Introduction

A major advancement in our understanding of the mechanisms of host defense against pathogen invasion is the identification that the innate immune system uses Toll family receptors to signal for the presence of microbes and initiate host defense (see reviews in 1, 2). Toll-like receptors (TLRs) act as the pattern recognition receptors to detect pathogen attack and trigger host defense responses, as envisioned by Dr Janeway 3. Since Janeway and Medzhitov cloned and identified the first human homolog of the Toll protein 4, more than 10 mammalian TLRs have been reported (see reviews in 5, 6, 7). The merging work on genes in the Drosophila that control dorsal-ventral patterning and the long sought after gene that controls lipopolysaccharide (LPS) responsiveness led to the discovery of the Toll family proteins that mediate pathogen recognition by macrophages 4, 8. The spectra of the ligands for each TLR have been reported extensively (see reviews in 5, 6, 7, 9). Understanding their complex role in host defense offers tremendous potential for developing novel therapeutic drugs for more effective treatment of infectious and immune diseases.

While the importance of Toll family receptors in sensing pathogens is well recognized, it is also plausible that they may function in non-infectious diseases since TLR expression is also regulated in conditions other than infection 10. Studies into the function of TLRs in non-infectious tissue injury have just begun. This review will focus on our recent understanding on TLR-matrix interactions in non-infectious diseases 11.

TLRs in innate immunity

TLR2 was initially suggested to be involved in LPS recognition and signaling 12, 13. However, later studies demonstrated that TLR2 is a major receptor to recognize Gram-positive cell-wall components, including peptidoglycan and lipoteichoic acid, as well as mycobacterial cell-wall components such as lipoarabinomannan and mycolylarabinogalactan, and yeast cell-wall zymosan 8, 14. In vitro, TLR2-deficient macrophages produced reduced tumor necrosis factor and interleukin-6 in response to Staphylococcus aureus when compared to wild-type macrophages 8, 14.

Tlr4 gene is the long-sought-after causative gene for the LPS hyporesponsiveness in C3H/HeJ and 10ScCr mice 15. C3H/HeJ mice have a defective response to bacterial endotoxin because of a homozygosity for a codominant allele, Lpsd 15. Akira's group generated TLR4-deficient mice and demonstrated that TLR4 on macrophages recognizes the Gram-negative bacterial wall component LPS 8. Destructive mutations of TLR4 predispose to the development of Gram-negative sepsis, leaving most aspects of immune function intact 8. LPS interacts with LPS-binding protein and CD14 to present LPS to TLR4, which activates inflammatory gene expression through NF-κB and MAPK signaling 16.

There are four TLRs that are intracellularly situated and can be stimulated by nucleic acids during viral infections to initiate antiviral immunity, including interferon production 17. TLR3 recognize double-stranded RNA 18 and West Nile virus 19, TLR7 and TLR8 can be stimulated by antiviral derivatives such as imidazoquinoline and ixoribine, and guanosine- and uridine-rich single-stranded RNA oligonucleotides derived from human immunodeficiency virus-1 20. DNA from bacteria has stimulatory effects on mammalian immune cells, which depend on the presence of unmethylated CpG dinucleotides in the bacterial DNA. In contrast, mammalian DNA has a very low frequency of CpG dinucleotides, and these are mostly methylated. CpG DNA induces a strong Th1-like inflammatory response, while mammalian DNA does not have immunostimulatory activity. The cellular response to CpG DNA is mediated by TLR9 21.

