IL-1β, and IL-1ra) have been studied in the pathogenesis of sepsis. IL-1α and IL-1β are agonists, whereas IL-1ra is a receptor antagonist. In humans, IL-1α is synthesized as a 31-kDa precursor protein (pro-IL-1α) that may function as an intracellular growth factor. IL-1α is a biologically active membrane-bound cytokine. In contrast, IL-1β is a secreted cytokine. The active IL-1β molecule is produced by enzymatic cleavage of a cytoplasmic precursor protein (pro-IL-1β). The enzyme responsible for the cleavage of the pro-IL-1β is a cysteine protease called the IL-1β-converting enzyme or ICE, also known as caspase-1. There are three IL-1 receptor chains. When IL-1 binds to the ubiquitously expressed type I receptor (IL-1RI), a complex is formed which then binds to the IL-1 receptor accessory protein (IL-1R-AcP), resulting in a high-affinity signal-transducing heterodimer. Conversely, binding of IL-1 to the type II receptor (IL-1RII), which is devoid of an intracellular signal-transducing domain, does not result in cell activation. Therefore, IL-1RII functions as a decoy receptor for IL-1β. IL-1 exerts biological activities similar to those of TNF. Acting synergistically, these two cytokines cause the classic symptoms and signs of sepsis and septic shock.

The third member of the IL-1 gene family is IL-1ra, a naturally occurring specific receptor antagonist. IL-1ra is induced by a variety of microbial products or pathogens (viruses, bacteria, and yeasts) and by the proinflammatory cytokines IL-1 and TNF. As its name suggests, IL-1ra competes with IL-1 by binding to the IL-1RI, but not to the IL-1R-AcP, and thus blocks IL-1-mediated effects. Even when injected at high concentrations, IL1-ra is devoid of agonist activity. Interested readers should refer to an outstanding review of the biological activity of IL-1, IL-1R, and IL-1ra (54).

Experimental Data.

Administration of IL-1β to mice or rabbits induced a shocklike state (123, 153). In contrast to mice with deletion of the TNF gene, mice with deletion of the IL-1β gene were not resistant to the lethal effects of endotoxin, indicating that IL-1β is not an essential mediator of the systemic responses to LPS (62). In a sheep model of systemic E. coli sepsis, prophylactic administration of anti-IL-1β antibodies improved hemodynamic performance but had no significant impact on mortality, possibly because of small sample size (Table 2). Treatment with IL-1ra improved survival in animal models of E. coli endotoxemia and systemic sepsis (Table 2).

Clinical Data.

High circulating concentrations of IL-1β and of IL-1ra were detected in patients with septic shock (38, 71). Notably, the concentrations of IL-1ra needed to completely inhibit the binding of IL-1 to target cells far exceed the concentrations of IL-1ra measured in patients with sepsis (usually less than 20 ng/ml). The impact of a 72-hour infusion of human IL-1ra on mortality of patients with severe sepsis and septic shock was investigated in three clinical studies (Table 3). In a phase II study that included 99 patients, IL-1ra was found to reduce 28-day mortality in a dose-dependent fashion (44% in the placebo group versus 32%, 25%, and 16% in the three IL-1ra treatment groups, respectively) (68). However, two large phase III clinical trials with 893 and 696 patients did not confirm the beneficial effects of IL-1ra (66, 124). In the first phase III trial, 28-day mortality was 34% in the placebo group, versus 31% and 29% in the two IL-1ra groups (66). In a post hoc analysis, it was observed that patients with a predicted mortality ≥24% benefited from IL-1ra. A confirmatory phase III study was discontinued for reasons of futility after an interim analysis. At that time, the mortality rates in the placebo group and IL-1ra group were 36% versus 33% (124).

IL-6

For many years, IL-6 was viewed as a prototypic proinflammatory cytokine (reviewed in references 17 and 82). This postulate was based on the fact that IL-6, like TNF and IL-1, is produced abundantly in response to LPS stimulation and circulates at high concentrations in patients with acute infections. On the other hand, unlike TNF and IL-1, IL-6 does not upregulate the expression of proinflammatory effector molecules such as nitric oxide or prostaglandins, or adhesion molecules such as ICAM-1. Also in contrast to TNF and IL-1, administration of large doses of IL-6 does not cause shock. Moreover, IL-6 inhibits TNF, IL-1, and chemokine production in vitro and in vivo and might therefore also be viewed as an anti-inflammatory cytokine. IL-6 is a critical inducer of acute-phase protein synthesis in hepatocytes and plays a key role in the differentiation of myeloid cells. IL-6 works in concert with IL-1 to induce proliferation of T cells and differentiation of B cells, and also promotes the secretion of immunoglobulins. IL-6 binds to the membrane-bound IL-6 receptor (IL-6R), composed of a ligand-binding chain (gp80) and a ubiquitously expressed signal-transducing peptide (gp130). IL-6 also binds to a soluble receptor (sIL-6R). In contrast to most other cytokines, sIL-6R forms an agonistic complex with IL-6 that also binds to gp130 to trigger cellular responses—an activity that is termed "IL-6 trans-signaling." Through trans-signaling, IL-6 is able to act on cells that do not express the membrane-bound IL-6R. Trans-signaling is selectively counteracted by a soluble form of gp130 that prevents signaling via membrane-bound gp130.

