The repertoire for pattern recognition of pathogens by the innate immune system is defined by cooperation between Toll-like receptors

Materials and Methods

Results and Discussion

An understanding of the role for TLRs in PAMP recognition by macrophages has been facilitated by the use of DN receptors. A natural DN mutation in TLR4 underlies LPS insensitivity in mice (19, 26, 27), whereas a similar mutation in TLR2 suppresses responses to cell wall components of Gram-positive bacteria, mycobacteria, and fungi (3, 4). TLR6 is related to both TLR2 and TLR4, although nothing is known about its ligands (28). We found that mutation of proline-691 to histidine generated a dominant negative form of TLR6 (DN-TLR6), as occurs when the homologous prolines are mutated in TLR2 and TLR4 (3, 29). RAW-TT10 mouse macrophages were transiently transfected with an expression vector encoding a bicistronic mRNA transcript that directed expression of both DN-TLR6 and GFP (3). The cells were stimulated with yeast cell wall particles (zymosan) or bacteria, and the production of TNF-α was measured by intracellular staining and flow cytometry. The expression of GFP and DN-TLR6 from the same mRNA transcript permitted a direct comparison of TNF-α production in DN-TLR6-expressing cells (GFP-positive) with cells in the same sample that did not express the inhibitory protein. Thus, cells that were stimulated with zymosan, and that did not express the inhibitory protein (GFP-negative cells), produced TNF-α (Fig. 1). By contrast, DN-TLR6 inhibited zymosan-stimulated TNF-α production to background levels (Fig. 1). DN-TLR6 also completely inhibited TNF-α production stimulated by the Gram-positive bacterium Staphylococcus aureus, whereas it had no effect on TNF-α production induced by the Gram-negative bacterium Salmonella minnesota. The spectrum of pathogens recognized by TLR6 mirrored the specificity of TLR2 (Fig. 1; ref. 3). In addition, expression of DN-TLR6 or DN-TLR2 completely inhibited TNF-α production elicited by other Gram-positive bacteria, including Streptococcus pneumoniae, Listeria monocytogenes, Lactobacillus crispatus, and Enterococcus faecalis (data not shown). By contrast, expression of DN-TLR4 solely inhibited TNF-α production elicited by Gram-negative bacteria, and had no effect on responses to Gram-positive bacteria (data not shown). The observation that DN-TLR6 and DN-TLR2 independently and completely inhibit TNF-α production elicited by whole Gram-positive bacteria suggests that together, these two receptors coordinate this proinflammatory signaling.

To test whether TLR6 and TLR2 signaling is initiated by identical PAMPs, we stimulated macrophages with the purified bacterial cell wall components, peptidoglycan, the generic synthetic tripalmitoylated lipopeptide (Pam₃CSK₄), and LPS (Fig. 2). To exclude the effect of differences in the sensitivity of the macrophages to the different stimuli, we stimulated the macrophages with the lowest dose that elicited a maximal TNF-α response (Fig. 2A). Peptidoglycan (10 μg/ml), a major surface component of Gram-positive bacteria, strongly induced TNF-α in macrophages (Fig. 2B). Expression of either DN-TLR6 or DN-TLR2 completely inhibited peptidoglycan-induced TNF-α production (Fig. 2B). However, the PAMP specificity of TLR6 and TLR2 was not completely overlapping. Bacterial lipoproteins are a family of proinflammatory cell wall components found in both Gram-positive and Gram-negative bacteria (11, 30). The stimulatory activity of a bacterial lipoprotein resides in its acylated amino terminus, and is mimicked by the Pam₃CSK₄ lipopeptide, which recently has been shown to activate the proinflammatory transcription factor, NF-κB, in a TLR2-dependent manner (8, 9, 11). Pam₃CSK₄ (100 ng/ml) stimulated TNF-α production in macrophages, and expression of DN-TLR2 completely inhibited this response (Fig. 2B). By contrast, expression (over a concentration range of 10³) of DN-TLR6 had no effect on Pam₃CSK₄-induced TNF-α production (Fig. 2B and data not shown). DN-TLR6 also had no effect on TNF-α production induced by lower doses of Pam₃CSK₄ (half-maximal stimulatory dose = 14 ng/ml; Fig. 2A). In addition, DN-TLR6 had no effect on Gram-negative bacterial LPS-induced TNF-α production, which was mediated by TLR4 (Fig. 2B). Together, these results demonstrate that TLR2, TLR4, and TLR6 each have a defined ligand specificity, with the specificities of TLR6 and TLR2 partially overlapping. Furthermore, the results with Pam₃CSK₄ and LPS confirm the specificity of the assay, and exclude the possibility that the effect of DN-TLR6 is to inhibit the activity of TLR2 or TLR4, respectively.

