Mammalian Toll-like receptors: to immunity and beyond

Abstract

Toll-like receptors (TLRs) constitute an archetypal pattern recognition system. Their sophisticated biology underpins the ability of innate immunity to discriminate between highly diverse microbial pathogens and self. However, the remarkable progress made in describing this biology has also revealed new immunological systems and processes previously hidden to investigators. In particular, TLRs appear to have a fundamental role in the generation of clonal adaptive immune responses, non-infectious disease pathogenesis and even in the maintenance of normal mammalian homeostasis. Although an understanding of TLRs has answered some fundamental questions at the host–pathogen interface, further issues, particularly regarding therapeutic modulation of these receptors, have yet to be resolved.

Introduction

Until the 1990s immunology was dominated by advances in adaptive immunity, a biological system restricted to vertebrate species. This led to developments in transplantation biology and vaccine design, together with an improved understanding of the pathogenesis of infectious and autoimmune disease. In contrast, innate immunity, a phylogenetically more ancient biological system, remained poorly characterized. Although some facets of this system such as phagocytosis, free radical killing and activation of complement had been explored, the fundamental means by which pathogens are first recognized by the host (and distinguished from commensal organisms or self-antigens) were entirely unknown.

In 1991 a remarkable sequence similarity between Toll, a transmembrane protein involved in Drosophila embryogenesis, and the human interleukin-1 receptor was described [1,2]. A mammalian Toll-like receptor homologue (TLR1) was then cloned and mapped to chromosome 4 before an immune role for both Drosophila toll and mammalian Toll-like receptors (TLRs) was confirmed [3–5]. A total of 13 mammalian TLR paralogues (11 are expressed in humans) have now been described, each responsible for the ‘pattern-recognition’ of distinct invariant microbial structures [6] (Table 1). Critically, the TLR members of the IL-1 receptor family not only play a pivotal role in innate defence, but also facilitate the non-clonal high-fidelity discrimination of self from non-self originally predicted by Janeway [7]. In this review, we will describe how advances in TLR biology have revealed previously hidden sophistication within innate defences and created new fundamental questions even beyond the field of immunology.

Table 1.

List of common abbreviations found in the Toll-like receptor (TLR) field.

dsRNA	Double-stranded RNA
ECSIT	Evolutionarily conserved signalling intermediate in Toll pathways
HSP	Heat-shock protein
IKK	IκB kinase
IRAK	Interleukin-1 receptor-associated kinase
JAK	Janus kinase
Jnk	Jun N-terminal Kinase
LPS	Lipopolysaccharide
LTA	Lipoteicoic acid
Mal	myD88-adapter-like (= TIRAP)
MAPK	Mitogen-activated protein kinase
mDC	Myeloid derived dendritic cell
MIF	Macrophage-inhibitory factor
MyD88	Myeloid differentiation factor 88
NF-κB	Nuclear factor-κB
PAMP	Pathogen-associated molecular pattern
pDC	Plasmacytoid dendritic cell
PGN	Peptidoglycan
PRR	Pattern recognition receptor
RIP	Receptor interacting protein
SIGGR	Single immunoglobulin IL-1R-related molecule
shRNA	Short hairpin RNA
siRNA	Short interfering RNA
SIRS	Systemic inflammatory response syndrome
SOCS	Suppressor of cytokine signalling
ssRNA	Single-stranded RNA
STAT	Signal transducer and activator of transcription
TANK	TRAF-family member-associated NFκB-activator
TAK1	Transforming growth factor β-activated kinase-1
TAB1	TAK-1 binding protein-1
TBK1	TANK-binding kinase 1
TIR	Toll/IL-1 receptor (domain)
TIRAP	TIR domain-containing adapter protein (= Mal)
TICAM	Toll-IL-1 receptor homology domain (TIR)-containing adapter molecule
TLR	Toll-like receptor
TNFR	Tumour necrosis factor (TNF) receptor
TRAF	TNFR-associated factor
TRAM	TRIF-related adapter molecule (= TICAM-2)
TRIF	TIR domain-containing adapter inducing interferon (IFN)-β (= TICAM-1)

Complex distribution of TLR pattern recognition

Not only do TLRs exhibit marked differential tissue activity, but their absolute levels within a discrete cell type can also be highly dynamic (Table 2). Critically, both expression axes provide an integral part of receptor function and prevent host injury from inappropriate triggering of powerful down-stream inflammatory responses or autoimmune phenomena.

