|Home | About | Journals | Submit | Contact Us | Français|
Summary: The mammalian immune system is optimized to cope effectively with the constant threat of pathogens. However, when the immune system overreacts, sepsis, severe sepsis, or septic shock can develop. Despite extensive research, these conditions remain the leading cause of death in intensive care units. The matrix metalloproteinases (MMPs) constitute a family of proteases that are expressed in developmental, physiological, and pathological processes and also in response to infections. Studies using MMP inhibitors and MMP knockout mice indicate that MMPs play essential roles in infection and in the host defense against infection. This review provides a brief introduction to some basic concepts of infections caused by gram-negative bacteria and reviews reports describing MMP expression and inhibition, as well as studies with MMP-deficient mice in models of infection caused by gram-negative bacteria and of septic shock. We discuss whether MMPs should be considered novel drug targets in infection and septic shock.
The matrix metalloproteinases (MMPs) constitute a family of at least 25 structurally and functionally related Ca2+-containing, Zn2+-dependent endopeptidases (Table (Table1)1) (119). MMPs, as indicated by their name, can cleave most if not all structural extracellular matrix (ECM) proteins, and researchers previously focused mainly on these matrix-remodeling properties. More recent research prompted expansion of this classical view. In addition to their functions as tissue-remodeling enzymes, MMPs also act as processing enzymes that selectively cleave a long and growing list of substrates. Their nonmatrix targets include cell surface receptors, cytokines, chemokines, cell-cell adhesion molecules, clotting factors, and other proteinases (141). Characterization of these newly discovered MMP substrates and generation of MMP mutant mouse strains have demonstrated the relevance of these enzymes in multiple processes. MMPs participate in fundamental processes, such as cell proliferation, differentiation, adhesion, migration, angiogenesis, apoptosis, and inflammation (69). The remarkable diversity of MMPs in both substrates and functions demands tight control over these enzymes in order to avoid undesired cleavage. This control is established by the need of MMPs for induction, secretion, and activation to achieve full activity. Compartmentalization and inhibition, for instance by their natural inhibitors tissue inhibitors of MMPs (TIMPs) or the acute-phase reactant α2-macroglobulin, form other levels of control (119). Loss of control leads to an imbalance in the expression or activities of MMPs, and this has been implicated in many disease processes. More details about MMPs and their structure, regulation, and function in health and disease can be found in other, recently published reviews (31, 69, 112, 119). This review will shed light on the role of MMPs in infection, sepsis, and septic shock.
Billions of individual microorganisms, collectively referred to as the normal flora, grow on or in the host, having developed intimate, beneficial, and sometimes essential relationships. Only a subset of microorganisms, called pathogens, can cause infection, defined as illness caused by microbial invasion. The outcome of an infection depends on the pathogenicity or virulence of the pathogen and on the susceptibility or resistance of the host to that pathogen. Neither the virulence nor the resistance of the host is a constant factor. Virulence is influenced by factors such as nutrients, temperature, and pH, whereas resistance of the host depends on factors such as diet, age, gender, the presence of other pathogens, and underlying diseases and their treatment (e.g., immunosuppressive drugs after organ transplantation or chemotherapy for cancer patients), as well as on genetic factors. The toxins produced by some pathogenic microorganisms are important virulence factors. These toxins act on specific host cells or molecules, resulting in specific impairment of a major host cell function (77). Bacteria are often classified as gram positive or gram negative, based on how they stain with Gram stain. (56). Gram-positive and gram-negative bacteria cause considerably different types of infections (21). This review will focus on gram-negative infections. Gram-negative bacteria are characterized by the presence of a unique outer membrane rich in lipopolysaccharides (LPS). Peptidoglycans and lipoproteins, as well as flagellin of flagellated bacteria, are other molecules that form part of the gram-negative cell wall. These molecules are all recognized by the immune system and hence contribute to the host response. These molecules are called pathogen-associated molecular patterns (PAMPs) (52).
