Cellular Recognition of Tri-/Di-palmitoylated Peptides Is Independent from a Domain Encompassing the N-terminal Seven Leucine-rich Repeat (LRR)/LRR-like Motifs of TLR2*

Toll-like receptors (TLRs) mediate microbial pattern recognition in vertebrates. A broad variety of agonists has been attributed to TLR2 and three TLRs, TLR4, TLR2, and TLR5, have been demonstrated to bind microbial products. Distinct agonists might interact with different subdomains of the TLR2 extracellular domain. The TLR2 extracellular domain sequence includes 10 canonical leucine-rich repeat (LRR) motifs and 8–10 additional and potentially functionally relevant LRR-like motifs. Thus, the transfection of TLR2 LRR/LRR-like motif deletion constructs in human embryonic kidney 293 cells and primary TLR2-deficient mouse fibroblasts was performed for analysis of the role of the regarding domains in specific pattern recognition. Preparations applied as agonists were highly purified soluble peptidoglycan, lipoteichoic acid, outer surface protein A from Borrelia burgdorferi, synthetic mycoplasmal macrophage-activating lipoprotein-2, tripalmitoyl-cysteinyl-seryl-(lysyl)3-lysine (P3CSK4), dipalmitoyl-CSK4 (P2-CSK4), and monopalmitoyl-CSK4 (PCSK4) as well as lipopolysaccharide and inactivated bacteria. We found that a block of the N-terminal seven LRR/LRR-like motifs was not involved in TLR2-mediated cell activation by P3CSK4 and P2CSK4 ligands mimicking triacylated and diacylated bacterial polypeptides, respectively. In contrast, the integrity of the TLR2 holoprotein was compulsory for effective cellular recognition of other TLR2 agonists applied, including PCSK4. The formation of a functionally relevant subdomain by a region including the N-terminal seven LRR/LRR-like motifs rather than by single LRRs is suggested by our results. They further imply that TLR2 contains multiple binding domains for ligands that may contribute to the characterization of its promiscuous molecular pattern recognition.

Immune responses toward microbes are preceded by their recognition. Pathogen-associated molecular patterns (PAMPs) 1 are cell constituents representing groups of microbes or parasites rather than single species. For instance, lipopolysaccharide (LPS) is a key PAMP of Gram-negative bacteria (1)(2)(3). Mannose-and LPS-binding protein, scavenger receptors, and CD14 as well as members of the toll-like receptor (TLR) family are examples of pattern recognition receptors mediating recognition of microbial products prior to the early phase of host defense (4). Many receptors of the innate immune system are expressed constitutively, thus enabling immediate-early responses (2) including the release of proinflammatory cytokines.
The human protein family of TLRs encompasses 10 members from which TLR4 was the first to be implicated in vertebrate immunity (2,(5)(6)(7)(8). Although in vitro evidence suggested involvement of TLR2 in LPS recognition, an analysis of specific rodent strains proved a role for TLR4 as the prime LPS signal transducer (3,4,9,10). More specifically, the identifications of an inactivating TLR4 point mutation and a TLR4 null allele in C3H/HeJ and B57BL/10ScCr mice, respectively, provided the initial evidence. The phenotype of consequently LPS-resistant gene targeted TLR4 Ϫ/Ϫ mice as well as identification of an inactivating point mutation in TLR2 in Chinese hamsters displaying normal LPS responsiveness (TLR2 d/d /TLR4 ϩ/ϩ ; d ϭ defect) further validate these findings. Peptidoglycan (sPGN) and lipoteichoic acid (LTA) from Gram-positive bacteria, lipoarabinomannan from mycobacteria, neisserial porins, bacterial tripalmitoylated, and mycoplasmal diacylated lipoproteins, as well as yeast products and glycosylphosphatidylinositol-anchored proteins of the protozoa Trypanosoma cruzi are examples of microbial PAMPs eliciting host responses via TLR2 (4). Principal differences in pattern recognition through TLRs such as the distinct necessity for intracellular PAMP uptake in the case of TLR9 have been demonstrated previously (11). In addition, for sPGN as well as a substructure thereof, muramyl dipeptide, the cytoplasmic nucleotide-binding oligomerization domain 2 protein has been identified as a signaling pattern recognition receptor (12,13).