TLR signaling pathway

TLR family receptors share structural similarity as well as a similar signaling pathway with interleukin-1 receptors. Upon ligand binding to TLR, the adaptor molecule MyD88 is recruited to TLR complex as a dimer 22, 23. Binding of MyD88 promotes association with interleukin-1 receptor-associated kinase 4 (IRAK4) and IRAK-1 24, 25, 26. Tumor necrosis factor-associated factor 6 (TRAF6) is recruited to IRAK-1 27. The complex IRAK-4/IRAK-1/TRAF6 dissociates from the receptor and then interacts with another complex consisting of transforming growth factor-β-activated kinase (TAK1), TAK1-binding protein 1 (TAB1), and TAB2 28, 29. TAK1 is subsequently activated in the cytoplasm, leading to the activation of IkB kinase kinases (IKKs) 28, 30. IKK activation leads to phosphorylation and degradation of IκB, and consequent release of NF-κB. Once translocated into the nucleus, NF-κB activates the expression of inflammatory chemokines and cytokines.

In addition to MyD88, other Toll/interleukin-1-receptor (TIR)-domain-containing adaptor proteins were identified and characterized. For example, TIRAP (TIR-domain-containing adaptor protein) 31, 32, TRIF (TIR-domain-containing adaptor protein inducing interferon-β) 33, and TRAM (TRIF-related adaptor molecule) 34 mediate MyD88-independent induction of interferon-β, which in turn activates the expression of interferon-inducible genes such as CXCL10.

Endogenous ligands for TLRs

In addition to the array of ligands that represent various pathogens, one cannot ignore the increasing amount of publications reporting endogenous ligands for TLRs. There are about four groups of putative endogenous ligands for TLR reported. First, several inflammatory proteins or peptides were reported to signal through TLRs. For example, heat-shock protein (Hsp)60 family chaperones were suggested as ligands for TLR2 and/or TLR4 in macrophages and B cells 35, 36, 37. High mobility group box 1 (HMGB1) is released extracellularly during acute inflammatory responses. Park and colleagues reported that stimulation of neutrophils, monocytes, or macrophages by HMGB1 required both TLR2 and TLR4 resulting in increased nuclear translocation of NF-κB and enhanced expression of proinflammatory cytokines 38. Murine β-defensin 2 acted directly on immature dendritic cells as an endogenous ligand for TLR-4, inducing up-regulation of costimulatory molecules and dendritic cell maturation 39. Second, some collectin molecules may signal through TLRs. For instance, the collectin surfactant protein-A is involved in innate host defense and the regulation of inflammatory processes in the lung. Surfactant protein-A-induced activation of the NF-κB signaling pathway and up-regulation of cytokine synthesis were reported to be critically dependent on the TLR4 functional complex 40. Third, nucleic acids of mammalian origin may act as endogenous ligands for TLRs and may promote systemic lupus erythematosus 41. In this study, Barrat and associates reported that mammalian DNA and RNA, when complexed with autoantibodies, are potent self-antigens for TLR9 and TLR7, respectively, and induce IFN-α production by plasmacytoid predendritic cells 41. The last group was the breakdown species of glycosaminoglycans. Oligosaccharides of hyaluronan (HA) activated dendritic cells and endothelial cells via TLR4 42, 43. We showed that HA fragments signal through TLR2 and TLR4 in macrophages 11. Maturation of dendritic cells by soluble heparan sulfate was reported to be mediated through TLR4 44.

