Domain Exchange between Human Toll-like Receptors 1 and 6 Reveals a Region Required for Lipopeptide Discrimination*

Among the 10 human Toll-like receptors (TLRs), TLR2 appears to be unique in its requirement for cooperation with other TLRs, namely TLR1 and TLR6, to mediate cell signaling. Through reconstitution experiments, we have defined more precisely the function of these human TLRs. Human colonic epithelial cells cotransfected with TLR1 and -2 preferentially respond to a synthetic tripalmitoylated bacterial lipopeptide analogue (Pam3CSK4). However, examination of a wide variety of lipopeptide derivatives indicates that recognition by human TLR1 and -2 does not strictly correlate with the number or position of the acyl chains on the modified cysteine residue. Conversely, human TLR2 and -6 exclusively respond to lipopeptides possessing a diacylglycerol group. Most surprisingly, we have found that an R stereoisomer of diacylated macrophage-activating lipopeptide 2 (MALP-2) exclusively activates epithelial cells through TLR6 and -2 but not through TLR1 and -2. These results suggest that the chirality of the central carbon of the diacylglycerol group of these agonists is a structural determinant for human TLR recognition. Examination of chimeric receptors, generated by domain exchange between TLR1 and -6, has revealed that leucine-rich repeats 9–12 of the extracellular domain enable these receptors to discriminate between structurally similar lipopeptides. However, additional chimeric constructs reveal that this region alone is not sufficient to generate receptors that can functionally cooperate with TLR2. Our results support the idea that TLR1 and TLR6 diverged during evolution to differentially recognize natural lipoprotein structures and that this function has been conserved with respect to the human receptors.

The innate immune system is an evolutionarily conserved defense system that comprises the first line of defense in response to an invading pathogen. In vertebrates, the cellular innate immune response generates a variety of pro-inflammatory mediators to combat infection while simultaneously activating the adaptive immune system. The primary triggers of this response are the Toll-like receptors (TLRs), 2 a family of cell surface innate immune sensors that exist in organisms ranging from lower invertebrates to higher mammals. Mammalian TLRs alert the host to the presence of infection through direct recognition of conserved structural components of viruses, bacteria, or fungi. For example, the agonists for TLR3-5 include molecules such as viral doublestranded RNA, enteric bacterial lipopolysaccharide, and bacterial flagellin, respectively (reviewed in Refs. 1 and 2).
In contrast to other TLRs, TLR2 appears to mediate cellular responses to a much wider variety of microbes and microbial components. In accord with this diverse recognition, TLR2-deficient mice exhibit differential outcomes in a vast array of infection models (reviewed in Ref. 3). At least part of the apparent versatility of TLR2 for ligand recognition is afforded by additional TLR family members, namely TLR1 or TLR6, that functionally cooperate with this receptor, an act which presumably extends the ligand repertoire of this receptor (4 -8).
It is well established that TLR2 mediates cellular responses to bacterial lipoproteins (9,10). In bacteria, production of triacyl-lipoproteins occurs through the addition of a diacylglycerol group to the sulfhydryl group of cysteine followed by cleavage of the signal sequence and addition of a third acyl chain to the now free N-terminal group of the cysteine (11). Many bacterial species do not possess the gene encoding the last enzyme in the pathway and therefore produce lipoproteins that lack the third acyl chain. MALP-2, a proinflammatory lipopeptide from Mycoplasma fermentans, is one such diacylated molecule (12). Studies using peritoneal macrophages derived from TLR-deficient mice have revealed that murine TLR1 and TLR6 predominantly mediate responses to triacylated and diacylated lipopeptides, respectively (13)(14)(15). Thus, murine TLR1 and TLR6 functionally cooperate with TLR2 and enable this receptor to discriminate between these structurally similar lipopeptides.
A number of studies have shown differences between murine and human TLR homologues in the recognition of pathogen-derived agonists. For example, murine and human TLR4 completely differ in their ability to respond to lipid IVa, a biosynthetic precursor of LPS lacking two secondary acyl chains, and to taxol, an LPS mimetic (16 -18). Additionally, murine and human TLR9 are found to differ in the sequence context of the CpG motifs that they recognize (19). With respect to TLR2, the murine receptor appears to recognize both tripalmitoylated and trilauroylated lipopeptides, whereas the human receptor only recognizes the former but not the latter compound (20). Together, these studies demonstrate that one cannot make strict inferences about specificity with respect to TLR recognition between mice and humans.