The cellular reactions induced by TLR2 agonists involve activation of nuclear factor (NF)-B and kinases such as p38, c-Jun N-terminal kinase (JNK), extracellular signal regulated kinase (ERK) 1/2, and Akt/protein kinase B. In this regard, TLR2 agonist largely resemble TLR4 agonist effects but differ in some aspects (4,14). In addition, cooperation between TLR1 and TLR2 in recognition of triacylated peptides, as well as cooperation between TLRs 6 and 2 for diacylated mycoplasmal peptides, has been reported previously (15)(16)(17). Whether heteromerization is obligatory for cellular recognition of specific PAMPs such as acylated proteins or whether it is required for all TLR2-mediated effects remains unknown. In addition, specific TLR homodimers/heterodimers might associate with further receptor chains such as CD14, MD-2, and/or MD-1/Rp105 as has been demonstrated for TLR4 (18,19).
Because the extracellular domain of TLR2 is considered to interact with various PAMPs (see above) (15), the question arises whether different parts of the ECD interact with these various ligands. TLR ECD sequences include arrays of leucinerich repeat (LRR) motifs. The LRR consensus sequence encompasses 24 -29 amino acid residues containing a highly conserved core region (LXXLXLXX(N/L)XLXXLXXL) and is implicated in protein-protein interaction (5,20). Crystal structures of multi-LRR domains of proteins such as ribonuclease inhibitor and internalin potentially provide a model of TLR ECD structure (20,21). The ribonuclease inhibitor and internalin crystal structures revealed that the LRR motifs are composed of ␤ strand-helix modules with the ␤ strands being oriented in parallel and positioned in close proximity. Based on these structural considerations, it might be expected that mutations of the extracellular domain of TLR2 could influence susceptibilities to infections. This has been implicated for a polymorphism of the TLR2 intracellular domain (ICD), which correlates with altered functionality of the receptor (22). Accordingly, we generated TLR2 ECD deletion mutants and compared the ability of the resulting protein constructs to mediate recognition of a variety of TLR2-specific PAMPs. We have found that cell activation by distinct TLR2-specific PAMPs requires different subdomains of the TLR2 ECD.
Cell Culture-The human embryonic kidney cell line (HEK) 293 as well as TLR2 Ϫ/Ϫ embryonic fibroblasts (MEFs) were applied for protein overexpression and functional analysis. TLR2 Ϫ/Ϫ mice were kindly provided by Tularik Inc. (South San Francisco, CA) (28). TLR2 Ϫ/Ϫ mouse embryonic fibroblasts (MEFs) were generated from embryos isolated at day 12 post-fertilization. Cells were grown under regular mammalian cell culture conditions in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum (Roche Applied Science), standard antibiotics (Invitrogen, Auckland, Scotland), and 50 M thioglycerol (Sigma). Cells were passaged and expanded for five times. Frozen stocks were thawed and cultured for experiments.
Reporter Gene Assay-3 ϫ 10 4 HEK293 cells or TLR2 Ϫ/Ϫ MEFs were cultured on single wells of 96-well plates. HEK293 cells were cotransfected with an NF-B-recruiting endothelial-leukocyte adhesion molecule-1 (CD62E) promoter luciferase construct and a Rous sarcoma virus promoter ␤-galactosidase reporter plasmid (29) as well as a cytomegalovirus promoter-regulated expression plasmid for human TLR2 by the calcium phosphate precipitation method (10,30). For equilibration of expression levels, DNA amounts used were adjusted and expression levels were analyzed by immunoblot analysis (data not shown). TLR2 Ϫ/Ϫ MEFs were transfected by electroporation at 960 microfarads and 260 mV (Gene Pulser II system, Bio-Rad). 7 h after medium change, preparations of bacterial products or analogues were added to transfected cells for 16 h. Cells were lysed for measurement of luciferase and ␤-galactosidase activities using reagents from Promega (Madison, WI) and PE Biosystems (Bedford, MA). Luciferase activities were related to ␤-galactosidase activities for normalization.