The host's first response to infection is local inflammation, characterized by the famous words of Celsus: calor (heat), rubor (redness), dolor (pain), and tumor (swelling). A fifth cardinal sign was added by Galen, namely, functio laesio (disturbance of function) (122). The innate immune system acts as the immediate line of defense against pathogens. It consists of surface barriers, inflammatory cells, cytokines, chemokines, proteases, and various other components. In most cases, these components together provide effective defenses against pathogens. However, if the immune system fails to remove these pathogens from the local invasion site, an overwhelming infection and immune response can develop, causing severe life-threatening symptoms. The archetypical example of an infection that becomes systemic is entry of bacteria into the bloodstream, a condition called bacteremia. The subsequent systemic inflammatory response is called sepsis, and if it originates from a gram-negative infection, it is known as gram-negative sepsis. The disease evolves into a severe sepsis when signs of organ dysfunction are apparent and into septic shock if hypotension persists despite adequate fluid resuscitation (101). It is estimated that 750,000 cases of severe sepsis occur annually in the United States, with a mean mortality rate of 28.6% (3). Of all sepsis cases, 38% are caused by gram-negative infections. Isolates of Pseudomonas and Escherichia coli are the most common in sepsis patients (154). Trauma and burns can also trigger an exaggerated inflammatory response and shock. When no infection is involved, the situation is more generally referred to as systemic inflammatory response syndrome (SIRS). In the clinic, SIRS is diagnosed when the patient suffers from more than one of the following clinical findings: fever, tachycardia, tachypnea, or leukocytosis. In SIRS, the signals initiating the inflammatory response originate from the host and are called alarmins. ATP, high-mobility group box 1, and DNA are well-studied examples of alarmins (10). Together, the endogenous alarmins and the exogenous PAMPs constitute the larger family of damage-associated molecular patterns (DAMPs).
Morbidity and mortality due to severe sepsis and severe SIRS are caused by the uncontrolled inflammatory response, not by the bacteria or other insults themselves as was previously thought (20). To study sepsis, animal models are used. One model is the administration of LPS. When this endotoxin induces an exaggerated inflammatory response, researchers speak more specifically of endotoxemia and endotoxic shock. The different definitions developed at the international sepsis forum conferences are summarized in Fig. Fig.11 (15).
Despite extensive research, there is still a lack of good therapies. The current treatment for sepsis, severe sepsis, and septic shock is control of infection and support of the failing organs. Control of infection is not only meant to eliminate the ongoing infection, e.g., by use of antibiotics, but also to prevent new infections. Support of impaired organ function consists of fluid resuscitation and use of vasopressors to normalize blood pressure, mechanical ventilation for respiratory insufficiency, and kidney dialysis for kidney failure (135). Organ dysfunction and septic shock are consequences of a dysregulated immune response, and therefore extensive research on modulation of the immune response is being carried out. After more than 30 phase III randomized trials with patients with severe sepsis, only one treatment was shown to be beneficial, namely, the treatment with drotrecogin-alfa (activated), i.e., human recombinant activated protein C (153). However, even this treatment remains controversial. Therefore (and this does not happen frequently), a new phase III trial is planned. The lack of an effective treatment, the high prevalence, the high mortality rate, the rapidity with which resistance to antibiotics develops, the proportional increase in the ageing population, and the associated high economic costs all underscore the need for further extensive studies: only by acquiring a better understanding of the fundamental processes involved in sepsis will we be able to define novel targets for new therapies.
Inflammation can be induced by many triggers, such as allergens, chemicals, and cytokines, as well as DAMPs, which are recognized by pattern recognition receptors (PRRs) (52). Three major protein families have been identified as DAMP sensors: Toll-like receptors (TLRs), Nod-like receptors, and RIG-like helicases. TLRs, which are localized at the cell membrane, and Nod-like receptors, which are cytosolic receptors, are involved in the detection of bacterial PAMPs (6). One of the best studied PAMPs of gram-negative bacteria is LPS, also called endotoxin. LPS is very potent, and minute amounts are sufficient to trigger the innate immune system (9). It is recognized by TLR4 and its coreceptors: LPS-binding protein, CD14, and myeloid differentiation 2 (147). Ligand binding induces signaling cascades leading to the activation of transcription factors, such as activator protein 1, nuclear factor κB (NF-κB), and interferon regulatory factor 3 (5, 111). These transcription factors induce de novo expression of multiple proinflammatory genes, leading to the release of inflammatory mediators, including cytokines, chemokines, adhesion molecules, and clotting factors (111). This proinflammatory response leads to endothelial alterations, recruitment and activation of inflammatory cells, and a hypercoagulation state (Fig. (Fig.2).2). Neutrophils and macrophages eliminate the pathogen by the release of toxic products (e.g., hydrogen peroxide, lysozyme, and MMPs) and phagocytosis. All these mediators and inflammatory cells are essential for a “normal” immune response but are very detrimental when their release is excessive and uncontrolled. Simultaneously, anti-inflammatory pathways are also activated, leading to the release of anti-inflammatory cytokines that dampen and terminate the inflammatory response (111). During septic and endotoxic shock, homeostasis is completely lost, inflammation dominates over anti-inflammatory pathways, and coagulation dominates over fibrinolysis. The result is tissue injury, organ failure, and very often also death (Fig. (Fig.22).