Here we have performed a systematic analysis of the structural features of lipopeptides important for human TLR2/1 versus TLR2/6 discrimination. As different bacterial species express different lipoprotein structures and active lipopeptides can be chemically synthesized, these molecules provide a means to examine the structural features of these agonists essential for recognition by either TLR2/1 or TLR2/6 combinations (21)(22)(23). Our studies have revealed that subtle changes in * This work was supported by start up funds from the University of Illinois and National Institutes of Health Grant AI052344 (to R. I. T.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  lipopeptide structure, including the chirality of the central carbon of the diacylglycerol group, can drastically alter human TLR recognition. Through the generation and testing of chimeric TLR1 and TLR6 constructs, we have defined a central region of the extracellular domain of these receptors required for lipopeptide discrimination. These studies have important implications for therapeutics designed to target septic shock as well as for the rational design of synthetic vaccine adjuvants of which the use of lipoproteins is an active area of investigation (22,24,25).

MATERIALS AND METHODS
Reagents-Racemic mixtures as well as pure R and S isomers of the synthetic bacterial lipoproteins were purchased from EMC Microcollections (Tuebingen, Germany): N-palmitoyl-S- [2,3- ). An additional R isomer of MALP-2 was purchased from Alexis Biochemicals (Lausen, Switzerland). Mouse monoclonal FLAG antibody was obtained from Sigma. Secondary antibody, (R)-phycoerythrin AffiniPure F(abЈ) 2 fragment donkey anti-mouse IgG (H ϩ L) was obtained from Jackson ImmunoResearch (West Grove, PA).
RT-PCR Analysis-Total RNA was extracted from each cell line using RNeasy (Qiagen, Valencia, CA) according to the manufacturer's instructions. Five micrograms of total RNA was used for cDNA synthesis with the SuperScript First-strand Synthesis System for RT-PCR (Invitrogen). PCR was performed with a Mastercycler gradient thermocycler (Eppendorf, Westbury, NY). The primers used for the detection of human TLRs 1-10 and human actin (used as an internal control) are listed in TABLE ONE.
RNase Protection Assay-RNA was isolated from cells and hybridized overnight with a radiolabeled RNA probe to different Toll-like receptors according to the manufacturer's instructions (Riboquant, BD Bio-sciences). Samples were treated with RNase, and protected hybrids were purified and then separated on a 4.75% acrylamide-urea gel followed by autoradiography.
Cloning-Construction of TLR1 and TLR6 Chimeras-All wild type and chimeric TLRs were generated as N-terminal FLAG-tagged constructs within pFLAG-CMV (Sigma). In order to construct chimeras, unique restriction sites were engineered in TLR1 and TLR6 by sitedirected mutagenesis. Primers were used to engineer an NheI site in TLR1 and TLR6 at the end of LRR5, an XbaI site at the end of LRR8, and another XbaI site at the end of LRR12. Primers are listed in TABLE TWO. Prior to creating the chimeric receptors, the constructs with engineered restriction sites were tested for activity in comparison to wild type TLR1 or TLR6. All were found to maintain wild type levels of activity (data not shown).
The LRR9 -12 exchanges were engineered without the use of restriction enzymes by using the technique of overlap extension PCR as described (26). Primers for this set of cloning are listed in TABLE THREE. The rest of the internal domain exchanges, LRR9 -17, LRR6 -12, and LRR6 -17, were constructed by digesting pairs of TLR1 and -6 with the appropriate restriction enzymes and re-ligating the appropriate fragments. All positive chimeric clones were verified by sequencing.
Transient Transfection and Luciferase Assays-SW620 cells were cotransfected with various combinations of TLRs along with a firefly luciferase gene driven by the IL-8 promoter (27) and pRL-null, a Renilla luciferase transfection control (Promega, Madison, WI). Transfections were performed using FuGENE 6 at a lipid to DNA ratio of 4:1 (Roche Applied Science). The total amount of transfected plasmid DNA was equalized by supplementing with empty vector, pFLAG-CMV. Two days post-transfection the cells were stimulated with the indicated agonists for 6 h, and cell lysates were collected. Firefly luciferase and Renilla luciferase enzyme activities were determined using the luciferase assay system (Promega, Madison, WI) according to the manufacturer's instructions. Firefly luciferase activity was normalized to that of Renilla luciferase activities to correct the transfection efficiency. After correcting for transfection efficiency, all values were normalized to those of unstimulated cells transfected with reporters and empty FLAG-CMV vector.