Immunoblot Analysis-HEK293 cells were lysed upon protein overexpression and stimulation for 30 min. Lysates from 2.5 ϫ 10 5 cells or immune complexes prepared from 3 ϫ 10 6 cells for each sample were prepared and analyzed by immunoblot analysis as described previously (31). For analysis of JNK phosphorylation, 3-fold amounts of total lysates (approximately 7.5 ϫ 10 5 cells) were applied. Rabbit polyclonal antisera specific for phosphorylated p38, ERK1/2, JNK, or Akt/protein kinase B were used (Cell Signaling). Specific epitopes were visualized by enhanced chemiluminescence (Western lightning, PerkinElmer Life Sciences).
Intracellular Staining-Pools of transfected HEK293 cell clones were grown on polylysine-coated glass carriers each with eight culture dishes with removable walls (BD Biosciences). Cells were washed with phosphate-buffered saline and incubated with 50 g/ml Alexa Fluor 488conjugated concanavalin A (Molecular Probes) in serum-free Dulbecco's modified Eagle's medium at 4°C for 15 min. The medium was removed, and the cells were washed with phosphate-buffered saline and fixed with 2% formalin for 20 min at room temperature. Cells were washed and blocked with phosphate-buffered saline containing 0.2% saponin and 3% bovine serum albumin for 30 min at room temperature. A first antibody, either anti-FLAG polyclonal rabbit antiserum (3 g/ml) from Sigma or mouse monoclonal anti-human TLR2 2.1 (5 g/ml) provided by Dr. Lien was applied prior to washing after 30 min of incubation. As a second antibody, Alexa Fluor 546-conjugated goat anti-rabbit/mouse IgG (4 g/ml) was applied for 30 min (Molecular Probes) and washed. Cells were sealed in the presence of mounting fluid (Chlamydia pneumoniae micro-IF, Labsystems Oy, Helsinki, Finland) for analysis with a laser-scanning microscope with documentation unit (LSM510, Carl Zeiss, Oberkochen, Germany).
Immunoprecipitation-For immunoprecipitation of transiently overexpressed proteins, 10 g of total expression plasmid DNA for the expression of the respective two proteins was transfected into 3 ϫ 10 6 HEK293 cells seeded on 100-mm dishes by the calcium phosphate precipitation method (30). Mutant constructs and controls applied were overexpressed as FLAG-tagged hybrid proteins while the co-expressed protein was Myc-tagged. FLAG mAb M2 beads were used for precipitation (Sigma) (31). Immunocomplexes were analyzed by application of polyclonal anti-Myc tag antiserum for immunoblot analysis (Santa Cruz Biotechnology, Santa Cruz, CA).

Mutagenesis-Structural information about TLR ECDs is
restricted to sequence-based domain assignment at this stage. Using this method, the presence of 19 LRRs in TLR2 has been described previously (5). In addition to the 10 canonical LRR motifs, we have assigned 10 LRR-like motifs to the TLR2 ECD previously (9). We deleted single or groups of LRR/LRR-like motifs from the TLR2 ECD with the assumption that the removal of entire LRR subdomains would not alter overall protein structure (Fig. 1). We used the resulting constructs for potential identification of domains distinctively involved in cellular PAMP recognition. All of the TLR2 constructs were expressed at similar levels as revealed by anti-FLAG tag immunoblot analysis of total lysates of HEK293 cells following transfection of equal amounts of specific expression plasmid DNA preparations. The sizes observed were in agreement with expected mutant protein sizes. DNA amounts were adjusted for transfection, and expression levels were controlled by immunoblot analysis (data not shown).