Mice are susceptible to different gram-negative bacteria, and infection can be easily established, for example, in their respiratory, gastrointestinal, or urinary tract. These routes of infection do not always lead to sepsis. To generate sepsis experimentally, bacteria are administered intravenously (i.v.) or intraperitoneally (i.p.) (91). A drawback of these models is that it is very hard to always reproduce the dose in different experiments. Also, very large loads of bacteria induce endotoxic rather than septic shock, because the outcome is determined more by the amount of LPS than by the growth and spread of the bacteria.
This brings us to another model mimicking sepsis: the i.v. or i.p. injection of LPS, which on its own reproduces many of the features of gram-negative sepsis (19, 120). This model induces endotoxemia, is very acute, enables precise dosing, and is highly reproducible, but no infection is involved, which makes the model somewhat less relevant to sepsis in humans. Nevertheless, this model is still frequently used. It is noteworthy that LPS preparations isolated from different gram-negative bacteria might differ in their biological effects, for example, with respect to release of cytokines (100). Furthermore, most commercial LPS preparations are not pure but contain a mixture of ligands that could activate cells through TLRs other than TLR4. Therefore, every study should clearly specify the source of LPS and its purity.
A third type of sepsis model relies on infection by endogenous flora and requires surgical intervention (20). The surgical models frequently used are cecal ligation and puncture (CLP) and colon ascendens stent peritonitis. Both models generate an acute inflammatory reaction caused by a continuous influx of enteric bacteria into the peritoneal cavity (78). These models are not easy to perform, and they are even more complex than the infection models because the septic focus generated consists of a mixture of different bacteria. This, however, can also be seen as an advantage, because it better represents the real situation in peritonitis.
Another frequently used approach is the simultaneous injection of animals with LPS and d-galactosamine, a hepatotoxic drug (30). This treatment increases the sensitivity of mice to the lethal effects of LPS. However, this model is irrelevant to sepsis, since the most prominent pathological feature is hepatic necrosis, which is rarely observed during sepsis in humans (88). Also, this model is mediated exclusively by tumor necrosis factor (TNF) (55), which is not the case in sepsis (18).
LPS can also be applied topically. This approach does not mirror the systemic effects of sepsis but is useful for studying the direct, local effects of LPS on specific organs. The most commonly used routes of injection are the lungs (leading to acute lung injury) and the brain (23, 159).
One must realize that there are considerable differences between different models. The effects of LPS injection are thought to be solely TLR4 dependent, whereas the bacterial infusion models also trigger other PRRs. Lipoproteins of the cell wall, flagellin, and bacterial DNA and RNA are recognized by TLR2, TLR5, TLR9, and TLR7, respectively (95). Another important difference is the time of exposure to the challenge. Bolus LPS is sudden and lasts only briefly, whereas the stimulus during infections develops gradually and persists over hours or days. These and other differences have been discussed in several reviews (13, 26, 123, 131). We cannot claim that a particular model is superior to the other models. Each model replicates some of the features of the disease process but fails to reproduce the whole complexity of human sepsis. In order to reach a meaningful conclusion about the role of a particular molecule in sepsis, we should collect all the data generated from the different models (83). The goal of this review is to do just that for the MMPs.
MMPs were initially believed to play a role exclusively in cancer. However, we now know that many pathological conditions are characterized by overexpression of MMPs, suggesting that tight regulation of MMP genes is critical for normal homeostasis. Unraveling the molecular mechanisms controlling MMP gene expression might identify new therapeutic targets. However, MMPs are also regulated posttranscriptionally. For instance, cytokines and growth factors can modulate the mRNA stability of several MMPs (27, 109). After translation, the majority of MMPs are secreted as latent zymogens that need to be activated, either in the pericellular milieu or at the cell membrane, before they can exert their function. Exceptions are the MMPs containing a furin-like enzyme recognition motif (MMP-11, MMP-28, and the membrane-type MMPs [MT-MMPs]). These MMPs can be processed into active enzymes intracellularly (73, 116, 117). However, the extracellularly activated MMPs might also have intracellular activities, as has been reported for MMP-2 (63) and suggested for MMP-7 (157). The amount of active MMP is further influenced by factors such as MMP catabolism, MMP clearance, and endogenous inhibitors. Hence, the expression of an MMP evaluated by measuring its mRNA does not necessarily mean that the enzyme is also active. In any case, it remains an indication that the protein might be involved in some way. Several expression analyses have shown that MMPs are indeed upregulated during infection with different gram-negative bacteria. As an example, Affymetrix gene chip analysis detected increased mmp transcripts in Peyer's patches after infection with Salmonella and Yersinia (40). Also, sepsis and septic shock were associated with higher MMP expression.
Ramsey's group investigated the expression of MMPs, and particularly MMP-9, in urogenital Chlamydia muridarum infection. They concluded that mice differing in their susceptibility to the development of chronic chlamydial disease also differ in the relative expression and activity of MMPs (121). Single-nucleotide polymorphism analysis revealed that the human Q279R mutation, located in exon 6 of the mmp-9 gene, reduces the risk for severe disease following eye infection with Chlamydia trachomatis (97).