Experimental Data.

The exact role of IL-6 in E. coli endotoxemia is yet to be determined, but IL-6 does not seem to play a major role. Induction of the acute-phase response by recombinant IL-6 had no effect on mortality in a mouse model of lethal endotoxemia. Two studies without apparent differences in experimental design examined the mortality of IL-6 knockout mice with E. coli endotoxemia and yielded conflicting results (Table 2). Mortality did not differ between knockout and wild-type mice in one study (50), whereas the other study found an increased mortality in the knockout mice (163). In E. coli peritonitis, the absence of IL-6 was shown to be detrimental; IL-6 knockout mice injected intraperitoneally with live E. coli had a higher mortality than control mice (Table 2). Whether the outcome in this model reflects the proinflammatory or the anti-inflammatory activities of IL-6, or its regulatory functions that balance pro- and anti-inflammatory responses, is uncertain.

Clinical Data.

Among all cytokines examined as prognostic factors in patients with sepsis, severe sepsis, or septic shock, IL-6 was found to be one of the best predictors of disease severity and patient outcome. Several studies demonstrated that high levels of IL-6 are associated with an increased risk for fatal outcome (43, 121, 129). Based on these findings, IL-6 levels have been used as a selection criterion for enrollment in sepsis trials (126, 129). It is not known at present whether high IL-6 levels directly contribute to the severity of disease or simply represent an indirect marker.

IFN-γ, IL-12, and IL-18

Gamma interferon (IFN-γ), IL-12, and IL-18 are functionally related cytokines. IFN-γ is a 17-kDA protein produced primarily by activated T lymphocytes and NK cells. Microbial toxins (LPS and gram-positive superantigenic exotoxins) are potent inducers of IFN-γ release. IFN-γ binds to a unique receptor (reviewed in reference 24). IFN-γ upregulates MHC class I and II and Fc receptor expression and exerts very powerful priming effects on monocytes and macrophages, strongly enhancing the production of cytokines, hydrogen peroxide, and nitric oxide after exposure to other proinflammatory stimuli. IFN-γ is thus an essential mediator of the microbicidal activity of macrophages.

IL-12 and IL-18 are two cytokines with potent IFN-γ-stimulating activities. IL-12 is a heterodimeric cytokine composed of two covalently linked subunits (p35 and p40) (reviewed in references 25 and 90). IL-12 is produced by monocytes, macrophages, dendritic cells, and B lymphocytes in response to microbial products or intracellular parasites. IL-12 binds to the IL-12 receptor, a member of the gp130-like cytokine receptor superfamily. IL-12 plays an important role in the initiation of the inflammatory response by upregulating the production of IFN-γ by NK and T cells, stimulating the proliferation of activated NK and T cells, and sustaining the generation of cytolytic T cells. In turn, IFN-γ promotes the release of IL-12 by macrophages, thereby inducing a critical positive feedback loop for the phagocytosis of pathogens and T-cell differentiation. Originally identified as an IFN-γ-inducing factor, IL-18 is expressed by a broad range of cells (including macrophages, T and B cells) after exposure to microbial products (reviewed in reference 9). Like IL-1β, IL-18 arises from a precursor pro-IL-18 molecule that is processed to the mature bioactive IL-18 by enzymatic cleavage mediated by ICE/caspase 1. Like IL-12, IL-18 stimulates IFN-γ production by NK and T cells. However, IFN-γ production by naive T cells induced by IL-18 first requires an upregulation of the IL-18 receptor by IL-12.

Experimental Data.