Table 2.

Location, modulation, ligands and function of mammalian Toll-like receptors (TLRs).

TLR	Location	Modulation	Ligands	Functions	Refs
TLR1	Ubiquitous		Soluble bacterial factors (Mycobacteria, Neisseria,Borrelia); Tri-acyl peptides	Heterodimers with TLR2	[23,57]
TLR2	Surface membrane and phagolysosomes Myeloid,mast, and NK cells; mDCsT cells (including gamma–delta)	Enhanced expression with LPS, dsRNA Enhanced expression with IL-2, IL-15, IL-1beta, IFN-gamma, TNF-alpha	Bacterial: lipoprotein stereoisomers, PGN, LTA, phenol-soluble modulin (S.epidermidis), porins (Neisseria); LPS (Leptospira, Pseudomonas Helicobacter); lipoarabino- mannan (M. tb)Fungal: zymosan Viral: HCV core and NS3 proteins, measles virus, human CMV, HSV-1 Parasitic: trypanosomal, treponemal phospholipids Endogenous: HSP70, HSP60, defensins; Cys3pam	Co-operation with Dectin-1 T helper 2 responsesCo-stimulatory receptor modulation Metalloproteinase inductionAlpha-defensin production from NK cells Modulation of apoptosis, cell repair Adrenal responses	[9,27,33,34,44,54,58,72,87,91,137,141,143,145,149]
TLR3	Intracellular in mDCs, NK cells		Viral/helminth: dsRNA Synthetic: polyinosinic–polycytidylic acid [poly (I:C)] siRNA, shRNAEndogenous: mRNA	Anti-viral: type I interferon production/cross-presentation Pathogenesis of West Nile virus	[10,14,41,43,152,173]
TLR4	Surface membrane of monocytes, mast cells and neutrophils Golgi in gut epithelial cells Regulatory/gamma–delta T cells Endothelial cells	Up-regulated IFN-gammaUp-regulated by RSVUp-regulated by bacterial superantigens Modulated by MIF	Bacterial: LPS; Pseudomonas exoS; C. pneumoniae, H.pylori HSP 60Viral: RSV F protein, MMTV envelope proteinParasitic: T. cruzi lipidsEndogenous: HSP 70, HSP 90, fibronectin, heparin, hyaluronic acid, fibrinogen, beta-defensin 2. Synthetic: taxol (murine); MPL (LPS mimetic)	Cell surface and intracellular recognition of LPSCo-ordination of neutrophil functionsDC maturation Apoptosis Endothelial functions (e.g. leucocyte roll)Atherosclerosis	[12,16–18,24,25,52,53,55,60–65,142,156]
TLR5	Surface epithelial, NK cells; mDCs monocytes		Bacterial: flagellinSynthetic: discontinuous 13 amino acid peptide	T helper 2 responses Lung/gut mucosal immunity	[11,150,161]
TLR6	Surface myeloid, mast, B cells		Diacyl lipopeptides (mycoplasma)	Heterodimers with TLR2	[23,57]
TLR7	Endosomal:pDCs, ?mDCs, B cells, eosinophils		Viral: ssRNA (e.g. influenza, VSV, HIV)Synthetic: imidazoquinolines	Eosinophil survival Defence against viral infection	[42]
TLR8	Endosomal: NK, T, myeloid cells		Viral: ssRNASynthetic:imidazoquinolines	Human not murine antiviral defence	[42]
TLR9	Endosomal: pDCs, B/NK cellsSurface of tonsillar cells	Down-regulated by CSF-1	Bacterial and viral DNA (HSV 1 and 2; murine CMV)Malaria schizontsHost chromatin	B cell cross-priming Pathogenesis of autoimmunityVaccine adjuvant	[26,39,45,86]
TLR10	Expressed in B cells, pDCs		Unknown	?TLR2 dimerization? role in asthma	[23,154]
TLR11	Murine uroepithelium		Unknown	Murine uroepithelial defence, ?humans	[13]