HSP60, produced in large amounts by chlamydiae during infection, induced several MMPs in a concentration- and time-dependent manner. This effect was LPS independent, because heat treatment abolished MMP production (61).
Helicobacter strains cause gastric injury by inducing several MMPs. Here again, it was shown that genetic variants of MMP genes are associated with the development of gastric ulcer in Helicobacter infection. Carriage of allele G of the MMP-7 promoter confers a 1.6-fold-increased risk of gastric ulcer in humans. Carriage of allele A of a coding single-nucleotide polymorphism in exon 6 of mmp-9 confers a 2.4-fold-increased risk. A Helicobacter pylori constituent that augments disease risk is the pathogenicity island (PAI), which encodes a secretion system that translocates bacterial effector molecules into host cells. This virulence factor selectively leads to the induction of some MMPs, such as MMP-7 (102), while other MMPs, such as MMP-9 and MMP-2, are also upregulated in strains lacking this PAI (62). Levels of MMP-9 specifically originating from macrophages are also increased in human and mouse Helicobacter-associated gastritis (7). Helicobacter infection also upregulates TIMP-1 and TIMP-3 in glandular epithelium and stroma (11). As already mentioned, induction of MMP-7 is dependent on an intact PAI and also seems to be regulated by p120 catenin, a component of adherens junctions, and by Kaiso, a transcription repressor (102, 158). Aberrant nuclear translocation of p120 in response to PAI-positive Helicobacter strains relieves the Kaiso-mediated transcriptional repression of MMP-7. The induction of MMP-7 plays a role in stimulating migration of gastric epithelial cells (158) and hyperproliferation of gastric epithelial cells. This hyperproliferation relies on the cleavage of the insulin growth factor-binding protein 5 by MMP-7, which contributes to the bioavailability of insulin growth factor II (86).
In patients with cystic fibrosis, Pseudomonas aeruginosa is the most common pathogen (36) and MMP-7 is markedly upregulated in the lungs (76). Pulmonary infection with Pseudomonas aeruginosa induced the expression of both MMP-7 and MMP-10 (57). Flagellin, not LPS, was identified as the inductive factor released by Pseudomonas aeruginosa that regulated MMP-7 expression (76). MMP-9 was sixfold upregulated in response to corneal Pseudomonas aeruginosa infection (87). Elastase, an enzyme produced by Pseudomonas aeruginosa, also could strongly activate pro-MMP-1, -8, and -9 under experimental conditions (104). Investigation of the role of this elastase in the repair of human airway epithelial cells in culture showed that this bacterial protease impedes closure of the airway epithelial wound by altering cell motility and causing an imbalance between the pro form of MMP-2 and its activated form (25). These reports exemplify the above-mentioned complexity of bacterial infections, in which many PAMPs (LPS, flagellin, and elastase) are present to trigger the immune system.
Gastrointestinal infection with Salmonella enterica serovar Typhimurium and Escherichia coli further demonstrated the importance of MMP-7 in host defense. In the mouse, MMP-7 is coexpressed with the α-defensins in the Paneth cells of the small intestinal crypts, and it was found that MMP-7 activates these defensins, enabling them to kill bacteria (157). Generally, exposure to bacteria seems to be the trigger for MMP-7 induction in epithelial cells. This hypothesis is further supported by the observation that MMP-7 was not expressed by germfree mice but was induced after colonization with Bacteroides thetaiotaomicron, a commensal bacterium in the intestine (75). Serine proteinases derived from E. coli are specific activators of pro-MMP-2, because phenylmethylsulfonyl fluoride, a serine protease inhibitor, completely interfered with the LPS-mediated activation of pro-MMP-2 (143).
The above-mentioned reports demonstrate that MMPs, for instance, MMP-7, might be beneficial (activation of defensins) or detrimental (gastric injury), depending on the stimulus and the organ. This duality should be taken into consideration when developing new therapies.