Several studies have shown that IFN-γ, IL-12, and IL-18 are important mediators of endotoxic shock (Table 4). Prophylactic treatment of endotoxemic mice with anti-IFN-γ, anti-IL-12, or anti-IL-18 antibodies improved survival. Likewise, anti-IFN-γ antibodies were found to reduce mortality in animal models of systemic E. coli sepsis (Table 4), even when treatment was started several hours after the onset of shock (89). As an exception to the "rule" that neutralization of proinflammatory cytokines has no effect or is harmful in focal E. coli sepsis, prophylactic immunoneutralization of IFN-γ was found to reduce mortality in E. coli peritonitis (87). We can only speculate why this was the case. IFN-γ might not be very critical in the early antimicrobial defensive response, so that its neutralization would not worsen prognosis through impairment of local infection control. This hypothesis is supported by the observation that IFN-γ becomes detectable in the serum only 4 hours after induction of peritonitis (87). However, if the infection disseminates and triggers a systemic proinflammatory cytokine cascade (including the production of IFN-γ), neutralization of IFN-γ with anti-INFγ antibodies might attenuate an overwhelming proinflammatory response and thus improve outcomes. In contrast to IFN-γ, but in keeping with observations with TNF, neutralization of IL-12 or IL-18 activity using antibodies or gene knockouts has proven to be harmful in E. coli peritonitis, as shown by a decrease of bacterial clearance, faster dissemination of bacteria into the bloodstream, and higher organ injury, although mortality was not affected (perhaps because of high mortality in the control group) (Table 4).

Table 4Evaluation of the role of IFN-γ, IL-12, IL-15, IL-18, MIF, and HMG-1 in animal models of Escherichia coli endotoxemia and sepsis

Clinical Data.

Apart from reports on circulating concentrations of IFN-γ, IL-12, and IL-18 in patients with sepsis, there have been few clinical trials of therapeutic interventions with these cytokines. Knowing that patients with sepsis who survive the initial proinflammatory phase enter an anti-inflammatory phase associated with immunosuppression and increased susceptibility to superinfection, two studies examined whether treatment with IFN-γ was beneficial in patients in the anti-inflammatory phase of sepsis, as demonstrated by reduced monocyte function and reduced levels of proinflammatory cytokines (55 , 88). IFN-γ therapy was indeed shown to normalize HLA-DR expression on monocytes and the production of proinflammatory cytokines. The impact of IFN-γ on patient outcome was difficult to evaluate, however, because IFN-γ-treated patients were compared with historical controls. Randomized, placebo-controlled studies are needed to assess whether IFN-γ treatment can improve outcomes in patients with sepsis.

IL-15

IL-15 is a pleiotropic proinflammatory cytokine produced by multiple cells and tissues, including monocytes/macrophages, dendritic cells, bone marrow stromal cells, epithelial and endothelial cells, and fibroblasts, in response to stimulation with LPS or bacteria (reviewed in reference 63). In vitro, IL-15 shares many biological effects with IL-2. IL-15 plays an important role in the development of NK cells (proliferation, differentiation, and cytotoxicity) and intestinal intraepithelial lymphocytes. IL-15 produced by macrophages in response to exposure to microbial products works in concert with IL-12 to induce optimal production of IFN-γ by NK cells. At very low concentrations, IL-15 appears to downregulate proinflammatory cytokine (TNF, IL-1, and IL-6) production by macrophages, but exerts the opposite effect at high concentrations. IL-15 is also an activator of human neutrophils and may therefore promote innate immune responses. This cytokine is a potent inhibitor of several apoptosis pathways in lymphocytes via induction of antiapoptotic molecules (e.g., Bcl-2) (97, 114), and also blocks TNFRI-mediated apoptosis of fibroblasts by inhibiting a very early step in the apoptosis-signaling cascade (37). IL-15 binds to the IL-15 receptor α-chain with high affinity, but a signal is only transduced in the presence of the IL-2/IL-15 receptor β- and γ-chains. IL-15 can also lead to signaling responses in the presence of the heterodimeric IL-2/IL-15 receptor β- and γ-chains in the absence of the IL-15 receptor α-chain.

Experimental Data.

A recent study has indicated that transgenic mice overexpressing IL-15 are resistant to an otherwise lethal intraperitoneal E. coli challenge (76) (Table 4). Bacterial burden and levels of TNF in serum were similar in transgenic and wild-type mice, but TNF-induced apoptosis in peritoneal cells, liver, spleen, and lungs was significantly suppressed in IL-15 transgenic mice. In wild-type mice, exogenous IL-15 also prevented TNF-induced apoptosis and improved survival after intraperitoneal injection of E. coli. These results are in line with studies of TNF: if neutralization of proinflammatory cytokines is harmful in E. coli peritonitis, then overexpression or exogenous administration of pro-inflammatory cytokines would be anticipated to be beneficial.