However, the view of endogenous TLR ligands faces some skepticism due largely to the agonist contamination issue since most “ligands” used in these studies were prepared in microbial systems 45, 46, 47, 48. Many candidate endogenous ligands were shown to activate through more than one TLR 37, 38, 49, 50. For example, Hsps were reported as endogenous ligands for TLRs 35, 49. However, subsequent studies identified that LPS 48 and non-LPS contaminants 47 in Hsp preparations were responsible for cytokine induction. Secondly, the lack of biochemical studies on endogenous ligands and their respective TLRs thus far impedes the acceptance of the concept of endogenous ligands for TLR. Although many of the studies showed that the candidate endogenous ligands stimulated cells to produce cytokines via TLRs in vitro, biochemical analysis demonstrating the direct binding between a TLR and its endogenous ligand would be the key to coin the concept. Recently, Schaefer and colleagues 51 demonstrated direct binding between TLR4 and biglycan by using gel filtration, coimmunoprecipitation and electrospray ionization tandem mass spectrometry. Paradoxically, even genuine TLR agonists can be contaminated with agonists of other TLRs. For example, the initial mistaken assignment of TLR2 as the endotoxin receptor was likely because the agonist used in the studies was contaminated 8, 12, 52. Care has to be taken to ensure that TLR activation by endogenous ligands is genuine and not attributable to the contamination of the agonist preparation by the presence of LPS 11, 51. More importantly, increasing amount of evidence from in vivo studies 11, 51, where ligand contamination is less likely to contribute to TLR signaling, supports the concept of the endogenous ligands for TLRs. For example, Schaefer and associates 51 demonstrated that biglycan signals through TLR4 and TLR2 in a set of elegant in vivo experiments, and a critical question would be how these ligands function in various disease states.

TLRs and lung injury

It is becoming clear that TLRs not only play a role in the recognition of pathogens and in the initiation of immune responses but also have a fundamental role in non-infectious disease pathogenesis. Recent studies suggested that TLR may regulate and contribute to the pathogenesis of diseases with non-infectious tissue injury, since elevated expression of TLRs has been described in such diseases or pathological states 10, 53, and TLR polymorphisms have been associated with the susceptibility to certain diseases 54, 55, 56, 57. For example, inactivation of the MyD88 pathway led to a reduction in atherosclerosis through a decrease in macrophage recruitment to the artery wall that was associated with reduced chemokine levels 58. TLR4 (Asp299Gly) polymorphism, which attenuates receptor signaling and diminishes the inflammatory response to Gram-negative pathogens, was associated with a decreased risk of atherosclerosis 59. These studies suggested that the activation of TLR pathway promotes the development of atherosclerosis. The biological roles of TLRs in non-infectious lung pathogenesis have been explored and can be classified into three areas (Table 1): (i) In host defense, TLR4 recognizes diesel exhaust particles 60 and pollutant ozone 61, supporting that TLR4 may be one of the recognition receptors to sense pollutants in the airways. (ii) In allergic diseases, TLR9 ligand-induced pulmonary indoleamine 2,3-dioxygenase activity by resident lung cells rather than by pulmonary dendritic cells inhibits T-helper cell (Th)2-driven lung inflammation and airway hyper-reactivity in the experimental asthma, providing evidence for the inverse relationship between microbial exposure and the prevalence of allergic asthma and autoimmune diseases in Westernized countries 62. This study also highlights the notion that activation of innate immunity can impede adaptive Th responses 62. (iii) In tissue injury and repair, several recent studies examined the expression patterns of TLR2 and/or TLR4 in smoke-induced lung injury and in patients with chronic obstructive pulmonary disease 63, 64, 65. Hemorrhage-induced lung TNF-α production, neutrophil accumulation, and protein permeability, but not NF-κB activation, were dependent on a functional TLR4 66. TLR4-TLR2 cross-talk activated a positive-feedback signal leading to alveolar macrophage priming and exaggerated lung inflammation in response to invading pathogens during hemorrhage-induced acute lung injury 67. On the other hand, a recent study demonstrated that TLR4 has a protective role in oxidant-mediated injury, by maintaining appropriate levels of antiapoptotic responses in the face of oxidant stress 68. This study is in accord with our study in a bleomycin-induced lung injury model 11.