LPS induces transcription of several MMP genes, and several groups have investigated the signal transduction pathways inducing their expression. The induction of some MMP genes in cell cultures seemed to depend on the activation of NF-κB (58, 125) and/or mitogen-activated protein kinases p38 (64, 127) and ERK1/2 (64). The increased expression of MMPs after an LPS challenge suggests that these proteases may influence the pathogenesis of endotoxemia. MMP-9 is released after infusion of bacterial LPS in healthy human volunteers (2). Accordingly, increased levels of pro-MMP-9 and pro-MMP-2, as well as activated forms of MMP-9, were found in the plasma of two patients with gram-negative sepsis. The levels of these MMPs were related to the severity of sepsis (118). In a clinical study of patients with septic shock, Nakamura et al. found that MMP-9 levels in nonsurvivors of severe sepsis were higher than those in survivors and healthy controls (96). Elevated levels of MMP-9 in critically ill patients were also observed in a study by Yassen et al. (161). A more recent study again confirms that patients with severe sepsis have higher levels of MMP-9, as well as TIMP-1 and TIMP-2. Patients with TIMP-1 levels of >3,200 ng/ml were 4.5 times more likely to die than those with lower levels. The researchers concluded that TIMP-1 might serve as a useful laboratory marker for predicting the clinical outcome for patients with severe sepsis (44). High levels of neutrophil MMP-8 also were found in the peritoneal fluid of critically ill patients with secondary peritonitis (43).
By analyzing the expression of some MMP and TIMP genes in livers, kidneys, spleens, and brains of mice at various time points after LPS injection, Pagenstecher's group demonstrated organ-specific and time-specific upregulation of several MMP genes and the TIMP-1 gene (113). This LPS-induced expression was dose dependent: a lethal dose of LPS induced more MMPs for longer periods of time than a nonlethal dose (113). Endotoxemia also induces rapid changes in MMP activity in the aortae, myocardia, and sera of LPS-injected rats (65, 66). Other studies demonstrate that the expression of MMPs following LPS challenge can be enhanced by catecholamines (140) or ethanol abuse (74).
Different cell types can produce MMPs in response to LPS, such as endothelial cells (58), fibroblasts (156), epithelial cells (156), inflammatory cells (monocytes and phagocytes) (14, 64, 125, 140), microglial cells (37), mast cells (144), and neutrophils (118).
MMPs can also be induced indirectly by LPS or bacteria. MMPs are induced or repressed by various signals, including cytokines, growth factors, hormones, and cell-ECM interactions. Many of these signals are generated during the LPS response or during infection. For instance, interleukin-1β (IL-1β) and TNF, both of which are important cytokines in endotoxemia and sepsis, can induce MMP genes (60, 129). In turn, MMPs can activate these cytokines and/or release them by shedding, generating a positive feedback loop. Since MMPs also destroy cytokines by proteolysis (51), a negative feedback loop can be created as well. Other molecules, such as glucocorticoids, retinoids, and progesterone, repress expression of MMPs. Interestingly, glucocorticoid treatment is one of the few available therapies that show some clinical efficacy (4). Repression of MMPs by these nuclear receptors is by direct action on the promoters of their genes to suppress trans-activation, as well as indirectly by inducing the transcription of the TIMPs or transforming growth factor β, which in turn suppress MMPs, such as MMP-7 (130). The production of MMP-8 and MMP-9 by neutrophils is also of interest in the context of inflammation. MMPs are normally secreted in the extracellular environment following their production. Neutrophil MMP-8 and MMP-9, however, are stored in the secondary and tertiary granules of neutrophils, respectively. Upon activation, neutrophils release the contents of their granules, rapidly leading to high levels of MMP-8 and MMP-9 without any need for de novo protein synthesis (118). Neutrophil MMP-8 plays a role in the pathogenesis and progress of LPS-induced acute lung injury. Lung MMP-8 levels not only were elevated but also correlated with pathological scores, the lung wet/dry weight ratio, and the number of neutrophils (160). MMP-3 and MMP-9 were also upregulated during LPS-induced neuroinflammation (93).
An obvious way to investigate whether MMPs are implicated in endotoxic shock is by inhibiting them. MMPs can be inhibited by synthetic or natural compounds, as well as by endogenously produced molecules, such as α2-macroglobulin (32), RECK (reversion-inducing cysteine-rich protein with Kazal motifs) (103), and the TIMPs (68). TIMPs have high affinities for MMPs, but their lack of selectivity and their possession of unique, MMP-independent biological activities disfavors their use as inhibitors because of the potential side effects (68).
The first generation of synthetic, broad-spectrum MMP inhibitors used hydroxamate as their zinc-binding group. The best known examples of this class are batimastat and marimastat. However, most hydroxamate inhibitors lacked specificity and also inhibited non-MMP zinc-based enzymes; consequently, new drugs that make use of alternative zinc-binding groups were developed (48).
Some antibiotics, such as the tetracyclines, also inhibit MMPs, not only by chelating the zinc and calcium ions but also by affecting the induction of the MMP genes. The chemically modified tetracyclines (CMT), which lack antibacterial activity, are most commonly used as MMP inhibitors because they have several advantages over conventional tetracyclines: they induce no gastrointestinal side effects or toxicities, they attain higher concentrations in plasma, and they cross the blood-brain barrier and blood-retina barrier (1). Polyphenols and catechins derived from green tea are well known examples of natural MMP inhibitors (28, 106).