Clinical Data.

No trials examining the impact of IL-15-directed treatment on mortality in patients with sepsis have been performed to date.

Macrophage Migration Inhibitory Factor (MIF)

Identified in the 1960s as a T-cell cytokine (28, 51), MIF emerged in the 1990s as an important component of the host antimicrobial defense and stress response that promotes proinflammatory functions of the immune system (44). Constitutively expressed by immune and endocrine cells, MIF is released in a hormonelike fashion by the pituitary and adrenal glands after exposure to endotoxin and stress (13, 20, 39, 120). In the same way, MIF is released by cells of the myeloid linage stimulated with microbial products or proinflammatory mediators (40, 45, 105, 119). MIF acts in an autocrine, paracrine, and endocrine manner to promote proinflammatory and immune responses. Work performed over the past decade has shown that MIF upregulates the expression of Toll-like receptor 4 to facilitate sensing of endotoxin-bearing bacteria (131) and sustains proinflammatory functions of macrophages by inhibiting p53-dependent apoptosis (108). Other biological activities ascribed to MIF include the activation of the extracellular signal-regulated kinase-1/2 (ERK-1/2) mitogen-activated protein kinase pathway (109) and the inhibition of the activity of JAB1/CSN5 as a coactivator of activator protein 1 (AP-1) (86). MIF was also found to counterregulate the immunosuppressive effects of glucocorticoids on immune cells (39). In agreement with its role as a modulator of inflammatory and innate immune reactions, MIF plays a pathogenic role in severe sepsis and septic shock (20, 34, 40, 42, 45), acute respiratory distress syndrome (57), rheumatoid arthritis (72, 112), glomerulonephritis (91, 92), and inflammatory bowel disease (52).

Experimental Data.

MIF is an important effector molecule of experimental E. coli endotoxemia in mice. Mortality was increased when LPS was injected concomitantly with recombinant MIF, whereas prophylactic treatment with anti-MIF antibodies reduced TNF production and improved survival (Table 4). Similar results were obtained in mice with deletion of the MIF gene (34). MIF is yet another exception to the rule that neutralization of proinflammatory cytokines has no effect or is harmful in initially localized E. coli infections: prophylactic anti-MIF antibodies were found to protect mice from lethal E. coli peritonitis (Table 4), whereas administration of recombinant MIF increased mortality (42). Improved survival was associated with a reduction of the TNF serum concentration. As for IFN-γ, we can speculate that MIF might not be crucial in the early first-line antimicrobial defense that keeps infections localized. On the other hand, MIF is a powerful stimulator of TNF and IL-1 production, the two key cytokines of sepsis and septic shock, and MIF also contributes to the pathogenesis of sepsis in the absence of TNF (42). It is thus conceivable that, without impairing local infection control, anti-MIF antibodies effectively block the harmful systemic proinflammatory cascade that ensues once a focal infection becomes generalized.

Clinical Data.

Several studies have documented elevated serum concentrations of MIF in patients with severe sepsis or septic shock caused by gram-negative or gram-positive bacteria. Consistent with the concept that high levels of MIF might be harmful in the context of an acute infection, serum concentrations of MIF were significantly higher in patients with septic shock and severe sepsis than in healthy individuals (42). Likewise, on admission to an intensive care unit, serum MIF levels were higher in patients with septic shock than in trauma patients or control subjects (19). In the subgroup of patients with septic shock, MIF concentrations were more elevated in nonsurvivors than in survivors and were correlated with levels of cortisol and IL-6. Confirming these observations, a recent study has demonstrated that high MIF levels are an early indicator of poor outcome (33).

High-Mobility Group Box 1 Protein (HMGB1)