JOURNAL OF BIOLOGICAL CHEMISTRY 36617
Flow Cytometry-By using a calcium chloride transfection method, 293T cells were transfected with a vector expressing a FLAG-tagged TLR construct or empty vector alone in 6-well plates. After 48 h, cells were removed from the plate with PBS (pH 7.4) and immediately transferred to 1.5-ml Eppendorf tubes. Cells were incubated on ice sequentially with anti-FLAG M2 antibody followed by PE-conjugated antimouse. After labeling, the cells were washed, fixed, and analyzed for surface TLR expression using a Coulter Epics XL instrument. Overlays were created using the Summit software program.

RESULTS
SW620 Epithelial Cells Are Deficient in TLR1, -2, and -6 Expression-Toward defining a functional role for human Toll-like receptors 1, 2, and 6, we screened a variety of cell lines for endogenous expression of these receptors by RT-PCR. All RT-PCR primer sequences were searched using BLAST against the NCBI human genome data base to ensure that they were unique to the individual TLR gene under study. Following the screening of a variety of cell lines, we discovered that SW620 cells, a human epithelial carcinoma cell line, expressed no detectable TLR1 or TLR2 by RT-PCR and that TLR6 message was only barely detectable after extended PCR cycles (Fig. 1). SW620 epithelial cells are fully LPS-responsive (28,29), and as expected, the TLR4 message was readily detectable. The expression profile of SW620 epithelial cells sharply contrasts that of THP-1 cells as the latter express all known human TLR family members with the exception of TLR3 (Fig. 1, left panel). All PCR products observed, including the actin control, were dependent upon reverse transcriptase and are therefore not generated from contaminating genomic DNA in the RNA preparations.
To obtain more quantitative TLR expression data, we performed RNase protection experiments on isolated cellular RNA using mixed radiolabeled probes to TLR1-6. (Fig. 1, right panel). THP-1 cells expressed all TLR transcripts except those encoding TLR3. In contrast, only TLR4 transcripts were detected in SW620 epithelial cells. Thus, the RNase protection data are in agreement with that obtained by RT-PCR analysis.
Human TLR1 and TLR6 Enable TLR2 to Discriminate Different Lipopeptide Structures-With the knowledge that SW620 epithelial cells are deficient in expression of TLR1, -2, and -6, we proceeded to examine the ability of these cells to respond to different synthetic lipopeptide agonists upon transient transfection with different combinations of these TLRs. Previous studies using peritoneal macrophages derived from TLR1-or TLR6-deficient mice have shown that TLR1 confers full sensitivity to the triacylated lipopeptide Pam 3 CSK 4 , whereas TLR6 is required for recognition of the diacylated lipopeptide MALP-2 (13,14). Additionally, TLR2-deficient mice are unresponsive to both Pam 3 CSK 4 and MALP-2 demonstrating the essential requirement for this receptor in mediating responses to all lipopeptides.
As shown in Fig. 2A, transfection of either TLR1, -2, or -6 alone does not enable SW620s to respond to Pam 3 CSK 4 . Cotransfection with TLR1

Chemical structure of lipopeptides utilized in this study
* Central chiral carbon. R, S, and/or racemic R/S-isomers tested in this study. ** Chiral carbon always "R"-isomer in this study. and TLR2, but not TLR6 and TLR2, renders SW620s more responsive to the Pam 3 CSK 4 agonist. Conversely, coexpression of TLR6 and TLR2, but not TLR1 and -2, enhances the response of SW620s to MALP-2 (R) (from Alexis Biochemicals) (Fig. 2B). The modest activation observed with cells transfected with TLR2 alone, or the TLR2/1 combination, is probably due to low levels of endogenous TLR6 expression as detected by RT-PCR (Fig. 1). This concept is supported by the fact that PamCys-PamSK 4 , a lipopeptide agonist that activates cells transfected with the TLR2/1 combination and not the TLR2/6 combination, exhibits no ability to activate SW620 epithelial cells transfected with TLR2 alone (Fig.   2C). In conclusion, the data strongly support the current idea that, as observed in murine systems, human TLR2 cooperates with either TLR1 or TLR6 to recognize and discriminate between triacylated and diacylated lipopeptides, respectively.