Different broad-spectrum inhibitors have been tested in different sepsis models, and all of them demonstrate that inhibition of MMPs confers protection against septic shock (Table (Table2).2). This is evidence that MMP inhibition might be of therapeutic interest. However, broad-spectrum inhibitors are not selective, and so the protection can also be attributed to inhibition of other metalloproteinases, such as the ADAMs (a disintegrin and metalloproteinase), because this family is also inhibited. Indeed, many reports attribute the attenuated TNF response after treatment with broad-spectrum inhibitors to the inhibition of ADAM-17, the major TNF-converting enzyme (TACE). However, several MMPs can also shed the membrane-anchored TNF (41, 90). Investigation of more selective inhibitors is needed, not only because of medical interest (a more selective inhibitor will normally reduce the side effects), but also for elucidating the specific roles of each MMP in septic shock.
In many studies, the inhibitor was administered for prophylaxis. A study by Milano et al. indicated that this prophylactic treatment is essential for protection, because no protective effect was seen if the inhibitor was injected 1 hour after LPS treatment (89). They concluded that inhibitors act on the early response to LPS. On the other hand, CMT-3, given 12 h after induction of CLP, could still prevent the sequelae of sepsis (39). Despite this discrepancy (which probably can be explained by the differences between the models), these studies combined prove that at least some MMPs and/or ADAMs might be involved in sepsis models. To identify the MMPs that mediate the morbidity and mortality associated with LPS challenge, more specific inhibitors or MMP mutant mouse strains are needed. Regasepin-1 and a metalloproteinase inhibitor that contains l-pyridylalanine inhibit MMP-8, MMP-9, and TACE, and these inhibitors could prevent the mortality associated with endotoxic shock (46, 47), indicating that at least one of these proteases might play a role.
Interestingly, Nenan's group provided evidence for the involvement of MMPs in the inflammatory response by using the reverse of inhibition. Instead of inhibiting MMPs, they instilled a peptide corresponding to the catalytic domain of recombinant human MMP-12 in the mouse airways and showed that MMP-12 itself can induce an early inflammatory response characterized by neutrophil infiltration, cytokine release, and gelatinase activation, followed by a delayed response consisting mainly of macrophage recruitment (98). Marimastat reduced both early and late responses (99).
The use of MMP inhibitors during infection also seems to be protective. CMT impeded ascension of Chlamydia muridarum into the upper genital tract, blunted acute inflammatory responses, and reduced the rate of chronic disease development (50).
Genetic knockouts of MMPs are very effective tools for identifying essential functions of MMPs in different conditions. Analysis of MMP knockouts revealed surprisingly subtle phenotypes, and only Mmp14-null mice show a severe developmental abnormality. Rare examples of Mmp14-null mice that survived until 10 weeks old showed severe dwarfism and craniofacial anomalies (45). However, after challenge, for example with LPS, remarkable phenotypes develop in many MMP knockout mice.
MMPs might indirectly contribute to the eradication of bacteria by activating antimicrobial proteins. The prototypical example is the activation of α-defensins by MMP-7 in Paneth cells of the small intestine. As a consequence, MMP-7-deficient mice were very sensitive to orally administered E. coli and Salmonella (157). MMPs could also be involved in the direct killing of bacteria. For instance, the C-terminal cathelicidin-like domain of MMP-12 is involved in the intracellular killing of bacteria by macrophages (42) (Fig. (Fig.3).3). For therapeutic purposes, these MMPs with antibacterial capacities should be considered for use as antitargets. Many other reports, however, clearly demonstrate that MMPs are interesting targets during infection.
The excessive MMP activity that is observed following infection is expected to cause tissue damage. Uncontrolled ECM cleavage by MMPs (for instance, of the endothelial cells or the blood-brain barrier) directly contributes to tissue damage (Fig. (Fig.3).3). A large neutrophil influx, which is also orchestrated by MMPs (see MMPs Modulate LPS-Induced Inflammatory Mediators below), also causes damage. Tissue damage not only harms the host directly, it also helps to disseminate the bacteria. Induction of apoptosis during infection is another mechanism by which the pathogen might harm the host. TNF and FasL are two important proapoptotic molecules that can be released from the cell membrane by MMPs (34). It is thus plausible that MMPs are involved in the apoptotic process during infection. However, apoptosis as a response to intracellular bacteria is a useful way for the host to eliminate infected cells, decreasing the spread of infection and preventing persistence of the pathogen (92). These examples demonstrate that it will be very difficult to predict whether or not an MMP should be inhibited during infections. Moreover, the outcome of an infection depends on many parameters, such as the bacterial species and its virulence (flagellated versus nonmotile, intracellular versus extracellular life cycle), the inoculum dose, the route of infection, and the immune status and age of the host. It is quite certain that the levels and activities of MMPs vary substantially in these different conditions, which might explain the conflicting results obtained with MMP mutant mice in infection studies. Mice challenged with different kinds of bacteria, different infection doses, or different infection routes can display different phenotypes. For instance, Lee et al. showed that MMP-9 plays a protective role in infection with Pseudomonas (70), whereas McClellan et al. concluded the opposite (87). Other reports of studies using gram-negative bacteria in MMP mutant mice are summarized in Table Table3.3. The question of whether MMPs are involved in infections has been answered. The challenge researchers now face is to elucidate when an MMP is beneficial or detrimental during a particular infection.