HMGB1 is an abundant, constitutively expressed, highly conserved and broadly distributed protein, with a sequence homology between rodents and humans of 98% (reviewed in reference 165). HMGB1 was identified almost 40 years ago as a nonhistone nuclear chromosomal protein. Over the years, HMGB1 was shown to function as a DNA-binding protein that stabilizes nucleosomes, facilitates gene transcription, and augments the transcriptional activity of members of the steroid hormone receptor family. The fact that HMGB1 knockout mice rapidly die after birth indicates that the protein exerts essential biological activities. In addition to being a nuclear protein, HMGB1 is also a membrane-bound protein involved in neurite outgrowth and a cytoplasmic protein released from cells in the context of inflammatory diseases. When HMGB1 is present in the extracellular milieu, it is recognized by the innate immune system as a "necrotic marker" indicating tissue damage (99). HMGB1 is either passively released in the extracellular space from damaged or necrotic cells (133) or can be actively secreted by immune cells (156). Monocytes, macrophages, and pituitary cells secrete HMGB1 after exposure to LPS, TNF, IL-1, and IFN-γ. Kinetic studies of HMGB1 production in vitro (by LPS-stimulated macrophages) and in vivo (in mice injected with LPS) revealed that HMGB1 is a late mediator, secreted up to 20 hours after LPS administration. Most of the initially released HMGB1 is derived from preformed intracellular pools; subsequently released HMGB1 is synthesized de novo (156). HMGB1 stimulates the transcription of several proinflammatory cytokine genes by monocytes/macrophages (11) and activates endothelial cells to express cellular adhesion molecules and tissue-type plasminogen activator (3), thus enhancing the proinflammatory host response. Structurally, HMGB1 is composed of three domains: two homologous DNA binding motifs, the A box and the B box, and a negatively charged C terminus. The A box is devoid of intrinsic proinflammatory activity and acts as a natural antagonist of the proinflammatory functions of the entire HMGB1 molecule. The A box was shown to inhibit HMGB1-induced TNF and IL-1 release from macrophages in a dose-dependent manner (164).

Experimental Data.

HMGB1 does not circulate in resting, unstimulated mice. Following LPS injection, serum concentrations of HMGB1 start to increase after a lag phase of a few hours, reaching a peak after 16 hours and remaining elevated for at least 15 hours thereafter. Repeated administration of anti-HMGB1 antibodies protected mice from lethal E. coli endotoxemia, even when the first dose was given 2 hours after LPS (followed by the administration of additional doses 12 and 36 hours after LPS) (Table 4). High levels of HMGB1 were found to be toxic. Indeed, HMGB1 increased mortality in mice given a sublethal dose of LPS. When HMGB1 was injected alone at high doses, mice developed the classic signs of endotoxemia-associated organ failure and died 1 to 2 days later (156). Treatment with the A box at 0, 12, and 24 hours after LPS significantly reduced mortality in a mouse model of lethal E. coli endotoxemia (Table 4). Unlike HMGB1, the A box was not found to be toxic (164). Experiments examining the role of HMGB1 in live systemic and focal E. coli sepsis have not yet been performed.

Clinical Data.

HMGB1 was not detectable in the serum of healthy subjects, but significant levels were observed in critically ill patients with sepsis. In septic patients, HMGB1 levels were higher in nonsurvivors than in survivors (156). A recent study found high serum levels of HMGB1 in patients with both severe sepsis and septic shock up to one week after admission, but without correlation to patient survival (142). Further studies are needed to elucidate the role of HMGB1 in sepsis-induced mortality and explore the therapeutic role of HMGB1 inhibitors in severe sepsis and septic shock.

ANTI-INFLAMMATORY CYTOKINES

Cytokines have been categorized into two main groups based on whether they exert mainly proinflammatory or anti-inflammatory activities. Although very convenient, this classification is an oversimplification that does not take into account the diversity of the biological properties of many cytokines. In fact, several cytokines, for example IL-6 and TGF-β, exert both proinflammatory and anti-inflammatory functions. IL-4, IL-10, and IL-13 are prototypic anti-inflammatory cytokines. Their classification as anti-inflammatory cytokines is based on the observation that these molecules inhibit the production of proinflammatory cytokines (primarily TNF and IL-1) and toxic oxygen and reactive nitrogen species by myeloid cells. However, it is important to keep in mind that these cytokines also exhibit immunostimulatory properties, for example, on B and T lymphocytes. Soluble cytokine receptors (i.e., soluble p55- and p75-TNFR) or cytokine receptor antagonists (i.e., IL-1ra) also act as anti-inflammatory molecules.

IL-10

Produced by monocytes, macrophages, and B lymphocytes, IL-10 is a prototypic anti-inflammatory cytokine. IL-10 inhibits cytokine production by activated macrophages, neutrophils, NK cells, and CD4 T-helper type 1 lymphocytes (Th1 cells) and also suppresses cell-mediated immune responses and macrophage-dependent proliferation of T cells. Originally, IL-10 was identified as a product of CD4 T-helper type 2 lymphocytes (Th2 cells) that inhibited the synthesis of cytokines (IL-2 and IFN-γ) by Th1 cells (reviewed in reference 79). Over the years, IL-10 has been shown to inhibit the production of numerous cytokines implicated in inflammatory and immune responses, including TNF, IL-1, -2, -3, -6, -8, -12, G-CSF, GM-CSF, MIP-1α, MIP-1β, and IFN-γ, to mention only a few (159). IL-10 also modulates cytokine receptor expression and was observed to downregulate the expression of TNF receptors and upregulate the release of soluble p55-TNFR and p75-TNFR. However, IL-10 also exhibits immunostimulatory features by promoting the proliferation of B cells and the differentiation and upregulation of MHC class II expression by B cells.