Table 1 Recent studies on the role of TLR in non-infectious lung injury and inflammation
Full size table

HA as a signaling molecule

During tissue injury, there is increased turnover of ECM. A variety of lung diseases such as asthma, emphysema and pulmonary fibrosis are associated with abnormal ECM turnover. In chronic lung diseases, the ECM is modified by the inflammatory milieu, and degradation products generated by oxidants and other mechanisms take on unique properties not attributable to the precursor molecules. HA is a major ECM component, a non-sulfated glycosaminoglycan composed of repeating polymeric disaccharides D-glucuronic acid and N-acetyl glucosamine 69. It is distributed widely in nature, such as the Streptococcus coat. In mammals, HA is abundant in heart valves, skin, skeletal tissues, the vitreous of the eye, umbilical cord, and synovial fluid. HA has the property of a biological lubricant, reducing friction during movement and providing resiliency under static conditions. Under physiologic conditions, HA exists as a high molecular weight polymer (>106 Da) and undergoes dynamic regulation resulting in the accumulation of lower molecular weight species following tissue injury 70, 71. One often compares its structure with LPS. Although they both contain repeating di- or oligosaccharides, HA does not contain lipids. Saccharide units in HA are simple D-glucuronic acid and N-acetyl glucosamine. Core polysaccharide in LPS consists of a short chain of sugars with variations, for example, 2-keto-3-deoxyoctonoic acid-heptose-heptose-glucose-galactose-glucose-N-acetyl glucosamine. The O antigen chain of LPS consists of repeating oligosaccharide subunits made up of 3-5 sugars. The individual chains vary in length ranging up to 40 repeat units 72.

A historical study by West and colleagues 73, 74 showed that oligosaccharides derived from high molecular weight HA stimulate angiogenesis in vivo and endothelial proliferation in vitro. We and others have demonstrated that HA degradation products generated in vitro in the 200 000 Da range induce the expression of a variety of genes, including chemokines, cytokines, growth factors, signal transduction molecules, and adhesion molecules in macrophages, eosinophils, dendritic cells, kidney epithelial cells, and fibroblasts, suggesting that endogenously generated HA fragments may be a dynamic molecule that influences cell behavior by regulating inflammatory processes 43, 75, 76, 77, 78, 79, 80, 81. These studies support a concept in which the modified matrix regulates inflammation and repair processes during tissue injury.

CD44 and lung injury

HA binds to numerous cell surface proteins. Among them, CD44 is a major cell surface receptor for HA 82. CD44 is a type 1 transmembrane glycoprotein that is expressed on most cell types. The interaction of HA and CD44 has been implicated in the regulation of a variety of biological processes, including tumor growth and metastasis, wound healing, T-cell recruitment to sites of inflammation, macrophage activation, neutrophil migration, and endothelial cell activation (see reviews in 83, 84). For example, HA promotes leukocyte trafficking to the inflammatory sites during tissue injury 85. A recent study demonstrated that gdT cells are required for HA deposition in the ECM and subsequent macrophage infiltration into wound sites 86. HA fragments can deliver maturational signals to antigen-presenting dendritic cells and can provide costimulatory signals to specific T cells in regulating dendritic cell maturation and trafficking 43. However, CD44 null mice were found to develop normally 87. In order to investigate the role of HA in lung injury and repair, we examined the inflammatory response in CD44-deficient mice 71. CD44-deficient mice and wild-type C57Bl/6 mice were instilled intratracheally with bleomycin, a widely used model for pulmonary injury and fibrosis. CD44-deficient mice displayed an increased mortality, and increased numbers of neutrophils, macrophages, and lymphocytes in bronchoalveolar lavage fluid 71. In addition, there was overwhelming HA accumulation in the injury lungs in the CD44-deficient mice relative to the wild-type control mice 71. Thus, the CD44-deficient state was associated with both an inability to resolve lung inflammation and a failure to properly remove HA degradation products from the lung tissue. These data suggest that the production of HA following acute tissue injury may serve the very important function of initiating the host innate immune response by providing an essential signal to macrophages, which then produce chemokines that recruit other leukocyte subsets required to debride the tissue injury and begin restoring tissue integrity 71.