Both the Acylation State and Chirality of the Central Carbon of the Diacylglycerol Group Are Discriminated by Human TLR1 and TLR6-To define further the structural features of lipopeptides that are important for recognition by human TLR1 versus TLR6, a variety of synthetic lipopeptides was examined for their ability to stimulate cells through these receptors. These synthetic lipopeptidesvariedinnumberandpositionofthepalmitoylgroupswithrespecttoboth the glycerol moiety and the N-terminal cysteine residue (TABLE THREE). We also examined lipopeptides containing two different primary amino acid sequences, either a synthetic peptide sequence CSKKKK or one derived from MALP-2 GNNDESNISFKEK. In addition, the contribution of the chiral configuration, R or S, the central carbon of the diacylglycerol group, was also assessed.
Initial experiments were performed using lipopeptide species that were synthesized as racemic (R/S) mixtures (obtained from EMC Microcollections, Tuebingen, Germany). As shown in Fig. 3, the triacylated lipopeptide Pam 3 CSK 4 (R/S) was a strong TLR2/1 agonist but at higher concentrations exhibited some activity toward TLR2/6. Most surprisingly, the diacylated lipopeptides Pam 2 CSK 4 (R/S) and MALP-2 (R/S) exhibited comparable activities toward both the human TLR2/1 and TLR2/6 pairs ( Fig. 3A and TABLE FOUR). This is in contrast to the earlier data obtained with MALP-2 (R) (Alexis Biochemicals) (see below). PamCSK 4 , a lipopeptide containing a single acyl chain on the N-terminal cysteine residue, was a weak agonist for TLR2/1 (EC 50 Ͼ100 ng/ml). A similar lipopeptide possessing the N-terminal acyl group and only one acyl group on the glycerol moiety, PamCysPamSK 4 , exhibited strong activity toward TLR2/1, comparable with that elicited by Pam 3 CSK 4 (R/S). Neither PamCSK 4 nor PamCysPamSK 4 exhibited any ability to activate SW620s transfected with the TLR2/6 combination.
Synthetic Lipopeptide R Isomers Are More Potent than the Corresponding S Isomers-Previous work using murine macrophages has shown that the R isomer of MALP-2 is 100 times more potent than the S stereoisomer (30). This prompted us to examine the contribution of the chirality of the diacylglycerol group's central carbon to cell activation of human TLRs. In all cases, the R isomer of Pam 3 CSK 4 , Pam 2 CSK 4 , and MALP-2 were the most potent agonists, requiring significantly lower lipopeptide concentrations than the corresponding S isomers (Fig. 3B and TABLE FIVE). Within these general trends, the triacylated lipopeptide Pam 3 CSK 4 was a more potent agonist for TLR2/1, whereas FIGURE 1. TLR expression profile of THP-1 and SW620 cells. Left panel, RT-PCR analysis. Total RNA was isolated from the cells and treated with DNase I. cDNA was prepared using an oligo(dT) primer in the presence (ϩ) or absence (Ϫ) of reverse transcriptase to confirm the absence of contaminating genomic DNA. PCR was performed using primers specific for actin and each human TLR (see TABLE TWO). Right panel, RNase protection assay. Total RNA was isolated from the cells and hybridized overnight with a mixed radiolabeled probe to various TLRs (BD Biosciences). The mixture was then treated with RNases to remove free probe. Protected probe was separated on a 4.75% polyacrylamide gel followed by phosphorimaging. Probes to L32 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were also included as housekeeping genes. Probe (P) and sample (S) are indicated. the diacylated lipopeptides, Pam 2 CSK 4 and MALP-2, were more potent agonists for TLR2/6. MALP-2 (R) from two sources (Alexis Biochemicals and EMC Microcollections) exhibited greater activity toward TLR2/6 than TLR2/1 ( Fig.   3B and TABLE FIVE). However, although MALP-2 (R) from EMC Microcollections exhibited significant activity for TLR2/1, the same compound from Alexis Biochemicals possessed no activity in the context of this receptor pair. The underlying reason for this difference is currently unknown.