Functional Analysis of TLR2 ECD Mutant Constructs-For all of the preparations of bacterial products used, wild-type TLR2 conferred NF-B dependent reporter gene activation and release of IL-8 in HEK293 cells (Tables 1 and 2). The mean values displayed in both tables represent the results of at least three independent experiments. The data in Table 1 were normalized by calculation of the ratio of NF-B-dependent and constitutive reporter gene activity. The significance of all of the values listed in the tables was analyzed through application of the Student's t test for unconnected samples upon relation to vector controls. sPGN and all of the additional non-tripalmitoylated TLR2 agonists used did not induce cellular activation through any of the TLR2 mutant constructs overexpressed at equal levels. Examples are constructs Mut2 to Mut4, which cover three-fourths of the entire LRR-rich region ( Fig. 1 and Tables 1 and 2). Mut1, however, mediated a weak signal upon application of P 3 CSK 4 (Tables 1 and 2). In contrast, MutA and MutB lacking the LRRs adjacent to the Mut1 deletion, single LRR6, or LRRs 6 and 7, respectively, were not functional. This was also true for MutC carrying a deletion that was limited to the C terminus of Mut1 (LRRs 4 and 5) and two constructs carrying C-terminally extended deletions, namely MutD (LRRs 4 -7) and MutE (LRRs 4 -9) ( Fig. 1 and Tables 1 and 2). Deletion of the N-terminal three LRRs abrogated cell activation as well (MutF). Notably, C-terminal extension of a deletion represented by Mut1 resulted in a successively increasing cell activation upon application of P 3 CSK 4 , OspA, and inactivated B. subtilis through MutG (LRRs 1-6) and MutH (LRRs 1-7), respectively ( Fig. 1 and Tables 1 and 2). However, further C-terminal extension of the deletion abrogated cell activation as revealed from overexpression and analysis of the constructs MutI and MutJ, which lack the eight and nine N-terminal LRRs, respectively ( Fig. 1 and Tables 1 and 2). Results from analysis of transiently and stably transfected HEK293 cells were similar (data not shown).
We further analyzed whether increased expression levels of MutH or other mutants would enable cellular recognition of a wider variety of PAMPs. Transfection of 50-fold amounts of expression plasmid for MutH as compared with wild-type TLR2 partially conferred cellular activation by the diacylated peptide R-MALP 2 in a dose-dependent manner. However, activation by application of other TLR2 agonists such as sPGN was barely detectable even upon application of very high amounts of stimulants ( Fig. 2A). None of the other mutants mediated responsiveness following either increased expression or through application of ligands at high concentrations ( Fig. 2A).
To assess the role of single palmitoylations for recognition of tripalmitoylated peptides by TLR2, two P 3 CSK 4 derivatives, P 2 CSK 4 and PCSK 4 , were used to challenge transiently transfected HEK293 cells expressing each of the TLR2 mutants that were generated (Fig. 2B). Wild-type TLR2 conferred NF-B activation upon application of all three derivatives. The constructs MutG and MutH mediated response to P 3 CSK 4 as well as P 2 CSK 4 to different degrees. In the case of MutH, P 2 CSK 4 induced a more robust NF-B activation compared with P 3 CSK 4 (Fig. 2B). PCSK 4 , albeit clearly activating cells expressing wild-type TLR2, did not elicit a significant signal through any of the mutant constructs analyzed (Fig. 2B). Similar results were obtained upon transfection of 50-fold amounts of mutant expression plasmid as compared with wild-type TLR2 plasmid as well as application of very high amounts of stimulants (Fig. 2C). Cotransfection of both wild-type TLR2 and each of the mutant DNA constructs in a ratio of 1:50 was performed for analysis of mutant effects on wild-type TLR2mediated cell activation. TLR2 deletion mutants inhibited wild-type TLR2-mediated cell activation when transfected cells were stimulated with heat-inactivated B. subtilis or P 3 CSK 4 with the exception of MutH (data not shown). Consistent with results from analysis of transfection of HEK293 cells, overex-  with asterisk), the LRR C-terminal (oval), and transmembrane (small rectangle) as well its intracellular domain (large rectangle) from its N terminus (left) to its C terminus (right) in relation to 18 deletion mutants generated thereof is shown. Deleted regions are depicted as dotted lines (⌬ amino acid residues). In wild-type and mutant constructs (with the exception of Muticd lacking a signal peptide), the original signal peptide was replaced by a heterologous signal and an N-terminal FLAG tag peptide.