Differential tissue and cellular expression of TLRs

Differential expression of TLRs is particularly marked at epithelial surfaces, with important consequences for both innate defence and microbial pathogenesis. For example, while TLRs 2, 3, 4 and 5 have specific roles in bronchial and gastrointestinal epithelial defence, murine TLR11 appears to facilitate urogenital immunity [8–13]. Other tissues, such as the blood–brain barrier and endothelium, also have distinct TLR expression patterns with specific biological consequences. While TLR3 appears to play a critical role in the penetration of the blood–brain barrier by West Nile virus, TLR 2 and 4 expression on endothelial cells leads to specific sensitivity to bacterial ligands and directs the movement of rolling leucocytes within the microcirculation and their subsequent migration into inflamed tissues, such as in the lung [14–17]. The distribution of TLRs on the gut epithelium is particularly sophisticated, with different anatomical compartments of the gastrointestinal tract exhibiting specific TLR expression patterns [18]. In addition, the polarized or intracellular anatomical location of TLR4 and TLR5 may allow discrimination of commensal microorganisms from invasive pathogens, although it is now thought that some constitutive contact between gastrointestinal microbiota and TLRs may be required to maintain gut integrity [11,12,19–21].

The differential expression of TLRs on myeloid and lymphoid cells also underpins many key immune responses [22,23]. Neutrophils, natural killer cells, mast cells, eosinophils, B-cells and monocytes all have characteristic TLR expression patterns with explicit immunological consequences [24–31]. Indeed, the discovery of TLRs on T cells, including those with gamma–delta receptors, has suggested the capacity for unexpected functional sophistication [32–34]. Importantly, TLR expression discriminates myeloid from plasmacytoid dendritic cells (DCs) and also defines their respective states of differentiation and activation [22,31,35]. The profound consequences that TLR-mediated signalling provokes in this context, particularly with respect to development of adaptive immunity, have been the subject of several seminal reviews [36,37].

The expression profile of TLRs in immune cells (and particularly DCs) also best illustrates the fundamental cellular compartmentalization of these receptors. While certain TLRs are trafficked to the cell surface for engagement of extracellular pathogens (TLRs 1, 2, 4, 5 and 6), others are almost exclusively found at intracellular locations (TLRs 3, 7, 8, 9). This key design feature within innate immunity allows targeted and powerful TLR-mediated effects to operate against intracellular pathogen-derived products, which closely resemble endogenous host antigens. Several elegant techniques have been designed to demonstrate this division, including the fluorescent tagging of TLRs or down-stream signalling elements (such as the adapter protein MyD88) and the tracking of various TLR chimeras to distinct subcellular compartments [38–40]. The intracellular location of TLRs 3, 7, 8 (non-functional in mice) and 9 means that their ligands (viral dsRNA, ssRNA and prokaryotic DNA) require internalization to the endosome before signalling pathways can become activated [39,41,42]. However, this protective compartmentalization may break down where there is tissue injury, when even endogenous mRNA may become immunoactive [43]. Moreover, as is exemplified by the intracellular location of TLRs 2, 4 and 5 in some circumstances and the cell surface expression of TLR9 in others, this seemingly fundamental division is not absolute [12,38,44,45]. Detailed structural analyses of cytoplasmic anchoring sequences and the identification of chaperoning proteins (such as gp96) have now provided an insight into the mechanisms responsible for determining the subcellular destination of different TLRs [12,46]. Finally, it should be noted that complete viral and bacterial responses may not occur unless TLR signals are received from both epithelial surfaces, where contact first occurs with a pathogen, and the leucocytes in which antigens from the pathogen are subsequently processed [47,48].

Dynamic expression of TLRs

In addition to the diverse tissue and cellular locations of TLRs, it is also now clear that their expression is not static. Not only does the global expression of TLRs gradually alter with age, but a more rapid modulation in levels of TLR mRNA or protein can occur following the exposure of cells to environmental stress, pathogenic microbes or host mediators such as cytokines [49–55]. Notably, functionally distinct TLRs may even influence one another's level of expression. In one model, the activation of neutrophils via TLR4 led to a co-ordinated up-regulation in TLR2 expression on co-incubated endothelial cells via free oxygen radical release [15]. Even following rapid trafficking to the cell surface (see above), TLR location remains fluid within the context of membrane lipid rafts [56]. This allows an ongoing dynamic interaction with other cell surface structures (including other TLRs) to facilitate extension of pattern-recognition repertoires or more complex immune functions [57,58]. The pattern-recognition of lipopolysaccharide (LPS) appears to require the formation of a particularly complex macromolecular structure [59].