LRR9 -12 of TLR1 and TLR6 Are Required for Lipoprotein Discrimination-Having defined lipopeptide specificity with respect to human TLR1 and TLR6, we then wished to define the region of the receptors responsible for mediating the observed lipopeptide discrimination. The extracellular domain of both TLR1 and -6 contains 19 sequential LRR motifs, each of which forms a short ␤-strand and loop structure. In other LRR proteins, the LRR solenoids stack on top of each other to form a higher order structure (reviewed in Ref. 31). In order to preserve this higher order structure, while defining the region of TLR1 and TLR6 required for lipoprotein recognition, chimeras between TLR1 and TLR6 were constructed by reciprocally exchanging increasing portions of the N-terminal extracellular domain between the two receptors.
The ability of the chimeric receptors to mediate responses to different lipoprotein agonists in conjunction with TLR2 was assessed in SW620 epithelial cells. As before, in cotransfections with TLR2, wild type TLR1 enabled SW620s to respond to PamCysPamSK 4 and PamCSK 4 . However, as an increasing number of N-terminal LRRs of TLR6 were replaced with those of TLR1, this receptor gradually lost the ability to mediate responses to these agonists in cooperation with TLR2 (Fig. 4A). When LRR1-8 of TLR1 was replaced with those of TLR6 (construct T6(1-8)/T1), responses to these agonists were still observed. However, when this N-terminal replacement was extended to LRR12 or -17, all activity toward these agonists was lost. The reverse chimeras show that replacement of LRR1-12 of TLR6 with those of TLR1 (construct T1(1-12)/T6) rescued cellular responses to PamCysPamSK 4 and PamCSK 4 , and further replacement with LRR1-17 fully rescued responses to both agonists. As shown in Fig. 4B, the data obtained with Pam 3 CSK 4 mirror

MALP-2-(R)-EMC MALP-2-(S)-EMC
These results indicate that LRR9 -12 of TLR1 are minimally required to mediate recognition of lipopeptides that are specific for this receptor. As shown in Fig. 4B, MALP-2 (R) from Alexis Biochemicals was found to be a highly specific agonist for the human TLR2/6 pair and not for the TLR2/1 pair. Replacement of the 12 N-terminal LRRs of TLR1 with those of TLR6 resulted in a receptor with some activity toward MALP-2 (R), whereas replacement of the 17 N-terminal LRRs fully recovered activity for this agonist (Fig. 4B). Conversely, although replacement of the first 8 LRRs of TLR6 with those of TLR1 greatly attenuated activity for MALP-2 (R), complete loss of activity was not observed until this replacement extended into the first 12 LRRs. Thus, similar to TLR1, LRR9 -12 of TLR6 are minimally required to mediate recognition of MALP-2 (R). These data show that receptor specificity for either TLR1 or TLR6 agonists is dependent upon the presence of LRR9 -12 of the corresponding receptor.
More importantly, all of the receptor chimeras exhibit activity toward at least one lipopeptide agonist demonstrating that none of them are inactive. To determine whether the observed partial losses of activity may reflect reduced receptor cell surface expression levels, we performed flow cytometry of 293T cells following transfection of our receptors, all of which possess a FLAG tag a the N terminus. As shown in Fig.  5, cell surface expression of all the TLR1 chimeric receptors is readily measurable and only slightly less than that observed for the wild type TLR1 receptor. These results suggest that the partial loss of activity exhibited by some of the chimeras is not due to loss of cell surface expression. Similar results were observed with respect to chimeras of TLR6. In summary, these results demonstrate that LRR9 -12 of TLR1 and TLR6 are essential for enabling these receptors to discriminate different lipopeptide structures.
LRR9 -12 Alone Are Not Sufficient for Lipoprotein-mediated Activation-Our studies thus far have shown that LRR9 -12 are required for TLR1-and TLR6-mediated lipopeptide discrimination. To address whether this region is sufficient for this activity, we generated chimeras in which this region alone was exchanged between these two receptors. As shown in Fig. 6, the TLR1 chimeric receptor containing LRR9 -12 of TLR6 was unable to mediate responses to MALP-2 (R). However, extending the exchanged region to incorporate either LRR9 -17 or LRR6 -12 of TLR6 partially rescued MALP-2 (R), again demonstrating that the common region encompassing LRR9 -12 is required for lipopeptide discrimination. A chimera encompassing LRR6 -17 fully restored activity for MALP-2 (R).