Experimental Data.

Prophylactic administration of IL-10 resulted in profound inhibition of cytokine production and protected mice from E. coli endotoxic shock. Conversely, neutralization of IL-10 activity with specific antibodies increased the mortality of experimental endotoxemia (Table 5). These findings are not surprising in view of what is known about proinflammatory cytokine production in E. coli endotoxemia: attenuation of the proinflammatory host response is clearly beneficial in E. coli endotoxemia, and this can be achieved either by neutralization of pronflammatory cytokines or by administration of anti-inflammatory cytokines. Once again, things are less straightforward for focal E. coli infections; if blockade of proinflammatory cytokine activity has no effect or is harmful in E. coli peritonitis, one might expect the blockade of anti-inflammatory cytokine activity to be beneficial. However, IL-10 knockout mice with E. coli peritonitis exhibited a higher mortality than wild-type mice despite improved bacterial clearance (Table 5). These data suggest that although IL-10 facilitates the dissemination of bacteria during E. coli peritonitis, it reduces mortality through attenuation of the systemic proinflammatory response. In view of all the data obtained thus far, it is evident that an optimal outcome of initially localized E. coli sepsis does not depend on either a strictly proinflammatory or a strictly anti-inflammatory state, but on a suitably balanced mix of the two.

Table 5Evaluation of the role of IL-4, IL-10, IL-13, and TGF-β1 in animal models of Escherichia coli endotoxemia and sepsis

Clinical Data.

Serum levels of IL-10 are elevated in patients with septic shock. Levels of IL-10 usually do not exceed 1 ng/ml in patients with sepsis, except in patients with meningococcemia, in whom concentrations as high as 10 to 100 ng/ml have been reported. IL-10 therapy has been investigated for the treatment of inflammatory diseases such as Crohn’s disease, psoriasis, and multiple sclerosis, but not for the treatment of sepsis.

IL-4 and IL-13

Originally discovered as a B-cell growth factor, IL-4 is produced by activated T cells, mast cells, and basophils. IL-4 is a critical cytokine for the differentiation of CD4 T cells into Th2 cells. IL-4 exhibits growth-promoting properties, increases the expression of MHC class II expression on B cells, and induces immunoglobulin class switching (IgG4 and IgE). Myeloid cell production of proinflammatory mediators, including TNF and IL-1, is suppressed by IL-4 (reviewed in reference 35). Cell activation is mediated by the high-affinity IL-4 type I receptor (IL-4RI), which is a heterodimer comprising a 140-kDa IL-4-specific ligand binding chain (IL-4Rα) and a signaling chain (γ-chain), also known as the common γ-chain (γ_c), that is shared with several other cytokines, including IL-2 and IL-7. A second receptor, the IL-4 type II receptor (IL-4RII), or IL-13 type I receptor (IL-13RI), is used by both IL-4 and IL-13. This receptor is composed of IL-4Rα and IL-13RαI, a IL-13-specific ligand binding chain (reviewed in reference 116).

Expressed predominantly by activated T cells (primarily Th2 cells, but also by Th1 and Th0 cells), IL-13 exerts the same effects as IL-4 on macrophages and B cells, but is devoid of activity on CD4 T cells (169). Moreover, IL-13 is known to activate eosinophils and mast cells. IL-13 binds to the IL-13RI and to a second receptor, the IL-13 type II receptor (IL-13RII), characterized by another IL-13-specific ligand binding chain, IL-13RαII. However, binding of IL-13 to IL-13RII does not cause cell activation. A soluble form of IL-13RII has also been described. These observations have led to the speculation that IL-13RII may be a decoy receptor (reviewed in reference 73).

Experimental Data.

Three studies demonstrated that treatment with IL-4 and IL-13 protects mice from lethal E. coli endotoxemia (Table 5). Two studies employed single injections of recombinant IL-13 (113, 118). Survival was increased in a dose-dependent manner; mortality was significantly lower even when rIL-13 was given 30 minutes after LPS injection (113). The third study used intraperitoneal gene transfer (cationic liposomes containing plasmids coding for IL-4 and IL-13) (18). Treatment was associated with improved functions of peritoneal macrophages. All three studies found significantly decreased serum levels of TNF in the treatment group. No studies have yet examined the role of IL-4 and IL-13 in systemic E. coli sepsis or E. coli peritonitis.