Complexity of TLR ligands

Unexpectedly, sophistication in host TLR biology has been found to be matched by complexity in their putative ligands. First, it has been suggested that TLRs not only recognize pathogen-associated molecular patterns (PAMPs) but also host proteins, particularly following tissue injury. Endogenous TLR ligands include extracellular matrix proteins (fibronectin, fibrinogen and hyaluronic acid) and other mediators of host proinflammatory responses such heparin sulphate, beta-defensins or heat-shock proteins [60–65]. Some reports have suggested that endotoxin contamination may be responsible for the activity of some of these ligands, particularly where recombinant products are used [66]. However, even where this confounder has been excluded, endogenous proteins have been found to have a highly sophisticated influence on host immune responses [67]. Conversely, the removal of TLR2 lipopeptide components from LPS by phenol re-extraction has also improved the interpretation of the resulting TLR signature [68]. Although such contamination has been responsible for the repeated misidentification of TLR2 as the pattern-recognition receptor for some species of LPS, this area remains complex and controversial [69–72].

Paradoxically, a complete reliance on highly pure TLR ligands may also not be desirable. This is because whole microbes can trigger quite different TLR response patterns to those induced by their individual component structures [73]. Indeed, different molecular domains of the same structural component from some pathogens can activate distinct TLRs. For example, the C- and N-terminals of the ExoS virulence factor from Pseudomonas induce differential activation of TLRs 2 and 4 [74]. Alternatively, different components of group B streptococci have been shown to induce parallel but distinct activation of different TLR pathways [75]. The net effect of TLR challenge with whole bacteria or fungi is the induction of a complex cumulative gene activation programme within cells such as macrophages and neutrophils [76,77]. This co-ordinates cellular responses as diverse as proinflammatory cytokine release, oxidative microbial killing and apoptosis [78].

Further complexity is introduced by the description of dose-dependent and time-dependent TLR ligand effects. While low doses of LPS provoke the T helper 2 responses seen in some asthma models, higher doses appear to induce DC-driven T helper 1 effects [79]. Not only does infection induce temporally distinct activation of TLRs, but their chronic stimulation can also invoke complex cellular processes [80]. While myeloid cells can become tolerant to repeated stimulation, thereby blunting proinflammatory responses (discussed below), the persistent presence of some ligands can overcome peripheral T cell tolerance [81]. Under other circumstances, repeated TLR ligation may lead to a more generalized disruption of lymphoid tissues and may be utilized by some pathogens to subvert normal innate defences [82,83].

Finally, the precise nature of mammalian TLR–ligand interactions remains less well delineated than in Drosophila[84]. Indeed, it has proved remarkably difficult to confirm direct physical contact. Nevertheless, the extraordinary high fidelity of TLR-mediated pattern-recognition is highly suggestive of sophisticated physical interactions. TLRs have been shown to be capable of differentiating subtle structural variants of individual PAMPs, including different species of LPS or RNA, plasmid DNAs of different methylation status and contrasting lipopeptide stereoisomers [71,72,85–87]. The demonstration of competitive antagonism between TLR ligands provides additional evidence for direct receptor–ligand interaction [88]. Manipulation of ligand structure, analysis with atomic force microscopy and the use of sophisticated transfection models have contributed towards the resolution of this issue [89–92]. In addition, X-ray analysis of the leucine-rich repeat (LRR) sequences in the extracellular portion of TLRs predicts a capacity for sophisticated and direct engagement of microbial PAMPs [93]. The basic framework for TLR pattern-recognition is likely to consist of a horseshoe-shaped solenoid that contains an extensive beta-sheet on its concave surface with numerous ligand-binding insertion points.