It is noteworthy that these latter TLR chimeras are readily expressed on the cell surface (Fig. 8). In conclusion, the internal exchange constructs confirm the requirement for LRR9 -12 for lipopeptide discrimination by TLR1 and -6; however, reciprocal exchange of this region alone does not in itself confer a switch in lipopeptide specificity for either receptor.

DISCUSSION
Here we have examined the structural requirements of lipopeptide recognition by human TLRs from the perspective of both agonist and receptor. As summarized in Fig. 9, human TLR2/1 recognizes racemic mixtures of Pam 3 CSK 4 , Pam 2 CSK 4 , PamCSK 4 , PamCysPamSK 4 , and MALP-2 (EMC Microcollections) and is therefore capable of mediating responses to monoacylated, diacylated, or triacylated lipoproteins. The potent agonist activity of PamCysPamSK 4 for TLR2/1 demonstrates that this receptor pair can fully tolerate loss of a single acyl chain from the diacylglycerol group if an N-terminal acyl chain is present on the cysteine residue. In contrast, the human TLR2/6 receptor pair stringently discriminates between the number and position of palmitoylated groups. The data demonstrate that human TLR2/6 recognition requires that lipopeptides minimally contain a diacylglycerol group. Lipopeptides missing either acyl chain completely lack TLR2/6 activity. That Pam 3 CSK 4 activity is greatly attenuated with respect to TLR2/6, despite possessing a diacylglycerol group, suggests that the N-terminal acyl chain acts to inhibit recognition by this receptor pair possibly through steric interference.
We have found a very good correlation between our results and those previously documented for murine TLRs with respect to lipopeptide FIGURE 4. LRR 9 -12 of TLR1 and TLR6 are required for lipoprotein discrimination. SW620 cells were cotransfected with TLR2 and either a TLR1 or a TLR6 N-terminal chimera, an IL-8-driven luciferase gene, and a Renilla transfection control. Cells were stimulated with 20 ng/ml PamCysPamCSK 4 (R/S) or 1 g/ml PamCSK 4 (A) or 20 ng/ml Pam 3 CSK 4 or 20 ng/ml of MALP-2-(R)-Alexis (B). Luciferase activity was measured as described previously. The data shown are the mean Ϯ S.D. of three replications in a given experiment. Chimeric constructs are indicated below with black areas indicating regions of TLR1 and shaded areas indicating regions of TLR6.
recognition by TLR2/1 versus TLR2/6 pairs. For example, we have observed that Pam 3 CSK 4 is a strong agonist for human TLR2/1 and a very weak agonist for human TLR2/6. This observation is in agreement with murine systems in which, compared with wild type mice, macrophages derived from TLR1-deficient mice possess a much weaker sensitivity toward this triacylated peptide (14). In support of this, mice deficient in either TLR1 or TLR2 produce low titers of antibodies against a lyme disease vaccine based on OspA, a triacylated lipopeptide for Borrelia bungderferi (15). Moreover, in this study, human nonresponders to this vaccine were shown to lack surface expression of TLR1.