pression of wild-type human TLR2 restored responsiveness toward LTA as well as P 3 CSK 4 in TLR2 Ϫ/Ϫ MEFs as indicated by NF-B-dependent reporter gene activation (Fig. 3A) and release of IL-6 ( Fig. 3B). In contrast, MutH-mediated cell activation was restricted to P 3 CSK 4 stimulation and further mutants such as MutJ were inactive (Fig. 3, A and B).
Cellular Localization of Wild-type and Mutant TLR2-Pools of six cell clones overexpressing wild-type TLR2, Mut1, MutF, MutG, MutH, or MutJ were analyzed immunocytochemically. Concavalin A was used for staining of the cell membrane. Although control HEK293 cells did not express a FLAG epitope, overexpression of wild-type FLAG-TLR2 revealed the localization of the tagged protein specifically at the cell membrane (Fig. 4). Overexpressed wild-type TLR2 and all of the mutant proteins analyzed were located at the cellular membrane and not within the cell as revealed by comparison with overexpressed FLAG-tagged IL-1 receptor-associated kinase 1, which represents a cytoplasmically located protein ( Fig. 4 and data not shown). We further applied an anti-human TLR2 monoclonal antibody (mAb 2.1) for analysis. The presence of wild-type TLR2, MutH, and MutG at the cell membrane was confirmed. None of the other mutants used was recognized by mAb 2.1 to a detectable degree, although most of them inevitably carried the domain forming the respective epitope in wild-type TLR2 and MutH (data not shown).

DNA Binding of NF-kB and Phosphorylation of Cellular Kinase Akt as Well as That of Mitogen-activated Protein Kinases p38, ERK1/2, and JNK Mediated by TLR2 and Mutant
Receptors-Controls as well as HEK293 cell clone pools stably expressing wild-type TLR2, MutH, or MutJ were subjected to molecular analyses of cell activation. Nuclear extracts as well as total lysates of cells were prepared 2 h or 30 min after the start of stimulation, respectively, with sPGN, P 3 CSK 4 , or PMA. Nuclear extracts were applied to EMSA and total lysates for analyses of cellular kinase Akt as well as mitogen-activated protein kinases p38, ERK1/2, and JNK phosphorylations by immunoblot analysis. EMSA revealed nuclear translocation and binding of NF-B to a canonical NF-B DNA recognition element as well as kinase phosphorylation in all of the clones upon PMA stimulation as compared to unstimulated cells (Fig.  5, A and B). NF-B activation and phosphorylation of kinases analyzed upon stimulation with sPGN depended on the expression of wild-type TLR2 but were absent in all of the other clones tested. P 3 CSK 4 induced activation of NF-B, and phosphorylation of cellular kinases was not restricted to HEK293 cells overexpressing wild-type TLR2 but was also observed in cells overexpressing MutH lacking the N-terminal seven LRRs. Control HEK293 cells as well as cells overexpressing mutant J did not respond to challenge with P 3 CSK 4 as revealed from NF-B EMSA and analysis of kinase phosphorylation (Fig. 5, A and B).
Interaction of Wild-type TLR2 with Mutant TLR2 Constructs-To analyze the role of the TLR2 ECD in homologous or heterologous interaction of TLR2, we performed immunoprecipitation experiments. FLAG-tagged wild-type TLR2, Mut1, Mut2, Mut3, Mut4, MutH, MutJ, MutCK, MutCD, or vector as negative control were cotransfected with Myc-tagged wild-type TLR2 or TLR1. FLAG-tagged mutant as well as wild-type proteins as indicated all coprecipitated with the Myc-tagged wild-    type TLR2 or TLR1 (Fig. 6 and data not shown). Even TLR2 mutant constructs either lacking both the entire LRR-rich domain and the LRR C-terminal C-rich (LRRCT, MutCD) domain or the LRR-rich domain only (MutCK) coprecipitated with the wild-type TLRs 2 and 1 ( Fig. 6 and data not shown).