Many mutant mouse strains, including some MMP-deficient mice, show either a decreased or an increased resistance to LPS challenge (132, 155). MMP-9 deficient mice are resistant to endotoxic shock, especially when they are young (29). Some groups may not have observed this resistance because they used adult mice (124). TIMP-3, on the other hand, plays a protective role, as TIMP-3-null mice were more susceptible to LPS (138). Treatment with a metalloproteinase inhibitor rescues the TIMP-3 knockout mice. This beneficial role for TIMP-3 was also reported in the CLP model by Martin et al. (84, 85). LPS instilled in the trachea led to greater accumulation of neutrophils in the alveolar space of MMP-8 knockouts (110), pointing to an anti-inflammatory role for MMP-8 in this setting. However, by using an air pouch model, Tester et al. demonstrated a proinflammatory role for MMP-8 (146). We also described earlier a proinflammatory role for MMP-8 in a TNF-induced hepatitis model (152). These results demonstrate that an MMP can have both pro- and anti-inflammatory functions.
By using MMP-3 knockout mice, Gurney et al. provided evidence that the LPS-induced opening of the blood-brain barrier is mediated by MMP-3, which degrades tight junction and basal lamina proteins and thereby facilitates neutrophil influx through this barrier (38). Table Table44 provides an overview of the individual MMPs that have been investigated in models of infection, endotoxemia, and sepsis.
LPS, together with LPS-binding protein, CD14, and myeloid differentiation 2, activate TLR4. CD14 can be proteolytically cleaved from the cell surface to form a soluble peptide. Senft et al. showed that MMP-9 and MMP-12 contribute to the shedding of CD14, at least in the lung (133). Soluble CD14 may substitute for mCD14 in the activation of cells lacking this accessory protein (33) (Fig. (Fig.3).3). TLR4 is critical for the responses to LPS, but mounting evidence indicates that it can also detect endogenous ligands. This might explain why SIRS can develop in the absence of an infection and why anti-LPS therapies failed (107). Many ECM fragments can act as endogenous TLR4 ligands, such as fibronectin (105), heparan sulfate (53), and biglycan, as well as heat shock proteins (149), hyaluronic acid (145), and fibrinogen (137). Interestingly, biglycan-null mice have a survival advantage in LPS-induced shock, supporting the notion that the release of ECM fragments is important in endotoxic shock (128). Furthermore, a SIRS response can also be induced by administration of elastase, the enzyme that cleaves and releases heparan sulfate proteoglycans (54). These examples highlight the importance of the ECM in the regulation of TLR4. It is possible that the ECM constrains the TLR4 function by holding TLR4 in a nonsignaling conformation (12). Degradation of the ECM by proteases produced during infection or tissue injury might relieve this constraint on TLR4 function and at the same time create agonists to trigger the receptor. Of course, large amounts of LPS might overcome this suppressive mechanism (Fig. (Fig.3).3). The crucial step in induction of innate immunity thus is not necessarily the stimulation of TLR4 but might be the release of TLR4 from constitutive inhibition by ECM (12). Since MMPs cleave almost all ECM components, they might play an important role in the regulation of TLR4. For therapeutic purposes, blocking MMPs might prevent septic shock by keeping the TLR4 receptor in a quiescent state.
MMPs are involved in all phases of inflammation. They are clearly implicated in the recruitment of inflammatory cells, starting from extravasation from the capillaries to migration through the ECM. The activity of cytokines, such as IL-1β and TNF, as well as the chemotactic potential of chemokines, such as LPS-induced CXC chemokine (LIX), can be altered by MMP processing (41, 129, 150, 151) (Fig. (Fig.3).3). MMPs can also activate or inactivate other proteinases, such as serine proteinases, which also play a role in these inflammatory processes (110). Moreover, MMPs are also implicated in the resolution of infection and in tissue repair. Thus, MMPs are crucial for a normal immune response, but excessive release of these proteinases leads to severe tissue damage. A more extensive overview of all functions performed by MMPs during inflammation can be found in recently published reviews (10, 17, 31, 69, 80, 82, 115, 151).
