DC-specific ICAM-3-grabbing nonintegrin
Heat shock protein
The innate immune system is essential for host defense and is responsible for early detection of potentially pathogenic microorganisms. Upon recognition of microbes by innate immune cells suchas macrophages and dendritic cells, diverse signaling pathways are activated that combine to define inflammatory responses that direct sterilization of the threat and/or orchestrate development of the adaptive immune response. Innate immune signaling must be carefully controlled, and regulation comes in part from interactions between activating and inhibiting signaling receptors. Toll-like receptors (TLR) have recently emerged as key receptors responsible for recognizing specific conserved components of microbes including lipopolysaccharides from Gram-negative bacteria, CpG DNA, and flagellin. Full activation of inflammatory responses by TLR may require the assembly of receptor signaling complexes including other transmembrane proteins that may influence signal transduction. In addition to TLR, many additional receptors participate in innate recognition of microbes, and recent studies demonstrate strong interactions between signaling through these receptors and signaling through TLR. Useful models for these interacting signaling pathways are now emerging and should pave the way for understanding the molecular mechanisms that drive the rich diversity of inflammatory responses.
The first step in innate immunity is the recognition of microbes by receptors that have evolved to recognize molecular structures that are not normally found in self-tissues. These structures are likely to be required by the microbe for survival and are thus not able to change rapidly under innate-immune selection pressure 1. Innate immune receptors may be transmembrane signaling receptors, as is the case for the Toll-like receptors (TLR), or may be soluble components that opsonize microbes, marking them for recognition by receptors such as complement receptors, Fc receptors (FcR) or integrins 2.
Innate immune receptors trigger a variety of responses depending on the receptor and the cell type: (1) they mediate internalization of the microbe by phagocytic cells; (2) they activate antimicrobial killing mechanisms such as the production of reactive nitrogen and oxygen; and (3) they stimulate the production of inflammatory cytokines and chemokines that recruit and activate other immune cells and orchestrate the development of adaptive immunity. Activation of these innate immune responses must by tightly regulated; too little response leaves the host susceptible to infection, and too much response may lead to lethal systemic inflammation or autoimmunity. Cooperation between innate immune receptors is key to regulating and shaping innate responses. This discussion will focus on two types of receptor cooperation. First, receptors can aggregate into larger signaling complexes in which multiple components besides the main ligand-binding receptor generate signals important to the overall effect of receptor activation. Second, receptors recognizing different ligands can independently activate signaling cascades that interact inside a cell to dictate the outcome.
Mammalian TLR are a family of ten microbe-recognition receptors that are central to effective innate immunity. TLR recognize a broad spectrum of ligands including modified lipids (LPS and bacterial lipoproteins), proteins (flagellin), and nucleic acids (DNA and double-stranded RNA). For more background on the TLR family and their ligands, the reader is referred to a number of recent reviews 3–5. The observed diversity in TLR-activating ligands has prompted the suggestion that different TLR may activate different downstream responses and that these differences may help tailor immune responses to be effective against specific organisms. The scope of these differences is just beginning to be elucidated, and it is clear that, for example, TLR4 induces responses that are not activated by TLR2 3–5. The central role for TLR in inflammatory responses to microbes is best illustrated in mice lacking specific TLR or signaling molecules. Mice naturally deficient in TLR4 (C3H/HeJ mice) or carrying a targeted deletion in TLR4 are non-responsive to LPS, and are more susceptible to infection by Gram-negative bacteria such as Salmonella typhimurium than their wild-type counterparts. Similarly, mice deficient in TLR2 are profoundly hyporesponsive to Gram-positive bacteria and are more susceptible to infection by Gram-positive organisms such as Staphylococcus aureus3–5.
However, given the nearly absolute role for specific TLR in the detection of their respective ligands, it is instructive to note that the effect of TLR deletion on microbial infection phenotypes is often not severe. For example, although Escherichia coli LPS is recognized by TLR4, mice deficient in TLR4 do not show enhanced susceptibility to intravenous or peritoneal infection with the bacteria 6, 7. Similarly, Listeria monocytogenes, a Gram-positive bacterium, can induce inflammatory signaling through TLR2, but mice deficient in TLR2 are not more susceptible to infection 8. These observations probably reflect redundancy in recognition since nearly all microbes are likely to be recognized by multiple TLRas well as by additional innate recognition systems. Redundancy is crucial for innate immune defense to prevent pathogens from easily subverting recognition. It also increases the odds that a microorganism will be detected early at a stage when moderate, controlled inflammatory responses can be effective without damaging the host. As discussed below, recent studies suggest that the magnitude and consequences of TLR activation are controlled both by assembly of TLR into activation clusters and by coactivation through other microbial recognition receptors.