DISCUSSION
Comparative mutational analysis of mouse and human TLR4 and analysis of TLR5 implicated particular ECD domains in species-specific recognition of LPS modifications and binding of flagellin, respectively (33,34). TLR2 and/or TLR4 binding of glucuronoxylomannan capsules of Cryptococcus neoformans as well as of LPS and sPGN have been reported previously (35)(36)(37). To date, evidence for direct binding of PAMPs to TLRs as well as recognition of a relatively large variety of PAMPs particularly through TLR2 is compelling and may imply the existence of different binding sites of various specific ligands. Here we used mutagenesis of the TLR2 ECD for its functional analysis.
We speculated that in addition to the 10 canonical LRR, 8 -10 LRR-like motifs present in the TLR2 ECD sequence might represent functionally relevant subdomains (5,9). The LRRs are evenly distributed throughout the TLR2 ECD, and we deleted them in four blocks, each containing five motifs (Fig.  1). We then focused on the 10 N-terminal LRRs by successive deletion of internal regions. In total, 14 mutant ECD TLR2 constructs were generated (Fig. 1). Specifically, we asked which of the ligands within a representative group of known agonists were able to induce cell activation through mutant constructs. We identified one class of agonists inducing signaling in the absence of the seven N-terminal LRRs. As such, our results

FIG. 2. TLR2 mutant-mediated stimulus and NF-B-dependent reporter gene activation in HEK293 cells.
Cells were cotransfected with reporter gene constructs as well as CD14 and wild-type or mutant TLR2 expression plasmids. 24 h after transfection started, cells were stimulated with the microbial products or synthetic derivatives as indicated for 16 h and lysed. Cell activation was measured as luciferase reporter activity in the lysates. Experiments were repeated at least twice. (A) dose kinetics of agonists applied to cells transfected with 50-fold amounts of expression plasmids for TLR2 mutants MutH and MutJ as compared to wild-type TLR2 (IL-1␣ as positive control). Amounts of agonists applied increased successively: P 3 CSK 4 , 1 ng/ml, 10 ng/ml, 100 ng/ml, and 1 g/ml; sPGN, 10 ng/ml, 100 ng/ml, 1 g/ml, and 10 g/ml; MALP-2, 10 pg/ml, 0.1 ng/ml, 1 ng/ml, and 10 ng/ml; and IL-1␣, 20 ng/ml. B, P 3 CSK 4 in comparison with P 2 CSK 4 and PCSK 4 induced cell activation through TLR2 and mutants as indicated. C, dose kinetics of agonists applied to cells transfected with 50-fold amounts of expression plasmids for TLR2 mutants MutH and MutJ as compared to wild-type TLR2. Amounts of agonists applied increased successively: P 3 CSK 4 and P 2 CSK 4 , 1 ng/ml, 10 ng/ml, 100 ng/ml, and 1 g/ml; PCSK 4 , 10 ng/ml, 100 ng/ml, 1 g/ml, and 10 g/ml. imply interaction of dipalmitoylated/tripalmitoylated peptides with the C-terminal region of the TLR2 ECD (38).
A trend of mutant activity became evident at equal expression levels of each TLR2-derived construct. Only MutG and MutH conferred cell activation to significant degrees upon challenge with P 3 CSK 4 , P 2 CSK 4, or OspA but not upon application of any of the other PAMPs and analogues as indicated ( Fig. 2 and Tables 1 and 2). Tripalmitoylation is a typical characteristic of bacterial proteins eliciting host responses through TLR2 (4,39). Additionally, increased overexpression in combination with application of increased amounts of synthetic R-MALP-2 rendered MutH-expressing cells responsive, whereas sPGNinduced cell activation was barely detectable (Fig. 2A). None of the mutants encompassing Mut2 to Mut4, MutA to MutF, as well as MutI and MutJ (Fig. 1) mediated activation of the signaling pathways analyzed upon application of any TLR2specific agonist (Tables 1 and 2).