High-fidelity TLR-mediated pattern recognition can be exploited by some pathogens. Salmonella adapts itself to the host microenvironment by adjusting the acylation state of its lipid A to down-regulate proinflammatory potential [94]. Pseudomonas species can also modulate the structure of their lipid A, through the addition of ethanolamine, aminoarabinose and palmitate moities, to resist killing and recognition while the presence of hexa-acylated LPS in patients with cystic fibrosis may increase proinflammatory injury [95,96]. Finally, bacteria, fungi and viruses can suppress host responses by induction of interleukin (IL)-10 through TLR-dependent pathways in addition to the selective production of phenotypes resistant to TLR recognition [97–99].

Complexity in TLR signalling pathways

A detailed description of the signalling pathways down-stream from TLRs has been the subject of several reviews and is also outlined in Figs 1 and 2 [100,101]. As with the biology of TLR–ligand interactions, some general themes regarding signal transduction are beginning to emerge. It is now clear that two intracellular adapter proteins called MyD88 and TRIF (TICAM-1) provide a central ‘platform’ for propagating the complex signals derived from TLRs via a direct cytoplasmic association with kinases (e.g. IRAKs, RIPs) and transcription factors [102–106]. Two additional gating adapters called TIRAP (Mal) and TRAM (TICAM-2) appear to be necessary for signal transduction from TLRs 2 and 4 [107–109]. The multiple homophilic interactions that occur between these varying signalling elements are conducted through a system of common modular structural moieties, including TIR and ‘death’ domains [100].

Fig. 1.

Complexity and consequences of Toll-like receptor (TLR)-mediated pattern recognition. Each TLR has an extracellular domain [containing multiple leucine-rich repeats, (LRRs)], and a Toll-interleukin (IL)-1 receptor (TIR) domain. Sophisticated interactions between different TLRs can occur both at the cell surface and at common nodes within signal transduction pathways. Cross-talk between TLRs in distinct cellular locations or on distant cell types can facilitate co-ordinated innate and adaptive immune responses and the modulation of other cellular processes such as apoptosis.

Fig. 2.

Toll-like receptor (TLR) signalling cascade and the locations for negative regulation and microbial subversion of these pathways. Two central adapter proteins MyD88 and TRIF propagate TLR signal transduction by interacting with TLRs via their respective Toll-interleukin (IL)-1 receptor (TIR) components and recruiting down-stream enzymes (e.g. IRAK4, TRAF6) through their ‘death’ domains. Subsequent modulation of transcriptional control elements is less well defined but occurs through linker molecules such as TAB1, TAK1 and TBK1. The powerful proinflammatory properties of Gram-negative lipopolysacharide (LPS) may be explained by TLR4-mediated recruitment of both major adapter proteins. Abbreviations are explained in Table 1 [100–120,126–132].

While some TLRs have the capacity to activate multiple adapters and their associated down-stream signalling pathways, others are more restricted. For example, while TLR4 ligation can result in the recruitment of all of the above adapter molecules, TLR3-mediated induction of NF-κB translocation may be entirely dependent on TRIF–RIP kinase interactions [110]. Further down-stream, while there may be considerable cross-talk between different TLR signal transduction pathways at transcription factors (such as IRF3, IRF7 and NF- κB), there is still the capacity for different TLR ligands to generate highly distinct protein phosphorylation patterns and gene activation programmes [111]. The resulting pathogen-specific cellular responses may explain observed differences in host immunity to Gram-positive and Gram-negative pathogens. In addition, different signalling molecules may be involved in different components of the host response to individual pathogens [112]. The specific kinetic characteristics inherent in different signalling pathways may enable the necessary co-ordination of such responses. For example, IRF-3 activation through TRIF is more rapid following TLR3 than TLR4 ligation, facilitating a rapid beta-interferon response to viral products, while NF-κB activation appears to be under duel phase control from early MyD88-mediated signalling and late MyD88-independent signalling [113].