2 Formation of receptor complexes
A surface receptor may transduce signals by either of two mechanisms: ligand binding might trigger multimerization of the receptor that is sufficient for signaling. Cytokine receptors such as the GM-CSF receptor function this way; GM-CSF binds to and dimerizes the receptor α and β chains, ligand-bound dimers come together, and signaling is activated through JAK/STAT family proteins 9. Alternately, ligand binding might trigger assembly of a multi-receptor complex in which several components besides the central ligand-binding receptor contribute to signaling. For example, the ligand-binding TCR α and β chains assemble with more than eight other transmembrane polypeptides to form the active TCR complex 10. The precise contributions of each protein to the overall signal are still being worked out. Assembly of a TCR activation complex permits exquisite regulation of the consequences of TCR engagement; information about affinity, duration and costimulation is integrated by the complex to signal cell death, survival, anergy, or activation. This mechanism has evolved to tightly regulate the generation of an appropriate T cell repertoire. Evolutionarily, the innate immune system has had even more time to fine-tune the regulatory mechanisms controlling inflammation, and it is likely that TLR signaling is similarly tightly constrained.
TLR have been generally modeled more like cytokine receptors, where ligand-induced dimerization is sufficient to trigger intracellular signaling. The comparison is appropriate since TLR signaltransduction is nearly identical to IL-1R signaling. The IL-1R and the IL-1R accessory protein dimerize to form a high-affinity binding site for IL-1. This dimerization triggers association with MyD88, an adaptor molecule that mediates downstream activation of IL-1R-associated kinases that, through interactions with TRAF6, lead to activation of NF-κB and MAP kinases and activate transcription of inflammatory cytokine genes 3–5.
As a first principle, these mechanisms of signaling hold true for TLR since artificial dimerization of TLR in a variety of myeloid and non-myeloid cells results in measurable activation of thetranscription factor NF-κB and production of cytokines such as TNF-α 11, 12. The sufficiency of TLR dimerization in signal transduction is supported bycell-free experiments indicating that TLR signaling domains interact directly with MyD88 13. MyD88 is required for TLR signaling, as demonstrated by the observation that MyD88 deletion in mice abrogates most TLR signaling 14. However, important differences in IL-1R- and TLR-induced responses are emerging. An additional adaptor molecule called TIRAP that has homology with MyD88 is required for MyD88-dependent signaling through TLR2 and TLR4, but not by the IL-1R (or other TLR such as TLR3, 5, 7 and 9) 15, 16. Also,a number of TLR-induced responses such as activation of IFN-β production and maturation of DC in response to LPS (through TLR4) occur in the absence of MyD88 or TIRAP, leading to the suggestion that additional MyD88-like adaptor molecules may associate with some TLR to mediate additional specific responses 15–17. Akira and co-workers have recently suggested that a molecule called TRIF may fulfill this role 18.
Although a TLR-dimerization model of signaling is straightforward and attractive, a few reports on the mechanism of TLR4 signaling (by far the most thoroughly studied TLR) have suggested that the receptor may associate with other surface molecules that could affect signaling (Fig. 1). It is likely that TLR4 and its associated molecules accumulate in membrane microdomains (lipid rafts) that are rich in glycosphingolipids and cholesterol. The formation of lipid rafts is essential for signaling through many receptors, including the TCR, by stabilizing interactions and recruiting necessary signaling molecules. Several components of the active TCR complex are constitutively associated with rafts, including CD3, CD4/CD8, LAT, Lck, and Fyn. TCR binding to MHC-peptide leads to recruitment of lipid raft domains and assembly of the active signaling complex 10.