TLR2 dependence of cell activation upon challenge with P 3 CSK 4 , P 2 CSK 4 lacking the amide-linked fatty acid, and PCSK 4 missing the two ester-bound fatty acids was in line with a recent report describing primary immune cell responses to two of these ligands ex vivo as TLR4-independent (39). Although stimulating activities of P 2 CSK 4 and P 3 CSK 4 were almost equal, those of PCSK 4 were only 30 -50% (Fig. 2B). On the other hand, palmitoylation of a peptide at the amino group of a terminal cysteine was sufficient for recognition through TLR2, yet dipalmitoylation increased the stimulatory potential of the peptide considerably. These results suggest that triacylation is not obligatory for TLR2-dependent stimulatory activity. It also raises questions for further aspects of structural properties, which recently have been addressed in the case of a diacylated peptide such as R-MALP-2 (17). Notably, the levels of P 2 CSK 4induced cell activation were similar when mediated through MutH or wild-type TLR2. However, PCSK 4 did not induce cell activation through any of the mutant TLR2 constructs (Fig.  2B). Thus, the additional palmitoylation of PCSK 4 confers independence of cellular recognition from the N-terminal third of the TLR2 LRR-rich domain. This was confirmed upon increased TLR2 mutant expression and amounts of stimulants applied (Fig. 2C). Interestingly, recognition of diacylated R-MALP-2 through MutH was detectable only at increased expression levels (Fig. 2, A and B).
Overexpression of mouse TLR2 (15) as well as that of human TLR2 (Fig. 3, A and B) complemented cellular responsiveness FIG. 4. Cellular localization of TLR2 mutant constructs upon overexpression in HEK293 cells. Pools of HEK293 cell clones stably overexpressing wild-type TLR2 or mutants MutH or MutJ as well as negative controls were analyzed for cellular localization of FLAGtagged proteins. Concavalin a (Con A) was applied for staining of cellular membranes (green), whereas FLAG epitope (␣-Flag) localization was analyzed by immunostaining (red). Overlay of both signals was performed (yellow). Bar in each picture represents distance of 10 m on the original slide.
FIG. 5. Stimulus-dependent cellular activation of TLR2 mutant overexpressing HEK293 clone pools as revealed by NF-B specific EMSA and phospho-ERK1/2, p38, JNK, and Akt-specific immunoblot analysis. Pools of HEK293 cell clones stably overexpressing wild-type TLR2 or mutants MutH or MutJ as well as negative controls were analyzed for NF-B DNA binding activity 2 h upon stimulation by EMSA (A), as well as after 30 min of stimulation by phospho-(p) Erk1/2, p38, JNK, and Akt-specific immunoblot analysis of total cell lysates (B): 1, unstimulated; 2, sPGN (10 g/ml); 3, P 3 CSK 4 (0.1 g/ml); and 4, PMA (1 g/ml). The NF-B-specific signal is marked by an arrow, whereas a signal beneath was nonspecific. Equal protein loading was controlled by application of antibodies specific for the regarding kinases independent from activation as indicated. Only pAkt analysis was controlled by application of an antibody specific for a distinct kinase (B, JNK).
to specific agonists of primary TLR2 Ϫ/Ϫ MEFs. Consistent with the above described results, overexpression of construct MutH in TLR2 Ϫ/Ϫ MEFs mediated P 3 CSK 4 but not LTA or sPGNinduced NF-B-dependent reporter gene activation and release of IL-6 ( Fig. 3, A and B). Our results further demonstrated comprehensive effects of specific LRR deletions on signal transduction in terms of nuclear translocation and DNA binding of NF-B as well as Akt, p38, ERK1/2, and JNK phosphorylation upon challenge with TLR2 agonist P 3 CSK 4 or sPGN (Fig. 5, A  and B).