TLRs and HA signaling

CD44 appears to be involved in HA binding in vitro since the blockade with anti-CD44 antibodies partially inhibits HA fragment signaling in macrophages 88. However, it is not sufficient to mediate HA signaling 11, 89. Thus, there must be another receptor system required for HA signaling. Since HA is a repeating disaccharide structure with features of “pathogen-associated molecular patterns”, and since recent reports suggest that HA oligomers can signal through TLR4 in dendritic cells 43 and endothelial cells 42, we investigated the potential role of TLRs in non-infectious lung injury 11 where matrix turnover is an important regulator of the inflammatory response 43, 86. Both TLR2 and TLR4 signaling pathways appear to require MyD88. MyD88-deficient mice (and macrophages) are unresponsive to LPS in vitro and in vivo 22, 23, 90. To investigate the potential role for TLRs in mediating HA signaling, we utilized elicited peritoneal macrophages from MyD88-deficient, TLR1-, TLR2-, TLR3-, TLR4-, TLR5-, and TLR9-deficient mice. Stimulation of chemokine gene expression by HA fragments is abolished in the MyD88-deficient macrophages 11. Chemokine MIP-2 expression is reduced, but remains present in both TLR2- and TLR4-deficient macrophages, respectively. TLR4- and TLR2-deficient mice were then crossed to generate TLR2- and TLR4-double-deficient (TLR2−/−TLR4−/−) mice. HA fragment-induced chemokine and cytokine expression was completely abolished in TLR2−/−TLR4−/− peritoneal macrophages 11. These data suggest that the polymeric properties of HA may mimic cell-wall components from both Gram-positive and Gram-negative organisms. In contrast, targeted deletion of TLR1, TLR3, TLR5, and TLR9 had no effect on HA fragment-induced chemokine expression. As our data demonstrated HA induces chemokine expression through TLR2 and TLR4, the next question would be: do HA fragments directly bind to TLR4 and TLR2? Biochemical studies such as photoaffinity labeling or direct binding should be performed to elucidate whether and how HA fragments bind to TLRs and to examine whether any other binding partners are involved.

To determine if these findings were relevant to acute lung injury in patients, we then examined circulating HA fragments produced in vivo and purified from the serum of patients with acute lung injury. The human HA degradation products were of similar molecular mass (peak 200 kDa) to the in vitro generated HA degradation products and stimulated chemokine production in wild-type, but neither TLR2−/−TLR4−/− nor MyD88−/− peritoneal macrophages 11.

TLR deficiency and lung injury

The observation that CD44 deficiency led to unremitting inflammation following lung injury clearly implicated an alternative signaling pathway by which macrophages responded to HA fragments. Our discovery that both TLR2 and TLR4 are required for macrophages to express inflammatory genes in response to HA fragments suggested that TLR2/TLR4 deficiency should lead to an impaired inflammatory response to non-infectious lung injury. To test this hypothesis, we challenged TLR2−/−TLR4−/− mice with bleomycin-induced lung injury. TLR2−/−TLR4−/− mice developed a reduced inflammatory response to lung injury with a decrease in transepithelial neutrophil migration and reduced expression of MIP-2 11. However, these mice displayed an increase in mortality. This surprising finding led us to inspect the degree of tissue injury and epithelial cell integrity. We found evidence to suggest increased tissue injury and epithelial cell apoptosis 11. Thus, TLR2/TLR4 deficiencies appear to protect the host from acute inflammation, but may impede epithelial cell repair processes important in tissue injury recovery.

HA-cell interaction in lung injury

Next, we tested whether the blockade of HA action with an HA-blocking peptide could produce a similar phenotype as in TLR2−/−TLR4−/−. An HA-blocking peptide has been previously shown to be effective in vivo in inhibiting HA-cell interactions 42, 85, 91. Inhibition of HA binding with the peptide in vivo recapitulated the phenotype observed in the TLR2−/−TLR4−/− mice after lung injury, enhanced lung injury and increased epithelial cell apoptosis, and decreased transmigration of neutrophils 11. Increasing production of high molecular weight HA by overexpression of HA synthase-2 under the control of the lung-specific CC10 promoter is protective against mortality, lung injury, epithelial cell apoptosis following high-dose bleomycin 11. Therefore, HA-cell interaction regulates the responses to lung injury.