It is possible that similar recruitment of rafts is required for TLR4 signaling. CD14, as a GPI-linked surface molecule, is constitutively associated with rafts 19, whereas it appears that TLR4 is not 19, 20. Upon LPS activation of macrophages, TLR4 translocates to rafts, and raft-disrupting drugs interfere with LPS-induced signaling 20. However, soluble CD14 and CD14 expressed as a type I transmembrane protein facilitate LPS signaling, suggesting either that constitutive localization of CD14 to rafts is not a prerequisite for signaling, or that other members of the signaling complex also drive localization to rafts 21.
The central role for TLR4/MD2/CD14 in LPS-induced signaling is clear, but several additional membrane proteins that associate with lipid rafts have been implicated in LPS binding and LPS-induced signal transduction and may play important ancillary roles in myeloid cell responses to LPS.
First, Mac-1 (also called complement receptor 3, CD11b–CD18), resides in detergent-insoluble lipid rafts in LPS-stimulated monocytes 19. Wright and Jong initially identified Mac-1 as an LPS-binding receptor and the role of the receptor in LPS-induced signaling has been enigmatic ever since 22. Although expression of Mac-1 in CHO cells (which express TLR4–MD-2, but not CD14) is sufficient to permit LPS-induced NF-κB activation 23, human patients lacking CD18 seem to respond normally to LPS 24. Further, CD14 antibodies inhibit LPS signaling in human cells 25, and mice lacking CD14 (and expressing normal levels of Mac-1) are profoundly deficient in recognition of LPS 26. Still, careful examination of LPS responses in macrophages from mice lacking CD18 revealed that whereas most LPS-induced responses are normal, specific responses including induction of mRNA for COX-2 and for IL-12 are diminished 27. The data suggest that although Mac-1 is not absolutely required for LPS recognition, it contributes to TLR4-mediated LPS-induced inflammatory signaling. Further studies will be required to fully elucidate the subtle role of Mac-1 in LPS signaling.
A series of reports by Triantafilou et al. 20, 28–30 have suggested that a complex containing heat shock protein (Hsp) 70, Hsp90 and growth differentiation factor (GDF) 5 associate with the transmembrane chemokine receptor CXCR4 to form an LPS-binding receptor. These proteins were identified by LPS immunoaffinity purification, and their role in LPS-induced signaling was supported by demonstrating that antibodies to each of these molecules blocked induction of TNF-α in human monocytes by LPS 30. Further studies using a technique called fluorescence resonance energy transfer (FRET), which measures close interactions between molecules in live or fixed cells, have confirmed direct interactions between these members of a putative LPS-recognition complex and support a model in which LPS first binds CD14 and then is transferred to a Hsp70–Hsp90 complex 30.
A study by Pfeiffer et al. 19, also making use of FRET, has further suggested that a host of additional molecules including FcγRI (CD64), FcγRII (CD32), FcγRIII (CD16), integrin-associated protein (CD47), and CD81 associate with CD14 during LPS recognition. Each of these molecules is found in lipid rafts, although there is little evidence yet to suggest that these molecules are functionally important for LPS-induced signaling. As discussed below, FcR ligation with antibodies has important effects on TLR4 signaling, but there is no evidence that FcR are activated in the LPS-induced clusters.
CD55, also known as decay-accelerating factor, is a GPI-linked surface protein expressed on a wide variety of cells that participates in inhibiting complement activation and protects cells from complement lysis. CD55 was identified biochemically as an LPS-binding protein and by FRET as a CD14-interacting protein 19, 31. As observed for Mac-1 above, CD55 expression in CHO cells facilitates LPS-induced NF-κB activation, suggesting that CD55 may play a role in coordinating TL4 signaling 32.
Together the data strongly suggest that LPS-induced recognition is accompanied by the formation of a multicomponent receptor complex. The data, however, have yet to coalesce into a clear model for how this complex functions. For example, it is not yet clear whether receptors other than TLR4 directly transduce signals in response to LPS that synergize or modify signaling through TLR4. It is possible that the role of the complex is to facilitate LPS presentation to TLR4 or coordinate the subcellular location of TLR4 or downstream signaling molecules. Since activation by LPS must be tightly regulated to prevent damaging the host, it is likely that these molecules (and perhaps many more yet to be described) are required for formation of a signaling complex akin to the active TCR complex. Although TLR4-based signaling is the core, the additional molecules are required to assemble the signaling complex, set the activation threshold, and to shape the output. Further studies aimed at fully defining the structure of resting and LPS-stimulated TLR4–MD2 protein complexes together with genetic screens for novel proteins involved in cellular responses to LPS will soon elucidate the complete mechanisms underlying LPS-induced signaling.