Whether MMPs perform all these functions in response to LPS has to be investigated, because the action of a certain MMP depends on the nature of the stimulus and is highly cell and tissue type specific. For example, MMP-9 is important in respiratory infection with Francisella tularensis (81) but has no obvious role in LPS-induced lung inflammation (8). Tester's group showed, by using an air pouch model, that MMP-8 was a critical mediator of neutrophil chemotaxis because it cleaves the LPS-induced chemokine LIX. MMP-8 is not the sole LIX activator. MMP-12 also can cleave LIX, and as with MMP-8-deficient mice, fewer neutrophils were recruited to the air pouches of MMP-12-deficient mice (24). Impaired neutrophil migration was also observed in the corneal stroma of MMP-8 deficient mice. However, in this LPS-induced model, the chemotactic molecule produced by MMP-8 seemed to be not LIX but the tripeptide Pro-Gly-Pro (72). In this model, MMP-9 did not play a role in neutrophil migration (72). Addition of TIMP-2 together with LPS also reduces the neutrophil influx (35). Therefore, MMPs are involved in neutrophil influx following LPS exposure, which forms part of the early phase of the host's response to LPS, by interfering with the chemokine activity.
In a later phase, anti-inflammatory mechanisms terminate the production of proinflammatory mediators and chemokines, and thus the inflammatory cell influx, and promote repair of tissue damage. By cleavage and inactivation of large amounts of CXC and CC chemokines, MMP-12 can destroy the recruitment signals produced by these chemokines. This hypothesis was based on the observations that the numbers of polymorphonuclear neutrophils and macrophages were not decreased in MMP-12-deficient mice compared to wild-type mice 72 h after LPS instillation in the lung (24). This report by Dean et al. is an elegant example of how a protease can have multiple activities (pro- and anti-inflammatory) during different phases of the host response.
MMP-7 might modulate IL-1β release indirectly, via the maturation of defensins, as suggested by Shi et al., who found that IL-1β release from LPS-activated macrophages is completely blocked by mature defensins (136).
Animal models of septic shock have delivered proof-of-concept that MMPs might be of therapeutic interest. However, the disappointing results obtained with MMP inhibitors in the cancer field raised serious questions about the clinical applicability of these inhibitors. Major concerns are the lack of selectivity and the severe side effects, such as musculoskeletal pain, that were observed after long-term use of MMP inhibitors. However, acute diseases, such as sepsis and septic shock, usually require treatment only briefly; homeostatic, beneficial MMPs will be inhibited only for a short time, and the side effects are probably milder. Nevertheless, the search for new, more selective inhibitors is crucial, because higher selectivity not only will further decrease the side effects but also will be useful in fundamental research. Selective inhibitors will help to unravel the important roles that individual MMPs play in inflammation, infection, and septic shock. Development of an effective treatment requires specific targeting of MMPs at the right time. For this purpose, we should clearly define the different phases of inflammation, sepsis, and septic shock and identify the specific MMP expression and activity profile of each phase. We should try to understand both the temporal control and the tissue specificity of the expression of MMPs as well as the importance of that expression to the outcome. When we know where and when a certain MMP plays a detrimental or beneficial role, we can try to dampen or augment it in that organ at a particular time. Of course, answering these research questions will demand a heavy investment, and translation of the results to the clinic is not guaranteed. After all, whether a treatment is effective can be answered only empirically in clinical trials.
This study was supported by FWO Vlaanderen, Belgium, and by the IAP-6/18 initiative of Belgian Science Policy. I.V. is a research assistant of the FWO Vlaanderen.
We thank Amin Bredan for editing the manuscript and Johan Decruyenaere for his help with the definitions of sepsis.
We report no conflicts of interest with regard to this paper.
Ineke Vanlaere (born 1981) graduated as a Master in Biotechnology in 2003 with the highest possible degrees. She was awarded a special prize by the examination jury and received a grant by the major Flemish funding agency (FWO) to start a Ph.D. in Professor Libert's laboratory. She is interested in acute inflammation, bacterial infections, and sepsis, especially the role of matrix metalloproteinases therein.
Claude Libert (born 1964) graduated as a Master of Sciences in 1987. He was trained by his mentor, Professor Fiers, in the fields of molecular biology and genetics. He obtained his Ph.D. in 1993. During 1994 to 1995, he was a guest at the molecular biology laboratory Istituto di Ricerca di Biologia Molecolare near Rome, Italy. In 1997, he became a group leader at Flanders Institute for Biotechnology (VIB) and in 2003 a professor at the University of Ghent, Belgium. He has received several awards and published over 80 peer-reviewed papers. His major interest lies in the study of the regulation of inflammation and infection using a mouse molecular genetic approach.