HA-TLR interaction on epithelial cells

Previous studies in kidney injury models have demonstrated that fragmented HA markedly stimulates chemokine MCP-1 production by renal tubular cells 92, triggers cell adhesion molecule expression through a mechanism involving activation of NF-κB and activating protein-1 93, and promotes proximal tubule cell migration and re-epithelialization, through MAPK activation 94. Quinn's group has shown that low molecular weight HA stimulated cytokine IL-8 production by lung epithelial cells, via JAK2 kinase, not MEK1/2, pathway 95. It is plausible that injured epithelial cells respond to HA fragments by producing chemokines and cytokines to recruit inflammatory cells to the injury sites. To investigate the hypothesis that HA and TLR interactions are important in lung injury and repair processes, we asked if HA on the epithelial cell surface plays a role in lung injury. Isolated lung alveolar epithelial cells have increased apoptosis at baseline and exhibit greater total apoptosis in response to bleomycin. The exogenous addition of high molecular weight HA is protective against apoptosis 11. We found that bleomycin induces both NF-κB activation and apoptosis in primary lung epithelial cells 11. It is likely that HA regulates basal NF-κB activation in epithelial cells. NF-κB is known to regulate apoptosis 96, 97 and we have previously shown that HA fragments can activate NF-κB in macrophages 77. Primary epithelial cells from TLR2−/−TLR4−/− mice were found to have a significant increase in spontaneous apoptosis relative to wild-type cells 11. Intriguingly, we also found that cell surface HA is severely abrogated in TLR2−/−TLR4−/− epithelial cells 11. The mechanism underlying the abolishment of cell surface HA in TLR2−/−TLR4−/− epithelial cells is unknown. Nevertheless, these data suggest that epithelial cell-surface HA promotes basal NF-κB activation in a TLR-dependent manner, and that this activation has a protective effect against injury 98.

Conclusion

As part of inflammatory responses, HA fragments are generated following tissue injury in both animal models and patients. The removal of HA fragments requires the HA receptor CD44. The failure to remove HA from the lung following injury results in unremitting inflammation. The persistent HA fragment presence leads to chronic inflammation and tissue damage. Here we propose that the interactions of HA and TLRs serve two critical functions in host defense against non-infectious lung injury (Figure 1). Soluble HA fragments can stimulate macrophages to produce chemokines and cytokines that recruit inflammatory cells to the site of injury. On the other hand, native high molecular weight HA limits the extent of lung epithelial cell injury by providing a basal NF-κB activation and inhibiting apoptosis and promoting the repair of parenchymal cell injury through a TLR-dependent mechanism. Shaping the balance of HA in favor of high molecular weight form could favor the host recovery from acute lung injury.

Figure 1
figure 1

HA signals through TLR2 and TLR4 in both epithelial cells and macrophages. (A) High molecular weight hyaluronan (HMW-HA) on the cell surface or surrounding epithelial cells signals through to TLR2 and TLR4, providing cells with basal NF-κB activation. In turn, the tonic NF-κB activity prevents epithelial cells from undergoing apoptosis upon injury. Thus, HMW-HA on epithelial cells provides cells with a survival signal. On other hand, low molecular weight hyaluronan (LMW-HA) fragments generated during tissue injury stimulate chemokine/cytokine expression in macrophages, leading to inflammatory responses. (B) The loss of TLR2 and TLR4, or MyD88 on epithelial cells causes the loss of basal NF-κB activation and leads to an increased apoptosis of epithelial cells upon injury. In macrophages, LMW-HA cannot transduce signals due to TLR deficiency. This leads to a decreased NF-κB activation and subsequently decreases inflammatory responses.

Full size image