3 Collaboration between TLR and other immune-recognition receptors
3.1 Microbes trigger a variety of innate immune receptors that may interact with TLR
TLR are not the only microbial recognition receptors. A wide variety of receptors have been implicated in cellular recognition of microbes during the innate immune response (Table 1). Each of these receptors generates intracellular signals that have the potential to elicit inflammatory responses or to modulate inflammatory responses triggered by other receptors. Any given microbe is likely to be recognized by a combination of receptors, and the inflammatory response is likely to be defined by the sum of the interactions of these receptors. For example, a complement-opsonized yeast will likely be recognized by (at a minimum) complement receptor 3, dectin-1, and TLR2. Simultaneous activation of multiple innate immune recognition receptors provides endless possibilities for tailoring inflammatory responses to be effective against specific microbial pathogens. A number of receptor combinations have recently been explored that will help elucidate the mechanisms of collaborative recognition.
Dectin-1 is a C-type lectin implicated in innate recognition of yeasts 33 that we have recently discovered cooperates with TLR2 in eliciting inflammatory responses to zymosan, a model yeast particle 34. Dectin-1 is a small receptor containing a single extracellular C-type lectin domain and a short 40 amino acid cytoplasmic tail containing a single ITAM-like activation motif. Dectin-1 is related to members of the lectin-like NK-cell receptors (it is closely related to NKG2D and LOX-1) and is encoded on human chromosome 12 in the NK gene complex 35, 36. The receptor was originally identified by Ariizumi et al. 37 as a protein selectively expressed by DC that binds to T cells and can provide costimulatory signals for T cell activation. Dectin-1 expression was subsequently confirmed on several related cell types including monocytes, macrophages and neutrophils, but it is not expressed in NK cells 35, 36, 38. In an expression-cloning search for phagocytic receptors for zymosan, Brown and Gordon identified dectin-1 as a major phagocytic receptor recognizing β-1,3-glucans 33. Expression of dectin-1 in non-phagocytic cells is sufficient to confer the ability to efficiently internalize zymosan.
Inflammatory signaling in response to zymosan requires TLR2 and TLR6 12, 39. Although phagocytosis of zymosan is mediated by dectin-1 (and probably additional phagocytic receptors), inflammatory responses such as activation of NF-κB and production of TNF-α require heterodimerization of TLR2 and TLR6. Both TLR molecules are recruited to zymosan phagosomes, and expression of inhibitory forms of either receptor blocks activation of inflammatory signaling, but does not inhibit phagocytosis 12, 39. Similarly, TLR2- and MyD88-deficient macrophages internalize zymosan particles efficiently but fail to produce cytokines 34. However, some inflammatory responses are intact in TLR-deficient cells. For example, during phagocytosis of zymosan, reactive oxygen is produced 34. Similarly, production of inflammatory cytokines appears to require more than just TLR activation since pure TLR2 agonists such as the synthetic bacterial lipopeptide PAM3CSK4 are poor inducers, compared with zymosan, of IL-12p40 in macrophages and DC, even though IL-12p40 production in response to zymosan is defective in TLR-deficient cells 34.
During macrophage or DC recognition of zymosan, both dectin-1 and TLR2 are recruited to phagosomes, where dectin-1 binds β-glucans. Dectin-1 ligation causes tyrosine phosphorylation of the receptor's ITAM-like signaling motif, leading us to examine whether dectin-1 activation could cooperate with TLR in defining the inflammatory response to zymosan. We observed that dectin-1 triggers intracellular signals that, in addition to mediating phagocytosis, induce the NADPH-oxidase, and potentiate the production of IL-12p40 34. Thus dectin-1 and TLR offer a good model of collaboration between innate immune recognition receptors in defining the inflammatory response to a microbe (Fig. 2).