Previous studies have shown that removal of one or two acyl chains from the diacylglycerol moiety (PamCysPamSK 4 or PamCSK 4 , respectively) results in dramatic loss of activity for human monocyte-derived dendritic cells (32). We have observed that removal of these acyl chains results in a complete loss TLR2/6-dependent activation. Again, this is in agreement with the observation that the latter compound maintains activity in TLR6-deficient mice (33) and that triacylated lipopeptide  SW620 cells were cotransfected with TLR2 and either TLR1 or TLR6 internal chimera, an IL-8-driven luciferase gene, and a Renilla transfection control. Cells were stimulated with MALP-2 (R) Alexis, and luciferase activity was measured as described previously. The error bars represent the S.D. of three replications in a given experiment. The chimeric constructs are indicated below with black areas indicating regions of TLR1 and shaded areas indicating regions of TLR6. compounds with shortened chains on the diacylglycerol moiety exhibit dramatically reduced activity for cells derived from TLR1-deficient mice (14). With respect to the amine-linked acyl chain, we have found that Pam 2 CSK 4 activates cells through both human TLR2/1 and human TLR2/6. This agonist has also been found to maintain activity for cells derived from TLR6-deficient mice (33). Although we have not system-atically examined the effect of peptide sequence on stimulatory activity, we have found that Pam 2 CSK 4 and MALP-2 from the same commercial source exhibit comparable activity for both human TLR2/1 and human TLR2/6. The idea that the number and position of the acyl chains is a primary determinant of TLR usage is further supported by the finding that addition of a third amine-linked palmitoyl group to MALP-2 (R)  . Schematic diagram of lipopeptide recognition by human TLR2/1 and TLR2/6. The ability of different synthetic lipopeptide structures to activate human TLR2/1 or TLR2/6 is indicated by the thickness of the arrows. The TLR2/6 pair is unresponsive to lipopeptides that contain a third cysteine linked acyl chain, whereas TLR2/1 is unresponsive to MALP-2 (R) (Alexis Biochemicals). The structure of each TLR extracellular domain was modeled based upon the known structures of other LRR proteins. The central LRRs indicated by boxes are those that are known to be involved in lipopeptide recognition based on the published work of Kirschning and co-workers (20) as well as this study. results in activity for murine peritoneal macrophages that is TLR6independent (34).
With respect to the diacylglycerol moiety, the R stereoisomer of MALP-2 has been reported to be at least 100 times more active than the S stereoisomer in mice (30). We have found in every case that R stereoisomers are significantly more active than the corresponding S compounds. Compared with the S isomers, the R isomers of Pam 3 CSK 4 exhibit heightened potency for both TLR2/1 and TLR2/6, while maintaining a preference for the former receptor pair. Both receptor pairs also exhibit greater sensitivity toward the R isomers of diacylated lipopeptides Pam 2 CSK 4 and MALP-2 (from EMC Microcollections) while exhibiting greater activity for the TLR2/6 pair. Most strikingly, we have found that MALP-2 (R) from Alexis Biochemicals is a highly potent agonist for human TLR2/6 while exhibiting no activity for human TLR2/1. This finding is in complete agreement with the observation that TLR6-deficient mice are unable to respond to this agonist (13,34). It is noteworthy that a recently generated antibody against human TLR6 has been shown to inhibit MALP-2-induced activation of human monocytes; however, the chiral form of this agonist is not stated in this study (35). At present we do not know why MALP-2 (R) from EMC Microcollections activates human TLR2/1, whereas the same compound from Alexis Biochemicals does not, but we believe this could reflect differences in methods of chemical synthesis.
A phylogenetic analysis of the human TLRs has shown that of all the TLRs, TLR1 and TLR6 are most homologous to each other with 69% identity and 80% similarity at the amino acid level. This homology, and the finding that these genes are tandemly arranged on chromosome four, strongly suggests that TLR1 and -6 arose from a gene duplication event. The fact that they have since been independently maintained further suggests that they have related but independent functions. Structural analyses of the extracellular domain of TLR1, -2, and -6, based on that of other LRR proteins, predicts that all three receptors contain 19 consecutive LRRs (31). Each 24-amino acid LRR forms a short ␤-strand and loop structure with the hydrophobic leucine side chains pointing toward the inside. In other LRR proteins, the individual repeats stack on top of one another in an ordered scaffold providing a potentially large surface area for protein-protein interactions and/or ligand binding. The recently published crystal structure of TLR3 (Protein Data Bank code 1ZIW) confirms this prediction (36), and the same structure has been successfully modeled for TLR2 (37).
It has been proposed that the large surface area of the TLR extracellular domain underlies the ability of these receptors to respond to many microbial and fungal agonists. We initially approached this hypothesis by generating deletion constructs. However, we found that the majority of these were not expressed on the cell surface; presumably due to improper folding of the receptor because of missing LRR solenoids (data not shown). To overcome this problem, we generated chimeric receptors in which various regions of the ECD were exchanged between human TLR1 and TLR6. This was based on the premise that these receptors are highly homologous yet exhibit different agonist specificity when coexpressed with TLR2. Through this approach we have found that the central region of the extracellular domain, composed of LRR9 -12 of both these receptors, confers lipopeptide discrimination. Most interestingly, this central region is the most divergent domain between these two receptors exhibiting only 45% amino acid sequence similarity.