Clinical Data.

IL-4 was detected infrequently in the bloodstream of patients with sepsis and generally in low concentrations (<1 ng/ml) (84, 148, 167). In most patients with sepsis, IL-13 is not detectable (147). There are no studies of adjunctive therapy with IL-4 or IL-13 in patients with sepsis.

Transforming Growth Factor Beta (TGF-β)

TGF-β, a member of a family of dimeric polypeptide growth factors, is produced by virtually every cell in the body, including epithelial, endothelial, hematopoietic, neuronal, and connective tissue cells (reviewed in reference 27). TGF-β is involved in the proliferation and differentiation of cells, embryonic development, angiogenesis, wound healing, and inflammatory responses. There are three isoforms of TGF-β, each of which is highly conserved in mammals, encoded by a distinct gene, and expressed in a tissue-specific and developmentally regulated fashion: TGF-β1, TGF-β2, and TGF-β3. TGF-β1 is the isoform most closely linked to inflammation. Like MIF and HMGB1, TGF-β is expressed constitutively. TGF-β regulates cellular processes by binding to a transmembrane hetero-tetrameric receptor complex (receptor I, receptor II, and receptor III) with serine-threonine kinase activity.

In general, TGF-β is considered to be an anti-inflammatory cytokine based on the observation that TGF-β1 knockout mice developed multiple lethal inflammatory foci shortly after birth, characterized by extensive leukocyte infiltration and overexpression of MHC class I and II molecules and proinflammatory cytokines (136). TGF-β is produced by all leukocytes and promotes leukocyte differentiation, but inhibits the proliferation and activation of these cells (reviewed in references 95 and 96). Moreover, TGF-β was shown to suppress synthesis of proinflammatory cytokines, nitric oxide, and reactive oxygen species in LPS-stimulated macrophages and to induce the release of TNFR and IL-1ra. However, TGF-β can also exert proinflammatory activities, such as leukocyte activation with enhanced production of proinflammatory cytokines, chemotaxis, and upregulation of adhesion molecules. In an attempt to reconcile these contradictory findings, one can hypothesize that TGF-β may function as a proinflammatory mediator as long as the inciting agent is present. Once the microorganism has been controlled, the anti-inflammatory modes of action of TGF-β are likely to predominate, limiting proinflammatory responses and inducing tissue repair.

Experimental Data.

Transgenic mice overexpressing TGF-β1 in the liver and exhibiting increased serum levels of TGF-β1 had an increased mortality compared with wild-type mice when challenged with a sublethal dose of E. coli LPS (Table 5). As discussed by the authors, the excess mortality of the transgenic mice, which suffer from progressive renal failure, might at least be partly due to an LPS-induced exacerbation of renal dysfunction. Compared with wild-type mice, TGF-β1 transgenic mice had lower serum nitrogen oxide levels but higher serum TNF levels. In a mouse model of E. coli pneumonia, prophylactic administration of TGF-β1 decreased blood and lung bacteria counts, but increased TNF serum levels, alveolar leukocyte recruitment, and lung injury scores without affecting mortality (Table 5). Given the extreme complexity of TGF-β biology, the results from these two studies are difficult to interpret, in particular, because the conditions in which either proinflammatory or anti-inflammatory activities of TGF-β predominate are not clearly defined.

Clinical Data.

Data on the role of TGF-β in patients with sepsis are limited. One study found that TGF-β1 serum levels are elevated at the time of diagnosis in patients with sepsis as compared with healthy controls, but did not correlate with clinical outcome (101). In another study, serum TGF-β1 was persistently elevated during the course of severe sepsis, with significantly higher baseline levels in survivors than in nonsurvivors (93).

OTHER POSSIBLE THERAPEUTIC TARGETS IN E. COLI SEPSIS

Nuclear Transcription Factor-Kappa B (NF-κB)

The transcription factor NF-κB is a pivotal regulator of the expression of numerous inflammatory and immune genes, including cytokines, chemokines, adhesion molecules, immune receptors, and enzymes implicated in the generation of proinflammatory mediators (reviewed in references 14 and 16). NF-κB is activated by a wide variety of stimuli, including LPS and gram-negative bacteria. NF-κB is a dimer, classically a heterodimer of p50 and p65 in human cells, which is found in the cytoplasm complexed to inhibitory kappa-B alpha (IκBα). After activation by a suitable stimulus, phosphorylation of IκBα results in dissociation of the NF-κB-IκBα complex and degradation of IκBα. Once dissociated from IκBα, NF-κB translocates to the nucleus, where it binds to consensus sequences in the promoter region of NF-κB-responsive genes and activates transcription of major proinflammatory cytokines such as TNF and IL-1.