The phagocytosis of particles is coupled to the production of inflammatory mediators. Like TLR2, TLR6 is expressed at the cell surface and is enriched on phagosomes containing zymosan (Fig. 3A; ref. 3). Interestingly, TLR2 and TLR6 both are recruited to phagosomes that contain additional PAMPs such as Gram-positive and Gram-negative bacteria (data not shown), as well as to phagosomes that do not contain PAMPs, such as those containing IgG-opsonized red blood cells (Fig. 3A). This suggests that TLRs are recruited to all phagosomes where they can sample the contents and potentially identify the nature of the pathogen. This would imply that additional TLRs also are recruited to phagosomes, a hypothesis that was supported by the observation that TLR1 was enriched on phagosomes containing either zymosan or IgG-opsonized sheep red blood cells (Fig. 3A).

Next, we examined whether TLR2 and TLR6 physically interact. CHO cells were cotransfected with epitope-tagged TLR6 and TLR2. Immunoprecipitation of V5-tagged TLR6 resulted in coprecipitation of HA-tagged TLR2, indicating that they were physically associated (Fig. 3B). The extent of the association between TLR6 and TLR2 was not influenced by the addition of stimulatory bacterial components (Fig. 3B). The association of TLR2 and TLR6 was mediated by their extracellular domains, because deletion of their cytoplasmic domains did not prevent their interaction (Fig. 3B and data not shown). Conversely, deleting the extracellular domain abrogated the association of TLRs. When the cytoplasmic and transmembrane domains of TLR1, -2, -4, or -6 were fused to the extracellular domain of CD4, they did not associate with wild-type TLR6 (Fig. 3C). Thus, in CHO cells, TLR2 and TLR6 readily associate with each other in a ligand-independent manner through their extracellular domains.

The observation that TLR2 and TLR6 were constitutively associated led us to examine their signaling capacity in CHO cells. Expression of either TLR2 or TLR6 did not lead to the activation of NF-κB, a transcription factor required for proinflammatory signaling (Fig. 4A). However, coexpression of TLR2 and TLR6 strongly activated NF-κB (Fig. 4A), under conditions in which these molecules physically associate (Fig. 3B). This ligand-independent constitutive activity required functional TLR2 and TLR6, because expression of either DN-TLR2 or DN-TLR6 did not result in NF-κB activation (Fig. 4A). In addition, these data demonstrate that DN-TLR6 is a signaling-deficient receptor, as was described for DN-TLR2 and DN-TLR4 (3, 29). Because coexpressed TLR2 and TLR6 were constitutively associated and active in CHO cells, we were unable to study the effects of ligands on TLR association or signaling. Golenbock and coworkers have demonstrated that CHO cells are genetically deficient in TLR2 (31), and they and others also have shown that ectopic expression of TLR2 in these cells permits reconstitution of signaling induced by peptidoglycan and lipopeptide (8, 9, 11, 12). We obtained identical results (Fig. 4A). However, expression of TLR6 in CHO cells failed to reconstitute peptidoglycan-induced signaling (Fig. 4A), further supporting a requirement for TLR2 in TLR6-mediated recognition of this PAMP. As expected, TLR6 also did not reconstitute bacterial lipopeptide-induced signaling in CHO cells (Fig. 4A).

When paired, TLR2 and TLR6 activate NF-κB; alone, however, these receptors do not signal (Fig. 4A). Because TLR2 is capable of signaling in a TLR6-independent manner when stimulated with lipopeptide (Fig. 2), we examined whether it formed a signaling complex with another TLR. Coexpression of TLR1 with TLR2 resulted in robust activation of the NF-κB pathway, whereas TLR1, alone, did not (Fig. 4A). Formation of this functional complex was specific, because coexpression of TLR1 and TLR6 did not signal. Thus, certain combinations of TLRs appear to form functional signaling complexes. This hypothesis was further examined in RAW macrophages, where production of an inflammatory cytokine, TNF-α, can be measured.