Another innate immune recognition molecule that magnifies TLR signaling is NOD2. NOD2 is a member of the cytosolic NOD family of proteins and has an N-terminal caspase-recruitment domain followed by a nucleotide-binding domain and a C-terminal leucine-rich-repeat region 40. This structure shares much in common with plant disease-resistance genes that are involved in host defense, suggesting that NOD proteins play a similar role in mammals 40. Indeed NOD1 has been implicated in cytoplasmic detection of LPS 41. Interest in NOD2 intensified when mutations in NOD2 were identified as the defect at the IBD1 locus, which is associated with chronic inflammatory bowel disease 42. Two recent reports now indicate that NOD2 recognizes muramyl dipeptide (MDP), a minimal repeating structure in bacterial peptidoglycans that is required for adjuvant activity 43, 44. Expression of NOD2 in cells that are unresponsive to MDP is sufficient to confer the activation of NF-κB in response to MDP 43, 44, and peripheral blood monocytes from patients who are homozygous for mutations in NOD2 fail to activate NF-κB and produce cytokines in response to MDP whereas responses to LPS are unaffected 44.
Long before the discovery of NOD2's role in recognition of MDP, it was appreciated that MDP can act synergistically with LPS and other TLR ligands to activate myeloid cells 45, 46. MDP strongly enhances cytokine responses to LPS through a mechanism that is independent of TLR and CD14 47. Still, microarray analyses of mRNA produced in primary human macrophages stimulated with MDP or assorted TLR agonists demonstrate that the activation program induced by MDP is highly similar to that of the TLR 48. Thus, NOD2 is likely to act as an innate immune recognition receptor whose activation produces inflammatory signaling directly and amplifies TLR-induced signaling. The potential for this cooperative system to preferentially amplify specific responses and thus shape the immune response differently from either receptor alone has not been fully explored.
The observation that some innate recognition receptors enhance induction of cytokines by TLR is reminiscent of the interactions between the BCR and CD19. The BCR is activated by antigen recognition, but the threshold for activation is markedly lowered by co-activation of CD19, which as a complex with complement receptor 2 recognizes complement C3d deposited on the antigen 49, 50. When complement deposition is mediated by innate mechanisms, this interaction between the BCR and CD19 represents a crucial interface between adaptive (gene-rearrangement-dependent) and innate immune mechanisms. The synergy between the BCR and CD19 is achieved through enhanced recruitment and retention of the activated BCR into lipid rafts 51 as well as through coactivation of key mediators of BCR signaling including Lyn and phosphatidylinositol-3 kinase 52, 53. Similar mechanisms may be at work in the regulation of TLR signaling by other immune receptors.
Two recent studies have directly explored collaboration between TLR and the BCR in activating B cells. Using a specific model for generation of autoreactive B cells by transgenic mice having an antigen-receptor that recognizes IgG2aa/j, Leadbetter and colleagues observed that DNA–antigen complexes strongly activate B cell proliferation through a mechanism that does not require complement 54. Activation required both the DNA and the antigen (IgG2a) in the complex. Further experiments suggested that stimulation of TLR9 by DNA potentiates BCR-induced B cellactivation 54. Bernasconi and coworkers have suggested a refinement of this model based on two observations. First, naive B cells express little TLR9 and do not proliferate in response to CpG DNA, although memory B cells express TLR9 and proliferate in response to CpG DNA. Second, activation of the BCR up-regulates TLR9 expression in naive B cells and permits proliferation in response to DNA. Thus, these authors have suggested a temporal collaboration between the BCR and TLR9 in which BCR activation primes B cells for subsequent activation by DNA 55.
Fcγ receptors that recognize IgG-opsonized microbes have been demonstrated to shape TLR-induced inflammatory signaling in macrophages and DC. FcR ligation causes reduced IL-12 production and increased IL-10 production in response to LPS 56, 57, while production of many other inflammatory cytokines including TNF-α and IL-1β is not altered 58, 59. In vitro, LPS-stimulated macrophages pulsed with IgG-opsonized ovalbumin activate T cells for IL-4 production, whereas cells pulsed with ovalbumin alone activate T cells for IFN-γ production, suggesting that the altered balance of IL-12/IL-10 produced after FcR ligation can bias immune responses from Th1 to Th2 60].