These complex signalling cascades also offer multiple ‘brake points’ for the negative regulation of innate immune activation (Fig. 2). Some negative regulators such as ST-2 appear to sequester adapter proteins (MyD88 or TIRAP/Mal), limiting their availability for down-stream signalling [114]. Other suppressor proteins, such as Tollip, MyD88s (a splice variant of MyD88), IRAK-M or SIGGR either target phosphorylation/kinase activity of key signalling intermediaries (e.g. IRAK-1, IRAK-4, TRAF-6), or disrupt the modular interaction domains [115–118]. Interestingly, there is the capacity to direct inhibitory molecules to specific transcription factors. For example, the protein SOCS has a powerful suppressive effect on type I interferon production while leaving the proinflammatory activity of NF-κB entirely intact [119]. In contrast, A20, a zinc finger containing cytoplasmic protein, suppresses the NF-κB activation pathway by deubiquitinization at the level of TRAF-6 [120]. It should be noted that the development of tolerance by many cell systems, when exposed to repeated TLR stimulation (often with endotoxin), is thought to be generated through modulation of these signalling controls rather than simple down-regulation of steady-state TLR expression [121–124].

Unsurprisingly, microbial pathogens have developed a multitude of strategies to exploit these signalling systems (Fig. 2). The bacterial species S. typhimurium and Yersinia utilize their type III effector proteins to suppress MAPK and NF-κB activation [125]. However, viruses appear to target the TLR signalling pathway even more directly. While the Vaccinia viral protein A52R resembles a signalling domain within TLRs, providing it with the capacity to subvert host proinflammatory responses through a direct interaction with IRAK2 and TRAF6, persistent intrahepatic hepatitis C infection may be facilitated by cleavage of TRIF [126,127]. Further down-stream Pox viruses also target the I-kappaB kinase complex to disrupt TLR signals to NF-κB and IRF3 [128]. Finally, viruses such as HIV can even directly utilize TLR signalling pathways to increase their intracellular expression [129].

Despite the clear advances in our understanding of TLR signalling, some fundamental questions still remain. Proximally, it appears that while some TLR-mediated effects require internalization of ligand–receptor complexes, others do not [38,130]. Down-stream, the processes linking the signal transduction cascade to transcriptional control elements also remain poorly defined. Although a number of signalling proteins such as RIP2, TAK1, and ECSIT have been identified, their detailed biology has yet to be determined [131–133]. Finally, the relative volume of signalling traffic through different adapter proteins remains unclear. Although all TLRs can operate through MyD88, the use of microarray expression profiling has suggested that following LPS stimulation of murine macrophages only 21·5% of LPS-responsive genes (n = 1055) are dependent on this adapter [100,134]. In addition, other important immune responses, such as the up-regulation of co-stimulatory receptors (CD40, CD80 and CD86), can operate in the absence of MyD88 activity [135].

Complex influence of TLRs in down-stream cellular and immune biology

As well as modulating the innate proinflammatory response through the production of cytokines and chemokines, TLRs are now known to participate in a number of other innate immune processes such as phagocytosis, the production of matrix metalloproteinases, iron sequestration (through induction of lipocalin 2) and defensin production [9,136–138]. However, TLRs are also now known to be crucial in a wide range of other more fundamental cellular processes (Fig. 3). These include actin polymerization, angiogenesis and the induction of apoptosis [139–142]. These specialized roles for TLRs not only underpin their more general role in tissue repair following injury at sites such as the gut and myocardium, but have also revealed new facets to host innate defence [19,21,143,144]. Finally, TLRs may even determine broader pathophysiology, such as the adrenal response to sepsis [145]. Importantly, these wide-ranging TLR-mediated effects are highly co-ordinated. This is demonstrated particularly well by the co-ordinated cell activation, migration and apoptosis of neutrophils and macrophages [15,21,24,47,48,146].

Fig. 3.

Summary of mammalian Toll-like receptor (TLR)-mediated biology. TLRs are now known to be pivotal in both immunological responses and more general cellular homeostasis. Their role in the pathogenesis of diverse disease processes may offer the prospect of new adjuvant immunological therapies in addition to improved vaccine design.