Experimental Data.

The impact of prevention of NF-κB activation by somatic gene transfer of a plasmid overexpressing IκBα was investigated in a mouse model of E. coli endotoxemia (31). Intravenous somatic IκBα gene transfer 5 days and 24 hours before LPS challenge was associated with a significant reduction of mortality. In a rat model of lethal E. coli peritonitis, treatment with microencapsulated antisense oligomers of NF-κB reduced serum levels of TNF and IL-1 and improved survival (49). Survival was increased even when the antisense oligomers were given 4 hours after the intraperitoneal injection of E. coli.

Clinical Data.

NF-κB activity was measured over a 10-day period in nuclear extracts of peripheral blood mononuclear cells obtained from 15 patients with sepsis (10 survivors and 5 nonsurvivors) (31). Results showed that the average NF-κB activity was higher in nonsurvivors than in survivors. All patients in whom NF-κB activity increased more than twofold above baseline died.

Cholinergic Anti-inflammatory Pathway

The cholinergic nervous system was shown to be an important regulatory pathway of the inflammatory response (26, 144). Recent investigations have revealed that the vagus nerve exerts anti-inflammatory effects through its principle parasympathetic neurotransmitter acetylcholine. Interaction of acetylcholine with nicotinic acetylcholine receptors (nAChRs) downregulates NF-κB activity and prevents cytoplasmic translocation and thus secretion of HMGB1 in LPS-stimulated macrophages (157). Activation of the cholinergic anti-inflammatory pathway by electrical stimulation of the efferent vagus in experimental animals inhibits production of TNF in various organs, including liver, spleen, and heart, and reduces circulating TNF in endotoxemic mice (32, 158).

Experimental Data.

Administration of the nAChR agonist nicotine protected mice from lethal E. coli endotoxemia when given shortly before LPS administration and then twice daily for 3 days. Treatment with nicotine was associated with significantly lower HMGB1 serum levels (157). In a mouse model of lethal E. coli peritonitis, unilateral cervical vagotomy performed before the onset of infection resulted in a marked increase of proinflammatory cytokines (TNF, IL-1, and IL-6) and liver damage, but did not affect survival (149). On the other hand, pretreatment with nicotine reduced the production of proinflammatory cytokines and liver damage, but impaired bacterial clearance and accelerated mortality.

Clinical Data.

No trials examining the role of the cholinergic anti-inflammatory pathway in patients with sepsis have been performed to date.

CONCLUSIONS

Recent work has clearly demonstrated that cytokines play a pivotal role in the pathogenesis of sepsis. Cytokines initiate the host inflammatory response and coordinate the cellular and humoral responses necessary to eliminate invading pathogens. However, overwhelming proinflammatory cytokine responses can induce shock, organ failure, and death, indicating that a tight control of cytokine production is essential for balanced immune responses. Severe sepsis and septic shock can thus be viewed as clinical manifestations of a failing innate immune response.

Animal models of E. coli endotoxemia and sepsis induced by intravenous injection of live E. coli are associated with a rapid systemic activation of the innate immune system characterized by an overwhelming proinflammatory cytokine storm. Neutralization of proinflammatory cytokines or administration of anti-inflammatory cytokines improves survival. For most cytokines, the protective effect is lost when treatment is given after the noxious agent. In models of focal E. coli sepsis such as peritonitis, anti-inflammatory therapy has yielded results with an apparent lack of consistency. The local antimicrobial host response is aimed at the eradication of the infectious focus. Proinflammatory cytokines are a key component of this defensive machinery, and inhibition of critical cytokines such as TNF is likely to have detrimental consequences. Should local antimicrobial defenses fail to control the invasive pathogen, the infection will disseminate and septic shock is likely to ensue. However, depending on the route of infection, the bacterial inoculum, and the pharmocokinetics of the targeted cytokine, anti-inflammatory therapy may not necessarily be associated with detrimental effects in the early (i.e., local) phase of sepsis and may even exert protective effects once the infection has disseminated and septic shock has occurred. These factors may explain the apparent discrepancy of results obtained with apparently similar anti-inflammatory treatment strategies.

ACKNOWLEDGMENTS

This work was supported by grants from the Swiss National Science Foundation (3100-066972), the Bristol-Myers Squibb Foundation, the Leenaards Foundation, and the Santos-Suarez Foundation for Medical Research.