The intracellular events that mediate FcR alteration of TLR-induced responses are not yet clear. It is possible that enhanced IL-10 production is directly responsible for the reduced IL-12 production in response to LPS 56. So then where specifically do the IL-10-potentiating signals arise? Murine macrophages express both activating (FcγRI and FcγRIII) and inhibitory (FcγRII) Fc receptors, raising the possibility that the alterations in TLR signaling after FcR ligation are due to the effects of the inhibitory ITIM-bearing FcγRII. However, macrophages from mice lacking FcγRII still strongly induce IL-10 in response to LPS after FcR activation 56. Conversely, macrophages from mice lacking the common γ chain required for expression and signaling of FcγRI and FcγRIII fail to show enhanced IL-10 production 56. Thus, the anti-inflammatory effect of FcR ligation is due to activation of ITAM-containing Fc-receptors. FcR ligation stimulates fluxes of intracellular calcium, and chelation of extracellular calcium blocks enhanced IL-10 production, whereas calcium ionophores are sufficient to potentiate induction of IL-10 by LPS, suggesting that other receptors that mobilize calcium might generally potentiate production of IL-10 in response to LPS 57.
FcR have been reported to associate with a variety of other surface molecules including Mac-1, CD9, and lipid rafts 61, 62, and it is possible that FcR ligation also causes activation of additional receptors that are responsible for reprogramming the cells for enhanced IL-10 and reduced IL-12 production. Complement-opsonized particles and Mac-1 antibodies also inhibit LPS-induced IL-12 production by macrophages, suggesting that Mac-1 activation triggers a similar effect to FcR ligation 57.
Another immune receptor that has been implicated in down-regulating induction of IL-12 by TLR is DC-SIGN (DC-specific ICAM-3-grabbing nonintegrin). DC-SIGN is a C-type lectin receptor that is expressed primarily on DC and whose primary function is to mediate adhesion to ICAM-2 on endothelial cells during extravasation and to bind ICAM-3 on T cells during antigen presentation 63, 64. DC-SIGN was first recognized as a receptor for HIV-1 envelope glycoprotein gp120, and it is important for HIV infection since it binds the virus and delivers it to lymph nodes, where the virus infects T cells 65. The receptor has a single extracellular carbohydrate-recognition domain followed by seven 23 amino acid repeats, a transmembrane domain, and a short 41 amino acid N-terminal intracellular tail. The extracellular carbohydrate-recognition domain binds to mannose-containing carbohydrates 66, and has recently beenshown to bind to mannose-capped lipoarabinomannan (ManLAM) from Mycobacterium tuberculosis and to mediate binding and internalization of the pathogen 67, 68. ManLAM inhibits IL-12 and stimulates IL-10 production by DC activated by TLR stimuli 68, 69, and DC-SIGN antibodies block the inhibition, demonstrating that ManLAM triggers anti-inflammatory signaling through DC-SIGN 68.
Many components from cell walls of M. tuberculosis activate inflammatory responses through TLR 70, and induction of IL-12 and Th1 type immunity is important for effective defense. Since the effect of DC-SIGN binding to ManLAM is to dampen effective responses, it is likely that the receptor has been co-opted by the pathogen as a survival strategy. It is not clear whether DC-SIGN has a natural role as a microbe-recognition receptor. It is possible that DC-SIGN potentiates anti-inflammatory signals as part of its role as a receptor for endogenous molecules.
A cell's response to an external stimulus is usually envisioned as ligand-induced activation of a receptor to stimulate a precise wave of interactions between intracellular signaling components that triggers transcription of a specific set of genes. In the context of innate immune recognition of microbes by phagocytes, it is increasingly clear that such simple models of activation are insufficient to describe the diversity of responses host cells may have. Individual receptors such as TLR associate with additional membrane proteins that may transduce signals that influence theinflammatory consequences of receptor activation. Further, microbial recognition is not likely to occur through recognition of a single microbial target molecule by a single innate immune recognition receptor. Instead, multiple receptors (TLR and others) recognize any given microbe and collaborate to define the net activation state of the host cell.
Rather than describe innate immune recognition as activation of an inflammatory signaling pathway, it is useful to think of innate immune recognition as activating networks of signaling pathways that integrate multiple types of information about the threat and combine to elicit an inflammatory output that evolution has selected to be effective against the threat. In order to improve our ability to intervene when immune responses to infection are insufficient to protect, we must define models for studying interactions in signaling networks activated by innate immune recognition receptors, and determine how effective responses are regulated.
D. M. Underhill is supported by National Institutes of Health Grant #GM62995.