In addition to its role in primary innate responses, differential TLR ligation is now also known to drive the development of subsequent adaptive immunity [36,37,147]. Although TLRs are involved in diverse B cell activity, including isotype switching and T independent cross-priming, it is their role in T cell immunology that dominates [29,30,36]. Through their role in dendritic cell antigen capture, differentiation and migration to lymph nodes (as described above), TLRs designate the T helper response. While TLR4 and 9 are most associated with T helper 1 responses, TLR2 and TLR5 have been found to instigate T helper 2 responses. The adapter protein MyD88 appears to play a particularly complex role in this process. While it has been reported recently that the adaptive response to lipopolysaccharide or Mycobacterium tuberculosis may be entirely intact in MyD88-null mice, others have suggested a prominent role for MyD88 in directing both T helper 1 and T helper 2 responses, depending on which TLR is ligated [134,147–150]. TLRs also direct DC-mediated viral/tumour antigen cross-presentation to cytotoxic T cells (largely TLR3 dependent), and modulate peripheral T cell tolerance and regulation [32,81,151–153]. Most critically, it is the TLR-mediated induction of co-stimulatory molecules (CD80, CD86) on DCs and T cells that determines the crucial capacity for T cell responses to be pathogen-specific [33,36].

Complex roles in disease and therapy

TLRs are complicit in the pathogenesis of a multitude of diseases (Fig. 3). Not only is the absolute level of TLR expression altered in a diverse range of pathological states, but there also appears a direct mechanistic link between TLR activation and the development of conditions such as atherosclerosis [8,154–156]. Indeed, the unexpected link between atherogenesis and sepsis provided by TLRs has been particularly revealing. While a number of microbial products may trigger plaque development, the release of oxidized phospholipids from this process may feedback to down-regulate the inflammation associated with active infection [157,158]. It is unsurprising that while the TLR4 D299G polymorphism increases host susceptibility to sepsis, it decreases the risk of atherosclerotic disease [154,159]. The role of TLRs in sepsis is particularly complex. Although disruption of TLR–ligand interactions can blunt inflammatory injury, the net effect in loss of innate recognition is often harmful to the host [160–162]. Furthermore, while various TLR polymorphisms have been associated with susceptibility to certain diseases or infections, such reports have been inconsistent [163,164].

Therapeutic strategies for both inhibiting and augmenting TLR pathways are being developed [165–167]. It is now possible to achieve TLR activation by specific RNA transfection, thereby avoiding other toxic effects from exogenous ligands [168]. Most impressively, TLR signalling pathways have been utilized to improve vaccine design and efficacy through use of CpG-based adjuvants, minimal epitopes and the bypassing of T cell help [29,30,169]. However, the targeting of TLRs in any therapeutic strategy will not only risk disrupting beneficial host immunity or homeostatic functions, but may also provoke damaging autoimmune phenomena or interfere with the action of concurrent drug therapy [170–172]. In addition, the innate system may circumvent artificial TLR modulation by directing responses through TLR-independent routes [59].

Future investigation of TLR biology

There remain several key difficulties with respect to experimental interpretation in systems designed to investigate TLRs. First, the expression of TLRs either at mRNA or protein levels cannot always be extrapolated to a functional phenotype. For example, NK cells express TLR9 mRNA but may not respond to CpG DNA, while in eosinophils, responses to TLR7 ligands appear to dominate despite the constitutive expression of multiple TLR mRNAs 1, 4, 7, 9 and 10 [23,28]. There are also fundamental differences in TLR biology between different mammals, and even between different cell types from the same organism [13,42,85,173].

Furthermore, TLR expression is so ubiquitous, and their functions so wide-ranging, that isolation of specific receptor-function relationships is highly problematic. This is exemplified by difficulties encountered due to both impure TLR ligands (see above) and heterogeneity within ex-vivo cell populations [24]. The ubiquitous nature of TLR activity may subvert highly specific molecular techniques such as siRNA gene silencing [174]. Finally, many experimental models that have been used to interrogate TLR function rely on gene over-expression or ‘knockout’. It still remains unclear how to prevent the covert biological compensation inherent in such systems from obscuring a result. The application of gene array bioinformatics in combination with advanced mathematical modelling may offer new strategies to analyse the complex biology of TLRs [77,175,176].

Conclusion

Toll-like receptors constitute the best-described group of pattern-recognition receptors described. In mammals, their biology has come full circle: not only do they direct the co-ordinated response of innate and adaptive immunity to infective pathogens, but just as in Drosophilae they are also now known to be involved in more fundamental host processes. The recent stellar advances that have been made within the TLR field have revealed a multitude of new questions. These will have to be answered before these receptors can be exploited successfully and safely with adjuvant therapies and genetic manipulation.