The extracellular domain of CD4 is known to promote homodimerization of the molecule (32, 33). Chimeras composed of the extracellular domain of CD4 fused with the transmembrane and intracellular region of TLR4 are constitutively active (6, 34). We generated chimeras in which the extracellular domains of TLR1, TLR2, TLR4, or TLR6 were replaced by the corresponding region of CD4, and we examined their signaling properties in macrophages. As reported (34), CD4-TLR4 induced TNF-α production in macrophages (Fig. 4B). By contrast, neither CD4-TLR1, nor CD4-TLR2, nor CD4-TLR6 induced TNF-α production (Fig. 4B), even though they were expressed at identical levels to CD4-TLR4 on the macrophage surface. These data indicate that the signaling domains of TLR1, TLR2, and TLR6 are not functionally equivalent to that of TLR4. The finding that CD4-TLR2 homodimers are not active suggests that TLR2 needs a signaling partner to activate TNF-α production. When CD4-TLR2 and CD4-TLR6 were coexpressed, TNF-α production was induced (Fig. 4B). Similarly, coexpression of CD4-TLR2 and CD4-TLR1 induced robust production of TNF-α (Fig. 4B). However, there was specificity in this interaction, because coexpression of CD4-TLR1 and CD4-TLR6 did not induce cytokine production (Fig. 4B). Thus, as seen in the CHO system, TLR2 can pair with either TLR6 or TLR1 to establish a functional signaling complex.

We have demonstrated that TLR6 cooperates with TLR2 in identifying a variety of PAMPs. In macrophages, expression of either DN-TLR2 or DN-TLR6 completely inhibited TNF-α production elicited by Gram-positive bacteria or yeast, and both receptors were recruited to phagosomes. In CHO cells, TLR2 physically associated with TLR6, leading to the constitutive activation of NF-κB. In macrophages, coexpression of CD4-TLR2 and CD4-TLR6 activated TNF-α production, whereas expression of either receptor alone was insufficient to induce TNF-α production. TLR2 and TLR6 cooperate in detecting bacterial components such as peptidoglycan, but they do not have identical specificity, as TLR2 detects bacterial lipopeptide without support from TLR6. This latter finding implicates another, as yet unidentified, TLR as a partner for TLR2 in the detection of bacterial lipopeptide, because our data indicate that CD4-TLR2 homodimers are not sufficient to activate TNF-α production in macrophages. We show that TLR2 is capable of forming a functional signaling complex with TLR1, indicating that TLR2 has multiple partners with which it can combine. However, we have not been able to demonstrate that TLR1/TLR2 complexes mediate the detection of lipopeptide (data not shown), suggesting that TLR2 has yet additional partners.

TLR2 has been shown to detect a wide array of PAMPs, and it has been unclear how a single molecule could accommodate such diverse ligands. Our data indicate that TLR2 forms functional complexes with TLR6, and perhaps with other TLRs, and thereby increases its range of detection. It is not yet clear whether TLRs recognize PAMPs directly, as suggested by recent data with TLR4 (19), or through endogenous ligands generated in response to PAMPs (35).

It is particularly interesting that the highly homologous cytoplasmic domains of TLRs are not functionally equivalent with respect to signaling. For example, whereas dimerization of the TLR4 cytoplasmic domain stimulates TNF-α production, similar dimerization of the cytoplasmic domains of TLRs 1, 2, and 6 does not. The cytoplasmic domains of TLR1 and TLR6, however, can form functional pairs with TLR2. One possible implication of this combinatorial diversity is that different TLR pairs may stimulate different signaling pathways. Such differential signaling would allow macrophages to tailor their responses to individual pathogens by producing specific patterns of inflammatory mediators.

The innate immune system has to recognize the large number of PAMPs found in nature. It must both distinguish these structures from self, and discriminate between different pathogens to mount an appropriate immune defense. To date, 10 human TLRs have been identified. Our data suggest that these molecules might form a combinatorial repertoire for innate immune recognition.