Localization of the overexpressed TLR2 mutant proteins at the cell membrane was observed ( Fig. 4 and data not shown), while no evidence for cytoplasmic localization was revealed that might have indicated non-functionality (40,41). Although we intended to minimize disruptions of protein structure, the malfunction of mutants such as MutA to MutD might result from disruption of a complex tertiary protein structure. Thus, a larger functionally important subdomain might be formed by the N-terminal third of the TLR2 ECD. This proposed subdomain might be structurally independent from the rest of the TLR2 ECD and possibly either consists of a LRR subgroup or is just coincidentally represented by the respective seven sequence motifs. Deletions within (Mut1 to Mut4, MutA to MutG) and deletions extending beyond this proposed N-terminal seven LRR subdomain (MutI, MutJ) might have caused severe structural changes biasing TLR2 mutant function. In contrast, the removal of the whole domain might have rather preserved the structure and function of the rest of the protein, thus retaining recognition of agonists that do not require its presence. These notions are further supported by results obtained from application of anti-FLAG antibodies for FACS analysis and of a human TLR2-specific monoclonal antibody (mAb T2.1) for cytochemical analysis, suggesting deletion-dependent disruption of wild-type TLR2 structure in most constructs, to a limited extent only in MutG but not in MutH (data not shown).
We found no evidence for involvement of the TLR2 ECD in receptor homodimerization or heteromerization. As revealed by immunoprecipitation upon overexpression, none of the deletions in the TLR2 ECD caused abrogation of interaction with wild-type TLR2 and TLR1 ( Fig. 6 and data not shown). This finding might explain dominant negative effects of mutants (data not shown) through their binding to the wild-type receptors. Because the TLR4 ECD mediates ligand specificity of receptor activation (42), TLR2 ECD mutants may interfere with homodimerization or heteromerization of wild-type TLR2 or TLR1 and prevent appropriate receptor complex activation upon specific extracellular challenge. Interaction between a mutant TLR2 construct lacking the entire ICD (TLR2⌬ICD) with wild-type or mutant TLR2 lacking the entire ECD (MutCD) was evident, whereas only a TLR2⌬ICD construct did not coprecipitate with the tagged cytoplasmic TLR2 domain (TLR2ICD, signal sequence deleted, data not shown). These results further indicate a role of the transmembrane domain rather than that of the TLR ECD in receptor dimerization/ oligomerization as has been proposed also by others previously (33,34) and that might be mediated by unknown proteins within a receptor complex.
Our data imply that the N-terminal 7 of 18 -20 LRRs are not involved in cellular recognition of triacylated and diacylated microbial polypeptides through TLR2. The activity of a respective TLR2 mutant (MutH) was only slightly diminished when compared with wild-type TLR2 (Tables 1 and 2 and Fig. 2B). In accordance with the findings of Mitsuzawa et al. (38) who demonstrated the involvement of the domain Ser-40 to Ile-64 as an sPGN-binding domain, our results suggest involvement of the N-terminal third of the TLR2 ECD in cellular recognition of sPGN and other TLR2 agonists applied (38). Thus, potential binding domains most probably differ for tripalmitoylated and diacylated polypeptides as compared with those of the other TLR2 ligands tested. These conclusions might contribute to elucidation of the molecular basis of TLR-mediated PAMP recognition including possible differences in cell activation triggered by distinct ligands or different doses of one ligand via one receptor (43,44). One possible explanation of our findings could be that there exist yet unknown recognition proteins for LTA and other TLR2 agonists, which differ from a potential P 3 CSK 4 recognition protein. If so, both types of endogenous proteins might function serum independently (39) and mediate cell activation by interacting with distinct regions of the TLR2 ECD. Future PAMP-TLR-binding and structural analyses will further clarify the perspective on pattern recognition receptor function of TLRs.