Many lines of evidence have shown that TLRs mediate cell activation through either receptor homodimers or heterodimers. In fact, the entire extracellular domains between TLR1 and TLR2 can be exchanged, and when coexpressed, the resulting chimeric receptors mediate cellular responses to Pam 3 CSK 4 (38). Thus, although both the extracellular and signaling domains of these two receptors need to be present to mediate responses to lipoprotein, they do not need to be contained within a single contiguous protein. Mutational analysis by Kirschning and coworkers (39) has determined that the first 7 LRRs of TLR2 are not involved in recognition of lipopeptides. Moreover, it has been shown that murine TLR2, and not human TLR2, mediates responses to trilauroylated lipopeptide and that this ability is conferred by LRR7-10 of this receptor (20,39). Taken together with our finding that LRR9 -12 of TLR1 and TLR6 are required for lipopeptide discrimination, these data strongly suggest that TLR heterodimers engage each other with the lipopeptide cooperatively sensed by the receptor pair within the centrally located LRR solenoids (Fig. 9). In this regard, a shared ligandbinding site composed of the concave ␤-sheet surfaces of the receptor pairs would be sterically problematic especially because TLR2 has been shown to be glycosylated at amino acid residues found within this concave surface (37). Thus, it is more likely that ligands are cooperatively bound by the outer loops of the LRR solenoids that make up the sides and convex back of a given receptor. Indeed, this configuration was recently modeled for the cooperative binding of double-stranded RNA to specific centrally located outer loops of TLR3 homodimers (36).
We have found that chimeric receptors of TLR1 and TLR6 generated by reciprocal exchange of LRR9 -12 are unable to mediate lipopeptide responses showing that additional LRRs, above or below this region, are required to functionally reconstitute responses in cooperation with TLR2. One possibility is that, although not directly involved in ligand recognition, additional LRRs outside this region contribute to activation through other mechanisms such as mediating specific contacts with TLR2 to generate a fully activated heterodimeric receptor complex.
A number of additional innate immune receptors have been shown to cooperate with TLRs in mediating responses to bacterial and fungal products (reviewed in Ref. 40). For example, the lectin receptor, Dectin-1, has been shown to enhance responses of TLR2 to fungal cell components, most notably ␤-glucans (41,42). Two independent studies have shown that the myeloid receptor CD14 enhances responses to lipopeptides (43,44) by mediating delivery of lipopeptides to the TLR2 complex (45,46). We have found that coexpression of CD14 with either TLR2/1 or TLR2/6 enhances responses of SW620s to different lipopeptides but does not alter agonist specificity with respect to the TLRs (data now shown). Most recently, a genetic screen in mice has revealed that the scavenger receptor CD36 enables macrophages to sense low concentrations of MALP-2 (R) and lipoteichoic acid but not MALP-2 (S), Pam 2 CSK 4 , Pam 3 CSK 4 , or zymosan particles (47). Taken together with our results, these studies suggest that CD36 is a specific coreceptor for the TLR2/6 pair. At present, we have not investigated the role of CD36 in our human cells lines; however, we have observed that lipoteichoic acid is a very poor agonist for SW620 epithelial cells expressing either TLR2/1 or TLR2/6 receptor pairs (data not shown).
The natural chiral form of the diacylglycerol group of lipoproteins that is produced by bacteria is currently unknown, but it is generally assumed that TLRs would have undergone selection to recognize the natural form or the R stereoisomer. The generality of our results is of considerable interest as, in addition to bacterial lipopeptides, many TLR2 agonists have in common a diacylglycerol motif including Trypanosoma cruzi-derived glycophosphatidylinositol anchors (48), mycobacterial phospholipomannan (49,50), Gram-positive bacterial lipoteichoic acid (51,52), and schistosomal lysophosphatidylserine (53,54). The involvement of either TLR1 or TLR6 in mediating responses to other TLR2 agonists that do not possess diacylglycerol groups, such as yeast ␤-glucans (41,42), neisserial porin (55), Yersinia V antigen (56), and Escherichia coli heat-labile enterotoxins (57), is another worthy area of investigation. These agonists may interact with other parts of the extracellular domains of these receptors and could shed light on their evolutionary divergence following gene duplication.